Time Varying Load Analysis To Reduce Distribution Losses Through Reconfiguration

IEEE Transactions on Power Delivery, Vol. 8, No. 1, January 1993

294

Time Varying Load Analysis To Reduce Distribution Losses Through Reconfiguration Robert P. Broadwater member

Asif H. Khan

Hesham E. Shaalan member

The Bradley Department of Electrical Engineering, Virginia Polytechnic Institute and State University Blacksburg, VA 24061

-

ABSTRACT An electrical distribution systam recoafiguration algorithm to reduce aystem loesee is presented. The algorithm

Robert E. Lee member Pennsylvania Power & Light, Allentown, Pennsylvania

circuit load profile makes reconfiguration possible to reduce 1.Thus, a circuit's load diversity is establishedby the daily calculates switching patterns am a function of time. Either electrical power demands of the various customer typee which are aerved by the circuit. The modeling of the diversified load seasonal or daily time studies may be performed. Both manual patterns d t h e various customer types is "idered. and automatic switchea are used to rexmntigura the system for Seasonal, daily, and hourly time variations of load provide seamnal studies, whereas only automatic switches are considered for daily studies. A load estkaation algorithm analysm pointa for the reconfiguration algorithm. Using thia provides load information for each time point to be analyzed. The load data the algorithm develops switching p a t t a m to reduce load estimation algorithm can incorporate any or all of the loosea, where time may vary Over a daily cycle and/or a seasonal following: spot loads, circuit meaaurementa, and customer timecycle. Ben&ita fmm eeasonal loee reduction can be accompbhed through manual switching, whereas b e d t a from the daily loes varying diversified load charactariotica. Voltage dependency of loads is considered at the circuit level. It is shown that switching reduction requirea automatic switching [l]. The reconfiguration of an electrical distribution system to at the system peak can reduce lossea but may cause a marginal increase in system peak. Data s t ~ ~ c used reduce lasses has a natural tendency to balance loading among t u to ~ ~model loads and to store switch configurations as a function of time are circuits. This balancing proceas placee the system in a better described. Example problem are provided to illustrate resulte. position to respond to emergency load transfem [2,3]. Liu, Lee, and Vu [4] derived a global optimality condition for Key words: reconfiguration, voltagedependency, switching the loas minimization problem and propoaed two algorithms: the f & ia based on the uniformly distributed load model and the patterns. second on a concentrated load model.The first algorithm obtains 1. INTRODUCTION the optimal solution when the minimum losa is obtained for every feeder pair. In the second algorithm, a similar procedure is An algorithm which analyzes distribution systems for loss performed by moving open pointa, one at a time, from nn actual reduction is dmcriied. Distribution systems consist of groups of switch position to another until no further loss reduction can be interconnected radial circuits. Their configurations may be acbieved. However, the authors left feeder load estimation for future work. varied with manual or automatic switching operations to Estimation of load distribution along a feeder is critical to transfer loads among the circuits. ,Ideally, the system configuration should produce minimum loeees. Starting with a succeclllful reco-ration analysis. The method presented customer information, and utilizes feeder measurements, given set of switch p i t i o m , the algorithm searches for a conatant spot loads. Customer information includes numbers and revised combination of switch poaitions to reduce loaaee. The algorithm converges when the search procadure cannot locate a classes of customers attached at a specified point in the circuit switch pair operation that further reducea loesee. model and time-varying diversified load characteristicsfor each Distribution system load varies seasonally by time and type of class. The algorithm accepts whatever load i d o m t i o n is day. Manual switching to reduce losees satisfies the utility's available. For inatance, if only spot laads are specified, then need to accommodate S W W M ~load variatione. The emergence of customer information and feeder measuremente are ignored. This work builds upon aspects derived from Civanlar's [5] distribution automation techology and equipment allows automatic switches to take advantage of daily and time of day and Huddleaton's [SIprevioue studieu. Civanlar's method does switching patterna to reduce h e s . The placement and not allow loads to be switched from a higher to lower voltage operation of switchea to reduce loaaes ia a complex task level Circuit and is baaed on a single switch pair operation per compounded by the time varying nature of loads. iteration [ 6 ] . Huddleeton solved the minimum loss " f i i t i o n problem utilizing a quadratic lms function and The load profile for a circuit is a function of customer typea. Laad profilea vary from circuit to circuit due to the mix and multiple switch pair operations per iteration. This algorithm utilizea Civanlar's voltage switching rule, a dispersion of customen, served. In distribution eysteme, circuit single switch pair operation per iteration, and a direct search peaks are noncoincident due to the diversity of load resulting method which incorporatee Huddleston's loss function. The from the categories of customer classes served. This variation in direct search method incorporate the constraints of specified voltage bandwidth and conductor and equipment ampacities. 92 WM 269-1 PWRD A paper recommended and approved by Depending upon the load-voltagedependency relationship, the I E E E Transmission and Distribution C o m m i t t e e of switching operatiom may cause individual switched loads t o the IEEE Power Engineering Society for presentation either increase or decrease in magnitude. At the system peak a t the IEEE/PES 1 9 9 2 Winter Meeting, New York, New condition, switching operatiom which reduce loeees may York, January 26 30, 1992. Manuscript submitted simultaneously increase the peak h l t s based on mch August 28, 1991; made available for printing conditions are presented. November 27, 1991. , A mathematical formulation is presented in the following section. Data structures wed in load modeling and storing reaulta of the algorithm are deacribed. The load model, 0885-8977/92/$3.W1993 IEEE

