Reduction Of Active Losses With Reconfiguration Of Electricity Distribution Networks

REDUCTION OF ACTIVE LOSSES WITH RECONFIGURATION OF ELECTRICITY DISTRIBUTION NETWORKS Gorazd KERB INEK

Miran H O R V A T , Joie VORSIC, Andrej O R G U L A N University of Maribor, Faculty of Electrical Engineering and Computer Science, 2000 Maribor, Slovenia

Abstract: Electricity distribution networks in densely populated urban areas are, in order to increase the reliability of supply, usually designed as meshed networks, though they

always operate in radial configuration. Reconfiguration of distribution networks is achieved through switching operations on switching devices of distribution network branches (switching odoff of a branch). It is often the case that the existing configuration is not the configuration with minimum power losses. With the new computer program, developed by our research team, it is possible to find the network configuration that enables operation with minimum losses. The program proposes the best network configuration to the dispatcher, and the latter makes his own decision whether to change the network configuration or not. Mathematical algorithm for the network reconfiguration is presented in the paper, as well as results of the reconfiguration of the electricity distribution network of the city of Maribor. Keywords: Distribution network, network reconfiguration, load flow, losses.

I. I N T R O D U C T I O N In the past the problem of costs caused by network losses was constantly neglecting. Gaining of independence of the Republic of Slovenia in 1991 and introduction of market economy are the reason that this issue is becoming more and more important. Electric power system components, i.e. generators, transformers, overhead lines, cables, substations and

0-7803-5515-6/99/$10.00 0 1999 IEEE

Slovenian Power Authority Vetrinjska 2, 2000 Maribor, Slovenia

consumer devices, are electrically and magnetically connected in order to supply consumers in a certain geographic region with quality electric power. The electric power system is subject to incessant changing, since every turning on or off in the system alters consumption and generation in the system, in many cases also the system configuration. All these changes have also impact on network losses. The objective of every electricity distributor is to operate the network with minimum losses and maximum reliability of operation. To comply with these objective it is necessary to apply one or more of the various network optimization methods. Searching for the network configuration with minimum power losses generally belongs to mixed-integer linear programming. Several papers have so far been written about this topic. Since the beginning of researching in this field, where Merlin and Back [ 11 are undoubtedly pioneers, several algorithms have been developed [2,3,4]. It is well known that distribution networks operate in radial configuration. Radial configuration is more suitable for distribution networks than meshed configuration, since it brings lower fault currents, easier switching of branches or consumers, and simpler protection of network components, although it does not ensure such reliability of supply as meshed configuration. Achieving higher reliability of supply is the reason why electricity distribution networks in densely populated urban areas are designed as meshed networks, though they always operate in radial configuration. The computer program finds the network configuration with minimum energy losses under all given constraints. This configuration can be achieved by changing of openlclosed state of switching devices on tie lines. This procedure is called network reconfiguration. Mathematical model, used in our computer program for network reconfiguration, is presented in the paper, as well as results of the reconfiguration of the electricity distribution network of the city of Maribor.

154

II. MATHEMATICAL MODEL OF NETWORK

b---I

U"

RECONFIGURATION

U,I

rlh

xo

ri I 9 xi I

U,

I

U,+/

Ut,

P L ~ + QL,+I I.

PL,,, QL,)

Pm Qo

The essence of the problem of searching the optimum network configuration with minimum energy losses is to find the minimum of energy losses for all active branches of the network (Fig. ij.

