USF Libraries
USF Digital Collections

An integrated optimal design method for utility power distribution systems

MISSING IMAGE

Material Information

Title:
An integrated optimal design method for utility power distribution systems
Physical Description:
Book
Language:
English
Creator:
Fehr, Ralph E
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Communication-based overcurrent detection
Distribution automation
Distribution substation
Intelligent sectionalizing
Power ring
Dissertations, Academic -- Electrical Engineering -- Doctoral -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: This dissertation presents a comprehensive and integrated design methodology to optimize both the electrical and the economic performance of a utility power distribution system. The proposal is structured to facilitate its adoption and incorporation into the existing utility infrastructure by allowing the various portions of the new design to be implemented gradually into the existing infrastructure without the need to abandon the portions of the existing system that are performing satisfactorily. The topology of the substation plays a vital role in determining both the reliability and the economy of the distribution system. The ring bus topology is offered as the best topology design, and its characteristics as seen at the distribution level are examined. A key concept presented in this dissertation is that the distribution system must be optimized as a whole, not subsystem by subsystem.Optimizing the substation and the primary feeder system separately does not assure an optimal system; in fact, independent design of the two subsystems is likely to produce a non-optimal system laden with operational problems. An integrated approach is essential to assure optimum performance, and the integration process requires an iterative approach. This iterative approach is presented using an example. Innovative changes to the protection strategy of the feeder system can greatly enhance the reliability of the distribution system. The use of communication-based overcurrent detection is presented. This transmission-like scheme, when applied at the distribution level, improves both the reliability and the economy of the system substantially over traditional time-coordinated overcurrent protection philosophies.An application of these proposed innovations leads to the design of a hypothetical system, which is in turn analyzed from both electrical and economic perspectives.
Thesis:
Thesis (Ph.D.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Ralph E. Fehr, III.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 148 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001709511
oclc - 68624101
usfldc doi - E14-SFE0001356
usfldc handle - e14.1356
System ID:
SFS0025676:00001


This item is only available as the following downloads:


Full Text

PAGE 1

An Integrated Optimal Design Method For Utility Power Distribution Systems by Ralph E. Fehr, III, P.E. A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Electrical Engineering Department of Electrical Engineering College of Engineering University of South Florida Co-Major Professor: Kenneth A. Buckle, Ph.D., P.E. Co-Major Professor: Alexander Domijan, Jr., Ph.D. Tapas K. Das, Ph.D. Wilfrido A. Moreno, Ph.D., P.E. Paris H. Wiley, Ph.D., P.E. Date of Approval: October 28, 2005 Keywords: Communication-Based Overcurre nt Detection, Dist ribution Automation, Distribution Substation, Intellig ent Sectionalizi ng, Power Ring Copyright 2005, Ralph E. Fehr, III, P.E.

PAGE 2

Acknowledgements I would like to thank Dr. Alex Domijan for his technical guidance throughout my time in the doctorate program. Also, I thank Drs. Ken Buckle, Tapas Das, Wilfrido Moreno, and Paris Wiley for their direction and assistance with this dissertation. Many colleagues have contributed ideas and in formation to this dissertation, including Messrs. David Darden and Yasodha Ratnasek era of Tampa Electric Company, Mr. Jack Stewart retired from Tampa Electric Company, Mr. John Raksany of the University of Wisconsin at Madison, Mr. Jim Hudock of Ge neral Electric Company, Mr. John Healy of Engineered Power Products, Mr. Somchai Songsi ri of the Provincial Electricity Authority of Thailand, and Professor Joseph Skal a of St. Petersburg [Florida] College. My father and my uncle, Mr. Ralph E. Fe hr, II and Professor Arthur R. Hill, both electrical engineers, have inspired me for as long as I can remember to learn about new things. Without that inspira tion, this work would have neve r been produced. Ms. Gayla Montgomery of the USF Electrical Engine ering department provided tremendous assistance in turning a dream into a degree. And most of all, I thank my wife Kare n and my boys Clayton and William for their understanding, support, and love throughout my time in the doctorate program.

PAGE 3

i Table of Contents List of Tables................................................................................................................. ....iv List of Figures................................................................................................................ .....v Abstract....................................................................................................................... .....viii Introduction................................................................................................................... ......1 Distribution Substation Design...........................................................................................5 Distribution Substation Topology...................................................................................5 Application of the Ring Bus at Distribution Voltages....................................................7 Applying the Ring Bus to Increase Reliability...............................................................9 Applying the Ring Bus to Simplify Operation.............................................................15 Applying the Ring Bus to F acilitate Maintenance........................................................20 Primary Feeder Design.....................................................................................................24 Construction Types.......................................................................................................24 Conductor Selection......................................................................................................27 Equipment Selection.....................................................................................................30 Communication-Based Over current Detection.............................................................31 Temporary Fault on Segment 1.................................................................................38 Temporary Fault on Segment 2.................................................................................39 Permanent Fault on Segment 1.................................................................................41 Permanent Fault on Segment 2.................................................................................41

PAGE 4

ii The Optimization Process.................................................................................................42 Assumptions..................................................................................................................42 Methodology.................................................................................................................45 Electrical Assessment...............................................................................................51 Economic Assessment..............................................................................................57 Consideration of Electrical Losses...............................................................................63 Impact to Reliability.....................................................................................................68 Distribution System Analysis...........................................................................................70 Assessment of Conductor Size.....................................................................................70 Utilization of Circuit Breaker Rating............................................................................72 Main Feeder Analysis...................................................................................................73 Quantification of Main Feeder Reliability....................................................................77 Looped Feeder Reliability Without Intelligent Se ctionalizing.....................................78 Looped Feeder Reliability With Intelligent Sect ionalizing..........................................81 Design Example................................................................................................................8 3 Conclusion..................................................................................................................... ...90 References..................................................................................................................... ....92 Bibliography................................................................................................................... ..94 Appendices..................................................................................................................... ...97 Appendix A: PowerWorld Feeder Model.....................................................................98 Appendix B: Design Example Output........................................................................100 Appendix C: Feeder Cost Estimate Assumptions and Methodology.........................125 Appendix D: Estimation of Energy Losses.................................................................127

PAGE 5

iii Appendix E: Implementation of Communica tion-Based Overcurrent Detection.......131 Appendix F: Assessment of Reliability with Intelligent Sectionalizing.....................135 About the Author...................................................................................................End Page

PAGE 6

iv List of Tables Table 1 – Typical Primary F eeder Construction Costs.....................................................25 Table 2 – AAC Conductor Data for Chosen Selection Set...............................................30 Table 3 – Maximum Servi ce Length Calculation.............................................................48 Table 4 – Effective Service Length Calculation...............................................................50 Table 5 – Summary of Analyzed Feeder Configurations.................................................53 Table 6 – Per-Mile Costs of Various Overhead Feeder Designs......................................62 Table 7 – Representative Equipment Costs......................................................................63 Table 8 – Comparison of 25 kV and 35 kV Equipment Costs..........................................84 Table 9 – Bus Records......................................................................................................98 Table 10 – Branch Records...............................................................................................99 Table 11 – Load Records..................................................................................................99 Table 12 – Resulting Outage Durations for Various Permanent Faults...........................136

PAGE 7

v List of Figures Figure 1 – Radial Bus.........................................................................................................6 Figure 2 – Power Ring......................................................................................................13 Figure 3 – Distribution Substati on Based on Power Ring Topology................................15 Figure 4 – Transfer Bus Topology Using Sw itches to Energize Transfer Bus.................17 Figure 5 – Transfer Bus Topology Using Circu it Breaker to Energize Transfer Bus......19 Figure 6 – Radial Bus with Bypass Switches Around Feeder Breakers...........................21 Figure 7 – Typical Time-Current Characteristic...............................................................33 Figure 8 – ITI Curve (Revised 2000)................................................................................34 Figure 9 – Feeder Configured for Communi cation-Based Overcurrent Protection..........37 Figure 10 – Typical Riser Pole.........................................................................................43 Figure 11 – Feeder Service Area......................................................................................46 Figure 12 – Electrical Assessment....................................................................................54 Figure 13 – Feeder Backup Configuration........................................................................55 Figure 14 – Load Duration Curve.....................................................................................64 Figure 15 – Piecewise Linear Feeder Loading Model......................................................65 Figure 16 – Economic Assessment...................................................................................67 Figure 17 – Layout of Main Feeder and Laterals.............................................................73 Figure 18 – Feeder Configuration.....................................................................................79 Figure 19 – 266.8 kcmil Laurel @ 14.4 kV Base Case...................................................101

PAGE 8

vi Figure 20 – 266.8 kcmil Laurel @ 14.4 kV Contingency Case.......................................102 Figure 21 – 397.5 kcmil Canna @ 14.4 kV Base Case....................................................103 Figure 22 – 397.5 kcmil Canna @ 14.4 kV Contingency Case.......................................104 Figure 23 – 556.5 kcmil Mistletoe @ 14.4 kV Base Case...............................................105 Figure 24 – 556.5 kcmil Mistletoe @ 14.4 kV Contingency Case..................................106 Figure 25 – 715.5 kcmil Violet @ 14.4 kV Base Case....................................................107 Figure 26 – 715.5 kcmil Violet @ 14.4 kV Contingency Case.......................................108 Figure 27 – 900.0 kcmil Cockscomb @ 14.4 kV Base Case...........................................109 Figure 28 – 900.0 kcmil Cockscomb @ 14.4 kV Contingency Case..............................110 Figure 29 – 1192.5 kcmil Hawthor n @ 14.4 kV Base Case............................................111 Figure 30 – 1192.5 kcmil Hawthorn @ 14.4 kV Contingency Case...............................112 Figure 31 – 266.8 kcmil Laurel @ 24.9 kV Base Case...................................................113 Figure 32 – 266.8 kcmil Laurel @ 24.9 kV Contingency Case.......................................114 Figure 33 – 397.5 kcmil Canna @ 24.9 kV Base Case....................................................115 Figure 34 – 397.5 kcmil Canna @ 24.9 kV Contingency Case.......................................116 Figure 35 – 556.5 kcmil Mistletoe @ 24.9 kV Base Case...............................................117 Figure 36 – 556.5 kcmil Mistletoe @ 24.9 kV Contingency Case..................................118 Figure 37 – 715.5 kcmil Violet @ 24.9 kV Base Case....................................................119 Figure 38 – 715.5 kcmil Violet @ 24.9 kV Contingency Case.......................................120 Figure 39 – 900.0 kcmil Cockscomb @ 24.9 kV Base Case...........................................121 Figure 40 – 900.0 kcmil Cockscomb @ 24.9 kV Contingency Case..............................122 Figure 41 – 1192.5 kcmil Hawthor n @ 24.9 kV Base Case............................................123 Figure 42 – 1192.5 kcmil Hawthorn @ 24.9 kV Contingency Case...............................124

PAGE 9

vii Figure 43 – Hourly Feeder Loading Data.......................................................................127 Figure 44 – Load Duration Curve...................................................................................128 Figure 45 – Average Load Level Durations....................................................................130 Figure 46 – Radial Feeder with One Secti onalizing Recloser a nd Backup Recloser.....131 Figure 47 – Radial Feeder with n Sectionalizing Recloser s and Backup Recloser........131 Figure 48 – Power Ring Logic for Communicat ion-Based Overcurrent Detection.......132 Figure 49 – Single Sectionalizing Reclos er Logic for Commu nication-Based Overcurrent Detection..................................................................................133 Figure 50 – Sectionalizing R ecloser 1 through n–1 Logi c for Communication-Based Overcurrent Detection.................................................................................134 Figure 51 – Feeder Sectionalized by N Reclosers..........................................................136

PAGE 10

viii An Integrated Optimal Design Method For Utility Power Distribution Systems Ralph E. Fehr, III, P.E. ABSTRACT This dissertation presents a comprehens ive and integrated design methodology to optimize both the electrical a nd the economic performance of a utility power distribution system. The proposal is structured to facili tate its adoption and in corporation into the existing utility infrastructure by allowing th e various portions of the new design to be implemented gradually into the existing infras tructure without the need to abandon the portions of the existing system th at are performing satisfactorily. The topology of the substation plays a vital ro le in determining bot h the reliability and the economy of the distribution system. The ring bus topology is offered as the best topology design, and its characteri stics as seen at the distribution level are examined. A key concept presented in this dissertation is that the distribution system must be optimized as a whole, not subsystem by subs ystem. Optimizing the substation and the primary feeder system separately does not assure an optimal system; in fact, independent design of the two subsystems is likely to produce a non-optimal system laden with operational problems. An integrated appr oach is essential to assure optimum

PAGE 11

ix performance, and the integration process requ ires an iterative approach. This iterative approach is presented using an example. Innovative changes to the protect ion strategy of the feeder sy stem can greatly enhance the reliability of the distribution system. The use of communication-based overcurrent detection is presented. This transmission-like scheme, when applied at the distribution level, improves both the reliability and the economy of the system substantially over traditional time-coordinated overcurrent protection philosophies. An application of these proposed innovati ons leads to the design of a hypothetical system, which is in turn analyzed from both electrical and economic perspectives.

PAGE 12

1 Introduction Since their inception in the late 19th century, power systems have been an instrumental component of contemporary life. It is im possible to imagine modern existence without the conveniences afforded by electricity. In many cases, these conveniences which our ancestors were able to do without, have become necessities for us. Fields such as health care, manufacturing, transportation, and reta il sales would be vastly different today without that precious comm odity we call electricity. As the uses for electricity became more num erous and diverse, so did the requirements placed on the power system by its users. At the start of the 20th century when electricity was used primarily for street lighting and indus trial (motor) applicati ons, the purity of the sinusoidal voltages provided by th e power company was of little concern. It was simply not important, since most of th e appliances in use at the time functioned equally as well when supplied by high-quality power as they did when fed from a source laden with voltage sags and swells, transient pert urbations, and large harmonic content. Development of new electrical appliances for ced the power system to evolve. Parameters that were previously of little or no concern became issues of paramount importance. The need to improve steady-state voltage c ontrol was among the first of many such adaptations required of the power system. As time went on, the number and complexity

PAGE 13

2 of the required adaptations grew, often at an expeditious rate. Unfortunately, most of the changes imposed on the power system were addressed indi vidually. This approach sometimes caused one issue to worsen as a nother was improved. Mo re often, it led to a solution that was perhaps sati sfactory, but seldom optimal. The power system infrastructure in use today in the United States is primarily the result of an architecture developed in the 1940s a nd 1950s, with an array of modernizations sprinkled throughout to address i ssues raised by the evolution of the uses of electricity and to raise efficiencies as much as prac tically possible (Willis, Welch, and Schreiber 2001, P. 2). The resulting system generally meets the major requirements placed upon it, but often does so in a suboptimal way. Incr easing the performance of the power delivery system, both technically and economically, would provide many benefits both to the utility companies who own and operate it and to the customer s who rely on it for a broad spectrum of uses. But this increase in perf ormance must be approached in a methodical way to yield an optimum result. The method of approach must be integrat ed, striving to optimize the power delivery system as a whole, not part by part. If one attempts to optimize the subtransmission system, the distribution substations, and the pr imary feeders separately the entire system will not be optimized; in fact, the entire syst em may not even functi on as intended (Willis 1997, P. 492). This is because the various components of a power delivery system are very interrelated. The design of the distribu tion substation influences the design of the primary feeders, and vice versa. Because of these interrelationships, an integrated

PAGE 14

3 approach must be taken when attempting to optimize the system as a whole. This was seldom the case in the past. Usually, indivi dual departments within a utility company would work on specific components of the power delivery system with little concern for the interrelated components. This approach led to a suboptimal system, which may have been acceptable at one time, but is sometimes not able to meet today’s requirements, and will undoubtedly fall short as future demands on the power system continue to grow in both number and complexity. Acknowledging and addressing practical issues is also critical to the success of any optimization effort. Equipment availability, i ndustry standards, and safety codes must be considered when proposing any system modifi cations. All major ch anges in layout or design of the power delivery system suggested by this or any other research must, in order to be seriously considered by the utility industry, be proposed in such a way that the changes can be gradually integrated into the existing infrastructure. This would allow a progressive transition from the present power delivery system to the proposed methodology. Without providing for a gradual tr ansition, it is unlikely that any new method, regardless of its benefits, would be accepted by the utility industry. As the existing power delivery infrastructure ages and becomes stressed beyond its capabilities, the timing for introducing a fresh perspective to deliver electricity to customers may never be better. The proposed infrastructure must also be able to incorporate new technologies and methods, which may be available today but are perhaps not yet mature enough for

PAGE 15

4 widespread use. Due to the immense e xpense and complex logistics involved in constructing a utility distribution system, a ny infrastructure built today will more than likely be in use fifty years or more in the fu ture. We have no idea what technologies will be available that far in the future, but our in frastructure must be flexible and adaptable enough to be viable that far into the future and beyond. The distribution substation serves as the sour ce for each distribution feeder. To assure adequate performance of the distribution system, the distributi on substation and its primary feeder system must be designed to provide the operating characteristics defined by the utility customers. The performance of the entire system should be optimized, and enough flexibility should be inherent to the sy stem to allow the future incorporation of new technologies and methods. Two aspects of the substation design will be explored in this dissertation: the substation topology a nd the selection and si zing of the substation equipment. A primary feeder system optima lly compatible with the substation design will also be developed, such that the substa tion-feeder combination achieves the greatest practical technical and economic performance wh ile allowing flexibility and adaptability for the incorporation of futu re technologies and methods.

PAGE 16

5 Distribution Substation Design Distribution Substation Topology The distribution substation provides the in terface between the high-voltage utility transmission system and the medium-voltage distribution feeder sy stem. It typically consists of at least one pow er transformer, highand medium-voltage bus work, highand medium-voltage protective devices (i.e., circuit breakers), a nd various auxiliary devices to support these major components. While many distribution substation topologies exist, the radial bus configuration, or a variant of it, is the standard topology for most distribution substations Radial buses have an attractive characteristic: only one circuit breaker is required per branch term inated on the bus. Minimizing the number of circuit breakers keeps the construction cost of the substation minimal, as circuit breakers tend to be costly components. Radial buses, although common in distribution substations, ar e seldom implemented at transmission voltages for numerous reasons th at adversely impact system reliability (Fehr 2002, P. 2). The most obvious system impact is caused by a bus fault. This scenario requires the tripping of every circuit breaker on the radial bus, which results in the deenergization of the entire bus. Similarly, the failure of a circuit breaker to trip during a line fault (breaker failure) also requires every breaker to trip, thus cl earing the entire bus.

