Application Assessments Of Distribution Network Minimum Loss Reconfiguration

1786

IEEE Transactions on Power Delivery, Vol. 12, No. 4, October 1997

APPLTCA Vesna Borozan, Member IEEE

Nikola Rajakovid, Senior Member IEEE

University "Sv. Kiril i Metodij", Faculty of Electrical Engineering, Skopje, Republic of Macedonia

University of Belgrade, Faculty of Electrical Engineering, Belgrade, Federal Republic of Yugoslavia

Abstract: Application aspects of optimal distribution network reconfiguration are highlighted in this paper. The problem is raised

number of papers in which the most of theoretical aspects of distribution reconfiguration are worked out. In its general frotn the theoretical solution level to the practical implementation formulation, finding a network configuration with minimum assessment level. For that purpose, a complex methodology for line losses is a mixed-integer nonlinear programming operational and short-term planning analyses of distribution problem. Since Merlin and Back's pioneer proposal .for systems is created. It manly regards non-automated or low level solving this problem 131, various algorithms have been automated distribution systems. Optimization of a distribution network configuration for a given loading condition is done by the suggested for reaching the optimum or near-optimum off a heuristic method presented in [I]. But, the methodology corresponding objective function [4]. In the most of appropriateness to the practical applications is attained by algorithms, the objective is minimum resistive line losses for introducing an original method for load estimation in distribution a given network loading condition 15-71, sometimes systems with minimum infonnation [2]. Furthermore, the combined with one or more additional objectives for optimal optimization results are evaluated throughout an appropriate operation of a distribution system [S-111. If this objective is costhenefit analysis. This way composed tool becomes flexible, consistently applied on a distribution network which load allowing appraisal of the network reconfiguration potential profiles continuously vary, the network configuration should benefits, as well as the economically approved frequency of be often changed, as well. Literature [12] suggests a switching. The results of the study on Skopje's distribution network reconfiguration algorithm capable to cover daily load are presented. variations, and to respectively achieve maximum lloss Keywords: Distribution system, Load estimation, Minimum loss reduction. Such a continuous reconfiguration is allowed by reconfiguration, Cost5enefit analysis. today's distribution automation and information technology and equipment [13], but the practical aspects of such an I. INTRODUCTION optimization remain to be carefully analyzed through costs, In the last decade the distribution system minimum loss transient effects and influence to system reliability. For most reconliguration has been proclaimed as a method for power of the real distribution systems performing frequent and energy saving achievement, at nearly no cost. The switching is impractical because of several more technical reconfiguration takes advantages of specific distribution and economic reasons. Therefore, there is the need for network structure and customer load varying nature. Most finding a cost effective coIlfiguration in which the network distribution networks operate radialy, even though there are would operate for some period of time. several interconnecting tie lines available to increase the Such a faced problem is usually solved by one of the system reliability. Additionally, the load profile in the following approaches: distribution network is a function of customer types served. - optimal configuration is obtained by minimizing Residential, commercial and industrial loads possess power losses for the system peak load 11, 3, 5-11]; differing daily and seasonal load curves. Load profiles vary optimal configuration is obtained by minimizing from feeder to feeder due to the mix and dispersion of energy losses for a given period 114-151. customers served. Using these tie lines, network The first optimization means that the distribution system configuration can be accommodated to the variation of load will be most efficient at times of peak demand and to achieve reduction of line losses. unnecessarily lossy during the off-peak times. On the other The popularization of this method has resulted in a large hand, the second approach guarantees maximum energy savings for a certain period, but not, as well, maximum cost PE-397-PWRD-0-01-1997 A paper recommended and approved benefit. Namely, potential cost savings may be missed by the IEEE Transmission and Distribution Committee of the IEEE because marginal costs of losses during the peak times, Power Engineering Society for publication in the IEEE Transactions which are very high, are not taken into account. on Power Delivery. Manuscript submitted July 30, 1996; made Considering that no priority should be given to system available for printing January 8, 1997. peak reduction, nor to energy loss reduction, we created an analytical tool for distribution system cost reduction via network reconfguration. Developed methodology and software contain three main steps: real-time load estimation, effective determination of minimum power loss configuration and costhenefit evaluation. Such a developed tool

