5_amit
TRANSCRIPT
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 1/7
1
Abstract -- Load forecasting has become a very crucial
technique for the efficient functioning of the power system. Thispaper presents a methodology for the short term load forecastingproblem using the similar day concept combined with fuzzy logicapproach and particle swarm optimization. To obtain the next-dayload forecast, fuzzy logic is used to modify the load curves of theselected similar days of the forecast previous day by generating thecorrection factors for them. These correction factors are then appliedto the similar days of the forecast day. The optimization of the fuzzyparameters is done using the particle swarm optimization techniqueon the training data set of the considered data set. A new Euclideannorm with weight factors is proposed for the selection of similar days.The proposed methodology is illustrated through the simulationresults on a typical data set.
Index Terms -- Euclidean norm, Fuzzy logic approach, Particleswarm optimization, Short term load forecasting, similar day
method .
I. INTRODUCTION
OAD forecasting has been an integral part in the efficientplanning, operation and maintenance of a non interruptingregulated power system. Short term load forecasting is
necessary for the control and scheduling operations of a powersystem and also acts as inputs to the power analysis functionssuch as load flow and contingency analysis [1]. Power utilitiesin India still depend on the averaging of set of similar days forshort term load forecasting, the results of which are very farfrom being accurate. Owing to this importance, variousmethods have been reported, that includes linear regression,exponential smoothing, stochastic process, ARMA models, anddata mining models [2]-[7]. Of late, artificial neural networkshave been widely employed for load forecasting. However,there exist large forecast errors using ANN when there arerapid fluctuations in load and temperatures [8]-[9]. In suchcases, forecasting methods using fuzzy logic approach havebeen employed. In this paper, we propose an approach forshort term load forecasting problem, using a hybrid techniquecombining the particle swarm optimization (PSO) and fuzzy
Dr Amit Jain, Research Assistant Professor, Head, Power SystemsResearch Center at International Institute of Information Technology,Hyderabad, Andhra Pradesh, India (e-mail: [email protected]).
M. Babita Jain is PhD student at Power Systems Research Center,International Institute of Information Technology, Hyderabad, AndhraPradesh, India (e-mail: [email protected]).
E. Srinivas MS by Research student, Power Systems Research Center atInternational Institute of Information Technology, Hyderabad, AndhraPradesh, India (e-mail: [email protected]).
logic approach. The PSO technique which roots from theconcept of Computational Intelligence capable of dealing withcomplex, dynamic and poorly defined problems that AI hasproblem with, has an advantage of dealing with the nonlinearparts of the forecasted load curves, and also has the ability to
deal with the abrupt change in the weather variables such astemperature, humidity and also including the impact of the daytype. In this method, we select similar days from the previousdays to the forecast day using Euclidean norm (based onweights, which are calculated such that they hold the essenceof the earlier data) with weather variables [10]. There may be asubstantial discrepancy between the load on the forecast dayand that on similar days, even though the selected days arevery similar to the forecast day with regard to weather and daytype. Therefore, the selected similar days cannot be averagedto obtain the load forecast. To avoid this problem, theevaluation of similarity between the load on the forecast dayand that on similar days is done using fuzzy logic combined
with particle swarm optimization. Many methods have beenreported for the load forecast using fuzzy logic [12]-[14]. Thisapproach evaluates the similarity using the information aboutthe previous forecast day and previous similar days. Thesimilarity resulted with a high MAPE (mean averagepercentage error) value. The fuzzy inference system is used forthe reduction of this MAPE and generation of suitablecorrection factors. The fuzzy system thus generates acorrection factor for the load curve on a similar day to theshape of that on the forecast day. The fuzzy parameters of thefuzzy inference system are optimized using the PSO techniqueimplemented in the MATLAB program of the short term loadforecasting. The PSO technique is thus used to evolve theinitial best parameters for the fuzzy inference system. Theoptimization of the fuzzy parameters using PSO is done usingthe training data set of the available data set and the fuzzyinference system with the optimized parameters is then used onthe testing data set. For the testing data set, first the correctionfactors of load curves on similar days are calculated using theoptimized fuzzy inference system for the similar days of theforecast previous day, then the forecast load is obtained byaveraging the corrected load curves on similar days using theprevious day’s correction factors. The correction factors areobtained using the training data set and then they are applied tothe testing data set. The approach suitability is verified by
applying it to a typical real data set. The overall MAPE wasfound to be very less (<0.03) for the considered data set. Thispaper contributes to the short term load forecasting, as it showshow the forecast can include the effect of weather variables,temperature as well as humidity and day type using a newEuclidean norm for the selection of the similar days and PSO
A Novel Hybrid Method for Short Term LoadForecasting using Fuzzy Logic and Particle
Swarm OptimizationAmit Jain, Member, IEEE , M. Babita Jain, Member, IEEE and E. Srinivas
L
2010 International Conference on Power System Technology
978-1-4244-5939-1/10/$26.00©2010 IEEE
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 2/7
2
optimized fuzzy logic approach for obtaining the correctionfactors to the similar days. The paper is organized as follows:section II deals with the data analysis; section III gives theoverview of the proposed forecasting method, discussing theselection of similar days using a new Euclidean norm and thePSO optimization of the fuzzy inference system is presented;
section IV presents the optimization of fuzzy parameters usingparticle swarm optimization; section V simulation results of theproposed forecasting methodology followed by conclusions insection VI.