PAGE 17

6 And, of course, a transformer fault either trip s the main breaker, if one is used, or all feeder breakers otherwise, thereby de-e nergizing the entire distribution bus. a) Bus Fault Clearing b) Line Fault Clearing with Breaker Failure Figure 1 – Radial Bus Clearing a transmission-voltage bus is unaccepta ble in virtually all cases because of the number of branches that are re moved from the network in the process. Most transmission systems are designed for singleor double-co ntingency operation, mean ing that a system with n branches must perform within expect ations (acceptable volta ges and flows) with 2. All Breakers Must Trip To Clear Fault 1. Bus Fault Occurs Trip To Clear Fault 3. All Non-Failed Breakers Must 1. Line Fault Occurs 2. Breaker Fails

PAGE 18

7 n–1 branches in service (single contingency) or with n–2 branches in service (double contingency). Clearing an entire bus remove s more than one or two branches from the network, thus necessitating mo re stringent (and less economical) planning criteria. The need to plan for higher-order continge ncies than n–2 can be mitigated by utilizing a substation bus topology that is more robust th an the radial bus. Such a topology would not require more than one unfaulted branch to be removed from the network during bus fault or breaker failure conditions. Several t opologies meet these requirements, but most require the use of more than one circuit br eaker per branch, such as the breaker-and-ahalf topology, which requires th ree circuit breakers for each pair of branches. The requirement for additional circuit breakers substantially increases the cost of the substation, so minimizing the number of circ uit breakers required in a substation is a fundamental goal of the substation design engineer. One topology, however, meets the above requirements of not de-energizing more than one unfaulted branch under all realistic contingency scenarios while mainta ining the economy of one circuit breaker per branch like the radial bus. This topology is the ring bus. Application of the Ring Bus at Distribution Voltages Power system engineers have developed a wi de variety of technical arguments over the years to justify using the ri ng bus in transmission substatio ns. These arguments range from increasing reliability to simplifying operation to facilitating maintenance (Fehr 2004, P. 2). All of these reasons form a sound rationale for applicatio n of the ring bus at

PAGE 19

8 transmission voltages. So why would thes e same arguments not apply at distribution voltages? The reasons used to justify applying the ring bus in transmission substations provide even more benefit when the ring bus is implemented at distribution voltages. This is because when applied on the distribution system, the reliability improvement measure (the ring bus) is being implemented at a point in the syst em closer to the customer than if it were applied on the transmission system (Brown 2002, P. 279). It is the customer who perceives power quality issues. Improving the quality of the system far from the customer will not be perceived by the customer to the same degree as a quality improvement made closer to the customer. So, power quality improvements will be more pronounced as measures are taken at points of the system closer to the customer. Operating flexibility and facilitation of main tenance also become more critical at the lower voltages, due to the lack of redundancy on the radial distribution system. By using a ring bus topology in the distribution substation, the network / radial interface is moved closer to the customer, thereby improving op erating flexibility by keeping a larger portion of the power system in a non-radial configuration. These factors make the selection of the radial bus as the de-facto st andard for distribution substations quite ironic when more robust options such as the ring bus are available at a comparable cost. This irony will be explored in the following sections.

PAGE 20

9 Applying the Ring Bus to Increase Reliability The term “reliability” can be very confusing and misleading if not adequately defined. In the general sense, the reliability of a power system includes both the availability of the energy supply as well as the quality of the power provided by the system (Brown 2002, P. 46). Since power quality requirements vary c onsiderably from customer to customer, and because utility system components do relativel y little to degrade power quality, this dissertation will functionally separate system availability from power quality, and will refer to the system availability component as “reliability.” Reliability will be measured in customer out age minutes. This raw value is perhaps the most comprehensive single measurement of reli ability for several reasons. The reliability of a power system is a perspective of the cust omer. It is the customers’ requirements and expectations that must be met for the system to be viewed as “relia ble.” Therefore, it follows that reliability measurements should be made from the customers’ viewpoint. Attempts to average customer outage minutes over portions of the system tend to dilute the magnitude of the outage as perceived by the customer, so raw customer outage minutes will be used to quantify reliability in this dissertation. Most reliability issues originate on the distribution system (Willis 1997, P. 155). There are several reasons for this, including the failure rates of distribution-class equipment compared to those of transmission-class eq uipment, substantially lower basic impulse level (BIL) ratings for distribution component s than for comparable transmission-class components, and circuit exposure (many more ci rcuit-miles of distribution circuits exist

PAGE 21

10 compared to transmission, thereby increasing the probability of a distribution outage). But perhaps the largest influence to reliability is the fact that the distribution system is usually radial whereas the transmission syst em is not. The networked nature of the transmission system boosts the reliability of that part of the system tremendously. While the distribution system remains mostly ra dial, use of the ring bus topology in the distribution substation moves the network-radial interface one step closer to the customer. In a ring bus distribution environment, only the feeders themselves are radial; the distribution substations are not. Consider a four-feeder di stribution substation servi ng a 50 MVA load. Assuming uniform circuit loading, each feeder serves 50 4 = 12.5 MVA of load. If the substation is configured as a radial bus, a bus fault, failu re of a feeder breaker, or transformer failure clears the bus, thereby de-energ izing the entire 50 MVA of load. When the substation is configured in a ring bus topology, no single event can clear the entire bus (Gnen 1986, P. 187). A bus fault would cause either tw o or three circuit breakers to trip, depending on the exact location of the fault. If only two breakers trip, one branch connected to the ri ng bus would be de-energized. If that branch is a feeder, the load lost is 12.5 MVA. Tripping thr ee breakers would de-e nergize two adjacent branches on the ring bus. If both branches were feeders, the load lost would total 25 MVA. Even this worst-case scenario keeps 50% of the substation load energized. This improvement in availability over the radial bus where any bus fault de-energizes the entire bus is especially significant where eith er momentary or sustai ned interruptions are

PAGE 22

11 of concern. And by carefully selecting wh ere on the ring bus various circuits are terminated, even more significant improvement s in reliability can be realized. For example, a feeder and the backup feeder for that feeder should not be terminated in adjacent ring bus positions, so that a single contingency cannot de-energize both. One of the branches lost when a pair (or three) circuit breaker s trip could be a transformer that supplies the ring bus. If that transformer is the only source to the ring bus, all loads supplied by the substation would be de-energ ized, but that conti ngency can be easily remedied. A second source, which is treated as simp ly another branch, can be added to the distribution ring bus for increase d reliability. With a second transformer, the loss of one transformer does not de-energize the bus. By carefully designing the ring bus, breaker failure would never de-energize both sour ces (the two transformers should not be terminated in adjacent ring bus positions). Use of more than one transformer in a subs tation is also common with the radial bus topology. Typically, each tr ansformer supplies one radi al bus, and the buses are connected together with normally-open tie breaker. When a transformer becomes disconnected from the distributi on bus, some means of source transfer must be executed. This is usually a break-before-make, or dead tr ansfer. Such a transfer results in a brief interruption of service to all f eeders on the bus normally served by the failed transformer. Although the transfer can be fair ly quick (a matter of seconds) and can be automatically

PAGE 23

12 implemented, a dead bus transfer can be quite objectionable when high power quality expectations exist. Unlike with the radial bus, a second source supplying a ring bus does not require a tie breaker or other additional components – ju st one more circuit breaker position in the ring bus. Both transformers would be operated in parallel, eliminating the need for a source transfer scheme as well as the accompanying momentary interruption. The parallel transformer operation increases fau lt current on the distri bution system. While higher fault current may require higher equi pment interrupting ratings, the magnitude of fault current can be kept manageable by carefully specifying transformer impedances. Higher fault current availability also improves system performance during the starting of large motors. This benefit can be substantial, especially when voltage dip issues are of concern, as is the case with many power distribution systems. When furnished with a second source, especi ally a source supplied from a transmission source independent of the one supplying the first source, the ring bus provides a very highly reliable distribution bus Total loss of s upply to the substation becomes a remote possibility when two independe nt sources are provided. Since the two transformers supplying the ring bus could be paralleled acr oss the transmission system, care must be exercised to assure the transformer loadings will be comparable and no excessive flows exist through one transformer, a portion of the ring bus, and back to the transmission system through the second transformer. This through-flow scenario would be more likely to occur during outages on the transmission system. Load flow analysis can predict

PAGE 24

13 Feeders Distribution Transmission System (Source) Transmission System (Source) POWER RING Feeders Distribution potential operating problems, and those problems can be resolved by judiciously specifying the impedances and de-energized tap settings of the transformers. Providing more than one sour ce to the distribution subs tation bus is one means of substantially increasing the re liability of the substation. Ring buses with two or more sources can be thought of as “ power rings” (Fehr 2004, P. 3). Figure 2 – Power Ring A power ring provides a reliability level comp arable to that of a transmission-voltage substation. As with the transm ission substation, loss of the entire bus is an unlikely scenario. Proper equipment sizing will allow operation of the substation with a single source. Outage restoration with a power ring topology is facilitated by the operations benefits described in the next section. The dual combinati on of higher availability and faster restoration in the event of a serv ice interruption makes the power ring topology very attractive in applications where high reliability is paramount.

PAGE 25

14 A detailed schematic diagram of a six-element power ring is shown in Fig. 3 on page 15. The circuits leaving the ring to the left a nd right, connected to th e ring by motor-operated line disconnect switches LD1 and LD4, ar e sources from substation transformers. Adjacent to the termination of each of these circuits on th e ring are a set of three bus potential transformers (PT1 and PT4) for voltage sensing. Each of the feeder circuits are also connected to the ring by motor-opera ted line disconnect switches (LD2, LD3, LD5, and LD6). The motor operators on the line disconnect switches allow both automatic and remote operation of the switches. Automatic opera tion of the line disconnect switches is necessary for intelligent sectionalizing the concept around which the feeder system protection methodology proposed by this dissertation is centered. Depending on the application, remote operation capability similar to that provided for a transmission-level facility by the utility’s System Control a nd Data Acquisition (SCADA) system may be desired to expedite switching and reconfigura tion procedures. Since all line disconnect switches are fitted with motor operators, remote operation can be easily implemented. All motor-operated disconnects are three-pole devices incapable of interrupting load. They can only be opened when no current is flowing through them. Each circuit breaker in the ring is equippe d with current transformers (CTs) on each bushing (CT#A and CT#B) for current measur ement. Breaker disconnect switches (BD#A and BD#B) are also instal led on either side of each breaker to allow isolation of the breaker from the bus for maintenance. These switches are single-pole, hook-stick

PAGE 26

15 operated devices incapable of interrupting lo ad. Since the breaker disconnect switches are only opened to perform maintenance on th e breakers, there in no need for remote operation. M LD6 CT6B BD6B CT6A BKR6MBD6A LD1 CT1B CT1A BD1A BD1B BKR1MLD2 M LD5 M M LD4 CT2B CT2A BD2ABD2B BKR2 LD3 CT3B CT3A BD3ABD3B BKR3 CT5B BD5B CT5A BKR5 BD5A CT4B BD4B CT4A BKR4 BD4A PT1PT4 Figure 3 – Distribution Substation Based on Power Ring Topology Applying the Ring Bus to Simplify Operation Due to the operating inflexibility inherent to the radial bus topology, many utilities opt to add a transfer bus and additional breakers and/ or switches to facili tate operation. While the transfer bus considerably increases the operating flexibilit y (and the cost) of the radial bus topology, it also complicates the operating procedures for the substation substantially. And, unfortunate ly, while enhancements can be made to the radial bus topology to increase operating flexibility, these enhancements do nothing to improve reliability.

PAGE 27

16 The transfer bus makes it possible to uninten tionally parallel two circuit breakers for a prolonged period of time. Paralleling two breakers renders the feeder protection ineffective, or at best, unpredictable. Th e system should remain in the state where two feeder breakers are paralleled only for short periods of ti me, such as during switching procedures. This practice makes the risks introduced by parallelin g circuit breakers acceptable by reducing the exposure time of this precarious configurat ion to a very short time (minutes). The possibility of improper operation and conf iguration of the transfer bus switches also exists. The switches used in the substation ar e typically incapable of breaking current. They can only be opened when no current is flowing through them. Attempting to break current with a switch not designed to do so ca n result in the failure of the switch and injury to the personnel operating the switch, as the failure of the switch can be quite violent. While procedures can be established to prevent improper operation of switches, the possibility always exists for the switches to be used improperly, and the consequences of improper operation can be catastrophic. Technical issues are also created by the incorp oration of the transfer bus, such as how to modify protection settings wh en a feeder is supplied from the transfer bus. Using switches to connect a source to the transfer bus requires that the protection for the feeder being supplied from the transfer bus be provi ded by the feeder breaker that energizes the transfer bus. Consider the tran sfer bus topology shown in Fig. 4.

PAGE 28

17 BD2B BD2A BKR2 Main Bus Transfer Bus BD3B BD3A BKR3 BD1B BD1A BKR1 BD4B BD4A BKR4 FDR1FDR2 FDR3 FDR4 TB1TB2 TB3 TB4 Source Figure 4 – Transfer Bus Topology Using Switches to Energize Transfer Bus Let us assume the objective of isolating BKR1 for maintenance. Initially, all transfer bus switches (TB#) are open and the tr ansfer bus is de-energized. Each feeder is supplied by its feeder breaker (BKR#) in a radial conf iguration. To isolate BKR1, FDR1 must be supplied by another source. Examining the a mmeters on feeders 2, 3, and 4 reveals that FDR4 has the lowest loading of the three. If the FDR1 loading plus the FDR4 loading is less than the continuous curr ent rating of BKR4, then BKR4 is an acceptable backup source for FDR1. If the combined loading of FDR1 and FDR4 exceed the rating of BKR4, then one or both of the feeders must be offloaded by field switching (transferring the load served by sections of the feeders to a source other than BKR1 or BKR4 by reconfiguring switches on the feeder system).

PAGE 29

18 When the combined loading of FDR1 and FDR4 is less than the continuous current rating of BKR4, then the switching pr ocedure at the substation may commence. First, TB4 is closed to energize the transfer bus from BK R4. Then, TB1 is clos ed, paralleling BKR1 and BKR4. This configuration with BK R1 and BKR4 paralle led is the condition previously described that must only be ma intained for a short time, because of the unpredictable behavior of the system protection while the sources are in parallel. It is a necessary configuration, unfortunately, to prev ent interruption of service to the load on FDR1 during the switching procedure. As quick ly as possible after the closing of TB1, BKR1 should be tripped. This puts the distribution system back into a radial configuration with bo th FDR1 and FDR4 supplied by BKR4. Now, BD1A and BD1B can be opened to isolate BKR1 and complete the switching procedure. With the increased load on BKR4, changes in the overcurrent rela y settings on that breaker will probably be necessary. And, of course, the original BKR4 relay settings must be restored when the FDR1 load is transferred back to BKR1 to assure proper protection of FDR4. Modern relays can be programmed with multiple groups of settings, which facilitate the changing of the setti ngs. But older relay technologies may be considerably more inconvenient to reset. The issue of changing relay settings can be avoided by using an additional circuit breaker to energize the transfer bus, as shown in Fig. 5.

PAGE 30

19 BD2B BD2A BKR2 Main Bus Transfer Bus BD3B BD3A BKR3 BD1B BD1A BKR1 BD4B BD4A BKR4 FDR1FDR2 FDR3 FDR4 TB1TB2 TB3 TB4 Source TBKR TBDA TBDB Figure 5 – Transfer Bus Topology Using Circuit Breaker to Energize Transfer Bus In this design, the relay settings on the feeder breakers remain unchanged when the transfer bus is in use since TBKR energizes th e transfer bus and picks up the load of the feeder connected to the transfer bus. But this design requires an additional circuit breaker (TBKR), which is used only when the transfer bus is in use. All of the above-mentioned ope rating concerns can be addr essed with de tailed operating procedures and adequate training of operati ng personnel. Despite these measures, the potential for operator error does exist. The ring bus topology avoids the issues centered around the transfer bus. Th e ring bus, though, does presen t its own set of operating issues, such as feeder reclosing procedures an d feeder protection stra tegies. These issues

PAGE 31

20 are arguably less complex than those posed by the transfer bus, a nd will be addressed later in this dissertation. Applying the Ring Bus to Facilitate Maintenance A major benefit realized by applying the ring bus configuration to th e distribution system is a simplification of the switching process to allow circuit breaker maintenance (Gnen 1986, P. 187). The feeder breaker isola tion switching procedure described for the transfer bus topology in the prev ious section requires that all of the load served by the breaker to be isolated must be transferred to another source. Even during light load periods, transferring the entire feeder load to another s ource may result in a lengthy and complicated switching procedure. This sw itching is not necessarily limited to the substation. The need to switch segments of feeders to alternat e sources, or field switching, is highly probable. As the load le vel increases, it may become impossible to serve the feeder’s entire load from other sources without vi olating operating criteria such as voltage limits and equipment loading levels. Installation of a bypass switch around the feeder breaker provides for easy maintenance of the breaker, but does so by compromising the protection of the system. Figure 6 shows a radial bus with bypass switches installed around each feeder breaker.

PAGE 32

21 BD2B BD2A BKR2 Main Bus BD3B BD3A BKR3 BD1B BD1A BKR1 BD4B BD4A BKR4 FDR1FDR2 FDR3 FDR4 Source BP1BP2 BP3 BP4 Figure 6 – Radial Bus with Bypass Switches Around Feeder Breakers Circuit breaker BKR1 can be isolated fo r maintenance by closing bypass switch BP1, then tripping BKR1 and opening breaker di sconnects BD1A and BD1B. But with the bypass switch closed, the only protection for FD R1 is the protection on the source to the main bus. This protection will probably be incapable of detecting many feeder faults, resulting in faults not being cl eared. If the main bus protecti on does detect a feeder fault, it will clear it by clearing the entire bus. And if careless switching results in the bypass switch remaining closed after BKR1 is returned to service, the system will remain in the same state of compromised protection as wh en BKR1 was out of service. For these reasons, the bypass switch design is u ndesirable for most applications.