t

t

t

'1

t

t

t

\

\

ss2

P L I~, Q L ~I

PL.~, QL,

Fig. 2: Radial electric network The number of open line switches defines the number of loops in the network. With switching these switches on and off it is possible to establish radial network configuration under condition that all network nodes (distribution transformed) are supplied. The higher is the possible number of switching operations, the higher is the number of possible combinations that can bring a reduction of power losses in the network. In the first step, for testing of the method applied in the described program, all possible combinations that give radial network configuration have been found for a small test network [4].For all these combinations load flows have been calculated. In this way the optimum network configuration with minimum power losses has been found. With the increasing number of possible switching operations, i.e. in larger networks, this method requires an enormous number of calculations and thus too much time. This is why the applicability of this method is very limited. It can be used only as a reference for testing other methods, while it is absolutely inadequate for the use in actual distribution networks. Switch Exchange Method has been chosen for the use in our computer program for network reconfiguration. The idea of this method lies in searching of optimum for each separate loop. Firstly an open branch (tie line) is closed. In this way a loop is closed in one part of the network. Then in such loop, one branch after another is opened, and for each such state load flows and losses are calculated. The position of open branch that yields the minimum power losses is remembered for the next step. This is repeated for all open branches of the original configuration. When this is done, the first step is completed. The calculated optimums from the first step are then used in the second one, when the whole procedure is repeated. Searching for optimum configuration is finished when two consecutive steps indicate the same optimum configuration. This method is simple and does not require too much computation time, though it allows only finding of a local minimum, which is very close to the global optimum, found by the very time consuming method described above. The procedure can also be accelerated by elimination of

9 I 10 I I 1 '-I 12 I' 13 I 14 I 'I h----

Fig. 1: Small electricity distribution network The following operational conditions have to be always fulfilled: all nodal voltages and all branch currents have to be within the prescribed limits, and network configuration has to remain radial. Minimum of the following function has to be found

under constraints:

i = l , 2,..., n where (Fig. 2): n - number of buses in the network

r; - resistance of branch i ii - current in branch i Pi - active power flow in branch i Qi - reactive power flow in branch i U,- voltage of node i

(3)

155

branches that do not contribute to the reduction of losses. This is an iterative method that with the application of a quick algorithm for load flow calculation in radial networks yields results after a very short computation time. The computer program on the basis of this method has been tested using t'ne data on the actual electricity distribution network of the city of Maribor. The optimum network configuration that enables operation with minimum losses has been found, taking into consideration all constraints (voltages and currents within the required limits, limited number of switching operations). Fig. 3 shows a flow diagram of the above described method.

Load Flow Calculation for Radial Networks A simple iterative method for load flow calculation in radial networks is used in our program for network reconfiguration, which is for such networks much faster than well known iterative method for meshed networks, such as Newton-Raphson or Gauss-Seidel methods. Fig. 4 shows a simple radial network.

Lcvcl I

Reading of network data

.c N = number of open lines K = 1 step

1

I7

Load flow calculation for existing configuration

4

closing of open line (forming of loop) M = number of lines in the loop

I I

Fig. 4:Branched radial network

I

All network branches are classified by levels, where the number of level increases with increasing distance from the point of supply. The ending nodes of every branch have always the same number as the respective branch. The point of supply (root node) is labeled with 0. At the beginning let's presume that voltage in all network nodes equals the voltage of the root node. (0)- U up --0

[M = M + 1

Load flow calculation Ploss,,emp < pI,, + optimum configuration

p = l , 2 ,...,n

(4)

where n is the number of nodes and govoltage of the root node. Then current injections are calculated for all nodes

N=N+1

Is this new optimum

K =K+1

I

1

where:

k - iteration number, 8,- power injection in node p ,

YES

Printing of optimum configuration

&, - total shunt admittance at node p .

I

This is followed by the calculation of branch currents, starting from the branches in the last level (having the highest branch number) and moving towards the branches connected to the root node.

Fig. 3: Flow diagram of searching the optimum network configuration

156

where: m - number of branches, q - ending node of the branch j , D - set of branches that start from the node q. - transformers

When all branch currents are calculated it is possible to determine the nodal voltages. They are calculated starting with the first level moving towards the last one.

Load flow calculation

Fault analysis - single pole - two pole -three pole

- Gauss-Seidel method - Newton Raphson method

- Voltage drops (radial networks) Finding of optimum configuration Network reconfiguration

where 5 is the impedance of the branch j , p and q are starting and ending node of the branchj.

I

I

The power mismatch at all network nodes equals:

AS - =-s , , ---Pu ( ~--P) . I ( ~ ) p * = 1 , 2 ,...,n The procedure of calculating of current injections in nodes, branch currents and node voltages is repeated until the power mismatch in all nodes is less than the specified tolerance the branch E. Load flows and network losses are calculated after the iterative procedure has been finished.