0885-8977/97/$10.00 0 1997 IEEE

1787

permits its flexible use in distribution system short-term planning, off-line decision-making for feeder reconfiguration, as well as, in on-line control of network operation. This paper presents that tool and the results of a study, performed in cooperation with Skopje's Utility, for optimal operation planning of their distribution network. The study results provided interesting answers to several long-standing questions. 11. STUDY NETWORK AND OBJECTIVES

The study network is a portion of Skopje's MV network (10 kV). This network is served by 3 substations, x/10 kV. About 79 MW of peak load, or 290 GWh energy in the year (22% of whole consumer demand) is supplied through this 3 substations. The network consists of mostly underground feeders (with total length of 172 km) and distribution transformers 10/0.4 kV. Installed capacity of distribution transformers is mostly 630 kVA, but there are few transformers with capacity of 400 kVA. Every network feeder is connected to 10 kV bus of the corresponding substation through remotely controlled circuit breakers. But, switching devices existing at the entrance and exit of a feeder in every distribution transformer are not remotely controlled. Consequently, every branch (part of feeder between two distribution transformers) in the network can be switched, but switching should be performed manually. Utility reconfigure the network twice a year, trying to adapt network's configuration to winter and summer variation of load. The "basic summer" radial configuration consists of 47 feeders, supplying power to 313 distribution transformers. There are 60 tie lines available for changing the network configuration. Hence, the "basic winter" radial configuration has 42 feeders, 269 distribution transformers and 48 tie lines. In the meantime, small network portions are reconfigured in emergency conditions following the fault, to isolate faulted branches, and, in normal conditions, to avoid overloaded network branches or reduce system losses. All of the present capacitors in the network are kept in a fixed position, so, they do not influence this reconfiguration study. The network serves a big part of Skopje's downtown and several residential areas with different standard of living. There are even some industq locations in the studied area. This mixture of customer types present in the consume was the chief reason for choosing the studied area. It was supposed that minimum loss reconfiguration would fully take advantages of load time variations. Main objectives of the performed study were: - to quanti@ the potential savings in losses and costs attainable by network reconfiguration, - to appraise the beneficial frequency of switching, and - to advice the need of investments for remotely controlled switches in the network.

111. METHODOLOGY Developed methodology contains three individual parts: load estimation, minimum power loss optimization and costhenefit evaluation. The estimation of network load is achieved by original algorithm [2], which economically and efficiently uses available remote measurements in a distribution system considering typical measurements from the past and a knowledge on load composition and load behavior at the distribution transformer level. Hence, the configuration with minimum resistive line losses is determined by an heuristic algorithm, [ 11, which efficiency advantages are already proved. Then, the cost effectiveness of studied network reconfiguration is evaluated through an appropriate analysis. The methodology composition and application capabilities can be observed on the chart of Fig. 1.

0 eratorl Decision for reconfiguration

~

PPanner

Continuous Reconfiguration Fig. 1. Flow diagram of the developed analytical tool

The main advantage of the proposed methodology is its flexibility. Load estimation algorithm assures real-time evaluation of a distribution network loading condition. Then, applying minimum power loss optimization algorithm, a new configuration which optimally relates the current loading condition can be found. If the optimal configuration differs from the current corfiguration, potential saving can be calculated through established costhenefit analysis. Making the decision for feeder reconfiguration is left to the system operator/planner. He is responsible to coordinate the decision to the general network operation strategy, equipment availability, security indices, and other relating factors. If this scheme is implemented in a particular time point when the system peak occurs, it would mean network configuration optimization to the peak power losses. Successive application of the methodology cycle at continuous time-of-day points would achieve network continuous recoilfiguration and maximum power and energy loss reduction. Simulating continuous reconfiguration in planning purposes, the cost effective frequency of switching could be