II. VARIABLES IMPACTING THE LOAD PATTERN
The analysis on the monthly load and weather data helps inunderstanding the variables, which may affect loadforecasting. The data analysis is carried out on data containinghourly values of load, temperature, and humidity of 7 months.In the analysis phase, the load curves are drawn and the
relationship between the load and weather variables isestablished [11]. Also, the week and the day of the week impact on the load is obtained.
A. Load Curves
The load curve for the month of May is shown in Fig. 1.The observations from the load curves are as follows:1. There exists weekly seasonality but the value of loadscales up and down.2. The load curves on week days are similar3. The load curves on the weekends are similar.4. Days are classified based on the o following categories:
a. Normal week days (Tuesday - Friday)b. Mondayc. Sundayd. Saturday
Monday is accounted to be different to weekdays so as to takecare for the difference in the load because of the previous dayto be weekend.
B. Variation of Load with Temperature
Temperature is the most important weather variable thataffects the load. The deviation of the temperature variablefrom a normal value results in a significant variation in theload. Fig. 2 and Fig. 3 show the relationship between the
temperature and load. Fig. 2 shows a plot between the averagetemperatures versus maximum demand. Fig. 3 shows the plotbetween the average temperatures versus average demand. Inthe plots the dot represents the actual values and the solid lineis the best fitted curve. The graphs show a positive correlation
between the load and temperature for the forecast month of July i.e. demand increases as the temperature increases.
C. Variation of Load with Humidity
Another weather variable that affects the load level ishumidity. To study the effect of this particular weathervariable on load we plot the maximum demand versus averagehumidity and the average demand versus average humiditygraphs as shown. Fig 4 shows the plot between the average
humidity versus maximum demand. Fig 5 shows the plotbetween the average humidity versus average demand. Fromthe graphs it can be seen that there exists a positive correlationbetween load and humidity for the forecast month of July i.e.demand increases as the humidity increases.
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 3/7
3
D. Autocorrelation of Load
It is known that the load at a given hour is dependent notonly on the load on the previous hour but also on the load atthe same hour of the previous day. Hence, it is assumed thatthe load curve is more or less similar to the load curve on theprevious day.
III. LOAD FORECASTING USING FUZZY LOGIC
A. Similar Day selection
In this paper, Euclidean norm with weight factors is used toevaluate the similarity between the forecast day and thesearched previous days. Euclidean norm makes us understandthe similarity by using the expression based on the concept of norm. Decrease in the Euclidean norm results in the betterevaluation of the similar days i.e., smaller the Euclidean normthe more similar are the days to the forecast day. In general,the Euclidean norm using maximum and minimum
temperatures along with the day type variable is used for theevaluation of the similar days. But, the norm using maximumand minimum temperatures is not efficient for the selection of the similar days because humidity is also an important weathervariable as also shown in section II C.
In the present work, we have proposed a new Euclideannorm to account for the humidity. The new Euclidean normuses average temperature, average humidity and day type withweight factors to evaluate the similarity of the searchedprevious days. The expression for the new Euclidean norm isas follows:
2 2 21 max 2 3( ) ( ) ( )avgEN w T w H w D (1)
max max max pT T T (2)
pavg avg avg H H H (3)
p D D D (4)
Where, Tavg and H avg are the forecast day averagetemperature and average humidity respectively. Also, Tpavg
and H pavgare the average temperature and average humidity of
the searched previous days and w1, w 2, w3 are the weightfactors determined by least squares method based on the
regression model constructed using historical data [2]. Thesimilar days are selected from the previous 30 days of theforecast day. The data selection is limited to account for theseasonality of the data. The day types considered for themethodology are 4(Tuesday-Friday), 3 (Monday), 2(Saturday),1(Sunday);
B. Fuzzy Inference System
The load forecasting at any given hour not only depends onthe load at the previous hour but also on the load at the givenhour on the previous day. Also, the Euclidean norm alone isnot sufficient for the load forecast as the selected similar daysfor the forecast day have considerably large mean absolutepercentage error (MAPE). Assuming same trends of relationships between the previous forecast day and previoussimilar days as that of the forecast day and its similar days, thesimilar days can thus be evaluated by analyzing the previousforecast day and its previous similar days.
