Novel Multiagent Based Load Restoration Algorithm For Microgrids

M.Syed Mohamed, EEE Dept/U.G.Scholar, Vels University, Chennai Abstract:

This paper presents a new application of genetic algorithms (GAs) optimization technique for solving the steady state loadshedding problem with cogeneration in operation. The objective is to minimize the sum of the squares of the difference between the connected active and reactive load and the supplied active and reactive power. The equality constraints associated with the problem of load shedding are the power flow equations. The inequality constraints are the voltage magnitudes and angles, active and reactive power generations and the limits on the line flows. The effects of GAs parameters and operators, such as population size, crossover, mutation and fitness function, are studied. Simulated results indicate that the proposed GA-based method is both feasible and effective to facilitate optimal load shedding planning. service area. To meet the service restoration, the alteration of topological structure of distribution system is done by changing the status of switches in distribution system satisfying electrical and structural constraints. For the modern day distribution system, it is hard to implement service restoration solely depending on experimental rules by human operators. To reduce the out of service area as efficiently as possible and the burden of operators, a computer aided decision supports assist the operators. The researchers have developed many methods to solve the service restoration problem in distribution systems [1]-[12]. Heuristic techniques [1]-[4] have been developed using heuristic rules. Because heuristic techniques rely on the knowledge of operator of distribution systems, a compromise between knowledge acquisition and performance of the solution is made. Expert systems [5]-[8] have been developed to quickly determine restoration plans and build look-up tables for distribution personal. Fuzzy set theory [9], network reduction technique [10] and the ranking based search method [11] have also been developed. The petri-net approach [12] has been developed. In [12] the knowledge or the configuration of the concern system in the form of graphic representation is expressed through a structured model. In [11] genetic algorithm is used to solve the service restoration problem. Before automation in distribution systems, only manual controlled switches were used in distribution system. After development of automatic controlled switches for the purpose of automation, these

Introduction:
FAULTED events are unavoidable in the huge and complex electrical power distribution systems. These faults affect the systems reliability and customers satisfaction. So the reduction in the effect of fault is necessary to maintain the systems reliability and customers satisfaction by restoring the service quickly in the area left unsupplied due to fault. After occurrence of the fault, the operator finds the location of fault, isolate the fault and then restore the service to the healthy components of the out of

manual controlled switches started to be replaced by automatic controlled switches. As a result, three categories of presently existing distribution systems can be seen. 1. Those have only manual controlled switches. 2. Those have both manual and automatic controlled switches. 3. Those have automatic controlled switches only. The operating time of both types of switches is different. Therefore, both type of switches should be considered separately. Sometimes service restoration for whole out of service area is not possible because power flow in the feeders goes beyond their power transfer capacity. Including the capacitor-controlled action can increase the power transfer capacity of the feeder and, therefore, enhances the chances of full service restoration. If full service restoration is not possible with capacitor control action also, including capacitor control action is definitely helpful to reduce out of service area. The distribution systems are required to operate in radial fashion for proper relay coordination and ease of fault location etc. So the structure of distribution system should remain radial after service restoration also. In any distribution system, there are always some loads, which are of highest priority (e.g. hospital). In the event of partial service restoration, the supply must be restored to highest priority customers and this fact should be reflected in the final solution of service restoration problem. In this paper the authors solved the service restorations problem using conventional genetic algorithm. During the optimization process of genetic algorithm, there is high probability of generation of better solution at every next generation. The string is represented by status of switches. In this work the original configuration is included in the initial population. Inclusion of original configuration in initial population reduces time to reach at optimal solution drastically. In this method, the following points are

considered. 1. Manually controlled switch operation and remotely controlled switch operations are considered separately. 2. Priority customers are considered 3. Out-of service, number of manual switch operation, number of remotely controlled switch operation and losses are minimized. 4. Voltage, current, rediality of the network and supply to priority customers are taken as the constraints. 5. To restore the service to the out of service area, capacitor switches are also considered along with tie switches and sectionalizing switches. In fact, this work is extension of method [11] developed by W.P. Luan and M.R. Irving. In method [11], remotely switches and manual switches are not considered separately and priority customers are also not considered.

2. Problem Formulation
When a fault occurs in a system, it is essential to restore the system with minimum time. The minimization of the time is directly related with the minimization of the number of switching operations. The objective of this work is to find the optimal network for restoration with minimum system loss, voltage drop and number of switching operations. The problem is solved and analyzed into two cases. 1)Optimal network without DG 2)Optimal network with DG The location and size of the DG are selected for minimum loss and voltage drop. The objective function and the constraints of the service restoration problem are described below.

2.1 Objective and Constraints

The total real power loss, the voltage drop and the number of switching operations respectively.

simplified manner, using only the parameters necessary for the implementation of AC flows. The electrical system components included are: generators, lines and / or cables, transformers and feeders.