PAGE 33

22 If a circuit breaker is used to connect the source to the bus, maintenance of this source circuit breaker is even more difficult, as the entire load of the bus must be supplied from other sources. This breaker, like the feed er breakers, could be equipped with a bypass switch, but then the protection of the main bus while the main breaker bypass switch is closed would rely on the ope ration of the next upstream br eaker, which is probably the transformer high-side breaker. It is unlikely that the overcurrent re lay on the transformer high-side breaker will be able to detect all fa ults for which it must operate with the main breaker bypass switch closed since the transf ormer impedance lies between the relay and the fault. This problem can be solved by a dding complexity to the protection scheme, but this may be an unwise decision. The complexities described in this s ection impose major constraints on system maintenance. At best, the constraints mean that circuit breaker maintenance can only be done at certain times, possibly at a higher-tha n-necessary cost due to complicated and time-consuming load switching. At worst, maintenance of critical circuit breakers, devices that could require a significant maintenance program because of their complicated mechanical nature, may be neglec ted. Compromising the integrity of circuit breakers will adversely impact both the sy stem reliability and the operating budget. The ring bus topology allows any circuit breaker even a source breaker, to be removed from service at any time, simply by tripping it and opening its disconnect switches. This is because each branch is served by not one but two circuit breakers under normal conditions. Only one breaker is necessary to ke ep a branch in service, so the other can be

PAGE 34

23 maintained without load switching. After remo val of a circuit break er from service, the ring topology is lost until the circuit breaker is returned to service. But even in this nonoptimal configuration, the reliability offered by the temporary open-ring bus configuration is no less reliable than that provided by the radial bus in its normal configuration. Not only does the elimination of load switching reduce the time and cost to switch the circuit breaker out of servi ce for maintenance, but it also reduces the probability of switching errors, which could lead to customer outages, equipment damage, or personnel injury. When devices other than circuit breakers must be maintained, line disconnect switches allow the ring topology to be restored after a branch is removed from service.

PAGE 35

24 Primary Feeder Design Construction Types Two general types of construction can be used for primary feeder construction: underground and overhead. Underground primary feeders are comprised of insulated, solid-dielectric cable, either directly buried in the eart h or installed in underground conduit. Faults occurring on underground cable s are of a permanent nature, since the breakdown of the solid-dielectri c insulation results in perman ent damage to the cable. The non-existence of temporary faults grea tly simplifies the opera ting methodology of an underground distribution system. When a fault is detected on an underground circuit, the circuit breaker supplying the ci rcuit is tripped and locked out – there is no need for automatic reclosing, since the fault cannot be of a temporary nature. Re-energization of the circuit cannot occur un til the faulted segment of cable is either elec trically isolated or replaced. Underground distribution system s are significantly more expensive to construct than comparable overhead systems. The primary cost difference is due to the cost of the solid dielectric cable compared to the cost of ba re aluminum wire, the most common conductor selection for overhead construc tion. The cost of construction varies with operating voltage and feeder ampacity. Table 1 compares bare overhe ad construction with direct-

PAGE 36

25 buried underground construction using typical U.S. construction cost da ta based on 2005 industry-average material and construction co sts. Please see Appendix C for calculation details, including design crit eria and pricing information. Table 1 – Typical Primary Feeder Construction Costs Ampacity Conductor 15 kV Class 25 kV Class 35 kV Class 443 266.8 kcmil Laurel $40,026/mi $47,570/mi $55,114/mi 639 477 kcmil Cosmos $54,540/mi $62,084/mi $69,628/mi 823 715.5 kcmil Violet $84,224/mi $96,822/mi $109,421/mi Overhead 1212 1351.5 kcmil Columbine$143,773/mi$163,915/mi $184,058/mi 415 350 kcmil Cu $132,578/mi$192,238/mi $298,300/mi 610 750 kcmil Cu $175,600/mi$254,620/mi $395,100/mi 830 2 x 350 kcmil Cu $225,383/mi$326,805/mi $507,112/mi Under g round 1220 2 x 750 kcmil Cu $298,520/mi$432,854/mi $671,670/mi Although underground distribution sy stems are used extensively in particular residential and commercial applications throughout th e world, most power distribution is accomplished using overhead construction. According to the Edison Electric Institute’s UDI database, of the 4,980,066.69 miles of distri bution circuits reported by U.S. utilities in 2004, 1,137,123.04 miles, or 23%, are unde rground circuits (EEI, 2004). Design details for overhead distribution lines vary gr eatly from system to system, influenced heavily by geographical factor s such as ambient temperature variation, precipitation

PAGE 37

26 levels, wind loading, lightning frequenc y, population density, terrain, and other topological features. For the purposes of analysis, a basic and so mewhat generic design is considered. Concrete poles supporting bare aluminum con ductor on polymer insulators is that basic design. Three commonly-used pole materials are used throughout the utility industry: wood, steel, and concrete. Wood has historically been the ma terial of choice for most applications because of its abundance, versatil ity, and durability. Its useful life, however, can be limited in damp or humid climates. Wood is also pron e to insect and bird damage. And the strength of wood may not be adequate for some applications. As forests become depleted, the availability of wood in certa in geographical areas is a growing concern. These concerns led to the use of alternativ e materials for distribution pole fabrication. Steel and concrete are common materials in us e today. Steel is very strong and relatively lightweight, but quite expensive compared to th e alternatives. Steel is also subject to corrosion. Painting and galvaniz ing retard the onset of co rrosion, but th ese measures seldom prevent corrosion from occurring at some point in time. Prestressed concrete is an outstanding structural material both physically and economically. It can be fabricated with ease a nd at a fairly low cost. It is durable in virtually all environments. On e of its few drawback s is weight. Larger poles can become very heavy, complicating transportation a nd erection. Fortunately most distribution applications involve poles sufficiently small su ch that weight is not a major concern. The

PAGE 38

27 manufacturing facility for the poles needs to be in relati vely close proximity to the installation location to keep tr ansportation costs reasonable. Polymer insulators have been in use by u tilities since the 1970s (Brewer 1994, P. 1). Prior to that, glass and porcelai n were the materials of choice for insulators. Glass and porcelain have excellent dielectr ic properties, but they have a low strength-to-weight ratio and are brittle. The polymer insulator is a fiberglass rod covered with an elastomeric coating such as ethylene propylene monome r (FPM), ethylene propylene diene monomer (EPDM), and silicone rubber. FPM and EPDM are resistant to erosion and tracking, while silicone rubber has excellent resistance to ultraviolet degradati on. Over the years, silicone has been alloyed with ethylene propylene rubber to produce an elastomeric coating material with outstanding electrical an d physical properties. It is this generation of polymer insulator in widespread use today. Polymer offers a weight reduction of close to 75% for 15kV distribution-class insulators over porcelain (Bernstorf 1992, P. 2). This adva ntage alone translates to significant cost savings when polymer insulators are used. The durability of the polymer design results in fewer insulator replacements – an expensive maintenance function on overhead distribution systems. Conductor Selection Three different types of conductor can be us ed on an overhead distribution system: bare aluminum wire, partially-insulated overhead cable, and fully-insulated spaced aerial

PAGE 39

28 cable. Bare wire is widely used throughout most of the wo rld. In areas where contact with conductors by trees or w ildlife is of high concern, th e consequent line-to-ground faults can be greatly reduced or even el iminated by covering the aluminum conductor with an insulation system. This substantiall y increases the cost, we ight, and diameter of the cable. The increase in weight and diameter has a substantial impact on structure cost and span length. The benefit of using insu lated cable is a significant decrease in temporary faults on the system. This not only improves reliability, but also lengthens equipment life, as the devices on the system are subjected to fewer through faults. Field experience of a major Asian utility indi cates operational problems associated with partially-insulated cable applied at 33kV (PEA 2002). These problems center about dielectric breakdown due to partial discharge. At 22kV, these problems still exist, but are far less prevalent. At both voltages, fu lly-insulated spaced aerial cable showed significantly better performance th an partially-insulated cable. Although spaced-aerial cable shows significa ntly improved performance over bare aluminum wire for distribution primary cons truction at both 22kV and 33kV on a large and diverse distribution system in Asia, the construction costs are difficult to justify for applications where reliability requirements are not extremely high (PEA 2003). For the purposes of analysis, bare aluminum wire – the most common overhead distribution primary conductor type by far in the United States today – will be considered. Spaced aerial cable, however, should be considered for applications where frequent interruptions due to temporary faults are unacceptable.

PAGE 40

29 Many varieties of bare aluminum conductor ha ve been developed. The use of aluminum alloys and steel reinforcing are two co mmon methods of improving the conductor’s performance. While these performance improve ments, such as increased tensile strength and capability of achieving higher operati ng temperatures, can be significant for transmission applications, the lower stringing tensions and shorter span lengths used in distribution applications diminishes the benef it of these improvements. Because of this and in the interest of economy, standard bare all-aluminum conductor (AAC) will be considered for the purposes of analysis. Many standard sizes of AAC are commercially available. A family of commonly-used American wire sizes has been defined as a sta ndard selection set in this dissertation. The sizes were selected to provide a range of ampacities from approximately 450 amperes (37.5% of typical circuit breaker rating) th rough 1200 amperes (100% of typical circuit breaker rating), in increments of approxima tely 70 amperes. The selection set of conductors is shown below in Table 2.

PAGE 41

30Table 2 – AAC Conductor Data for Chosen Selection Set Conductor temperature 75C, ambient temperature 25C, emissivity 0.5, wind 2 ft./sec. in sun. Equipment Selection A fundamental decision that must be made prior to specifying the major equipment components for a distribution system is that of operating voltage. Although many operating voltages are used on u tility distribution systems worldwide, only three voltage classes of outdoor power distri bution equipment are available: 15 kV, 25 kV, and 35 kV. Some manufacturers also provide 5 kV equipment, but this vo ltage class is not considered as a viable choice for utility power distribu tion due to its very s hort feeder reach. Usually, for a given operating voltage, the equipment voltage class is selected as the lowest voltage that is greater than or equal to the operating voltage. Exceptions to this rule are made when extra insulation strength is desired for proper insulation coordination. In these cases, the next higher equipment rating may be used. Size (kcmil) Codeword Ampacity* 266.8 Laurel 443 336.4 Tulip 513 397.5 Canna 570 477 Cosmos 639 556.5 Mistletoe 704 636 Orchid 765 715.5 Violet 823 795 Arbutus 878 900 Cockscomb948 1033.5 Larkspur 1031 1192.5 Hawthorn 1124 1351.5 Columbine 1212

PAGE 42

31 Any devices used on the power system that provide insulation to ground carry a voltage rating. These devices include transformers, circuit breakers, insu lators, arresters, insulated cable, reclosers, sectionalizers, and fuses. Substation buswork and bare overhead conductor have no voltage rating, bu t their physical means of support, the insulators, do. Phase-to-phase clearance of uninsulated conductors does depend on operating voltage, and is stipulated by the National Electrical Safety Code (IEEE C22002). Communication-Based Ov ercurrent Detection Typical distribution system protection consis ts of overcurrent-type protective devices placed in series at various points along the f eeder. At the substation, instantaneous and time-delayed overcurrent relays combined with the feeder circuit breaker form the first protective tier. At particular points along the feeder, othe r protective devices such as reclosers, sectionalizer s, and fuses are installed to mi nimize the amount of circuit that must be de-energized in the event of a fau lt. Since all of the protective devices on the feeder sense the fault as soon as it occurs, time delays must be used to coordinate the times at which the devices operate. Th is is done using different time-current characteristics (TCCs) for the various devi ces, then assuring pr oper coordination by graphing the TCCs between the minimum antic ipated fault curren t magnitude and the maximum fault current for the feeder and veri fying that the appropriate device operates first for all realistic fault scenarios.

PAGE 43

32 This general principle has been in widespread use for many decades. Its advantages include simplicity and the fact that each protec tive device is independent of the others. A specific device operates at a particular time based on the current magnitude sensed by the device. It is up to the engineer to make su re that the operating time is such to achieve proper coordination. There are disadvantages to the time-coordi nated overcurrent protection method. To achieve coordination, time delays must sometim es be built into the protection scheme. Referencing the time-current characteristic shown in Fig. 7, the pickup time for a time overcurrent relay may have to be delayed to 2 seconds for a current magnitude of 10 times the pickup current (time dial 5) to coor dinate with downstream devices. The relay could operate in as little as just over a half second for the same current magnitude if configured with a time dial setting of 1.

PAGE 44

33 Multiple of Pickup Current (Per-Unit) 0.1 0.1 1 1Time in Seconds10 100 1000 10 6 5 4 3 2 1 10 8 9 7 100Time Dial Figure 7 – Typical Time-Current Characteristic Slow fault clearing is a well-understood probl em on transmission systems. With longer fault durations come reduced transient st ability and increased equipment damage resulting from the fault current Stability is not an issue at the distribution level, but equipment damage, particularly to transformers as a result of the through-fault current, is. Maintaining motor load becomes problematic as the fault duration on the distribution system increases. The motor contactors drop out when the system voltage falls below a

PAGE 45

34 threshold defined by the design of the contact or. This explains why a motor may shut down when the facility where the motor is located does not experience an interruption in power. If a fault on an adjacent feeder depresses th e voltage low enough and long enough, the contactor will drop out. The ITI cu rve shown in Fig. 8 provides a guideline as to the acceptable magnitude a nd duration of voltage excursions (CBEMA 2000). 0.001c 0. 1 c 10 c 10 0c 10 00c 10,0 0 0c 1 ms 0.01 c 3 ms 20 ms 0.5 s 10 sNo Damage Region Prohibited Region 500 400 300 200 100 120 140 80 70 No Interruption Region1 cPercent of Nominal VoltageDuration in 60-Hertz Cycles or Seconds 90 110Percent of Nominal Voltage Figure 8 – ITI Curve (Revised 2000), Published by the Information Technology Industry Council

PAGE 46

35 The dropping out of motors frequently cause s a process shutdown, requiring an expensive and time-consuming restart. In many industr ies, frequent restarts are unacceptable. The disadvantages associated with time-c oordinated overcurrent protection could be overcome if the protective devices had the ability to communicate with each other. Utilizing blocking or permissive indication from other devices, robust, flexible, and reliable protection schemes can be developed. These schemes are similar to those used on transmission systems for many years. The sa me basic rationale will be used to justify the use of “transmission-type” protection sche mes on the distribution system as was used previously in this dissertation to advocat e the use of “transmission-type” substation topologies (i.e., the ring bus ) at the distribution level: if the method improves the performance of the transmission system, shouldn’t it also be c onsidered at the distribution level, where benefits such as enhanced re liability would be even greater (Fehr 2004, P. 4)? The optimum communication-based protection sc heme for a specific application depends heavily on the type of communi cation system used. Of th e many types of communication systems commercially available for and in use on utility power systems, direct fiber-optic channels have many benefits and few drawb acks (Moxley and Fodero, P. 21). Channel unavailability tends to be very low on direct fiber-optic ch annels, and the durations of channel failures when they occur is very short (in the milliseconds) (Moxley and Fodero, P. 23). The probability of channel failure durin g an electrical fault is low, and the ability to endure environmental elements is exceeded only by spread spectrum or licensed radio

PAGE 47

36 communication (Moxley and Fodero, P. 23). The terminal cost is low, but the cost to establish the channel is high. With this hi gh cost comes the ability to support very high data rates (in excess of 4 Gbps ), and very high communication speeds (less than 0.1 ms). These characteristics suggest sharing the unused fibers with a non-electrical utility application such as digital telephone or Internet commun ication to defray the high installation cost. The characteristics of direct fiber-optic ch annels are very compatible with a blockingtype protection scheme (Schwe itzer, Behrendt, and Lee, 1998, P. 4). In this type of protection system, an interrupting device that senses a fault outside (further downstream) its zone of protection will receive a blocki ng signal from the device that will clear the fault. If the blocking signal is not receive d, as would be the case if the communication channel fails, the result is overtripping, since the upstream device that senses the fault would trip in addition to the device closer to the fault. While not desirable, overtripping is preferable to failure to trip. A basic form of communication-based overcurrent protection utilizing blocking signals is explained as follows. Consider th e feeder diagram shown in Fig. 9.

PAGE 48

37 Seg. 2 POWER RING Seg. 1 R N.C. Feeder R N.O. Source Backup (Normal Source) Figure 9 – Feeder Configured for Communication-Based Overcurrent Protection The feeder configuration shown in Fig. 9 is typical of ma ny existing feeder configurations with a few subtle differences. The source for the f eeder is a power ring instead of a typical radial bus. The two segm ents are separated by an electronic recloser capable of communicating with other devices. Typically, this sectionalizing device would be a recloser without communication ca pability. At the end of the feeder is another electronic recloser, configured nor mally-open, instead of a manually-operated normally-open switch. Additional sectionalizi ng reclosers could be used. The optimum number and location of sectionalizing reclosers will be analyzed later in this dissertation. A two-segment feeder will be used to explain the concepts of communication-based overcurrent detection. The feeder shown in Fig. 9 will now be an alyzed being protected using communicationbased overcurrent detection. Th e key to this protection method is to make all protective devices on the feeder aware of the conditions sensed by the other protective devices on

PAGE 49

38 the feeder. This communication allows sophist icated logic to be pe rformed, allowing the fault to be cleared in the most appropria te manner, depending on its location (segment number) and nature (temporary or permanent) Four scenarios will be analyzed: a temporary (self-clearing) fault in each of th e two segments, and a permanent fault in each of the two segments. Appendix E details the logic used to implement communicationbased overcurrent detection. Temporary Fault on Segment 1 When a fault occurs on Segment 1, only the relays on the power ring breakers sense the fault current, since the fault occurs upstream of the sectionalizing recloser. When those relays sense a current magnitude in excess of their pickup se tting (fault current) and no block signal is received from the sectionalizing recloser se parating segments 1 and 2, the power ring breakers trip. This action clears the fault and de-e nergized the entire feeder. After the power ring breakers are open, a trip si gnal is sent to the sectionalizing recloser. Opening the sectionalizing recloser becomes significant when attempting to reclose the power ring breakers. When the fault is cleared and the entire lengt h of the feeder is de-energized, a reclose attempt will be made. One of the power ring breakers will reclose, energizing Segment 1 of the feeder. Note that Segment 2 rema ins de-energized, sinc e the sectionalizing recloser is open. This is to prevent a mome ntary interruption to Segment 2 in the event of a failed reclose attempt. Only one power ring breaker needs to be reclosed to ascertain

PAGE 50

39 whether the fault is still on the feeder. Fo r consistency, the reclosing breaker will be designated as the breaker adjacent to the fa ulted feeder in the counterclockwise position when the power ring is viewed from above. If the reclose is successful, service is restor ed to the load on Segment 1, but Segment 2 is still de-energized because of the open secti onalizing recloser. Afte r the reclosing breaker on the power ring remains closed for short predetermined period of time, the second power ring breaker and the se ctionalizing recloser is auto matically closed, restoring service to all load and restoring the system to its normal configuration. But if the reclose fails, the reclosing power ring breaker trips again, blinking the lights of the customers served from Segment 1. The customers on Segment 2 remain de-energized during the failed reclose attempt. At this poi nt, a time delay of 10 to 30 seconds can be executed, allowing more time for the fault to clear itself from the feeder. Then, another attempt can be made to reclos e the power ring recl osing breaker. If this second reclose attempt succeeds, the restora tion procedure detailed in the preceding paragraph can be executed. If the second reclose attempt fails, th e fault must be assumed to be permanent, so the procedure detailed in the Perman ent Fault on Segment 1 section is followed. Temporary Fault on Segment 2 When a fault occurs on Segment 2, both the relays on the power ring breakers and the sectionalizing recloser sense the fault current. Instead of relying on time-based

PAGE 51

40 coordination to assure that the recloser tr ips before the power ring breakers, a block signal is sent to the relays on the power ri ng by the sectionalizing re closer. This will allow the fault to be cleared by the recloser instead of by the power ring breakers. Several benefits result from clearing a Segment 2 fault with the sectionalizing recloser instead of with the power ring breakers. Although customers supplied from Segment 1 w ill still experience the voltage dip caused by the fault on Segment 2, they will not experi ence a total interruption of service. This benefit increases the reliability of the feeder by reducing the number of customers exposed to the outage. Impr ovements in power quality ar e also more easily obtained when a dip in voltage is the issue to be addre ssed as opposed to an inte rruption in service. From an equipment maintenance perspective, it is more desirable to have one recloser operate to clear a fault than two circuit breakers to reduce incremental maintenance requirements. After the fault current is interrupted, a recl ose attempt may be made. If reclosing the sectionalizing recloser succeeds, all load is supplied and the system is back in its normal configuration. But if the reclose attempt fails another reclose attempt may be made after a predetermined time delay. Failure of this second attempt may be interpreted as the fault being permanent, in which case, the pro cedure detailed in the Permanent Fault on Segment 2 section would be followed.