1788

appraised. Furthermore, a similar study analysis would point to the optimal number and locations of remote switches in the network. Energy loss reduction effectiveness of reconfiguration can be also considered by successive running of the load estimation module, for substation forecasted load in a given period T. The developed reconfiguration tool possesses all algorithmic performances for on-line application, as part of DMS. If such a need arises, a software decision-making aid should be also developed to include minimum loss reconfiguration reasons into the other aspects of optimal network operation. A. Load Estimation

In the analyses on optimal operation of a distribution system, information for load on feeder sections and laterals is vital. Unfortunately, for most of the distribution systems, the only information available is the total feeder current recorded at substations. Therefore, it was essential to devise an approach to estimate the real-time system load. This section shortly presents recently developed load estimation method [2], capable to supply input to control functions algorithms, like feeder reconfiguration, for a practical distribution system with very low level of monitoring and automation. The proposed method for load estimation is capable of using every available remote measurement in a distribution system in a simple and efficient manner. The method is load flow based like the analysis for feeder reconfiguration. Consequently, it was easy to fit it into the algorithm of this analysis. The procedure of load estimation begins with checking a real-time system model. The task of this step is examination of the existing scheme configuration by identifying the positions of normally open branches and checking the availability of various links and corresponding equipment. This step is especially practical for the load flow based applications, because by its execution, the radial (or weakly meshed) network is prepared for load flow solution. In the initial load flow solution of the feeder, distribution transformers' load values are taken from the typical diversified load curves or from a load forecast data, whichever is available. This application results in a load flow different from the actual circuit. Therefore, the final step is to adjust the load of the entire circuit until the simulation matches the measured values. Adjustment is made according to assumption of codirmed load where the individual loads change in a proportion of their peak values. The available remote measurements in the system are the current magnitudes at the beginning of each main feeder and the voltage magnitude on the low voltage side of the substations. Distribution transformers' historical load data consists only of the seasonal peak loads. There is not forecasting of load at the distribution transformer level. In order to construct the diversified load curves of distribution transformers, a load research in the study

consume was carried out. We classified the load into four customer classes: Residential I (electrically heated homes), Residential I1 (non electrically heated homes), Commercial and Industrial. Then, we determined typical normalized load curves per customer class on the basis of load measurements in the system and the method proposed in literature [16]. Each customer class differs typical load curves for three representative seasons: winter, fallhpring and summer; and for three representative days of the week: average weekday (Monday to Friday), Saturday and Sunday, for each season. Normalized diversified load curve, which represents the behavior of a distribution transformer load, is composed on the basis of knowledge on percentage consumption of different customer classes in the transformer load and corresponding normalized typical load curves. Typical curves are normalized according to their peak values. So, the approximate load of a distribution transformer in specified time interval of a day is calculated by multiplying the value of its seasonal peak load with the corespondent value from the normalized diversified load curve. The voltage dependence of load, in absence of experimental results, has been modeled on the basis of knowledge on the composition of customer classes consumption. The load of Residential I, Residential I1 and Commercial customer classes is modeled as a constant impedance and the load of Industrial customer class as a constant power. The accuracy of the method is examined on two testsystems [&lo] and verified by field measurements in the distribution system under consideration. Fig. 2 illustrates the appraised method accuracy on the bases of simultaneous measurements of currents at the distribution transformers of two sample feeders in the study network. The information on Fig. 2 represents a distribution of percentage estimation errors. Error of every examined case was determined as a percentage deviation of estimated load value from the corresponding measured value. Then, errors were grouped in 5% long intervals. At Fig.2, the number of estimation errors in one group is presented in percentage of all examined cases. So, it is evident that in 23% of estimated load cases, the error was smaller than 5%; or, that in 59% of cases the estimation error was smaller than 15%. Such an evaluated 25

g. h

*

s

20

0 Proposed estirmtlon method error(%)

.?