III. PROGRAM PACKAGE FOR NETWORK ANALYSIS The flow diagram of the program package for distribution network analysis is shown in Fig. 5. It enables load flow and fault analysis, as well as network reconfiguration. The results are presented in textual and graphical form. Graphic presentation of the results can be done schematically or geographically supported with the use of Geographic Information Services (GIS). The program package is written in Fortran. It works on personal computers in MS Windows 95/98 or MS Windows NT 3.5 1/4.0 operating systems. The only limitations of the program are computer RAM and available space on hard disk.

Reestablishment of network after faults

Printing and plotting of results

Fig. 5 : Flow diagram of network analysis

Calculations of network reconfiguration require network parameters, such as data on lines, transformers and consumers. For easier data input the typical data on lines, such as impedance's, capacities, voltage levels and maximum current or load flow capacity, are stored in a separate database. Lines or branches are therefore described only with cable type and its length, which makes the work with the program more transparent and eventual changing easier. Since some branches in the network cannot be opened, each line also has to be described with a parameter indicating whether the line can be opened or not. Individual nodes or transformer stations are described with the nominal voltage, consumption of active and reactive power and, if it exists, data on the power factor correction device. For plotting of the results geographical coordinates have also to be given. Instead of power consumption in nodes it is also possible to input the energy consumed in a certain period, together with the load curve for this period. This approach allows defining the typical consumers, such as households, industry, mixed consumers, etc. Once the network data are given to the computer, it is possible to perform load flow and fault analysis and to determine the optimum operating network configuration. The results are given in textual and graphic form. They can also be exported to existing Geographic Information Services (GIs), where the results become more transparent.

157

IV.ANALYSIS OF MARIBOR NETWORK AND RESULTS Data from 1997 have been used in the analysis of Maribor distribution network. The network contains four large 110/10 kV substation and one smaller 35/10 kV substation. These five substations are supplied from two bulk substations via 110 kV overhead lines. The major part of the 10 kV distribution network (about 200 km) is made of underground cables. The analysis took into account 37 1 transformer stations (nodes) and 453 lines between them (branches). The most common nominal capacity of the transformer stations is 630 kVA, while the others have nominal capacities 400, 250 or 100 kVA. The majority of the transformer stations is not remotely controlled. They supply about 50,000 consumers that in 1997 consumed some 400GWh of electricity and reached 65 MW peak power. This network supplies several types of consumers. Three types of consumers have been considered in the analysis: households, shops and industry. Energy consumption of these typical consumers has been monitored for years. Thus it has been possible to obtain their load curves for working days (Monday to Friday) and weekends (Saturday and Sunday).

Table 1 and Table 2 show the results of the results of network reconfiguration simulation for winter period. The tables show loss reduction and monthly energy saving (energy required for switching operations has been neglected) with regard to the existing network configuration. For each month 21 working and 10 nonworking (weekends) days have been considered. In the real circumstances it is impossible to reconfigure network three or four times a day, since the transformer stations are not remotely controlled. Because of this problem another simulation has been done. In this simulation only two configurations have been used, one for winter and one for summer period. Average consumption in these seasons has been used. Annual saving due to loss reduction would in this case be about $18,000. The results are given in Table 3. Table 1: Reconfiguration for working day

I

Day. period -

0:OO-6:OO 6:OO - 17:OO 17:OO - 22:OO 22:OO - 24:OO

I I

Loss reduction (kW) 28 39 68 16

I I

I

Monthlyenergy saving (kWh) 3.528 9.009 7.140 672

I

Table 2: Reconfiguration for weekend Daily load curve for households (working days -winter)

saving (kWh)

p(P4

1.oo

8:OO - 14:00 14:OO - 22~00 22:OO - 24:OO

0.75

3.120 2.240

340

-

0.50

Table 3: Proposed reconfiguration

0.25 0.00 0o:oo

I 06:oo

12:oo

18:OO

Period

0o:OO

Time

Winter Summer

I

Loss reduction (kW) 43 24

I Yearly energy saving I

IC

Daily load curve for households (weekends -winter) fJ

I

- (kWhj 163.860 105.120 268.980

I

(PU)