PAGE 52

41 Permanent Fault on Segment 1 When the fault on Segment 1 is determined to be permanent, the motor-operated feeder disconnect switch is automatically opened, an d the power ring break ers are closed to restore the integrity of the ring bus. Finally, the normally-open recloser at the end of the feeder is closed, which energizes the Segment 2 load from the backup source. Permanent Fault on Segment 2 When the fault on Segment 2 is determined to be permanent, the sect ionalizing recloser is locked out, and Segment 2 remains de-energiz ed until the cause of the fault can be found and removed from the system.

PAGE 53

42 The Optimization Process Assumptions The sizing of the components in a power di stribution system is a very complex and highly constrained problem primarily driven by the design of the feeder system. The design of the feeder system is also constraine d by practical considerations, and is heavily dependent on the nature of the service area for that feeder, such as the amount of load (kVA), its classification (residential, commercia l, industrial), its lo ad factor, its power factor, and many other technical characteristics (Willis, 1997, P. 285). A fundamental decision that must be made ear ly in the design process is the type and number of feeders to be supplied by each s ubstation. This dissertation will assume that all feeders exiting the substation will be of overhead construction, exiting the substation via underground feeder getaways comprised of insulated cable in conduit. These underground cables will be tran sitioned to overhead circuits using riser poles located as close to the substation as pract ical. A typical ri ser pole arrangement is shown in Fig. 10. The underground getaways alleviate safety issues and overhead clearance concerns caused by overhead distribution circuits within the substation. While the distribution circuits leaving the substation could con tinue as underground feeders, the underground getaway / overhead feeder configuration is th e more general case to analyze, and is in

PAGE 54

43 widespread use by American utilities at this time. Proposing a feeder construction similar to that which is already in us e by most utilities w ill facilitate the adop tion by utilities of the innovations proposed in this dissertation. Cable Terminations N eutral Conducto r Conduit Riser Pole Overhead Phase Conductors Solid-Blade Disconnect Switch Deadend Insulator Lightning Arrester Lightning Arrester Guy Wires Figure 10 – Typical Riser Pole The overhead conductor can be of three types: bare wire, pa rtially-insulated cable, or fully-insulated spaced-aerial cab le. The conductor type sel ection has major impacts on both the construction cost and th e reliability of the feeder. The initial assessment will be done for bare wire, as this is the conductor type predominantly used by U.S. utilities

PAGE 55

44 today. A differential analysis can be perfor med to determine the effects on reliability and cost of changing to partially or fully-insulated cable. Addressing the number of feeders to be supplied by each subs tation, a pragmatic approach will be taken. The logistics of ex iting a feeder from a substation can be quite complicated. The vicinity just outside the s ubstation becomes very congested with riser poles and overhead circuits as the number of circuits leaving the substation increases. This congestion is not only a design complic ation, but can pose safety and aesthetic concerns. The safety concerns involve utilit y workers having to maintain circuits having minimal safety clearances. These concerns must be considered as urgent issues and need to be eliminated if at all possible to protec t the people working on the system. Aesthetics, while usually not of primary engineering co ncern, also need to be acknowledged as pressures from the public to improve the ap pearance of power system infrastructure increase. Another practical consideration involves the routing of the f eeders. In developed areas, feeder construction is usually limited to stre et rights-of-way. Sin ce roads are typically built in a rectangular grid configuration, four feeder routes utilizi ng the street rights-ofway can be defined for each substation. For si mplicity, the four routes can be considered as north, south, east, and west. Double-circuiting distribution f eeders along these four routes may be attractive from a circuit routing viewpoint to enab le more feeders to be exited from the substation, but this

PAGE 56

45 practice can have very detrimental effects on re liability. Losing a single structure of a double-circuit line will result in the outage of both circuits. It should be remembered that distribution circuits, because of their relative size and le sser design requirements, are much more susceptible to failure than transm ission structures. An automobile collision will frequently bring down a distribution pole, but se ldom seriously damages a transmission structure. This makes the doubl e-circuiting of distri bution circuits quite undesirable from a reliability standpoint. To avoid double-c ircuiting and to facilitate circuit routing out of the substation, a total of four distribution feeder s will be considered for each substation. This assumption allows single feeders to be constructed in each of the routes named by the four compass direct ions, providing a robust infrastructure while maintaining a clean and simple routing and configuration. Knowing the total number of feed ers leaving the substation, the next decision to make is the ampacity requirement of each feeder. The required ampacity of a feeder is a function of the service area of the feeder, but the service area of the feeder depends on the feeder’s ampacity. This circular process requires a somewhat iterative approach to solve. Methodology The optimization method consists of two stag es which are performed sequentially. The first is an electrical assessment of the proposed configuration. That is followed by an economic assessment of the design. Before the electrical assessment can be done, some preliminary steps must be taken. The first preliminary step is the determination of the ampacity required for the feeder.

PAGE 57

46 A good starting point for determining feeder am pacity is to determine the average load density in kVA/mi2 for the feeder’s service territory. The load density for an existing area is readily available from hi storical metering data, but it is wise to adjust this known value for anticipated future changes in the lo ad – say over a 5-7 year horizon. Let this adjusted average load density be a constant value through out the entire feeder service area for illustrative pu rposes. In practice, non-uniform load densities and spot loads could be incorporated, but these considerations tend to complicate th e process, and will be avoided in this example. Assuming a realistic power factor (0.90) and that 1.0 per-unit voltage will be available at all points along the feeder (most likely in accurate assumptions, but acceptable for a starting point of an iterative pr ocess), the kVA load that can be served by a feeder equals the product of the nominal voltage in kV, the rated current of the feeder conductor, and the square root of three, or 3) I ( ) kV ( kVArated al min no feeder Eq. 1 Figure 11 – Feeder Service Area l 2l

PAGE 58

47 Referring to Fig. 11, the gray shaded triangle represents the service area for the feeder exiting the substation in the center of the figure and heading towa rd the right. The service area can be calculated as (2l) (l) = l 2. The load in the service area is l2d, where d is the average load density in kVA / mi2. Setting that load equal to th e load able to be served by the feeder based on conductor rating from Eq. 1 gives 32) I ( ) kV ( d lrated nominal. Eq. 2 From Eq. 2, feeder length l can be obtained as d ) I ( ) kV ( lrated nominal3 Eq. 3 Since this initial value for optimal feeder length is a f unction of the conductor current rating, it will be different fo r each size conductor. It will also vary inversely with the square root of the load density d. For illustrative purposes, an average load density of 1500 kVA/mi2 is used. Applying the feeder lengt h estimate shown in Eq. 3 over the selection set of conductors from Table 2, c onsidering several commonly used distribution voltages, an initial approximation of optimum feeder length can be determined. This length will be called the maximum service length of the feeder at the assumed load density. Any feeder length in excess of th e maximum service length will result in a thermal overload of the conductor. The results are summarized in Table 3.

PAGE 59

48Table 3 – Maximum Service Length Calculation – 1,500 kVA / sq. mi. Load Density A B C D E F G H I J K 1 AAC conductor Ampacity @ 75 deg C Resistance (60Hz @ 75 deg C) ohms per mile 12.47 kV 13.2 kV 13.8 kV 14.4 kV 22 kV 24.9 kV 34.5 kV 2 kcmil codeword Maximum Service Length (miles) 3 266.8 Laurel 444 0.4187 2.5 2.6 2.7 2.7 3.4 3.6 4.2 4 336.4 Tulip 513 0.3326 2.7 2.8 2.9 2.9 3.6 3.8 4.5 5 397.5 Canna 570 0.2820 2.9 2.9 3.0 3.1 3.8 4.0 4.8 6 477.0 Cosmos 639 0.2350 3.0 3.1 3.2 3.3 4.0 4.3 5.0 7 556.5 Mistletoe 704 0.2017 3.2 3.3 3.3 3.4 4.2 4.5 5.3 8 636.0 Orchid 765 0.1769 3.3 3.4 3.5 3.6 4.4 4.7 5.5 9 715.5 Violet 823 0.1579 3.4 3.5 3.6 3.7 4.6 4.9 5.7 10 795.0 Arbutus 878 0.1426 3.6 3.7 3.7 3.8 4.7 5.0 5.9 11 900.0 Cockscomb 948 0.1262 3.7 3.8 3.9 4.0 4.9 5.2 6.1 12 1033.5 Larkspur 1032 0.1109 3.9 4.0 4.1 4.1 5.1 5.4 6.4 13 1192.5 Hawthorn 1124 0.0966 4.0 4.1 4.2 4.3 5.3 5.7 6.7 14 1351.5 Columbine 1212 0.0861 4.2 4.3 4.4 4.5 5.5 5.9 6.9 15 16 Load Density: 1,500 kVA per square mile 17 kV>>> 12.5 13.2 13.8 14.4 22 24.9 34.5 Table 3 can be implemented as an electronic sp readsheet to facilitate the computation of the maximum feeder service le ngth. Row numbers and column letters are shown in Table 3 for reference. The data in columns C and D were reproduced from the Full-Line Product Catalog by Southwire Company (Sout hwire, 2003, P. 11-2). The maximum service length calculation as implemented in cell E3 is shown below. =SQRT(E$17*$C3*SQRT(3)/$D$16) Eq. 4 The load density in cell D16 can be varied to observe the e ffect of the load density of the maximum feeder service length.

PAGE 60

49 In an attempt to achieve an optimum design, the highest operating voltage allowable for each equipment class will be considered. Th e 15 kV equipment class will be operated at 14.4 kV, the 25 kV class will be operated at 24.9 kV, and the 35 kV class will be operated at 34.5 kV. Other common operati ng voltages are shown in Table 3 because utilities with these operating voltages alr eady installed may choose to keep their distribution systems operating at their pr esent voltages to avoid the expense and complication of replacing every service transf ormer on the system. Use of these other voltages may be slightly less than optimal, but they represent a practical optimum. The maximum feeder service length allows no contingency backup capacity. This limitation renders a feeder incapable of serving any load other than it s own at the time of peak loading. While it can be argued that the annual peak loading of a feeder lasts a relatively short time, typical load duration cu rves show that feeder loadings can remain close to their peak values for hundreds of hours per year. Due to these high exposure periods of not being able to serve load from adjacent feeders, some contingency capacity must be allowed. The amount of continge ncy capacity to be al lowed is a topic of considerable interest, since reserving contin gency capacity represents a stranded financial cost, but also allows more versatile operat ion of the system. The largest amount of contingency capacity that would ever need to be reserved is 50% (Gnen 1986, P. 235). At a 50% contingency capacity reserve, any feeder would be able to backup another entire feeder at any loading level without thermal overload. While this 50% capacity reserve is very conservative, it will be assumed in this analysis. Reducing the contingency capacity reserve below 50% increa ses the amount of time every year to a

PAGE 61

50 value greater than zero that another feeder cannot be backed by the feeder under consideration without thermally overloading that feeder. To reserve 50% contingency capacity for a feed er, its length must be limited to 70.7% of its maximum service length. This value will be called the effective service length of the feeder. If c denotes the desired conti ngency capacity in percen t, the effective service length esl is related to the maximum service length msl as shown below. msl c esl 100 100 Eq. 5 Table 3 is modified below to show the esl for 50% contingency capacity. Table 4 – Effective Service Length Calculation – 50% contingency capacity and 1,500 kVA / sq. mi. Load Density A B C D E F G H I J K 1 AAC conductor Ampacity @ 75 deg C Resistance (60Hz @ 75 deg C) ohms per mile 12.47 kV 13.2 kV 13.8 kV 14.4 kV 22 kV 24.9 kV 34.5 kV 2 kcmil codeword Effective Service Length (miles) for 50% Contingency Capacity 3 266.8 Laurel 444 0.4187 1.77 1.84 1.91 1.91 2.40 2.55 2.97 4 336.4 Tulip 513 0.3326 1.91 1.98 2.05 2.05 2.55 2.69 3.18 5 397.5 Canna 570 0.2820 2.05 2.05 2.12 2.19 2.69 2.83 3.39 6 477.0 Cosmos 639 0.2350 2.12 2.19 2.26 2.33 2.83 3.04 3.54 7 556.5 Mistletoe 704 0.2017 2.26 2.33 2.33 2.40 2.97 3.18 3.75 8 636.0 Orchid 765 0.1769 2.33 2.40 2.47 2.55 3.11 3.32 3.89 9 715.5 Violet 823 0.1579 2.40 2.47 2.55 2.62 3.25 3.46 4.03 10 795.0 Arbutus 878 0.1426 2.55 2.62 2.62 2.69 3.32 3.53 4.17 11 900.0 Cockscomb 948 0.1262 2.62 2.69 2.76 2.83 3.46 3.68 4.31 12 1033.5 Larkspur 1032 0.1109 2.76 2.83 2.90 2.90 3.61 3.82 4.53 13 1192.5 Hawthorn 1124 0.0966 2.83 2.90 2.97 3.04 3.75 4.03 4.74 14 1351.5 Columbine 1212 0.0861 2.97 3.04 3.11 3.18 3.89 4.17 4.88 15 16 Load Density: 1,500 kVA per square mile 17 kV>>> 12.5 13.2 13.8 14.4 22 24.9 34.5

PAGE 62

51 Electrical Assessment When the effective service length is known for each conductor size and operating voltage, the electrical assessment process can begin. The feeder load needs to be determined based on the effective service le ngth. This is done by multiplying the load density in kVA/mi2 by the effective service length, in miles, squared, as shown below. 2esl density load load feeder Eq. 6 Next, a load flow model is used to determin e the electrical performa nce of the feeder. A model was constructed using Version 8 of the PowerWorld Simulator (PowerWorld 2001), and is shown in Appendix A. The model consists of a representation of the main feeder, with the total feeder load lumped into ten equally-spaced nodes along the feeder. A single sectionalizi ng recloser is also modeled at the midpoint of the feeder. The main feeder terminates at a normally-open tie recl oser, which backs up to a feeder identical to the main feeder. A generator and swing bus with a scheduled voltage of 1.05 per-unit supplies the main feeder. A similar genera tor and swing bus arrangement supplies the backup feeder. A shunt capacitor bank is mode led at the first node downstream from the sectionalizing recloser on each feeder for contro l of the voltage profile of the feeder. The main feeder is defined as Z one 1 and the backup feeder as Zone 2, to facilitate the calculation of losses by feeder. Per-unit voltages and angles ar e displayed at each feeder node, and the current in amperes is shown for each feeder segment. The watts and vars delivered by the source is displayed for ea ch feeder. The total load and losses are tabulated for each feeder on the one line diag ram. The graphical representation of the feeder model can be seen in Appendix B.

PAGE 63

52 The load flow model is solved for the base ca se (the main feeder supplied from its normal source, the backup feeder suppl ied from its normal source, and the tie recloser open). When solved, a verification of acceptable volt age along the entire feed er length is made. Acceptable voltage is considered to be 5% of nominal. Since the source end of the feeder is modeled as the maximum allowabl e voltage (1.05 per-unit), then the minimum acceptable voltage at the end of the feeder is 0.95 per-unit, meaning that a 10% voltage drop can occur along the feeder. If a voltage drop in excess of 10% occurs along the feeder, the shunt capacitor can be used to correct the situation. When the base case shows satisfactory electric al performance, the contingency cases are studied. For the model under consideration, two contingencies exist. The first is for the backup feeder to supply its load along with the lo ad of the second segment of the main feeder. The first segment of the main feeder is supplied from its normal source. This scenario represents an outage of the sect ionalizing recloser on the main feeder. The second contingency is for the backup feeder to supply its entire load plus the en tire load of the main feeder. This mo re severe contingency represen ts the failure of the main feeder’s normal source. During contingencies, 10% of nominal vol tage is allowed. The second studied contingency produces large voltage drops since both the feeder load and the feeder length are twice that of the normal configuration. Here, the s hunt capacitors, which may have been unnecessary in the base cas e, are instrumental in maintaining adequate voltage. The second contingency also draws considerably more reactive power from the source than

PAGE 64

53 does the base case, so the current loading on th e first feeder segment must be monitored. In some cases, slight overloads in the first f eeder section are observe d. Overloads of up to 5% are considered to be acceptable, as th ey only exist on a very short length of the feeder. In reality, after the first connected load is tapped from the main feeder, these overloads will no longer exist. As voltage an d/or overload violations are observed, the reactive configuration of the system can be adjusted to alleviate them. Table 5 summarizes the feeder designs analyzed as part of the Design Example later in this dissertation, along with the pr oper reactive configurations to produce acceptable voltage and loading values. The losses for the base ca se are also noted. The computer simulation results of the electrical asse ssment are shown in Appendix B. Table 5 – Summary of Analyzed Feeder Configurations Load Base Case Losses % MW Operating Voltage Conductor Shunt Capacitor MW MVAR MW MVAR Losses 14.4 kV 266.8 kcmil 2400 kVAR 5.002.400.07 0.34 1.40% 14.4 kV 397.5 kcmil 3600 kVAR 6.403.100.09 0.45 1.41% 14.4 kV 556.5 kcmil 3600 kVAR 7.903.800.10 0.51 1.27% 14.4 kV 715.5 kcmil 4800 kVAR 9.204.500.12 0.61 1.30% 14.4 kV 900 kcmil 4800 kVAR 10.605.200.13 0.67 1.23% 14.4 kV 1192.5 kcmil 6000 kVAR 12.606.100.19 0.74 1.51% 24.9 kV 266.8 kcmil 1800 kVAR 8.604.200.07 0.34 0.81% 24.9 kV 397.5 kcmil 1800 kVAR 11.105.400.09 0.45 0.81% 24.9 kV 556.5 kcmil 1800 kVAR 13.706.600.11 0.55 0.80% 24.9 kV 715.5 kcmil 1800 kVAR 16.007.700.13 0.65 0.81% 24.9 kV 900 kcmil 1800 kVAR 18.408.900.15 0.75 0.82% 24.9 kV 1192.5 kcmil 1800 kVAR The process for performing the electrical assessment is shown in Fig. 12.