-

1

15

10

E % t i 5

0

3

Fig. 2. Distribution of estimation errors

1789

method accuracy was appraised satisfactory taking into consideration the availability of input information. Method Qualification for Reconfimration Study When the load estimation is studied with an intention to be used in a precise application, it is necessary to include considerations on the method validity for that application. For this reason, a sensitivity analysis was carried out. It was performed to show how input data variations influence the minimum loss configuration results. The analysis was in order to appraise the effect of load estimation errors on the recodiguration study results. The determination of the network minimum loss configuration for a given loading condition was performed by the algorithm [ 11, which will be discussed in the next section. This analysis was done for 11 basic loading conditions of the studied network in its "basic winter" configuration. The basic loading conditions were generated multiplying distribution transformers' peak load by some coefficient, in the interval from 0.2 to 1.2. The minimum loss configuration for every basic loading condition was found. The percentage values of losses in these optimal configurations are given in Table I. It is assumed that optimization input data (consisting of distribution transformers' load) has been estimated by the proposed estimation method; then, there are remote measurements of load at the beginning of every feeder in the network; and, consequently, the estimation errors has the same distribution as the results at Fig.2. For every basic loading condition, which distribution transformers' load values were considered exact, 50 possible solutions of the load estimation algorithm were generated by random variation of the basic load. The random generator's distribution was equal to the error distribution from Fig.2. Hence, it was watched for the sum of generated load values to match the basic load value at the beginning of each feeder. Then, for each of the 50 possible loading conditions, related to a basic loading condition, the configuration with minimum resistive line losses was found. Resulting configurations were the same with a corresponding basic optimal configuration (in 13,6% of cases), or they were TABLE I. INFLUENCE OF THE ESTIMATION ERRORS TO RECONmGURATION RESULTS

I

I

I

1

Base case / multiplier

I

0.3 0.4

I

I

I

0.5

0.6 0.7

Power losses for base

I

case (%)

0.4993

I

0.6681

I

1.1

I

I

0.1059

1

0.0907

,

I

2.0658

I

0.1066

1.0096 1.1822

0.0959

0,1039

1.8862

I

0.0834 0.1153

1.7082 1.2

I

(%)

0.8382

I

-

Averare loss deviation

I

I I

0.1119

I

0.0916

I

different in one branch (in 77,7% of cases), or different in two branches (in 8,7% of cases). It is interesting to notice that the maximal difference of two branches does not mean maximal difference in loss values, nor the Same configurations have equal losses. Therefore, the obtained set of optimal configurations was systematized relating to loss value deviation from the basic case loss value. The average percentage deviations of case loss values, from the corresponding basic loss value, are given in Table I. As the information in Table I shows, the deviations are not too much dependent on the network loading level. Taking this into consideration, we can average the percentage deviations in all of the cases. The result is an approximate error of optimization solution of 0. I%, which is consequence of the load estimation inaccuracy. The error value of 0.1% is comparable with the used optimization algorithm error value, and even smaller than the other heuristic reconfiguration algorithms' errors [I]. Moreover, when the practical application aspects of the proposed load estimation algorithm are under consideration, the optimization result uncertainty of 0.1% is assumed very small, too. Namely, in a distribution system with low level of monitoring and automation, the potential savings in losses should be much higher than 0.1% for reconfiguration to take place. All of this verify the used load estimation method for feeder reconfiguration studies.

B. Minimum Power Loss Optimization The determination of the network minimum loss coilfiguration for a given loading condition is performed by earlier developed algorithm [l]. The solution procedure starts with all the network switches closed, (except for those that should not be closed for any reasons). So, we should treat a weakly meshed network instead of a radial network. Network loads are represented by voltage dependent current injections at the network nodes. Load flow method used for nodal voltage and current injections calculation is a compensation based. It is specialized for solving distribution networks using good numerical performances of radial network solution algorithms, which efficiency is based on the oriented branch numbering scheme. Knowing the current injections at the network nodes, currents in the loop branches can be adjusted to minimize resistive line losses. Adjustment is made according to "optimal flow pattern" algorithm [3]. Branch currents for the optimal flow pattern are assured by simultaneous solution of Kirchhoff s current and voltage laws for resistive model of the network. In that model each branch of the network is represented by its resistance. At each stage of the process, when optimal flow pattern is reached, among all loops of the network, the branch having the lowest current has to be opened. The result of this act is elimination of one of the network loops that cause minimum disturbance in the optimum flow pattern.