The weighting factors are selected such that W1+W2 + W3 The numbers of switching operations are calculated using the expression given by Yogendra et al. (Yogendra Kumar, Biswarup Das, and Jaydev Sharma., 2005) swj is the status of jth switch just after the fault and k j is the status of jth switch in the restored network. The constraints are as follows. Radial network should be maintained. Bus voltages should be within the acceptable range, which is given by equation below

Line currents should be within the acceptable range, which is given by equation below

The generators are modeled as a typical bus in any load flow, with the active power and voltage defined as known data. In transformers, series resistance and the magnetizing branch are neglected leaving only the series reactance The load buses are modeled as constant active power regardless of the reactive power In the line model, series resistance, shunt conductance and shunt susceptance are neglected leaving only the series reactance. The graph plot for voltage vs active and reactive power is shown.

MODELING OF THE ELECTRICAL NETWORK

The electrical system is modeled in a

IV.GENETIC ALGORITHMS
A. Introduction
Genetic algorithms are adaptive methods generally used in search and optimization data problems, based on sexual reproduction and the principle of survival of the fittest.

exchanges the subsequence before and after the crossover point between individual parents to form children. - Mutation: This operator randomly changes one bit of the inhabitant given by the probability of mutation.

V. FITNESS FUNCTION
It must be define a fitness function that qualifies each inhabitant of the population. After that, the population can be sorted by ranking. The first thing to take into consideration is the load connected to the network. It is a desire to have connected load as much as possible. Therefore, the function must reward inhabitants with more load connected. For this, two variables are created to quantify the load remains connected, the first is the amount of load remains connected ,the second is the number of buses that remains on service. The fitness function is implemented to be an increasing function in these variables. Another variable to consider is the overloaded equipments. The main target is to eliminate every current violation. Therefore, the function fitness must assign smaller fitness values for the most overloaded network. In this case, two variables are defined to represent how overloaded is the network: sob, number of overloaded branches (or transformers) and Smax, is the ratio between the line load of the most heavily loaded line and the thermal capacity of the line. The following equation for the fitness function is proposed.

B. Adjustment function (fitness)

The fitness in the nature of an organism is defined as the probability that the organism survives to reproductive age and reproduces. The fitness function can be think as a numerical measure of profit, utility or goodness of a solution that can be maximize.

C. Operators
The simplest form of genetic algorithms involves three types of operations: - Selection: select people in the population for reproduction. The selection of an individual is related to its adjustment value. The biggest adjustment has an inhabitant the more likely is to be chosen for reproduction. The easiest to create is a roulette wheel where each individual in the population is assigned a portion of the wheel, proportional to its fitness value. The better fitness has an inhabitant, the greater the portion of the wheel to be assigned, ergo, the more likely is to be chosen in the draw. - Crossover: After being selected, individuals are crossed to produce offspring that are inserted into the next generation. This operator randomly chooses a crossover point and

indicates that the substation remains in service and is not disconnected in that combination. The value 0 indicates that the substation is part of the load shedding for that combination. Ultimately, each inhabitant represents the state of all network stations in the post-contingency situation. An iteration start, where in each step the program creates a new generation doing the following: 1.The program checks for fixed stations. If so, an amendment is created in the population by setting the value 1in positions corresponding to stations that must remain connected in every inhabitant. 2.Next, it calculates the fitness of each inhabitant of the population. 3.Next, it calculates the overloaded lines and transformes. Where sob, is the number of overloaded branches in the system. This function accomplishes all the requirements. For further details in the development of the function, see [1]. 4.Next, it calculates the fitness of every one with best fitness for every generation. 5.The selection operations performs the selection procedure. 6.Next, the Cross over and Mutation process takes place simultaneously. The flow char for the genetic algorithm is shown. Fig[1].

VI. MAIN PROGRAM

This section discusses the implementation of major program that calculates the optimal load shedding for the various contingencies that may occur in the network using power flow and genetic algorithms. Initially, the program creates a certain amount of individuals to form the initial population. Each inhabitant represents a possible combination of substations to disconnect. Each individual is represented by a vector, with length equal to the number of substations in the network. Each vector element is associated with a single substation and can take only two possible values, 0 or 1. The value 1

overloaded line or transformer expected if the solution suggested is implemented. Results and Discussion In this work, the IEEE 14-Bus power system is considered for testing the developed algorithm. For the 14-Bus system, the population size, crossover and mutations are 14, 0.8, and 0.01 respectively. The minimum and maximum values of voltage are selected as 13.8KV and 735KV respectively.