-

295

estimation procedure, and voltagedependency modeling of loads are explained. Next, the loes function evaluation coupled with a direct search procedure is presented. The handling of voltage and current constraints is explained. Finally, three examples which illustrate the concepts are solved. Subecripts and variablea are defined in Tabla Aand B of the Appendix.

2. PROBLEM STATEMENT The objective is to minimize system real loeses as a function of circuit topology for a given point in time. The circuit topology is a function of switch positions The problem to be solved can be mathematically expressed as follows:

=.

ALCORITHM RESULTS The switching configuration Data structure stores switch status for each switch in the eyatem. The set of switch positions stored in each data structure io unique, and is given a unique identifier referred to as a Configuration Identification. The Operations Data strudure Storee season, type d day, t h e of day, and the C o d i t i o n Identification for the associated point in time. Two-hundred and seventeen data &ruch.lres of type Operations Data are utilized. Of these two-hundred and seventeen ,"st two-hundred and sixteen correspond to the twenty-four hours in a day, three typea d days, and three SBBZJOM. The ~mainingstructure corresponds to the base case system configurationthat exists a t the start of the solution.

minimize f @ o m T ) } o v e r 3 subject to:

4.

(3) (4)

kWPedT5 max_kwPedT,p

(5)

The objective fundion is subject to a set of nonlinear equations and linear inequality constraints. For a given time T, the vector of complex loa& is estimated at nominal voltage. The vector ZT includes both automatic and manual switches or just automatic switchea depending upon the study to be performed (i.e. daily or seasonal). The nonlinear functions described by Eqns. (1)-(2)are representations of the system power flow equations which are a function of time and switch position. The inequality constraints repreeent voltage bandwidth, equipment and conductor ampacitiee, and system peak concerns. The algorithm which implements the solution of the problem does not necessarily reach the global minimum, but doea search for switch pair operations which reduce loesee while simultaneously satisfying system constraints. When no switch pair operation which d u m loeees can be found (i.e. regarding the currently open switches), the direct search method converges. The algorithm generates switching pathma v e m time, where each time point to be analyzed correaponda to a different loading condition. "he algorithm is implemented on Personal Computers running the OSE extended operating system and the included database management system.

3. DATA STRUCTURES Load information and m d i g u r a t i o n algorithm d t s are stored in memory in data s t " a and in tables in a database on the hard disk.

LOAD INFORMATION The Diversified Customer Load Data Structure stores information associated with diversified time-varying customer load characteristics. Elements of this data structure include customer class, season, type of day, time of day, hourly diversified KW usage, power factor, and cold load pickup factor. The Circuit Measurements Data Structure stores information about feeder power flow at the substation. Elements of this data structure include feeder name, feeder loss factor, season,type of day, time of day, hourly KW and power factor measurements for each phase.

LOAD ESTIMATION

The algorithm can uee available system measurements, load statistics b a d on cuetomer class, and/or load estimates. If available, the algorithm takes the following into account feeder circuit meamremenk, the number of customem by customer class throughout the circuik diversified cuetomer load curves. If none of the preceding load information is readily available, traditional load estimates may be used. Diversified load curvea may be developed from load research data or experiments such as those performed in the Athens distribution automation project [a]. The information flow for the load estimation process is indicated in Figure 1.