1790 The entire process, starting with the load flow solution for the actual meshed network configuration up to the opening of the switch in the branch carrying the smallest current, is repeated until the network becomes radial. Every time when one branch is switched off, the loop number decreases and the network gels new configuration. Increased efficiency of the method is achieved by implementation of the algorithms for partial reordering of network branches and loop impedance matrix re-evaluation [ 11. The network configuration determined that way is optimum or near optimum solution of the problem. The algorithm accuracy is appraised on test-networks [5,8,10] and results are presented in literature [l]. We consider that result uncertainty of the used minimum loss configuration method, for the study system, is not higher than 0.1%. C. Cost Evaluation

The objective in distribution network operation is to minimize costs, while the rel~abilityremains at least on the same level. To evaluate the costshenefits of minimum loss reconfiguration the following has to be taken into consideration: i) costs of losses; ii) costs of switching; and iii) outage costs. We calculate losses in a distribution network with the prices that distribution utility pays for taken power and energy. Such a valuation of losses comes from the basic aim of the proposed methodology. Namely, the methodology is developed as an optimization tool which should help the utility to decrease costs for losses. The final effect of proposed optimization would be decreasing of energy and peak power quantities taken from the transmission. Evaluating the costs of losses, it should be paid attention to the fact that the optimization of network configuration for maximum reduction of peak power losses does not mean maximum reduction of energy losses; and nor does the opposite. When the optimization on peak power losses results with a different network configuration than the optimization on energy losses, an analysis on the costeffectiveness of these two configurations should be carried out. Result depends mostly on existing prices of kWh, and peak kW and on quantities of loss reduction achieved by the different optimizations. Usually, the price of peak kW is 300-500 time higher in value, than the kWh price. It is estimated that the minimum peak power optimization takes advantage if the ratio of positive difference of energy losses and positive difference of peak power losses in two optimal configurations i s higher than 300-500. The cost of switching depends mostly on the level of automated control in a particular distribution network. If the recoIlfiguration is performed by remote switching, the cost would be evaluated by considering the price of switching devices and the influence of frequency of switching on the life span of these devices. Hence, when the manual switching is under consideration, the labor costs and costs of the possible interruption of service should be added. Anyway, the utility should provide a methodical strategy of feeder reconfiguration to ensure minimal interruption of

client service, and inviolation of functional ability of the equipment. The switching sequence must be selected such that the transient effect is minimized and no temporary line constrains are violated. An optimal switching strategy must be developed prior to an efficient implementation of a reconfiguration. The outage costs indicate the inconvenience to the utility or the customer caused by intemptions in a distribution network. The utility costs include the loss of revenue from the customers not served and increased expenditure due to maintenance and repairs. We account for these costs in our study. Types of customers are not equally sensitive to frequency and duration of interruptions in energy supply. So, the cost evaluation met by customers differs in dependence on customers' category. It is very delicate to find out direct effects of feeder reconfiguration to the customer service and system reliability. Most outage costs are due to faults and they can be reduced using advanced methods for fault location and restoration. In normal operation the expected outage costs can be taken into account by improving the reliability of the network. The direct reflection of a frequent reconfiguration to the system reliability would be a task of another detailed study. Our considerations are, besides the risks of equipment damage and switches' life span shortening that negatively influence the system reliability, that the reconfiguration for loss reduction contributes to the load balancing of feeders, that, on the other hand, improves security indices. Which of these influences will prevail depends on system characteristics, like existing reliability indices, state of operational equipment, etc. IV. STUDY ANALYSES AND RESULTS The performed analyses in order to plan the network configuration for January '96 are demonstrated in the following. January is the coldest month in the climate area. The system is usually heavy loaded and many system's substations reach their yearly peak load during this month. We used the historical information for this period of the year to simulate network loading condition. The corresponding available information was: - seasonal peak load of distribution transformers occurred during the winter '95, - remote meters' records for the average working day in January '95, and remote meters' records for the average non-working day in January '95. January '96 had 23 working and 8 non-working days. Studied network was in its "basic winter" configuration. The application of the load estimation algorithm on recorde load for this configuration showed that power losses would be in the interval from 0.5% to 1.3%, depending on daily load variations. There would not be any voltage limit violation in the network. Ap~roximatingthat every working day in January '96 is the same with the average working day in January '95, and every non-working day in January '96 with the average non-working day in January '95, then, total