Results for 14-Bus system:

If a fault occurs on the bus 9, the load across bus 9 is in out of service. The objective is to restore the loads connected to the bus in the out of service area. There may be a number of possible ways to restore the system by reconfiguring the network. In this case, the load can be reenergized by closing the switches across bus 9 and bus 11 or by opening the switches. This configuration gives the minimum loss, minimum voltage drop and minimum number of switching operations out of all possible configurations to restore the system. By this way the the load restoration is carried out in an efficient manner. The values of voltage, frequency, Active and Reactive power are tabulated[1].
Volta ge (Vrms ) e3 1 2 3 4 5 6 7 8 9 10 13.8 735 735 13.8 13.8 735 735 13.8 735 575 60 60 60 60 60 60 60 60 60 60

Fig1. Flow chart for Genetic Algorithm

This process is repeated until the total number of generations is reached. The inhabitant of best fitness of all generations is taken as a partial solution. For this inhabitant, implements a final adjustment to the partial solution to ensure that the combination selected do not disconnect excessive and unnecessary load. The setting procedure is described below: The program goes through every bit of the inhabitant changing each zero by one and re -run the power flow. In each test, if the new inhabitant has no overloaded elements is stored in an array. The resulting matrix contains all the changes successful. The inhabitant of the matrix with lower disconnected load is chosen and the rest are discarded. With this new individual, the process is repeated. When every additional change is unsuccessful, the adjustment is terminated. The adjusted inhabitant is considered the best solution found. Through the user interface, the program shows the final combination of station to be disconnected, the active power associated and the most

S. No

Freque ncy (Hz)

Active Power (W) e6 275 220 220 110 100 110 660 135 220 440

Reacti ve Power (var) e6 0 660 660 0 0 330 198 0 660 132

11 12 13 14

735 735 735 13.8

60 60 60 60

0 440 330 250

1.2 132 990 0

TABLE1: Voltage, Frequency, Active and Reactive power.

voltage drop for this configuration during restoration. Moreover the number of switching operations is reduced. The fault can be considered for all the buses one by one and the networks for restoration can be determined. In all the cases there is a lot of reduction in the loss, voltage drop and the number of switching operations. The simulation is done using MATLAB R2012a using Simulink blocks .

Conclusion:
A Novel Multiagent based load Restoration Algorithm for micro grids is analysed. The performance index used in this method is most suitable and valid for incorporating new modern distribution systems. The introduction of GA reduces the system time and system complexity to a higher level. Due to the elimination of additional other parameters, the execution time is reduced enormously. The convergence is guaranteed for any micro grids due to the application of genetic algorithm on load flow. The proposed method is capable of solving large scale systems.

When the fault is present in the system the restoration is carried out by considering the faults on the lines one by one. The results are given in Table 3. The network for restoration is determined for minimum loss, voltage drop and minimum switching operations.. It is observed from the results that the loss is reduced and

Kanezashi, A new algorithm for service restoration in distribution systems, Reference: IEEE trans. Power delivery. 1989 vol. 4, no. 3, pp.1832-1839. [1] K. Akoi , H. Kuwabara T. Satoh M. Kanezashi, [4] C. C. liu S. J. Lee, S.S. Venkata An expert system operational aid for Outage state optimal load allocation by automatic restoration and loss reduction of sectionalizing switches operation in distribution systems, IEEE trans. Power delivery.distribution systems, IEEE trans. Power syst.1988 vol. 3, no. 2, pp. 619-626 1987,vol. 2, no.4, pp. [5] C. E. Lin et al., A distribution system 1177-1185 [2]K.Akoi , H. Kuwabara T. Satoh Voltage outage dispatch by data base method with real time revision, IEEE trans. M. Kanezashi, drop constrained restoration of supply by switchPower Delivery, vol. 4. Jan. 1989. operation in distribution [6] Imamura et al., An application of fuzzy reasoning for service restoration (in Japanese) Trans. IEE Jpn., vol. 4, July 1989. [7] H. Mori et al., Parallel simulated system, IEEE trans. Power delivery. [3] K. Akoi, H. Kuwabara T. Satoh M. annealing for power system

decomposition, IEEE trans. Power system vol. 4, May 1994. [8] S. Toune, H. Fudo, T. Genji, Y. Fukuyama, Y. Nakanishi, A reactive tabu search for service restoration in electric power distribution system, IEEE International conference on evolutionary computation, Anchorage Alaska, May 4-11 1998. [9] Y. Fukuyama and H. D. chaing, A parallel genetic algorithm for service restoration in electric power systems, in proceeding IEEE FUZZ/FES Conference, Yokohama, Japan, Mar. 1995. [10] A. Augugliaro, L.Dusonchet, E.R. Sanseverino, Service restoration in compensated distribution networks using a hybrid genetic algorithm, Electric power syst. Research, 1998, vol. 46, 5966. [11] W.P. Luan, M.R.Irving, J.S. Daniel, Genetic algorithm for supply restoration and optimal load shedding in power system distribution networks, IEE Proc. Gener. Transm. Distrib., 2002, vol. 149, no. 2, pp. 145-151.
[12] Y. T. Hsiao and C. Y. Chien, Enhancement of restoration service in distribution systems using a combination of fuzzy-GA method, IEEE

trans. Power system vol. 15, November 2000.