PAGE 65

54 Determine Feeder Load based on Voltage and Solve Base Case Select Operating Conductor Size Effective Service Length Insert Feeder Impedance and Load Quantities into PowerWorld model Voltage Range OK? No Yes Adjust Capacitor Sizes and Configuration Transfer Main Backup Feeder Feeder to Solve Contingency Case Voltage Range OK? No Yes Adjust Capacitor Sizes and Configuration Current in 1st Segment OK? No Yes Acceptable Design Do Economic Assessment Figure 12 – Electrical Assessment

PAGE 66

55 One last consideration must be made before proceeding to the economic assessment. The load on a feeder decreases as one moves farthe r from the source. The ampacity required in the first segments of the feeder is not required along the entire length of the feeder. This suggests that a smaller c onductor size can be used star ting at some point a given distance from the substation. Doing so woul d reduce the cost to construct the feeder system. Before reducing conductor size, however, cont ingency switching configurations must be considered. The simplest backup procedure w ould be to supply one feeder by connecting it to the end of another feeder This is easily accomplishe d by installing a normally-open recloser at the end of each f eeder. Closing this recloser ties the two feeders together. This is shown in Fig. 13. for Fdr. 1) (Normal Source POWER RING 1 Fdr. 1 POWER RING 2 (Backup Source for Fdr. 1 and N.O. Fdr. 2 Normal Source for Fdr. 2) N.C. N.C. line disconnect switch line disconnect switch tie recloser Figure 13 – Feeder Backup Configuration

PAGE 67

56 Feeder 1 can be fed from Power Ring 2 by clos ing the tie recloser, but this puts the two power rings in parallel, a condition that needs to be avoided since pr otective devices can operate in erratic ways while sources are pa ralleled. Also, the short circuit current available on the feeder is increased on the order of 100% with paralleled sources. The paralleled sources must be tolerated fo r a short period (seconds) during switching procedures to avoid interrupti ng service to the customers on F eeder 1. But as soon as the tie switch is closed, the line disconnect sw itch at Power Ring 1 can be opened to reestablish a radial system with both feeders supplied from Power Ring 2. In practice, before the line disconnect switch can be ope ned, the current flowing through that switch must be interrupted. This is accomplished by tripping both power ri ng circuit breakers adjacent to Feeder 1. Afte r the breakers open, then the line disconnect can be opened, and finally the power ring brea kers can be reclosed. Wh ile this may seem like a complicated procedure, it is similar to the sw itching procedures in use today to transfer sources, and if implemented using automati c equipment, can be performed within seconds without interrupting se rvice to any customer (De La Ree, Elizondo, Depablos, and Stoupis 2002, P. 1). Referring to Fig. 13, the portion of Feeder 1 cl osest to the tie switch carries very little load under normal conditions, but during contingency conditions when Power Ring 2 must supply Feeder 1, that portion of Feeder 1 must carry all of Feeder 1’s load. The segments closer to the power ring may have to carry the load of both the main and backup feeders during contingencies. This i llustrates how conductor size can be reduced as a function of distance from the substation.

PAGE 68

57 Some practical considerations regarding chan ging conductor size along the feeder need to be mentioned. The use of too many different wire sizes on a feeder complicates the maintenance of the feeder substantially. Ma ny utilities use one c onductor size for the main feeder and a second (smaller) size for laterals. Two differe nt wire sizes is a manageable situation since conductor, hardware, and tool s for only two conductor sizes must be inventoried in the warehouses and st ocked on the line trucks. If the number of conductor sizes in use grows much beyond two, the logistics of maintaining a working inventory becomes complex. The electrical assessment process illustrated in Fig. 12 can be repeated using a smaller conductor size on both the main and backup feeder from the sectionalizing recloser to the end of the feeder. While the smaller c onductor on half of the feeder will save construction cost, it will increase losses a nd increase voltage drop. These negative changes, however, are typically rather sm all, and are easily quantified by loadflow analysis. Economic Assessment After a set of viable feeder configurations are identifi ed, they can be assessed from an economic viewpoint. The outcome of the eco nomic assessment wi ll be the optimum feeder configuration for the se t of constraints considered.

PAGE 69

58 The first economic component to be determined is the cost of the substation to supply the feeder system. As described earlier in this dissertation, the subs tation will be based on the power ring topology. The grea test cost variable in the su bstation design is the cost of the transformers. It is assumed for this evaluation that the subtransmission voltage supplying the substation is 138 kV The secondary voltage has a great effect on the cost of the transformer, as does the MVA rating. These two parameters differ among the feeder configurations. The transformers should be sized such that a single transformer feeding the power ring could supply all feeders while operating at it s maximum output, plus an assumed level of contingency loading, as conti ngencies may occur at or near the maximum loading level of the substation. As an example, a 20 MVA p eak loading per feeder requires 80 MVA of peak capacity for the four-feeder substation. If the continge ncy loading is assumed to be that of another entire feeder, an additional 20 MVA of transformer capacity, or 100 MVA total, is required. The assumptions leading to this required capacity of 100 MVA are rather conservative, and much more conservative than the philosophies presently used by most U.S. utilities. Based on the shape of the load duration curve, it may be determined that the time at which the feeder system operates at peak load is small enough to represent a minimal exposure risk. If the start of the load durat ion curve is sufficiently steep, perhaps a lowerthan-peak load level, say 90% of peak, can be used to determine transformer capacity. This assumption would reduce the required su bstation transformer kVA by as much as

PAGE 70

59 10%, depending on the contingency capacity re quirements considered. If a contingency does occur at loading levels a bove this reduced load level, the resulting ove rloads could be remedied by shedding load. Load shedding is a measure of last resort to alleviate an overload, but if the probability of having to do so is sufficiently (and acceptably) small, the resulting transformer sizing should be c onsidered, as the cost reduction could be substantial. Allowing a level of contingency capacity equal to that of a nother entire feeder is also much more conservative than is typically found on U.S. utility systems. This assumption, however, is critical for proper in telligent sectionalization of the feeder system as it is presented in this dissertation. The conti ngency capacity requirement for the substation transformer could be redu ced by increasing the comple xity (and cost) of the sectionalizing equipment on the feeder system. At first observation, the cost of additional transformer capacity appears to be considerably less than the additional costs incurred to implement a considerably more complex f eeder system, but this topic should be thoroughly researched to as certain that speculation. Standard power transformer designs produce units with four MVA ratings according to the following criteria:

PAGE 71

60 Rating = W / X / Y / Z MVA, where W = self-cooled (base) rating at 55C rise, X = rating with one stage of forced cooling at 55C rise, Y = rating with two stages of forced cooling at 55C rise, Z = rating with two stages of forced cooling at 65C rise, and X = 1.333 W, Y = 1.667 W, Z = 1.875 W. To provide 80 MVA at maximum output, th e transformer rating would have to be 42.667 / 56.889 / 71.112 / 80.000 MVA. Under normal conditions, each transformer would be loaded to 40 80.000 = 50% of its 65C rise rating or 40 42.667 = 93.75% of its base rating. While this loading may seem overly conservative to some engineers, the increase in transformer life and flexibility during contingency switching needs to be considered. While the effect of overloads decreasing transformer life is well understood, a less understood phenomenon is the change in useful life as a result of limiting transformer loading to a value less than the rated value (IEEE C57.92-1981, P. 10). Monte Carlo analysis indicates that transformer failure rates will be reduced as loading is dropped

PAGE 72

61 below the nameplate rating (Fu, McCalley, and Vittal 2001, P. 347). This observation strengthens the argument to limit the loading of substation transformers to lower values than those commonly used throughout the industry. Once the proper transformer sizes are determ ined, the cost of the major substation equipment can be tabulated. It is not necessa ry to include all costs, particularly those costs appearing in all substation designs, such as protection equipment, grounding, foundations, etc. It is nece ssary to capture all costs th at differ between the designs. After the substation cost is determined, the cost of the four feeders can be added. Estimates of feeder construction are done on a dollars per mile basis. The design of a specific feeder can deviate substantially fr om the design which produced the average cost. This could be due to geographic constraints or environmental conditions. Such deviations are insignif icant for estimation purposes, as a ll the feeder designs for that service area would involve si milar deviations, and thus similar additional costs. Dividing the cost of the substa tion and the four feeders by 4l2 where l represents the effective service length of the f eeder gives a facilities cost in dollars per square mile. This facilities cost is a major co mponent of the economic assessment. As the maximum feeder service length increases, so does the distance between substations. While a greater distance be tween substations reduces the number of

PAGE 73

62 substations required to supply a region of give n size, the cost of each substation rises due to the higher transformer capacity required. The feeder construction costs for each conduc tor type and voltage class are summarized in a spreadsheet. Its template is shown in Ta ble 6, and the assumptions made to generate the costs are detailed in Appendix C. Substa tion transformer, circuit breaker, switch, and recloser costs for various kVA or current and voltage ratings are summarized in Table 7. Table 6 – Per-Mile Costs of Various Overhead Feeder Designs AAC conductor 12.47 kV 13.2 kV 13.8 kV 14.4 kV 22 kV 24.9 kV 34.5 kV kcmil codeword Typical Feeder Cost ($/mile) 266.8 Laurel 40,026 40,026 40,026 40,026 47,570 47,570 55,114 336.4 Tulip 44,250 44,250 44,250 44,250 52,327 52,327 58,800 397.5 Canna 49,100 49,100 49,100 49,100 58,062 58,062 63,700 477 Cosmos 54,540 54,540 54,540 54,540 62,084 62,084 69,628 556.5 Mistletoe 62,750 62,750 62,750 62,750 71,396 71,396 80,250 636 Orchid 74,420 74,420 74,420 74,420 84,675 84,675 96,150 715.5 Violet 84,224 84,224 84,224 84,224 96,822 96,822 109,421 795 Arbutus 95,660 95,660 95,660 95,660 109,409 109,409 126,945 900 Cockscomb 108,150 108,150 108,150 108,150 123,632 123,632 143,890 1033.5 Larkspur 122,760 122,760 122,760 122,760 140,322 140,322 163,340 1192.5 Hawthorn 130,390 130,390 130,390 130,390 151,550 151,550 172,685 1351.5 Columbine 143,773 143,773 143,773 143,773 163,915 163,915 184,058

PAGE 74

63Table 7 – Representative Equipment Costs Transformers Base Rating Price Vacuum Circuit Breakers 1200A 2000A 3000A 15-24 MVA $325,000 15.5 kV, 25 kA $16,500 $16,900 $22,500 138/14.4 kV 27-37 MVA $375,000 15.5 kV, 40 kA $18,500 $18,900 $24,500 27 kV, 25 kA $19,000 $19,900 $24,000 25-32 MVA $375,000 27 kV, 40 kA $21,500 $22,400 $26,500 40-48 MVA $450,000 138/24.9 kV 50-64 MVA $500,000 Breaker Disconnect Switches 1200A 2000A 3000A 15 kV $1,245 $1,545 $1,800 25 kV $2,150 $2,670 $3,050 Reclosers V, I(rated), I(int) Price 15kV, 800A, 12 kA $16,500 Line Disconnect Switches 1200A 2000A 3000A 25kV, 800A, 12 kA $18,500 15 kV $7,185 $9,155 $9,575 25 kV $7,255 $9,230 $9,660 It should be mentioned that the feeder leng ths and loadings determined by this process are substantially greater than those currently utilized by many utilitie s. This is due to consideration of various reliability aspects, which are decreased as feeder length and load increases. This decrease in reliability on long and heavily-lo aded feeders can be mitigated by changing the protection philos ophy applied to the feeder system. A novel protection concept using communi cation-based overcurrent detect ion is presented in this dissertation to allow the implementation of l onger feeders without a dversely affecting the reliability of the system. Consideration of Electrical Losses The final component of the economic assessment is the cost of the el ectrical losses on the system. These losses consist of demand and energy losses, and occur throughout the system, particularly in the transformers and the primary feeder conductor. Demand losses can easily be determined, as they are the kilowatt losses occu rring at the time of

PAGE 75

64 peak loading, as reported by th e load flow simulation. Energy losses are more difficult to determine, as loading levels vary over time. A load duration curve can be used to estimate energy losses over a given period. 0%10%20%30%40%50%60%70%80%90%100% % of Time 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%% of Feeder Peak Load Figure 14 – Load Duration Curve Fig. 14 shows a typical load dur ation curve for a distribution f eeder. Based on the curve, an estimation of energy losses for the feeder can be determined. This can be done by representing the load duration curve as a piecewise-linear model, then rescaling the model so that the y-axis shows kilowatts of load instead of percent of peak load. Applying this technique to Fig. 14 produces the piecewise-linear feeder loading model shown in Fig. 15.

PAGE 76

65 This % of Time This Cumulative % of Time 0%10%20%30%40%50%60%70%80%90%100% % of Time 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%% of Feeder Peak Load95%90%85%80%75%70%65%60% 6.3%12.1%15.4%12.2%14.6%16.7%13.9%8.8% 6.3%18.4%33.8%46.0%60.6%77.3%91.2%100.0% for or (in Percent of Feeder Peak Load) and +5% Between this Load Level Figure 15 – Piecewise Linear Feeder Loading Model Now, the model in Fig. 15 can be viewed one load level at a time, and the energy losses for each load level can be determined by usin g loadflow simulation. The load on the feeder is first uniformly scaled to represen t the first load level of the piecewise-linear model. Then, the loadflow is solved, and the calculated demand losses are multiplied by the duration, in hours, for wh ich that loading level persis ts, producing an average energy loss value for that load level. That resulting average energy loss value is saved. After each load level is analyzed in this way, the average energy losses are summed to estimate the total energy loss for the feeder over an entire year. But the economic assessment requires that th e demand and energy losse s be specified in dollars, not kilowatts and kilowa tt-hours. Some means of de termining the financial value of the losses must be determined. Ma ny methodologies, produci ng widely varying

PAGE 77

66 results, exist for determining these rates. These methodologies incl ude using a wholesale rate, replacement cost, or past investment data (Willis 1997, P. 32). If th e utility has excess generation capacity, a zero cost for lo sses may even be justified. The most theoretically-sound met hodology involves using marginal, or incremental, cost. This method involves calculating th e cost to produce one more kW of demand or one more kWh of energy. While this sounds relative ly simple, it is actually quite involved. The data for calculating incr emental cost can be obtained from the utility’s economic dispatch algorithm of the automatic genera tion control (AGC) system. This value changes continually, but monthly, seasonal, or annual averages can be determined. These average values will be used to assess system losses. When the cost of the losses is determined fo r a year, that value must be combined with the facilities cost. To put the facilities cost on an annual basis, an annual cost must be calculated. This calculation requires the de finition of a time period representing the life of the facilities, and the speci fication of a discount rate. Both of these values may be difficult to determine, but their numeric values are not of primary importance. The same numeric values are used to evaluate all confi gurations, and it is this commonality that is important. It is common practice to use comp any-specified values for equipment life and discount rate. The economic assessment process is summarized in Fig. 16.

PAGE 78

67 Figure 16 – Economic Assessment Determine Substation Select Feeder Design Cost Determine Cost for the Four Feeders Select Lowest Total Annual Cost Calculate Facilities Cost ($/mi )2Cost = Cost + 4 CostSub Feeder42l Determine Annual Cost of Facilities per mi Discount 30 yrs @ 10% Determine Demand Loss Determine Energy Loss Per Year Per Year Determine Annual Cost of Losses per mi Demand Cost + Energy Cost242l2 Determine Total Annual Cost for System per mi Facilities Cost + Loss Cost2

PAGE 79

68 Impact to Reliability Care must be taken so that system reliabili ty is not compromised by the increased feeder length proposed by this optimization method. As the feeder length increases, more customers are supplied by that feeder. The more customers supplied by a feeder, the more customer outage minutes will occur in the event of a feeder outage. This effect can be mitigated by making changes to the protec tion strategy applied to the feeder. At present, the protection strate gy used on virtually all distri bution feeders is based on overcurrent detection. The ci rcuit breaker in the substation is the first protective device on the feeder. It is common to use downs tream sectionalizing devices to reduce the number of customers impacted by a downstr eam fault. As overc urrent devices are applied in series, the operating times of the devices must be coordi nated to assure that one specific device operates more quickly th an the other devices. The fastest device provides the primary protection, and the slower devices serve in a backup capacity in the event the primary device fails to clear the fault. As microprocessor and communication technol ogies have evolved, many computerized protective devices capable of communicating with each other have been introduced into the utility industry. It is the recommendation of this dissertation to utilize these devices in innovative ways to protect the distribution feeders. Inst ead of adopting a conventional coordinated overcurrent prot ection scheme, a communicationbased overcurrent detection methodology was introduced earlier in this dissertation, which not only optimizes reliability levels, but also minimizes ci rcuit breaker operations, thereby decreasing maintenance requirements.

PAGE 80

69 The method introduced earlier in this dissertation of automati cally reconfiguring reclosers to isolate faults, called intelligent sectionalizing, greatly reduces customer outage minutes compared to traditional coordination-based fa ult isolation methods which typically rely on manual system reconfiguration. Because of intelligent sectiona lizing made possible by communication-based overcurre nt detection, longer and la rger capacity feeders than are currently acceptable by most U.S. utilitie s can be used without adversely impacting reliability. The assessment of feeder reliability is addressed in Appendix F.