-

1791

monthly energy losses would be 424,560 kWh, or 1.05% of total delivered energy. The computer simulation of continuous network reconfiguration was carried out in order to examine the maximum possible savings in losses (simulation SI). In the simulation, the decisive role of the operator/planner was substituted by the rule: network reconJguration is performed each time when the power losses in the obtained optimal configuration for certain loading condition are for at least 0.2% smaller than those in the current conJguration. The value of 0.2% was determined by summing up the intervals of result uncertainty of used estimation and optimization algorithms. In order to generate as many minimum power loss configurations for further analyses as possible, the costs of switching were not considered in this stage. The results of S1 simulation for the average working and non-working day are shown in Table I1 and Table 111, respectively. Different optimal configurations got their marks in dependence on the type of day and hour of their occurrence. Table I1 and Table 111, for each configuration, contain the number of branches different than those in the previous configuration, then, the power loss savings in the period of daily peak and the appraised monthly energy loss savings, in case if the configuration stays in operation for the whole month. Loss savings are calculated in relation to losses in the network "basic winter" configuration. It should be noticed that, for the working day peak hour occurs at 5 p.m. and for the non-working day at 1 p.m. In comparison to the "basic winter" configuration, by performing such an ideal continuous reconfiguration, the following loss savings could be accomplished: 38,570 kWh of energy losses for the month (or, 8.57% of total monthly energy losses), and 69 kW of peak power (or, 7.53% of peak demand). With the current averaged price of energy and power of $0.042/kWh, these monthly savings are appraised to $1,620. This is a noticeable amount, but the continuous reconfiguration can hardly be proved if costs of switching are taken into consideration. Namely, one line switch costs about $2,000, and it can be operated 1000 times during the life cycle. Besides, the average labor cost to operate a switch is $21. Therefore, in our study, the total cost of one switching is approximately $23. The number of switching operations for performing the continuous recotfiguration during the month is 1,698. So, the amount of monthly benefits for saved losses is incomparable in regard to the amount of monthly costs for switching ($39,054). Even, if the switches were operated remotely, the cost for switching would be predominant. The same simulation was repeated, but under condition of performing network reconfiguration each time when loss savings higher than 0,5% ware achieved, (simulation S2). The results of this (S2) are shown in Table IV and Table V. The information in these two tables corresponds to the information in Tables I1 and Tables 111. It can be noticed that the number of switching (816 switch operations for whole month) decreased significantly. Such a continuous reconfiguration could accomplish savings of 37,857 kWh energy

TABLE U. RECONFIGURATION S 1 FOR AVERAGE WORKING DAY

35,604 35,651

I

9am-2pm

I

W9am

I

4

1

64

3om-10om 10pm-12pm

I

~ 3 I~ WlOpm

I

m5 2

I

I

69 68

I

No.of different branches

I

Peak power loss reduction (kW)

I

Configu

Period

-ration

3am-6am

N3am N6am NSnm N7nm

6am-5pm 5pm-7om 7om-11~m

I

6 8 11

I

Configu

Period

-ration

Oam-6am

WlODm

6am-l0pm

W6am

~

I

3

I

No.of different branches I 4

I

36,348 35,604

I

I

Monthly energy loss reduction (kWh)