TT Diversifid Load curves &

customer

hformafkn

1

Figure 1. Information Flow Diagram for Circuit Load Estimation Feeder power flow and power factors are modeled as a function of the type of day, time of day, season,and phase for each circuit and are defmed aa five dimensional arrays,

The diver~SiedKW curves for customer classes are used to allocate feeder measurements a t the modeled load points. If feeder measurements are unavailable, then the load may be directly calculated a t each load point from the number of customers and the diversified load curves. The diversified load

296

value for a line section is determined by multiplying the number of customem on a particular line section by the customer class divedied load curve value. A maximum of twelve custom= classes may be modeled at each line &ion. For clarity of presentation and d d p t i o n , it is assumed that all data are available. The portion of the measured power allocated to customer class "j"is the ratio cf the total divemifii load for customer class "j" to the total diversified demand, which ia independent of cuetomer class, multiplied by the measured KW for feeder 'c",

Feeder loesea and customer load scaling are accounted for by multiplying Eqn.6 by a loss factor.

PrnL~ssj,~ = lsj * Plos * Pmcusj,c.

fractional change in load amp

VDF, = volt

VDF, may be specified as a positive, negative, or zero value. A positive voltagedependent factor simulates constant impedance type loads, whereas a negative voltagedependent factor simulates constant power type loads. If Vi,p,~,crepreeents the complex voltage value at the ith load point for phase "p" and circuit "c" which is deviated from ita nominal value, then the voltagedependent load current is given by, (12)

(7)

The portion of the meaaured KW allocated to a single customer of class "j" is given by dividing Eqn. (7) by the total number of customers of class "j"for phase "p" on feeder "c",

The estimated KW load at the ith load point for each phase of feeder "c" and for customer class "j"baaed on time of day, type of day, and seaclon is obtained by multi lying Eqn. (8) by the number of customers of class "j"at the it!l load point,

P-est i,pj,T,c = Pma ij,c * a s i,pj,c.

The voltagedependency load factor (VDF,) for a particular circuit is be defined an

(9)

The number of customem at the ith load point may be specified interactively by the user or may come from a Customer Information Database interface. The total estimated load at the ith load point is given by,

(10)

An equation similar to Eqn. (10) is used for the reactive power load eatimate.

5. VOLTAGE DEPENDENCY FACTORS

Loads are modeled as voltagedependent. The voltage

6. LOSS FONCTION EVALUATION A Direct search method based on concept6 presented in Reference [7] evaluates changes in k e a predicted by Huddleston's lose function [6].For a given system codiguration the loss functions are evaluated for all switch pair operations adjacent to open switches, where only transfere from lower to higher voltage circuits are considered. Consider a distribution system with 'n' open switches. Results fmm power flow analysis are ueed to develop loss functions for each circuit [6,91. Circuit loas functions are evaluated at the nominal witching current values, where the nominal switching current ia the sum of the nominal load currents that are being switched [SI.The loss functions associated with all circuita are summed and the bane case loss model is given by,

N

P h S B = P-bap,. (13) c=o The loas functions are sequentially evaluated for each possible switching operation. Loas functions associated with all circuits are summed for each possible switching operation, developing additionallose models for each open switch, as given by, Ploes,, =

N

2

P-lassc,

C=O

where BW = 1,2, ...,Maximum number of open switches.

dependency factor allows system analysis to be easily performed using various degreea of traditional constant power, conatant current, and constant impedance load modele. For a given phaee "p", the nominal load current at the ith load point is given by,

If a daily r e c o d i t i o n study is being performed, then open manual switch- are not considered. The loas function for each allowable switch pair operation is compared to the base case lasa function and the switchbg operations that produce the maximum decrease in system loas are selected as the switching operations to be implemented at the current iteration, as given by,

A voltage dependency factor is associated with each circuit. Th-e factors can be determined experimentally by varying the substation bus voltage coupled with rapid sampling of feeder voltage and curremta. Experience at Pennsylvania Power & Light has shown that a tbree percent voltage reluction does not appear to reduce load while a decrease of five percent pmducea an approximate two and one-half percent drop in load. However, a load tends to regain ita previous value after about thirty

If A P h s 5 0.0, then the algorithm has converged since no switch pair operation further reducedthe system loss. Voltage and current comtrainta are checked before the recommended switching operation is implemented. If a constraint vidation occum, then the switching operation is not allowed, and the resulting open switch is tagged as not being available for U I J ~in further reconfiguration evaluation at that time point. If constraints are violated, then Eqn. (15) is used to

minutes.