PAGE 81

70 Distribution System Analysis Assessment of Conductor Size The size of the conductor used in a power circuit has many implications. It not only determines the ampacity of the circuit, but al so defines electrical ch aracteristics such as voltage drop and losses. Pr oper conductor sizing is certainly of prim ary importance in the design of any power circuit. When that power circuit is a di stribution feeder of substantial length, the importance of proper conductor selection becomes even greater. As with most system components, temperatur e rise is the most significant determining factor for the ampacity rating of a conductor. It is not the only factor, as conductor clearance (sag) may also be an important i nput to determining ampacity, but the operating temperature criteria is important to understand. Excessive heat tends to anneal metals, which causes permanent deformation (stretchin g) of the conductor as well as a decrease in tensile strength. This damage is irrevers ible. Keeping the opera ting temperature of the conductor below the design maximum temperature is essential to maintain the design life of the circuit (Westinghouse 1964, P. 47). As the metallurgical composition of the c onductor is varied, so too is its maximum allowable operating temperature. The most commonly used bare distribution conductors are those of the AAC (all-aluminum conduc tor) family. These conductors, made of

PAGE 82

71 aluminum alloy 1350-H19 wires concentrically laid, have a maximum design temperature of 75C. By using steel reinforcement or ex otic alloys, the operati ng temperature can be raised significantly above 75C, but these elevated operating temperatures come at a significant cost premium. Non-AAC conductors may be desirable for special applications, particularly for long spans requi red to cross rivers or highways, but AAC conductors are ideal for distribut ion feeder use, both for econ omy and ease of installation. Although the ampacities of AAC conductors may be lower for a given wire size than for other conductor families, loadings approach the thermal rating only during the most severe contingencies. Most of the time, c onductor loadings are below 50% of the thermal limit. Some engineers may view this low averag e loading of the distribution feeder as underutilized equipment or excessive unused capacity. It must be emphasized that the thermal wire rating is but one parameter that determines the optimal ampacity of the circuit. As an extreme example, consid er a 500 kV transmission line built with 2156 kcmil ACSR (bluebird) conductor bundled 3-per-phase. Although 2156 kcmil ACSR has an ampacity of 1622 amperes (at 75C wire temperature), it would be unrealistic and absurd to propose a rating of 080 214 4 3 ) 3 1622 ( 500 kVA, or 4,214 MVA, for the circuit. This is because the thermal rating is only one of several rating parameters for the transmission line. Long before reaching the thermal rating of the conductor, conductor sag or system stability will limit the circuit’s rating. Similarly, while thermal rating may be a controlling design parame ter under contingency loading, it is insignificant when considering normal load leve ls. Electrical losses are very significant

PAGE 83

72 under normal loading, as is demonstrated by the economic assessment process. Low (compared to the thermal limit) average loadings are instrumental in keeping losses small. Utilization of Circuit Breaker Rating The circuit breaker rating is in significant as a parameter to determine the ampacity of a distribution feeder. The minimum continuous current rating for me dium-voltage outdoor distribution-class circuit break ers available today is 1200 ampe res. A feeder rating of 1200 amperes would require prohibitively-la rge conductors. To limit the conductor temperature to 75C when the ambient conditions include 25C air temperature, sea level elevation, a wind velocity of 2 feet per sec ond, full sun exposure, an emissivity of 0.5, and a solar absorption coefficient of 0.5, an all-aluminum conductor would require a minimum cross-sectional area of 1351.5 kcmil to achieve an ampacity of 1200 amperes. At a weight in excess of 1.2 pounds per foot and a diameter of 1.34 inches, 1351.5 kcmil AAC would require very short span lengths a nd high-strength poles to meet the necessary structural requirements. These design criter ia make the construction of such a circuit prohibitively expensive when compared to the cost of a circuit utilizing smaller conductor, as shown in Table 6 on Page 62. Allowing conductor size to driv e the feeder rating, a reasona ble conductor size selection of 795 kcmil AAC provides an ampacity of approximately 880 amperes. An argument can be made that a 1200-ampere breaker loaded to only 880 amperes (at most) is underutilized. This line of reasoning can lead to poor system design. Contingency

PAGE 84

73 capacity must be reserved to allow for opera ting flexibility. The amount of contingency capacity maintained on the distribution system va ries from utility to utility, ranging from close to 50% (which allows backup of an entir e comparably-sized feed er) to close to zero (which makes contingency switching almost impossible). Because of the need to reserve capacity for contingency loading, making full use of the conductor’s thermal capacity at normal loading levels is not necessary, or ev en feasible, to assure an economical and efficient design. Main Feeder Analysis Let us assume that each substation has four feeders, one built in each direction of the compass. Let us also assume that the substations are arranged in a rectangular grid pattern of uniform spacing, and that the load density is uniform. These assumptions define a service area for each f eeder in the shape of an isos celes right triangle, where the height of the triangle is the feeder length (l) and the base is 2l. The service area for the east feeder is shaded in Fig. 17. Three-phase laterals, built perpendicular to the main feeder and constructed with a smaller conductor, serve the load in regions located too far from the main feeder to be served directly. Those laterals can be analyzed using the same methods as on the main feeder. Figure 17 – Layout of Main Feeder and Laterals l 2l

PAGE 85

74 Letting x denote the distance, in meters, along the feeder from the substation breaker, the change in current per length of feeder can be represented as x I k dx x dI 0 ) ( Eq. 7 where I (0) is the current at th e substation breaker, and –k is a constant to account for the linearly-distributed load along the feeder. To determine k integrate Eq. 7. lldx x I k x dI00) 0 ( ) (, Eq. 8 which yields l lx I k x I0 2 02 ) 0 ( ) ( Eq. 9 Evaluating limits, 2 ) 0 ( ) 0 ( ) (2l I k I l I Eq. 10 Since the current at th e end of the feeder I ( l ) is zero,

PAGE 86

75 2 ) 0 ( ) 0 ( 02l I k I Eq. 11 Dividing through by I (0), 2 12l k or 22 l k Eq. 12 Substituting this value of k into Eq. 7 gives x I l dx x dI 0 2 ) (2 Eq. 13 Integrating to express current as a function of distance down the feeder, X Xdx x I l x dI0 2 00 2 ) (, Eq. 14 which yields X Xl x I x I0 2 2 0) 0 ( ) ( Eq. 15 Evaluating limits, 2 2) 0 ( ) 0 ( ) ( l X I I X I Eq. 16

PAGE 87

76 Solving for I(X) 2 2 21 ) 0 ( ) 0 ( ) 0 ( ) ( l X I l X I I X I Eq. 17 The voltage drop across a differential length segment at some point along the feeder is the product of the current flowing through the feeder at that point, I ( x ), and the series impedance of the feed er per unit length, z dx z x I x dVdrop) ( ) ( Eq. 18 Substituting the expression for I(X) and integrating to find Vdrop(X), 2 2 00 2 23 1 ) 0 ( ) 0 ( ) 0 ( ) ( l X X z I dx l x z I dx z I X VXX drop. Eq. 19 The real power loss for the main feeder can be expressed as l l lossdx r l x l x I dx r l x I P0 4 4 2 2 2 0 2 2 2 22 1 ) 0 ( 1 ) 0 (. Eq. 20

PAGE 88

77 Integrating and evaluating limits, 2 2 0 4 5 2 3 2) 0 ( 15 8 5 1 3 2 1 ) 0 ( 5 3 2 ) 0 ( I l r l r I l x l x x r Il Eq. 21 Quantification of Main Feeder Reliability The reliability of a distribution system can be quantified by measuring customer outage minutes (COM). COM is simply the product of the number of failures over a period of time, the number of customers affected by each outage, and the time required to restore service to the outaged customers. On a year ly basis, the annual customer outage minutes (ACOM) for a distribution feeder can be expressed as ACOM = (number of failures per year) Eq. 22 (number of customers interrupted per failure) (service restoration time). The first term, representing the failure rate of the feeder, is best estimated by tracking historical data. An underlying assumption of this research is to work with existing infrastructure as opposed to a greenfield project. This assumption is crucial to assure that any findings to improve system reliability ca n be economically and feasibly integrated into the present system. Of course, any ch anges in equipment or technology that will reduce the failure rate of the feeder will proportionately reduce ACOM too.

PAGE 89

78 The second and third terms are the number of customers affected by each interruption of service and the time necessary to restore service to them. These terms are determined by the feeder topology, specifically how the feeder is sectiona lized. Looking at a simple example of a feeder with no sectionalizing, a ll faults on that feeder result in an outage that affects every customer on the feeder, a nd no customer’s service is restored until the problem causing the fault is repaired. As sectionalizing devices are introduced to the feeder, each outage potentially affects fewer customers and for a shorter duration. And the sectionalizing can be actuated either manually or automatically. Automatic feeder sectionalization will be referred to as intelligent sectionalization and can be achieved by implementing communication-based overcurrent de tection. Each type of sectionalizing has a significant impact on the reliability of th e feeder. These relationships are quantified in the next sections. Looped Feeder Reliability Wit hout Intelligent Sectionalizing Consider a main feeder of length L with a single normally-closed sectionalizing device, such as a recloser, located at a distance m per-unit of the feeder length from the power ring. The feeder terminates at a normallyopen recloser that ties to a similar feeder supplied from another power ring, making it a looped feeder with a normally-open point in the loop to make it radial. Assume that the feeder serves C customers, but the load density is not necessarily uniform. A per-unit multiplier n defines the portion of the customers served by the first segment (the segment closer to the power ring) as nC The remaining (1– n ) C customers are served from Segment 2 of the feeder.

PAGE 90

79 Seg. 2 {Length = (1-m) L} Customers Served: (1-n) C POWER RING for Fdr. 1) (Normal Source Fdr. 1 {Length = L} Customers Served: nC Seg. 1 {Length = mL} R N.C. POWER RING (Backup Source for Fdr. 1 and Normal Source for Fdr. 2) R N.C. Fdr. 2 N.O. R Figure 18 – Feeder Configuration Without intelligent sectionalizing, the ACOM of the feeder is the sum of the ACOM values for each feeder segment, or ACOM Fdr w/o IS = ACOM Seg 1 + ACOM Seg 2 Eq. 23 The two feeder segments experience different restoration times after a fault occurs. Although the entire feeder is interrupted wh en a fault occurs, customers on the unfaulted segment can be restored as quickly as the s ectionalization can be performed. Customers on the faulted segment, however, cannot be restored until the problem causing the fault is repaired. These two times may differ by a large amount. Defining S to be the time necessary to se ctionalize the faulted segment and R to be the time required to repair the damage causing th e fault, the ACOM for the feeder can be expressed as

PAGE 91

80 ACOM Fdr w/o IS = f m L [ n C R + (1 – n ) C S ] Eq. 24 + f (1 – m ) L [(1 – n ) C R + n C S ] The first term represents the ACOM due to faults occurring on Segment 1, while the second term accounts for outage minutes due to faults occurring on Segment 2. The factors of the first term outsi de the brackets represent the nu mber of failures on the first segment. The first term inside the brackets of the first term accounts for the outage minutes of the customers on Segment 1 due to faults on Segment 1, while the second term inside the brackets of the first term re presents the outage mi nutes of customers on Segment 2 affected by a fault on Segment 1. The second term corresponds to the same components for faults occurring on Segment 2. Expanding Eq. 24 and recombining terms and factoring, ACOM Fdr w/o IS = f m L [ n C R + (1 – n ) C S ] Eq. 25 + f (1 – m ) L [(1 – n ) C R + n C S ] = f m L n C R + f m L C S – f m L n C S + ( f L – f m L ) ( C R – n C R + n C S ) = f m L n C R + f m L C S – f m L n C S + f L C R – f L n C R + f L n C S – f m L C R + f m L n C R – f m L n C S

PAGE 92

81 = f L C ( m n R + m S – m n S + R – n R + n S – m R + m n R – m n S ) = f L C [ R (1 + 2 m n – m – n ) + S ( m – 2 m n + n )]. Looped Feeder Reliability With Intelligent Sectionalizing When the sectionalizing devices on the main feeder are provided with “intelligence” to determine how and when to operate, faulted feeder segments can be sectionalized by automatic device operation. Automatic sec tionalization can be done quite quickly — in well less than a minute. Since one minute is the threshold for accumulating outage time, as defined by IEEE Std. 1366-1998, automatic sectionalization can greatly improve a feeder’s reliability indices. Of course, mome ntary interruptions will still exist, and the consequences of momentary outages can be of major concern, but out age duration will be reduced significantly when intelligent sectionalization is implemented. Implementing intelligent sectionalization essentially reduces switching time for feeder reconfiguration to zero. Adjusting Eq. 21 to reflect S = 0 yields ACOM Fdr w/ IS = f L C [ R (1 + 2 m n – m – n ) + 0 ( m – 2 m n + n )] Eq. 26 = f L C R (1 + 2 m n – m – n ).

PAGE 93

82 S n n m m n m n m R n n m m S n m n m R C L f n n m m S C L f ACOM ACOM ACOM ACOM ACOM ACOMIS o w Fdr Fdr IS o w Fdr IS w Fdr IS o w Fdr Fdr 2 2 1 100 % 100 2 2 1 2 % 100 % 100 %/ / / /The change is annual customer outage minutes as a result of intelligent sectionalizing can be expressed as Eq. 27 Eq. 27 clearly shows that the improvement in annual customer outage minutes for the feeder is a function of repair time ( R ), sectionalizing time ( S ), and the location of the sectionalizing device on th e feeder (which determines m and n ). The improvement in ACOM for the feeder due to intelligent sectionalizing in not a function of the feeder failure rate ( f ), the length of the feeder ( L ), or the number of customers served by the feeder ( C ), since these quantities are not changed by the addition of intelligence to the sectionalizing devices.

PAGE 94

83 Design Example The design optimization methods proposed in th is dissertation will now be demonstrated in an example. A service area with a presen t average load density of 720 kVA per square mile will be considered. A good planning practice is to consider future load development in the area when designing the infrastructure. Load modeli ng in this hypothetical example indicates another 7% of growth is likely before load sa turation occurs. So, the initial load density of 720 kVA/mi2 is multiplied by 1.07 to yield 770 kVA/mi2. A complete analysis of the design options w ould involve consideration of all available operating voltages. The 25 kV cl ass of equipment, shown in columns I and J of Table 3, has a marked advantage over the 15 kV class equipment because of the lower currents required at the higher voltages. Indeed the 35 kV class shows an even larger decrease in current over the 15 kV class than does the 25 kV class, but along with this decrease in current comes a substantial increase in cost.

PAGE 95

84Table 8 – Comparison of 25 kV and 35 kV Equipment Costs Transformers Base Rating Price Vacuum Circuit Breakers 1200A 2000A 3000A 25-32 MVA $375,000 27 kV, 25 kA $19,000 $19,900 $24,000 40-48 MVA $450,000 27 kV, 40 kA $21,500 $22,400 $26,500 138/24.9 kV 50-64 MVA $500,000 38 kV, 25 kA $26,000 $27,500 $34,000 38 kV, 40 kA $27,000 $29,000 $35,500 25-32 MVA $390,000 40-48 MVA $470,000 Breaker Disconnect Switches 1200A 2000A 3000A 138/34.5 kV 50-64 MVA $520,000 25 kV $2,150 $2,670 $3,050 35 kV $3,800 $4,350 $4,700 Reclosers Vrated, Irated, Iint Price 25kV, 800A, 12 kA $18,500 Line Disconnect Switches 1200A 2000A 3000A 35kV, 800A, 12 kA $24,000 25 kV $7,255 $9,230 $9,660 35 kV $11,000 $11,800 $12,350 Industry experience with 35 kV-class equi pment also shows a decrease in many reliability categories when compared to 25 kV-class equipment. For those reasons, this example will focus on the 15 kV and 25 kV equipment classes. To further reduce the number of configuratio ns to analyze, the most optimum operating voltage for each voltage clas s will be the only voltages considered: 14.4 kV for the 15 kV class of equipment, and 24.9 kV for the 25 kV e quipment class. In an actual application, the utility’s present distribution system operating voltage would also be assessed. The set of conductors listed in Table 2 on Page 30 are the conductor sizes that will be considered in this example. In an actual application, other conduc tor sizes presently in use by the utility may also be considered.

PAGE 96

85 Following the procedure illustrated in Fig. 12, the electrical assessment is performed. The first conductor and operating voltage anal yzed is 266.8 kcmil Laurel operating at 14.4 kV. Using Eq. 3, the maximum service lengt h of this feeder is found to be 3.8 miles when the load density is 770 kVA/mi2. miles mi / kVA A kV density load I kV mslrated al min no79 3 770 3 444 4 14 32 Eq. 28 In order for this feeder to be able to bac kup another feeder of similar size, a contingency capacity of 50% is required. Using Eq. 5, the effective service le ngth is found to be miles miles msl cc esl 68 2 79 3 100 50 100 100 100 Eq. 29 Based on this effective service length, Eq. 6 can be used to determine the load to be served by this feeder. kVA miles mi / kva esl density load load feeder 5530 68 2 7702 2 2 Eq. 30 Next, a load power factor must be assumed so that the watt and var components of the feeder load can be calculated. This ex ample assumes a load power factor of 90% lagging. This leads to fe eder load components of

PAGE 97

86 kW kVA loadreal2977 9 0 5530 and Eq. 31 kVAR cos sin kVA loadreactive2410 9 0 55301 Eq. 32 One-tenth of the real and reactive feeder lo ad components are modeled at each of the ten nodes of the main feeder and the backup f eeder. The conductor resistance in ohms per mile is multiplied by the feeder length in m iles to find the total feeder resistance. The resistance between two adjacent feeder nodes is one-tenth of the total. Since the reactance is determined by the specific geometry of the feeder design (the distance between each of the three conductors), a re asonable X/R ratio for a circuit of this operating voltage is assumed. This value is ta ken to be 5. The reactance, then, between two adjacent nodes is five times the resistance between those nodes. The main feeder is modeled as Zone 1 of the load flow network, and the backup feeder is modeled as Zone 2 to facilitate the com putation of losses for each feeder. Each substation bus is defined to be a swing bus with a scheduled voltage of 1.05 per-unit. The sectionalizing reclosers are modeled closed and the tie recloser is modeled open. Appendix A shows the details of the PowerWorld model. At this point, the base case is solved. The out put is shown in Fig. 19 of Appendix B. The voltage at the end of each f eeder is within acceptable li mits (0.9536 per-unit), and the current in the first segment of each feeder is well below the conductor ampacity (196 amperes, or 44.2%). This indicates a feasible base case.