65

35,045

63

I

34,443 34,978 35.769

66 68

Peak power (kW) 68

Monthly energy loss reduction (kWh) 35.604

66

35,385

Peak power loss reduction (kW) 68

Monthly energy loss reduction (kWh) 35,604 34,526

loss reduction

1

I

34,264

WlOom

No.of different branches I

4"1-6am

N4"n

6

65

6am-7pm 7om-1 lnm

N6am

8

63

34,443

N7om

8

68

35.769

Confgu

Period

-ration

Oam-4am

I

losses for the month (or, 8.42% of total monthly energy losses), and 66 kW peak power losses (or, 7.24% of peak demand). This savings are equivalent to $1,590. However, the costs of switching ($18,768) still exceed the benefits of this continuous reconfiguration. Contrary to the results of simulated continuous reconfigurations, almost every of sub-optimal configurations in Tables I1 - V offers pretty much the same savings, but without any switching during the month. In this fact, we saw the potential benefit of the distribution network minimum loss reconfiguration for our study. The system operation planner should make a choice of network configuration that would stay in operation at least for a month in one season. All of the offered cotfigurations are very close to each other by their potential loss savings. Therefore, the configuration satisfying remaining technical criteria for well operation of a distribution network or/and some practical

1792

criteria known to utility personnel, would take the advantage. Nevertheless, in order to complete this example of the study network configuration planning, we chose the configuration W3pm. This is a configuration that is both optimal at peak hours of an average working day and such that brings the greatest saving of energy losses. Implementation of this configuration, in comparison with the current configuration, would achieve the following savings: 36,348 kWh of energy losses for the month (or, 8.13% of total monthly energy losses), and 69 kW of peak power (or, 7.53% of peak demand). The financial effect of this implementation would be $1,540 for the month. V. CONCLUSION AND DISCUSSION

This paper has proposed a systematic methodology for distribution network minimum loss reconfiguration analyses. The methodology contains three main parts: real-time load estimation, effective determination of minimum power loss configuration and costhenefit evaluation. Such a construction permits its flexible use in study analyses on optimal operation, as well as, in short-term planning of distribution systems. It can be also used in off-line decision making for feeder reconfiguration, or even in on-line control of network operation, if corresponding computer applications are developed. By using the proposed methodology, a study on practical minimum loss reconfiguration benefits was carried out. The study was performed in cooperation with Skopje's Utility, for optimal operation planning of their distribution network. Part of the study analyses is presented in this paper. The results of the study contribute to the highlighting of the practical minimum loss reconfiguration benefits. Skopje's distribution network does not lead itself into reconfiguration, not even in the theoretically most suitable load conditions for reduction of network losses by this method. The reconfiguration on the time-of-day basis is still far from being economically proved. Even an automated switching would not bring better cost effectiveness to frequent network reconfiguration. The network configuration planning should be studied seasonally, trying to introduce some smaller corrections at least once a month. Such a distribution network planning can easily obtain great cost reduction. VI. ACKNOWLEDGMENTS Financial support given to this work by the Ministry of Science of the Republic of Macedonia is appreciated. The authors wish to thank Skopje's utility company for providing system data associated with the study. Especially, we wish to thank A. Sekerinski, G. Janevski and S. Ilievska for all of their support.