297 The low functione are evaluated for the base case and for each possible switch pair operation. The din& search method then determines the switch pair operation that prodthe maximum decrease in system losses. This switch pair operation is performed and power flow calculations are re-run on the circuits that changed. If no constraint violations OCCUT. the loas functions of the W e d circuita are updated. A new base case at this time point has now been created, and further comparisons are made against this base case. When switch pair operations p d u c e no further loss mduction, the algorithm has converged for this time point. If a new switching pattern has been generated, meults are s t o d to the Switching Codiguration Data Structure. The Configuration Identification, system loa-, and load level for the time point are atored in the Operations Data Structure. This procedure is repeated for all time pinta under atudy.

select the switch operation that reaulta in the next greatest decrease in loeses.

7. RECONFIGURATION ALGORITHM The flow of the Reconfiguration Algorithm is illustrated in Figure 2. For the fmt or next time point to be analyzed, program variablea related to constraint violations (SCE) and convergepce (PFE) are initialized. Using available information from the Divelaified Customer Load and Circuit Measurements data structurea, the loading condition for the given time point is evaluated. A power flow calculation is then performed for each circuit. If constraint violations OCCUT on the fmt iteration, those violations are flagged and the algorithm pnxeeds to the next time point to be analyzed. Otherwise, the low functions are calculated.

P Start

I

B

Load Model

I

Powor Flow

GFl V

Check 0

alnts

I

I

Exit on SCE

=

1

I

N

Switch Pair Operation

B

Q

t7 Store Results

I Figure 2. Recontiguration Algorithm Logic

I

I

298

8. IMAMPLEPROBLEMS

Ezamde 2:

Three example problems are considered which d e m ~ ~ t mme te of the load models. The fust example uees spot loa& only. Modeling in the second example is based on reaidenth1 and small commercial divernified load curves along with a spot load. The third example indudea feeder meam"enta along with cusltomsr divemified load curven and spot loads, and illuntratea the &e& of voltage dependency. Feeder voltage level is assumed to be 13.2 KV for all three ~p~~~mples. For eaee of manual calculation to verify algorithm multa, short line aectians were assumed to be C O M ~ X U C wi~th~ conductors and conductor configuratio~~ with an impedance of one ohm per mile. While loss r e d d o n s are shown to be small,they are demonstrated to eorist and are small due to the choice of line impedance and length. Examde 1: The main featurea of the r e c o n f i i t i o n algorithm are illustrated in this example. The problem wao deaigned to test the algoritm and therefore the solution is known in advance. Figure 3 shown the three circuita used. There are 126 possible switching configurations. Voltage dependency is assumed to be zero (i.e. conatant current load). Circuit loading and the "Ita of the analysis are shown in Table 1. The time point analyzed is a " m e r weekday at 6 p.m. The r e c o d i t i o n algorithm converged in twelve secondo. The results show that reconfiguration tends to balance loads, losses, and voltage levels. As a sanity check the load before and &er reeonfiguration remains the same at 198OOKW, while the losses are reduced by 26 KW. Operation of the four highlighted switchea in Figure 3 produce the calculated renulb.

s

The second example analyzes m e r weekday and weekend loading while aomming the voltage dependency factor to be zero. The sptem of two cbmita shown in Figure 4 has a midpoint constant spot load of 300 KW which is used to de"ta the benefita available by taking advantage of load divernity between m e r e n t customisr claesea. Circuit 1has 1000 all el& residential c-untomen~ per phase while Circuit 2 has 341 small commercial customers per phase. These numbem were chceen such that the noncoincident circuit peak loading is equal for a s u " weekday, but unequal for the weekend. Figure 6 shows the load curves for both customer classee. Recommended switching c h a w nhift the epat load to the circuit with the smaller load. Table 2 shows the switch pooitions, the recommended switching operation time points, and the circuit lOading. "he conatant spot load for the mummer weekday is served 13 houm per day by Circuit 1 and 11 h m per day by Circuit 2, whereas for the " m e r weekend Circuit 1 only supplies the spot load for 1hour. Furthemore, the total reduction in loesee for the " m e r a w n is 2141 KWh aeeUming 66 weskdaye and 26 weekends. Although not presented, and in contrast to the summer reaulta, multa for wintar weekdays and weekends reveal that the apot l a d is alwaye served by Circuit 2.

Figum 4. Example 2 System of Circuits

I .