PAGE 98

87 Next, the most severe contingency must be anal yzed. This is when the entire main feeder is supplied by the backup feeder. Opening th e main breaker of the main feeder (between Nodes 1 and 2), and closing the tie recloser si mulates this contingency. When the model is solved, voltage violations are indicated fr om Node 19, where the voltage is 0.8881 perunit, to the end of the feeder, where the vo ltage deteriorates to 0.5264 per-unit. The current in the first segment of feeder (696 amperes) is also greater that the conductor ampacity by 57.11%, due to the large reacti ve component of the current. Both the voltage and overload problems can be remedied by adding shunt capacitors to the model. A trial-and-error approach, adding capaci tance in 600 kVAR increments, leads to capacitor values of 2400 kVAR, one installed at Node 7 and th e other at Node 16. This output is shown in Fig. 20 of Appendix B. By using no more than one capacitor bank per f eeder segment, the possibility of back-toback switching problems is eliminated. Adding the capacitance in multiples of 600 kVAR assures equipment availability, as 600 kVAR is a standard capacitor unit size. The next step is to model the 2400 kVAR capac itors and rerun the base case to show the effect of the capacitors on the voltage prof ile of the feeder system in its normal configuration. The electrical parameters in this scenario remain within acceptable limits (1.0120 per-unit voltage at the end of the feed er). Based on the analysis with the 2400 kVAR capacitors in place, this contingenc y configuration is c onsidered feasible.

PAGE 99

88 Before proceeding to the economic assessment, an attempt would normally be made to reduce the conductor size between the sectionalizing recloser and the tie recloser on each feeder. But this case utilizes the smallest co nductor in the selection set, so reducing the conductor size is not a possibility. An example of reducing the conductor size downs tream of the sectionalizing recloser will be shown now considering the 14.4 kV implementation of 1192.5 kcmil Hawthorn conductor. Based on the highest currents flow ing in the second segm ent of each feeder during the most severe contingency, 795 kcmil Arbutus is considered as a possibility for that conductor, as it is the smallest conductor with sufficient ampacity to handle the current during contingency loading. First, an electrical assessment must be done to assure that the 795 kcmil Arbutus performs ad equately from a voltage drop standpoint. If not, the next larger conductor must be te sted. When the minimum acceptable conductor is determined, each conductor size between that minimum size and the conductor size used on the first feeder segment (1192.5 kcmil Hawthorn in this case) must be analyzed, as these calculated loss values must be assessed economically. When the 795 kcmil Arbutus is modeled between nodes 7 and 16 of the PowerWorld model, the base case analysis is acceptable with no shunt cap acitors modeled. But during the contingency case, such seve re voltage drops occur that the load flow algorithm fails to converge with no capacitors. The reactive power requirement of the feeder in this configuration is very high. Adding shunt cap acitors allows the algorithm to converge, but very large capacitors are required. Banks of 7.2 MVAR must be added to each feeder

PAGE 100

89 to bring the end-of-the-feeder voltage at Node 2 close to the acceptable range (0.8893 per-unit) during the contingency case. Such heavy compensation is ill-advised for many reasons. A single capacitor bank of such large size produces a very large voltage change when switched (6.29% in this example). This magnitude of voltage change is unacceptable, as voltage-sensitive loads would be adversely affected. Dividing the total capacitance into several smaller banks not only increases both installation and maintena nce costs, but also poses the possibility of back-to-back switching problems if the di stance between the banks is not sufficiently large. The next larger conductor, 900 kcmil Cock scomb, performs satisfactorily under the contingency case with considerably less reac tive compensation. This is a far better engineering solution than the heavily-compe nsated Arbutus conductor, so it will be considered a viable configuration. Now, the economic assessment can be performed. The process outlined in Fig. 16 is followed for each electrically-viable design optio n. This straightforward process is very data intensive, so computer-based analysis is advisable. A total annua l cost is calculated for each design option. The constrained optimal solution is identified by the lowest total annual cost.

PAGE 101

90 Conclusion Electric power distribution sy stems are made up of many components arranged into subsystems such as the subtransmission ne twork, the distribution substations, and the primary distribution feeders. While the s ubsystems can be addressed individually for many purposes, the planning and design of the distribution system needs to be approached considering the system as a whol e, not on a subsystem by subsystem basis. This integrated design approach is a requisite for reaching optimum conditions. The distribution substation and primary feed ers are two subsystems whose designs are highly dependent on each other. The substa tion topology has a pronoun ced effect of the reliability of the system. It also influences not only the design, but also the operation of the distribution feeders. A specifi c ring bus configuration, termed a power ring is analyzed and recommended as a superior topology for distribution substations. Since feeder operation so greatly influences the reliability of the system, special consideration must be made to assure a s ubstation / primary feeder combination that facilitates prompt restoration of service after an outage. Intelligent sectionalizing is implemented using a communication-based overcu rrent detection system This method of system protection is similar to scheme s used for years on transmission systems. Because of the proliferation of micropr ocessor-based protective devices in the

PAGE 102

91 distribution substation and the technical and economic improvements in fiber optic communications, applying sophisticated pr otection and automation methods on the distribution is not only feasible, but in many cases, economically attractive, particularly when the financial impacts of power outages are severe. The implementation of the innovations proposed in this work is only feasible if the improvements can be made in stages, on as as-needed basis. The methods proposed can coexist with existing technologi es. In fact, many substati ons today contain elements recommended in this dissertation. Many pr imary feeder designs are also generally compatible with the recommendations of th is research. By gr adually upgrading the existing system in stages on an as-needed basis, the economics of the upgrade become attractive, with reasonable payback periods and affordable costs. The distribution system described in this di ssertation not only meets the needs of today’s electric loads, but shows enough versatility to adapt to future load requirements that we cannot quantify today. New t echnologies to enhance power quality, improve restoration characteristics, and raise the efficiency of th e system can be integrated into the system design, allowing for a long useful life. By applying an integrated optimal design method, the resulting distribution promises to meet th e needs placed on the system today and well into the future.

PAGE 103

92 References Bernstorf, R. A Brief History of Non-Ceramic, Polymer, Composite Insulators Presented to the Southeastern Electric Exchange May 28, 1992. Brewer, S. Polymer Solutions to Co ntaminated Environments Ohio Brass. Presented to the Southeastern Electric Exchange – September 13, 1994. Brown, R. Electric Power Distribution Reliability Marcel Dekker. 2002. ISBN 0-82470798-2. CBEMA Curve Application Note Technical Committee #3 (TC-3) of the Information Technology Industry Council (ITI), 2000. De La Ree, J., Elizondo, D., Depablos, J., and Stoupis, J. An Adaptive Protection Scheme for Power Distribution Systems presented at Power System and Communication for the Future, Beijing, September 2002. Edison Electric Institute (EEI) Utilit y Data Institute (UDI) database, 2004. Electrical Transmission and Di stribution Reference Book Westinghouse Electric Corporation. 1964. Evaluation of Partially-Insulated Cable at 33 kV, Provincial Electricity Authority internal report, Bangkok, Thailand, 2002. Fehr, R. A High-Performance Distribution Substation Bus Topology Proceedings of the Seventh IASTED International Conf erence on Power and Energy Systems Clearwater Beach, Florida, Nov. 28-Dec. 1, 2004. Fehr, R. Industrial Power Distribution Prentice Hall 2002. ISBN 0-13-066462-6. Fu, W., McCalley, J., and Vittal, V. Risk Assessment for Transformer Loading. IEEE Transactions on Power Systems, Vol. 16, No. 3, August 2001. Gnen, T. Electric Power Distribution System Engineering McGraw-Hill, 1986. IEEE Guide for Loading Mineral-Oil-Immersed Power Transformers Up to and Including 100MVA with 55 or 65 average winding rise IEEE/ANSI C57.92-1981 (Reaffirmed Jan. 1992).

PAGE 104

93 Moxley, R. and Fodero, K. High-Speed Distribution Protection Made Easy: Communications-Assisted Protection Sc hemes for Distribution Applications IEEE, 2005. National Electrical Safety Code. ANSI/IEEE C2-2002. PowerWorld Simulator, Version 8.0, Build 11/02/01, PowerWorld Corporation, Urbana, Illinois, 2001. Primary Cable Installation Procedures, Provinc ial Electricity Authority internal report, Bangkok, Thailand, 2003. Schweitzer, E., Behrendt, K., and Lee, T. Digital Communications for Power System Protection: Security, Availability, and Speed presented at the 25th Annual Western Protective Relay Confer ence, Spokane, Washington, 1998. Southwire Company, Full-Line Product Catalog 2003. Willis, H., Welch, G., and S chreiber, R. Aging Power Delivery Infrastructures Marcel Dekker Inc., 2001. Willis, H. Power Distribution Planning Reference Book Marcel Dekker Inc., 1997.

PAGE 105

94 Bibliography Anderson, P. Analysis of Faulted Power Systems Iowa State University Press. 1973. ISBN 0-8138-1270-4. Apostolov, A. Improving Power Quality by Optimizing the Protection of Distribution Systems The Institution of Electrical Engineers, 2004. Apostolov, A. and Muschlitz, B. Practical Realities and Future Potential – Implementation Benefit s of DNP3 and IEC61850 DistribuTECH Conference and Exhibition, January 2004. Application of Peer-to-Peer Commu nications for Protective Relaying IEEE PSRC, WG H5, web published report, available at http://www.pes-psrc.org/h/H5doc.zip. Benner, C. and Russell, B. Practical High-Impedance Fault Detection on Distribution Feeders IEEE Transactions on Industry Appl ications, Vol. 33, No. 3, May/June 1997. Burke, J. Power Distribution Engineering Fundamental and Application Dekker Editions. Carroll, D., Dorfner, J., Lee, T ., Fodero, K., and Huntley, C. Resolving Digital Line Current Differential Relay Security and Dependability Problems: A Case History presented at the 29th Annual Western Protective Relay Conference, Spokane, Washington, 2002. Clarke, E. Circuit Analysis of A-C Power Systems John Wiley and Sons. 1943. Depablos, J. Internet Peer-to-Peer Communicati on Based Distribution Loop Control System Master’s Thesis: Virginia Polyt echnic Institute and State University, Department of Electrical and Comp uter Engineering, May 6, 2003. Digital Communications for Relay Protection Working Group H9 of the IEEE Power System Relaying Committee. Distribution Line Protection Pr actices – Industry Survey Results IEEE Power System Relaying Committee Report, IEEE Transactions on Power Delivery, Vol. 10, No. 1, January 1995.

PAGE 106

95 Domijan, A. and Song, Z. Flexible, Reliable, and Intellig ent Electrical eNergy Delivery Systems CRC Press LLC, Handbook of Power Electronics, 2002. ISBN 0-84937336-0. Domijan, A., Nara, K., and Hasegawa, J. FRIENDS: A Strategic Roadmap for Electric Utility Infrastructure and Services in a Competitive Environment Proceedings of the 2nd International Meeting on Systems Technology for Unbundled Power Quality Services, Feb. 5-7, 1999, Gainesville, Florida, pp. 5-14. Electric Utility Engineering Refe rence Book: Distribution Systems Vol. 3. Westinghouse Electric Corporation. 1965. Electrical Distribution: System Protection 3rd Edition. Cooper Power Systems. 1990. Bulletin 90020. Fairman, J., Zimmerman, K., Gr egory, J., and Niemira, J. International Drive Distribution Automation and Protection presented at the 55th Annual Georgia Tech Protective Relaying Conf erence, Atlanta, Georgia, 2001. Featheringill, W. Power Transformer Loading, IEEE Trans. Industry Applications vol. 19, no. 1, pp. 21–27, Jan./Feb. 1983. Feltis, M., Kumm, J. Applying the SEL-501 Relay for Fast Bus Trip and Simple Bus Breaker Failure Protection SEL Application Guide AG94-11, Schweitzer Engineering Laboratories, Inc. Galea, M. Rapid Spanning Tree in Industrial Networks available at http://www.ruggedcom.com/Pdf/Wpapers /RapidSpanningTreei nIndustrialNetwor ks.pdf. GE Industrial Power Systems Data Book General Electric Co. 1953. Guzman, A., Roberts, J., Zimmerman, K. Applying the SEL-321 Relay to Permissive Overreaching Transfer Trip (POTT) Schemes SEL Application Guide AG95-29, Schweitzer Engineering Laboratories, Inc. IEEE Guide for Loading Mineral-Oil-Immersed Overhead and Pad-Mounted Distribution Transformers Rated at 500k VA and Less with 55 or 65 Average Winding Rise IEEE/ANSI C57.91-1981 (Reaffirmed Jan. 1992). IEEE Guide for Loading Mineral-Oil-Immersed Power Transformers Rated in Excess of 100MVA IEEE/ANSI C57.115-1991. IEEE Recommended Practice for the Design of Reliable Industrial and Commercial Power Systems IEEE Standard 493-1990 (IEEE Gold Book), Sept. 1990.

PAGE 107

96 IEEE Standard 1366-1998: Trial Use Guide for Electric Power Distribution Reliability Indices. Lat, M. Application Guide for Surge Arre sters on Distribution Systems. CEA Contract Number 077-D-1 84A. Lenk, D., Koepfinger, J., and Sakich, J. Utilization of Polymer-Enclosed IntermediateClass Arresters to Improve the Performance of Modern Power Systems IEEE T&D Transactions Dallas Meeting 1991. Moxley, R. Analyze Relay Fault Data to Improve Service Reliability presented at the 30th Annual Western Protective Relay C onference, Spokane, Washington, 2003. Perez, L., Sorrentino, E., Hernandez, A., Urdaneta, A., and Bermudez, J. Applications of Adaptative Reclosers to Auto matic Distribution Systems 0-7803-2672-5/95. IEEE. 1995. Roberts, J., Stulo, T., and Reyes, A. Sympathetic Tripping Problem Analysis and Solutions presented at the 24th Annual West ern Protective Relay Conference, Spokane, Washington, 1997. Rosselli, G. and Fodero, K. How to Use Current Differential Relaying over Digital Phone Lines Presented at the 31st Annual West ern Protective Relay Conference, Spokane, Washington, 2004. Scheer, G. Answering Substation Automation Ques tions through Fault Tree Analysis presented at the Fourth Annual Substa tion Automation Conference, Texas A&M University, College Station, Texas, 1998. SEL-421 Relay User’s Guide Schweitzer Engineering Laboratories, Inc., available at http://www.selinc.com/instru ction_manual/421/421_UG_20040602.pdf. Soudi, F. and Tomsovic, K. Optimal Trade-Offs in Dist ribution Protection Design IEEE Transactions on Power Delivery, Vol. 16, No. 2, April 2001. Soudi, F. and Tomsovic, K. Optimized Distribution Protection Using Binary Programming IEEE Transactions on Power Deliv ery, Vol. 13, No. 1, January 1998. Transmission Line Reference Book Electric Power Resear ch Institute. 1978.

PAGE 108

97 Appendices

PAGE 109

98 Appendix A: PowerWorld Feeder Model The data shown in this appendix represents the PowerWorld feeder model used for the Design Example 14.4 kV operating voltage case for 266.8 kcmil Laurel conductor. Table 9 – Bus Records Number Area PU Volt Volt (kV) Angle (Deg) Load MW Load Mvar Gen MW Gen Mvar 1 1 1.05 15.12 0 5.12 0.35 2 1 1.04544 15.054 -0.95 0.05 0.02 3 1 1.04123 14.994 -1.9 0.15 0.07 4 1 1.03764 14.942 -2.82 0.25 0.12 5 1 1.03491 14.903 -3.71 0.35 0.17 6 1 1.03325 14.879 -4.55 0.45 0.22 7 1 1.03287 14.873 -5.3 0.55 0.26 8 1 1.02498 14.76 -5.87 0.65 0.31 9 1 1.01872 14.67 -6.32 0.75 0.36 10 1 1.01431 14.606 -6.65 0.85 0.41 11 1 1.01199 14.573 -6.82 0.95 0.46 12 1 0.95356 13.731 -6.68 0.95 0.46 13 1 0.95603 13.767 -6.49 0.85 0.41 14 1 0.96071 13.834 -6.12 0.75 0.36 15 1 0.96736 13.93 -5.61 0.65 0.31 16 1 0.97575 14.051 -4.97 0.55 0.26 17 1 0.98564 14.193 -4.25 0.45 0.22 18 1 0.99685 14.355 -3.45 0.35 0.17 19 1 1.00913 14.532 -2.6 0.25 0.12 20 1 1.02225 14.72 -1.73 0.15 0.07 21 1 1.03595 14.918 -0.86 0.05 0.02 22 1 1.05 15.12 0 5.14 3.09

PAGE 110

99 Appendix A – (continued) Table 10 – Branch Records From Bus To Bus Circuit Status Xfrmr R X C 1 2 1 Closed No 0.07215 0.36072 0 2 3 1 Closed No 0.07215 0.36072 0 3 4 1 Closed No 0.07215 0.36072 0 4 5 1 Closed No 0.07215 0.36072 0 5 6 1 Closed No 0.07215 0.36072 0 6 7 1 Closed No 0.07215 0.36072 0 7 8 1 Closed No 0.07215 0.36072 0 8 9 1 Closed No 0.07215 0.36072 0 9 10 1 Closed No 0.07215 0.36072 0 10 11 1 Closed No 0.07215 0.36072 0 11 12 1 Open No 0.07215 0.36072 0 12 13 1 Closed No 0.07215 0.36072 0 13 14 1 Closed No 0.07215 0.36072 0 14 15 1 Closed No 0.07215 0.36072 0 15 16 1 Closed No 0.07215 0.36072 0 16 17 1 Closed No 0.07215 0.36072 0 17 18 1 Closed No 0.07215 0.36072 0 18 19 1 Closed No 0.07215 0.36072 0 19 20 1 Closed No 0.07215 0.36072 0 20 21 1 Closed No 0.07215 0.36072 0 21 22 1 Closed No 0.07215 0.36072 0 Table 11 – Load Records Number ID Status MW Mvar MVA 2 1 Closed 0.050.020.05 3 1 Closed 0.150.070.17 4 1 Closed 0.250.120.28 5 1 Closed 0.350.170.39 6 1 Closed 0.450.220.5 7 1 Closed 0.550.260.61 8 1 Closed 0.650.310.72 9 1 Closed 0.750.360.83 10 1 Closed 0.850.410.94 11 1 Closed 0.950.461.06 12 1 Closed 0.950.461.06 13 1 Closed 0.850.410.94 14 1 Closed 0.750.360.83 15 1 Closed 0.650.310.72 16 1 Closed 0.550.260.61 17 1 Closed 0.450.220.5 18 1 Closed 0.350.170.39 19 1 Closed 0.250.120.28 20 1 Closed 0.150.070.17 21 1 Closed 0.050.020.05

PAGE 111

100 Appendix B: Design Example Output The figures in this appendix show the one line diagrams generated by the PowerWorld software depicting the results of each base case and contingency case simulation. The voltage magnitude and angle at each node is displayed. Also shown is the ampere loading in each feeder segment and the load at each node in megawatts and megavars. Open sectionalizing devices are shown as hollow squares, and closed sectionalizing devices are shown as filled squares. Below the oneline diagram in each figure is a section of tabular output recapping the nominal operating voltage, conduc tor size, and length of both the main feeder and the backup feeder. Also tabulated ar e the total feeder lo ad and the total feed er losses, in both megawatts and megavars.