VII. REFERENCES V.Borozan, D.RajiCid, RACkovski, "Improved Method for Loss Minimization in Distribution Networks", IEEE Trans. on PWRS, Vol. PWRS-10, NO. 3, August 1995, pp. 1420-1425. V.Borozan, N.RajakoviC, "Distribution System Load Estimation for Feeder Reconfiguration Studies", Proc. of V International Conference Tesla IIIMillennium, October 1996, Belgrade, Yugoslavia, pp. AMerlin, H.Back "Search for a Minimal-Loss Operating Spanning Tree Configuration for an LJrhan Power Distribution System", Proc. ofPSOC, Cambridge 1975, paper 1.216. R.J.Sa& M.M.A.Salama, A.Y.Chikhani, "A Survey of the State of the Art in Distribution System Reconfiguration for System Loss Reduction", Electric Power Systems Research, Vo1.31, No. 1, Oct 1994, pp.61-70. S.Civanlar, J.J.Grainger, H.Yin, S.S.H.Lee, "Distribution Feeder Recontiguration for Loss Reduction," IEEE Trans. on PI.z'ipD, Vol. PWRD-3, NO. 3, July 1988, pp. 1217-1223. D.Shirmohammadi, H.W.Hong "Reconfiguration of Electric Distribution Networks for Resistive Line Losses Reduction", IEEE Trans. on Pu/RD, Vol. PWRD-4, No. 2, April 1989, pp. 1492-1498. C.C.Liu, S.J.Lee, K.Vu, "Loss Minimization of Distribution Feeders: Optimality and Algorithms", IEEE Trans. on PWRD, Vol. PRWD-4, No. 4, April 1989, pp. 1281-1289. M.Baran, F.F.Wu, "Network Reconfiguration in Distribution Systems for Loss Reduction and Load Balancing", IEEE Trans. on PWRD, Vol. PRWD-4, No. 2, April 1989, pp. 1401-1407. H.D.Chiang, R.Jean-Jumeau, "Optimal Network Reconfigurations in Distribution Systems: Part 1: A New Formulation and A Solution Methodology", IEEE Trans. on PWRD, Vol. PRWD-5, No. 4, November 1990, pp. 1902-1909. H.D.Chiang, R.Jean-Jumeau, "Optimal Network Reconfigurations in Distribution Systems: Part 2: Solution Algorithms and Numerical Results", IEEE Trans. on PWRD, Vol. PRWD-5, No. 3, July 1990, pp. 1568-1575. LRoytelman, V.Melnik, S.S.H.Lee, R.L.Lugtu,"Multiobjective Feeder Reconfiguration by Distribution Management System", IEEE Trans. on PWRS, Vol. PRWS-11, No. 2, May 1996, pp. . R.P.Broadwater. A.H.Khlan, H.E.Shaalnn, R.E.Lee, "Time Varying Load Analysis to Reduce Distribution Losses Through Reconfiguration", IEEE Trans. onPkJRLl, Vol. PRWD-8, No. 1, Janu'uy 1993, pp. 294-300. J.B.Patton, D.T.Ritzy, J.S.Lowler, "Application SoRware for Modeling Distribution Automation Operations on the Athens Utilites Board", IEEE Trans. onPU'RD, Vol. PRWD-5, 1990, pp. 1019-1025. C.S.Chen, M.Y.Cho, "Energy Loss Reduction by Critical Switches", IEEE Trans. on P W , Vol. PRWD-8, No. 3, July 1993, pp. 12461252. G.J.Peponis, M.P.Papadopoulos, N.D.Hatziargyriou, "Optimal Operation of Distribution Networks", IEEE Trans. on P W S , Vol. PRWS-11, NO. 1, February 1996, pp. 59-67. E.Handschin, Ch. Dornemann, "Bus Load Modelling and Forecasting", Proc. ofPICA '87, May 1987, Montreal, Que.

.

Vesnn P.Borozan (M '92) was born in 1962. She received her B.S. and M.S. degrees in Electrical Engineering from the University "Sv. Kiril i Metodij" in Skopje, and her Ph.D. degree &om the University of Belgrade in 1996. In 1987 she joined the Liniversity "Sv. Kiril i Metodij" in Skopje, and presently she is a teaching and reseweh waistant in Power Systems at the same University. Her subjects of interest are computer applications in power and distribution system analyses. Nikola Lj. Rujakovid (M '89, SM '94) was bom in 1952 in Yugoslavia. He

received his Ph.D. degree from the University of Belgrade, Yugoslavia in 1983. Presently he is a full professor at the same University. He has published over seventy papers (15 in refereed journals), and three textbooks. He has also worked in nunierous power system projects. His research interests include steady-shte analysis of power systems, power system optimization and harmonic modeling. He is a member of the CIGRE and chairman of the PES Chapter in Yugoslavia.