FigLlre 3. Example 1System of Circuite Table 1. b d i g u r a t i o n Results of Example 1 circuit1

Before Reconfiguration: Total KW Load KwhLoesee Voltage h v e l ARer Reconfiguration: Total KW h d KwhLoMee Voltage Level

circuit2

7800

6000

698.3 116

413.2 118

6600

600

6600 600

117

117

circuit3

6000 413.2 118 6600 600 117

os'

. . . . . . 8. . . . . . 11

0

s . . .

.

I

.

18

.

.

.

-dPq

Figure 6.Load Curves For Summer Weekday

.

.

2;

299

Table 2. b u l t a of Example 2 Hourof Switch1 Switch2 KWLoading day status status circuit1 Circuit2 Summer Weekday: 12a.m. 2 a.m. 7 a.m. 9 a.m. 5 p.m.

open closed open closed open

closed open closed open closed

4410 3150 3210 4200 5280

3840 3161 3297 3678 5394

Summer Weekend 12a.m. 5 a.m. 6 a.m.

open closed open

c l d open closed

4590 2850 2670

3614 2578 2786

circuit 2

Example 3: Example three is designed to demonstrate the need for automatic switching to take advantage of daily loss reduction. A system of three circuita shown in Figure 6 contains forty-one line sections, seven protective devices, three line voltage regulators, and six switches. Many loads are found to be voltagedependent [SI. Due to the voltage dependency of loads, system total load varies with calculated voltage levels. Total system load also changea when switching operations are performed with voltage dependency factors unequal to zero. The effects of voltagedependency on system loading and loeaea is shown in Tables 3 and 4. The results presented are for a summer weekday at 2 a.m. with an arbitrary selection of voltage dependency factors of -0.03, -0.01, 0.0, 0.01, and 0.03. Positive voltagedependency factors simulate constant impedance load behavior and negative voltagedependency factors simulate constant power load behavior. From Table 3 it can be seen that as the voltagedependency fador increases in the positive direction the total load demeasea, and Table 4 illustrates similar behavior for lowea. Table 3 also indicak that for a load mix with a positive voltagedependency factor (i.e. constant impedance load), system load increases as lossen decrease. Without a coordinated voltage control system, this effect could be detrimental at the aystem peak time point. For the constant current base case system load, the total system lossea represent only 2.8 percent. However, the algorithm demonstrates that a 15.4 percent loas reduction is still available for this time point. Additional analysis not shown here indicates that the status of switches swl and aw2 are a function of daily load variation, whereas switches sw3 and sw5 change their status b a d on seasonal load variations. Hence, from a loss reduction perapedive switches swl and sw2 should be chosen as automatic, with the reat of the switches being manually operated.

Figure 6. System of Three Circuita Table 3 Load In KW For Three Circuita As A Function Of Voltage Dependency Factor For Summer Weekday at 2 a.m. -0.03 Base Case After Switchin

Change In Loa3

Voltage Dependent Facto-0.01 0.00 0.01

0.03

11161.8 11029.8 10965.8 10903.0 10781.0 11123.2 11017.4 10965.8 10914.9 10815.6 38.6 -34.6 12.4 0.0 -11.9 Table 4

Loaa In KW For Three Circuita As A Function Of Voltage Dependency Factor For Summer Weekday at 2 a.m. -0.03 Base Case Afterswitching Loss Saving

Voltage Dependent Factors -0.01 0.00 0.01 0.03

317.5 307.4 302.6 265.6 259.2 256.1 46.5 51.9 48.2

297.9 253.1 44.8

288.9 247.3 41.6

9. CONCLUSIONS A reconfiguration algorithm which analyzes time varying load patterns is presented. It is shown that seasonal, daily, and time of day system reconfiguration based on time varying loads and customer class diversities will reduce system losses. Furthermore, system reconfiguration has a tendency to balance loading and voltage levels among circuita. Depending upon the voltagedependency of the load mix, the system peak could increase while loss reductions are being realized. However, d t a indicate that the increase in total system load was very small. Extrapolating theae resulta to system peak conditions, on peak switching for reduced lossea should have little &e& load.