PAGE 112

101 Appendix B – (continued) Figure 19 – 266.8 kcmil Laurel @ 14.4 kV Base Case

PAGE 113

102 Appendix B – (continued) Figure 20 – 266.8 kcmil Laurel @ 14.4 kV Contingency Case

PAGE 114

103 Appendix B – (continued) Figure 21 – 397.5 kcmil Canna @ 14.4 kV Base Case

PAGE 115

104 Appendix B – (continued) Figure 22 – 397.5 kcmil Canna @ 14.4 kV Contingency Case

PAGE 116

105 Appendix B – (continued) Figure 23 – 556.5 kcmil Mistletoe @ 14.4 kV Base Case

PAGE 117

106 Appendix B – (continued) Figure 24 – 556.5 kcmil Mistletoe @ 14.4 kV Contingency Case

PAGE 118

107 Appendix B – (continued) Figure 25 – 715.5 kcmil Violet @ 14.4 kV Base Case

PAGE 119

108 Appendix B – (continued) Figure 26 – 715.5 kcmil Violet @ 14.4 kV Contingency Case

PAGE 120

109 Appendix B – (continued) Figure 27 – 900 kcmil Cockscomb @ 14.4 kV Base Case

PAGE 121

110 Appendix B – (continued) Figure 28 – 900 kcmil Cockscomb @ 14.4 kV Contingency Case

PAGE 122

111 Appendix B – (continued) Figure 29 – 1192.5 kcmil Hawthorn @ 14.4 kV Base Case

PAGE 123

112 Appendix B – (continued) Figure 30 – 1192.5 kcmil Hawthorn @ 14.4 kV Contingency Case

PAGE 124

113 Appendix B – (continued) Figure 31 – 266.8 kcmil Laurel @ 24.9 kV Base Case

PAGE 125

114 Appendix B – (continued) Figure 32 – 266.8 kcmil Laurel @ 24.9 kV Contingency Case

PAGE 126

115 Appendix B – (continued) Figure 33 – 397.5 kcmil Canna @ 24.9 kV Base Case

PAGE 127

116 Appendix B – (continued) Figure 34 – 397.5 kcmil Canna @ 24.9 kV Contingency Case

PAGE 128

117 Appendix B – (continued) Figure 35 – 556.5 kcmil Mistletoe @ 24.9 kV Base Case

PAGE 129

118 Appendix B – (continued) Figure 36 – 556.5 kcmil Mistletoe @ 24.9 kV Contingency Case

PAGE 130

119 Appendix B – (continued) Figure 37 – 715.5 kcmil Violet @ 24.9 kV Base Case

PAGE 131

120 Appendix B – (continued) Figure 38 – 715.5 kcmil Violet @ 24.9 kV Contingency Case

PAGE 132

121 Appendix B – (continued) Figure 39 – 900 kcmil Cockscomb @ 24.9 kV Base Case

PAGE 133

122 Appendix B – (continued) Figure 40 – 900 kcmil Cockscomb @ 24.9 kV Contingency Case

PAGE 134

123 Appendix B – (continued) Figure 41 – 1192.5 kcmil Hawthorn @ 24.9 kV Base Case

PAGE 135

124 Appendix B – (continued) Figure 42 – 1192.5 kcmil Hawthorn @ 24.9 kV Contingency Case

PAGE 136

125 Appendix C: Feeder Cost Estimate Assumptions and Methodology A typical construction cost pe r mile was required for ea ch conductor size and voltage rating considered in this dissertation. In actual cases, construction costs can vary significantly for the same conductor size and vol tage rating due to other factors. These factors include routing considerations overhead and underground obstructions, restrictions on span length or right-of-way width, and whet her the distribution feeder is built as a separate linear facility or u nderbuilt on transmission line structures. Although a detailed cost estimate is required to determine the construction cost for a specific distribution feeder, a typical cost per mile can be developed by making some reasonable assumptions. Such estimates work well for assessment purposes, since the same nonstandard conditions that would for ce the actual construction cost for one specific design to deviate substantially fr om the typical would exist for all design variations. A sound comparison between options can still be made using the typical cost per mile values. For the purposes of developing overhead feeder construction costs for this dissertation, the following method was used. A detailed cost estimate was done for a feeder built with 336.4 kcmil Tulip phase conductors and a 2/0 AWG neutral at 13.2 kV. Uniform span lengths of 175 feet were used, and two dead end poles (to simulate two 90 degree turns) were included in the design. That cost wa s tabulated in Table 6. Then, a differential costing process was applied to generate ty pical costs for the other configurations.

PAGE 137

126 Appendix C – (continued) Changes in wire sizes produced a differential cost for the conductor. As pole strength needed to be increased, a larger pole cost was utilized. Typical cost estimates for underground constr uction originated from a detailed design based on three 1000 kcmil copper conductor 15 kV cables utilizing a one-third sized concentric neutral. Nine 600-foot pulls were made through concrete-encased PVC conduit utilizing eight pullboxes, and two cable terminations per mile were included. Although a more precise cost could be de veloped for the other design options by performing a detailed design for each, such prec ision is unwarranted. This is because the underlying assumptions of 175-foot spans and tw o deadends per mile are likely to cause more of a deviation between the estimated and actual costs than the relatively small error introduced by the differential costing method. Differential costing is a quick, simple, and practical means of generating t ypical cost data. Such methods are essential to keep the complexity of a design method to a manageable level.

PAGE 138

127 Appendix D: Estimation of Energy Losses Energy losses on a power system can be determined on a real-time basis by determining the instantaneous power (demand) loss and in tegrating this value over time. This approach, unfortunately, cannot be implemente d with load flow software as used in planning environments. Energy losses, however, can be estimated reasonably well provided that some fairly detailed load data are available. Since energy losses are proportional to the squa re of the current, and the current flowing in any part of a power distribution system continually varies in magnitude, some approximation of average loading must be made. This can be done by obtaining historical loading data for a portion of the syst em, as is shown in Fig. 43. This data is typically available for existing feeders on an hourly basis. Data for undeveloped service areas can be estimated using data from similar developed areas. 3000 4000 5000 6000 7000 8000 Day 1Day 2Day 3Day 4Feeder Load in kVA Figure 43 – Hourly Feeder Loading Data

PAGE 139

128 Appendix D – (continued) After hourly loading data is obtained, it can be sorted in descending order and graphed. Scaling the x-axis as percen t of time instead of hours a nd normalizing the y-axis as a percentage of maximum feeder peak load produces a load duration curve as shown in Fig. 44. 0%10%20%30%40%50%60%70%80%90%100% % of Time 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%% of Feeder Peak Load Figure 44 – Load Duration Curve All feeder load duration curves share some si milarities. The first portion of the curve falls away from 100% very steeply, since peak loading on a feeder typically occurs for a very short period of time. After the slope d ecreases, a fairly linear segment occurs. The steep slope at the beginning of the curve i llustrates that the highest feeder loadings last

PAGE 140

129 Appendix D – (continued) for short periods, suggesting that the distribut ion system can be designed to handle peak loads slightly less than the actual peak loads and perform satisf actorily a very high percentage of the time. By exploiting this characteristic of the load duration curve, substantial cost savings may be realized. Pe rhaps the feeder design would be no different that that which would be developed for the 100% peak case, but the timing of the feeder construction may be delayed for some time peri od, resulting in savings due to deferral of expenditure. Energy losses can be estimated by remodeling the load duration curve as a piecewise linear approximation. This can be done using various increment sizes. Fig. 45 illustrates the principle using 5% increments. Each rectangular region represents energy, so rescaling the feeder loading to match the height of the rectangle, then resolving the load flow model will determine the demand losses at that average load level. Multiplying that demand loss by the number of hours at which the feeder operates at that load level is a good estimate of the energy loss during that period. Repeating this process for each rectangle, then summing the energy losses w ill approximate the energy losses for a year. This annual energy loss figure is a require d value in the economic assessment of the system design.

PAGE 141

130 Appendix D – (continued) This % of Time This Cumulative % of Time 0%10%20%30%40%50%60%70%80%90%100% % of Time 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%% of Feeder Peak Load95%90%85%80%75%70%65%60% 6.3%12.1%15.4%12.2%14.6%16.7%13.9%8.8% 6.3%18.4%33.8%46.0%60.6%77.3%91.2%100.0% At or Above This Load Level for or (in Percent of Feeder Peak Load) Figure 45 – Avergae Load Level Durations

PAGE 142

131 Appendix E: Implementation of Communi cation-Based Overcurrent Detection Consider a radial feeder divi ded into two segments by a no rmally-closed sectionalizing recloser, and backed up to another fe eder with a normally-open recloser. Seg. 2 POWER RING Seg. 1 R N.C. Feeder R N.O. Backup Source Figure 46 – Radial Feeder with One Sectionalizing Recloser and Backup Recloser The flowcharts in Figs. 48 and 49 detail th e logic needed to implement a communicationbased overcurrent detection scheme on the feed er as shown in Fig. 46 at the power ring and at the sectionalizing recloser, respectively. As additional sectionalizing reclosers are added, as in Fig. 47, the logic for the power ring breaker remains the same, except for additional blocking signal inputs (one from each downstream recloser). The logic for the downstream-most recloser remains as shown in Fi g. 49, except that references to “Seg. 2” become “Seg. n .” The upstream reclosers adopt a l ogic scheme as shown in Fig. 50. Seg. 2 POWER RING Seg. 1 R N.C. Feeder R N.C. (Normal Source) R N.O. Backup Source Sectionalizing Recloser #1 Sectionalizing Recloser #2 Backup Recloser Seg. 3 through Seg. n Sec. Recloser #3 through Sec. Recloser #n-1 Figure 47 – Radial Feeder with n Sectionalizing Reclosers and Backup Recloser

PAGE 143

132 Appendix E – (continued) Fault Current Detected Monitor Current Blocking Signal Received From Sectionalizing Recloser? Fault Located on Segment 2 Yes Do Not Trip No Trip Power Ring Breakers Wait for Sectionalizing Recloser to Trip Fault Current Still Present? Yes Sectionalizing Recloser Failed No Trip Power Ring Breakers Lockout Feeder Open Feeder Disconnect Switch Close Power Ring Breakers Wait for Feeder Reset Fault Current Sensed? Yes No Successful Reclose Restore Normal Configuration Start System Trip Reclosing Breaker on Power Ring Execute Time Delay then Breaker Again Close Reclosing Fault Current Sensed? No Yes Trip Reclosing Breaker on Power Ring Sectionalizing Recloser Close Remaining Power Ring Breaker Close Sectionalizing Recloser Transfer-Trip Reclose Sequence Close Reclosing Breaker on Power Ring Close N.O. Recloser to Backup Source Segment 2 Energized from Backup Source Fault Located on Segment 1 (or comm. channel failed) Figure 48 – Power Ring Logic for Communication-Based Overcurrent Detection

PAGE 144

133 Appendix E – (continued) Fault Current Detected Monitor Current Fault Located on Segment 2 Lockout Segment 2 Wait for Feeder Reset Fault Current Sensed? Yes No Successful Reclose Start System Trip Sectionalizing Recloser Execute Time Delay then Recloser Again Trip Sectionalizing Fault Current Sensed? No Yes Sectionalizing Recloser Trip Reclose Sequence Sectionalizing Recloser Reclose Figure 49 – Single Sectionalizing Recloser Logic for Communication-Based Overcurrent Detection

PAGE 145

134 Appendix E – (continued) Fault Current Detected Monitor Current Blocking Signal Received From Downstream Sectionalizing Recloser? Fault Located past Segment k+1 Yes Do Not Trip No Trip Sec. Recloser k Wait for Downstream Sec. Recloser to Trip Fault Current Still Present? Yes Downstream Sec. Recloser Failed No Trip Sec. Recloser k Lockout Seg. k+1 Fault Current Sensed? Yes No Successful Reclose Restore Normal Configuration Start System Execute Time Delay then Again Close Recloser k Fault Current Sensed? No Yes Trip Sectionalizing Recloser k Sectionalizing Recloser k+1 Close Sectionalizing Recloser k+1 Transfer-Trip Reclose Sequence Reclose Sectionalizing Recloser k Close N.O. Recloser to Backup Source to End of Feeder Energized from Backup Source Fault Located on Segment k+1 (or comm. channel failed) (Sec. Recloser k) Wait for Sec. Recloser k Reset Trip Sec. Recloser k Segment k+2 Figure 50 – Sectionalizing Recloser 1 through n–1 Logic for Communication-Based Overcurrent Detection

PAGE 146

135 Appendix F: Assessment of Reliability with Intelligent Sectionalizing The concept of intelligent sectionalizing is to use sectionalizing reclosers capable of communicating amongst themselves to efficien tly isolate faults that occur on the distribution system. The basic concept is il lustrated in this disse rtation using a single sectionalizing recloser located at a distance m per-unit of feeder le ngth from the power ring. The reliability provided by this sectionalizing reclos er is quantified. As more sectionalizing reclosers are added to the feeder, the total customer outage minutes experienced by the customers on the feeder will diminish, as more customers will experience either no interruption of service at all, or an amount quantified by sectionalizing time (seconds) instead of the tim e required to physically repair the problem that caused the fault (typically hours or even longe r). This substantial improvement in reliability is offset by an increased total cost for the distribution feeder, including both capital cost (to purchase and install the additional sectionalizing reclosers) and maintenance cost (to keep the additional reclosers in proper operating condition). When locating sectionalizing reclosers along the primary feeder, it is logical to sectionalize equal amounts of load. If the load distribution assumptions made in this dissertation are applied (uniformly-distributed load in a service area described by Fig. 11 on P. 46), the sectionalizing reclosers can be optimally located as shown in Eq. 33.

PAGE 147

136 Appendix F – (continued) L N N k dN | k Eq. 33 where dk | N = the distance from the power ring to the kth of N sectionalizing reclosers L = total feeder length Fig. 51 shows a primary feeder divided into N+1 segments by N sectionalizing reclosers. Seg. 2 POWER RING Seg. 1 R N.C. Feeder R N.C. (Normal Source) R N.O. Backup Source Sectionalizing Recloser #1 Sectionalizing Recloser #2 Tie Recloser Seg. 3 through Seg. n Sec. Recloser #3 through Sec. Recloser #n-1 Seg. n+1 R N.C. Sectionalizing Recloser #n Figure 51 – Primary F eeder Sectionalized by N Reclosers Table 12 shows the outage duration resulti ng on each segment for a permanent fault occurring at various points on the feeder. Table 12 – Resulting Outage Durations for Various Permanent Faults Permanent Fault on Segment # 1 2 3 … n–1 n 1 R – – ... – – 2 S R – ... – – 3 S S R ... – – … ... ... ... ... ... ... n–2 S S S ... – – n–1 S S S ... R – Resulting Outage Duration for Segment # n S S S ... S R Legend R = Repair time S = Sectionalization Time –=No Outa g e Time

PAGE 148

About the Author Ralph Edward Fehr, III received a Bachelor’s Degree in Electrical Engineering from the Pennsylvania State University in 1983 and a Master’s Degree in El ectrical Engineering from the University of Colorado in 1987. Mr. Fehr worked in the electric power industry for over 20 years, and has been an instructor at the University of S outh Florida since 1997. He authored Industrial Power Distribution (Prentice Hall 2002) and numerous techni cal papers. He is a member of Tau Beta Pi and Eta Kappa Nu, a senior member of the Institute of Electrical and Electronics Engineers, and a registered professional engineer in Florida and New Mexico.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001709511
003 fts
005 20060614112139.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 060510s2005 flua sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001356
035
(OCoLC)68624101
SFE0001356
040
FHM
c FHM
049
FHMM
090
TK145 (Online)
1 100
Fehr, Ralph E.
3 245
An integrated optimal design method for utility power distribution systems
h [electronic resource] /
by Ralph E. Fehr, III.
260
[Tampa, Fla.] :
b University of South Florida,
2005.
502
Thesis (Ph.D.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 148 pages.
Includes vita.
520
ABSTRACT: This dissertation presents a comprehensive and integrated design methodology to optimize both the electrical and the economic performance of a utility power distribution system. The proposal is structured to facilitate its adoption and incorporation into the existing utility infrastructure by allowing the various portions of the new design to be implemented gradually into the existing infrastructure without the need to abandon the portions of the existing system that are performing satisfactorily. The topology of the substation plays a vital role in determining both the reliability and the economy of the distribution system. The ring bus topology is offered as the best topology design, and its characteristics as seen at the distribution level are examined. A key concept presented in this dissertation is that the distribution system must be optimized as a whole, not subsystem by subsystem.Optimizing the substation and the primary feeder system separately does not assure an optimal system; in fact, independent design of the two subsystems is likely to produce a non-optimal system laden with operational problems. An integrated approach is essential to assure optimum performance, and the integration process requires an iterative approach. This iterative approach is presented using an example. Innovative changes to the protection strategy of the feeder system can greatly enhance the reliability of the distribution system. The use of communication-based overcurrent detection is presented. This transmission-like scheme, when applied at the distribution level, improves both the reliability and the economy of the system substantially over traditional time-coordinated overcurrent protection philosophies.An application of these proposed innovations leads to the design of a hypothetical system, which is in turn analyzed from both electrical and economic perspectives.
590
Adviser: Alexander Domijan, Jr., Ph.D.
Co-adviser: Kenneth A. Buckle, Ph.D., P.E.
653
Communication-based overcurrent detection.
Distribution automation.
Distribution substation.
Intelligent sectionalizing.
Power ring.
0 690
Dissertations, Academic
z USF
x Electrical Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1356