V. CONCLUSIONS

0000

0600

12 00

18 00

0000

Time

Fig. 6: Daily load curves for households in winter

Finding of the optimum network reconfiguration represents, regardless of the applied method, mathematically very complex and comprehensive problem. The progress in the computer technology enables the use of personal computers in solving these problems. Since the consumption in nodes is changing all the time, it would be necessary to reconfigure the network all the time just as the current optimum 158

configuration is changing with consumption. This requires a lot of data on current consumption in all nodes, simultaneous calculation of reconfiguration followed by immediate realization of all newly required switching operations in the network. This would be possible only if fully automated control system of the network existed. Building of such a system would be too expensive. One should also keep in mind that each switching operation requires some energy and that each switching shortens the switch's lifetime. This is why it is reasonable to change configuration for a larger time period, such as week, month, or season, like summer and winter. Results of the simulations for the Maribor electricity distribution system clearly show that, in spite of nonautomated network, it would be reasonable to manually reconfigure the network at least twice a year (in summer and winter). With regard to the current state of the network and because of increasing prices of electricity the savings due to loss reduction would be relatively high. The second benefit from the actively used program package for network analysis is reestablishment of the network after faults. Such a program package can also be efficiently used for planning purposes. The final conclusion of this study is that the investments in hardware and software would be quickly repaid in the new market conditions of the power system operation, which are to come in force in Europe in 1999, when it will be reasonable to optimize the distribution network configuration in real time. W e shall see what will bring the new market era.

VI. ACKNOWLEDGEMENTS The authors wish to thank the Ministry of Science and

Technology of the Republic of Slovenia for the financial support in the realization of the study on network reconfiguration. They also greatly acknowledge the support of "Elektro Maribor", the electricity Distribution Company of the city of Maribor, in collection of all data required for this research.

VU. REFERENCES A. Merlin and H. Back, "Search for a Minimal-Loss Operating Spaning Tree Configuration in Urban Power Distribution Systems", Proc. of 5'" Power Systems Comp. Con., Cambridge, U.K., Sept. 1-5, 1975. S. Civanlar, J.J. Grainger, H.Yin, S.S.H. Lee, "Distribution Feeder Reconfiguration for Loss

Reduction", IEEE Trans. on PWRD, Vol. PWRD-3, No. 3, July 1988, pp. 1217-1223. [31 D. Shirmohammadi, H.W. Hong, "Reconfiguration of Electric Distribution Networks for Resistive Line Losses Reduction", IEEE Trans. on PWRD, Vol. PWRD-4, No. 2, April 1989, pp. 1492-1498. [41 S.K. Goswami, S.K. Basu, "A new algorithm for the reconfiguration of Distribution feeders for loss minimization", IEEE Trans. on PWRD, Vol. PWRD-7, NO.3, July 1992, pp. 1484-1491.

VIU. BIOGRAPHY Miran Horvat was born in Maribor, Slovenia, on March 31, 1965. He received his B.S degree from Faculty of Electrical Engineering and Computer Science in Maribor in 1989. In 1989 he joined the Faculty of Electrical engineering and Computer Science in Maribor, and presently he is a teaching assistant in Power Systems at the same University. His current area of interest is analysis of transmission and distribution systems. Joie VorSiE was born in Maribor, Slovenia, on November 24, 1946. He received his B.S. degree from the University of Ljubljana in 1972, his M.S. degree from the University of Zagreb in 1982, and his Ph.D. degree from the University of Maribor in 1983, all in electrical engineering. In 1972 he was engaged at the University of Ljubljana as a research assistant. His work was focused on the research of the quality of electrical energy. In 1974 he joined the University of Maribor as an assistant. At present, he is Professor of Power Systems Engineering. From 1976 he was an active member of the self managerial governing bodies of the Faculty of Technical Sciences, and later at the Educational Community of R Slovenia and the Ministry of Science and Technology. Andrej Orgulan was born in Maribor, Slovenia on October 28, 1960. He received his B.S. degree in electrical engineering from the University of Maribor in 1986, his M.S. degree from the University Maribor in 1997. In 1989 he joined the University of Maribor as a research assistant. He works in researching of the Quality of Electrical Energy. Gorazd Skerbinek was born in Maribor, Slovenia, on June 28, 1965. He received his B.S degree in electrical engineering from the University of Maribor in 1989. In 1989 he joined the University of Maribor, where he worked as researcher in the field of analysis of transmission and distribution systems. In 1993 he joined the Slovenian Power Authority as development engineer for power system expansion planning.

159