300

Table B Definition of Variables

REFERENCES [I] R. E. Lee, C. L. Brooks, "A Method and Its Appliition to Evaluate Automated Distribution Control", IEEE Transactions on Power Delivery, Vol. 3, No. 1,July 1988, pp. 1232-1240. [2] C.H. Cantro, J. B. Bunch, T. M. Topka, 'Gemeralized Algorithum for Distribution Feeder Deployment and Sectionalizing, " IEEE Transactions on Power Apparatue and Systems, April 1980,pp. 649-667. 131 R. E. Lee, R. H. Osborn, V. F. W h k e r , M. T. Bishop, "Analysieof Time Varying Distribution Circuit Current and Laas C h a r a ~ c s "IEEE , Transactions on Power Delivery, Vol. 2, NO.4, October, 1987, pp. 1249-1264. [4] ChenChing Liu, Seung J. Lee, Khoi Vu, "h Minimization of Distribution Feeders: Optimality and Algoritbma," IEEE Tramactionon Power Delivery, April 1989, pp. 1281-1289. [6] S. Civanlar, J. J. Graingw, H. Yin,S. S. Ise,"Distribution Feeder Reconfiguration for Loss Muction,"IEEE Trans. on Power Delivery, Vol. 3, No. 3, July 1988, pp. 1217-1223. [q C. T. Huddleaton, R. P. Broadwater, A. Chandnrsekaran, "Reconfiguration Algorithm for Minimizing Losaw in Radial Electric Distribution Systems," Electric Power Systems Research Journal, Vol. 18, NO.1,1990, pp. 67-67. [7] M.J. Box, D. Davis, W.H. Swann,"Non-Linear Optimization Techniquw,' Monograph No.6, Oliver & Boyd, 1969. [8] P. A. Gnadt, J. S. Lawler, " A u t m " a g Electric Utility Distribution Systems: The Athens Automation and Control Experiment," Prentice-Hall, Inc., 1990. 191 R. P. Broadwater, A. Chandraeekaran,C. T.Huddleston, and A. H. Khan, "Power Flow Analyuia of Unbalanced Multiphase Radial Distribution System,"Electric Power Reclearch Journal, January, 1988, pp. 23-33.

Variable

P-mea Pmloss PmS PF-mea

-S

SW

tcus Tp_est V Ynom Vmin VmaX

Definition number of customers equipment current vector load current equipment current rating vector voltagedependent load current ayatempeak in kw customer type load scaling factor maximumvalue ofsystean peak in KW circuit KW loss fador total diversified KW load normalized diversified lnnr load constant KW load cuatomer baaed KW estimate quadratic Iwsfunction base w e loss model loea model bawd on load tranafer measured power allocated to cuetomer type feeder KW flow measurement Pmcue excluding line losses m e a d power allocated to single customer feeder power factor measurement load vector switch poeition vector total number of customers total kw estimate node voltage vedor a t time T nominal voltaga lower bound for syatem voltage upper bound for aystem voltage

Appendix: Definition of Subscripts and Variables Table A Definition of subscripts Subscript i m 9

Definition

...

a

load point i=1,2, load point m=1,2,... load point q=l,2,... segment a=1,2, number d open switchcircuits c=1,2, number of ckuita p=a,b,c ' Z m e r type time of day trpeofhY

Pk

repre8ent.a time of day, type of day, and season { t, d, Se } peak

s n c,cl,cZ

N

P

3 t

Se T

... ...

season

Biographies Robert P. Broadwater is an associate profegeor of electrical i Tech. His primary intereat is in electrical engineering at V distribution system analysis, deeign, and automation. He currently sewea as the aswciate editor of distribution and utilization systems for the Electric Power System Journal. He has worked with GE, B&W, TVA, TVF'PA, Oak Ridge National Lab, PP&L, and AP&L. A d H. Khan is pzwently puzauing a Ph.D. in electrical engineering at V i n i a Tech. He received his BE in electronica from NED University, Karachi, Pakistan. He received his MS from Tennessee Tech. His primary area of intereet ia software development for electrical diatribution systems. Heaham E. Shaalan is a Ph.D. candidate in Electrical Engineering a t Virginia Tech. He received hie BSEE and MEE from The Univemity Of Houston in 1986 and 1987, respectively. He is a member of Eta Kappa Nu Honor Society and IEEE Power Engineering Society. Hie reeeamh involves electric utility economic evaluation and distribution automation. Robert E. Lee began employment at PP&L in 1961 after graduation f " Drexel Univemity. Progreasing through distribution engineering, his most recent position has been supervising distribution dand reliability activities since 1982. He is Industry Advisor to two EPRI Distribution Automation projecta and is active in the IEEE/PES Transformer Committee and the IEEE/PES Distribution Subcommittee.