The fuzzy inference system is used to evaluate thesimilarity between the previous forecast days and previoussimilar days resulting in correction factors, used to correct thesimilar days of the forecast day to obtain the load forecast. Toevaluate this degree of similarity, three fuzzy input variablesfor the fuzzy inference system are defined [10].
k k L p ps E L L (5)k k T p ps E T T (6)k k
H p ps E H H (7)
Where, L p and L ps are the average load of the previousforecast day and the previous kth similar day, T p, T ps, Hp, H ps
show the value corresponding to temperature and humidityrespectively. E L, ET, E H take three fuzzy set values; Low (L),Medium (M), High (H).The membership functions of theinput variables and output variable are as shown in Figs 6 -7 .The fuzzy rules for the inference system for the given fuzzyvariables are based on the generalized knowledge of the effectof each variable on the load curve [11]. If the membership of EL is UEL, that of E T is UET and that of E H is UEH, the firingstrength, u, of the premise is calculated based on the minoperator. The firing strength of each rule is calculated asfollows:
min( , , ) L T H i E i E i E i (8)
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 4/7
4
The membership function of an inferred fuzzy outputvariable is calculated using a fuzzy centroid defuzzificationscheme to translate fuzzy output statements into a crisp outputvalue, Wk.
27 27
1 1 /
k k k i i i
i iW
(9)
The output value is expressed by W k which is the correctionfactor for the load curve on the k* similar day to the shape onthe forecast day. W k is applied to each similar day andcorrects the load curve on similar days. The forecast next dayload curve L (t) is then given by averaging the corrected loadson similar days.
k 1
1( ) (1 ) ( )
N k s
k
L t W L t N
(10)
Where Lk s (t), is the power load at t 'o clock on thekth corrected similar day, N is the number of similar days andt is hourly time from 1 to 24.
IV. OPTIMIZATION OF FUZZY PARAMETERS USING PARTICLE SWARM OPTIMIZATION
PSO is a population-based optimization method firstproposed by Eberhart and Colleagues [15, 16]. Some of theattractive features of PSO include the ease of implementationand the fact that no gradient information is required. It can beused to solve a wide array of different optimization problems.Like evolutionary algorithms, PSO technique conducts searchusing a population of particles, corresponding toindividuals. Each particle represents a candidate solution tothe problem at hand. In a PSO system, particles changetheir positions by flying around in a multidimensional searchspace until computational limitations are exceeded.
The PSO technique is an evolutionary computationtechnique, but it differs from other well-known evolutionarycomputation algorithms such as the genetic algorithms.Although a population is used for searching the search space,there are no operators inspired by the human DNA proceduresapplied on the population. Instead, in PSO, the populationdynamics simulates a 'bird flock's' behavior, where socialsharing of information takes place and individuals can profitfrom the discoveries and previous experience of all the othercompanions during the search for food. Thus, eachcompanion, called particle, in the population, which is calledswarm, is assumed to 'fly' over the search space in orderto find promising regions of the landscape. For example, inthe minimization case, such regions possess lower functionvalues than other, visited previously. In this context, eachparticle is treated as a point in a d-dimensional space, whichadjusts its own 'flying' according to its flying experience aswell as the flying experience of other particles (companions).In PSO, a particle is defined as a moving point in hyperspace.
For each particle, at the current time step, a record is kept of the position, velocity, and the best position found in the searchspace so far.
The assumption is a basic concept of PSO [16]. In the PSOalgorithm, instead of using evolutionary operators such asmutation and crossover, to manipulate algorithms, for a
d-variables optimization problem, a flock of particles areput into the d-dimensional search space with randomly chosenvelocities and positions knowing their best values so far(Pbest) and the position.
In the d-dimensional space, the velocity of each particle,adjusted according to its own flying experience and theother par tic le's flyi ng experience. For example, the ithparticle is represented as xi = (x i1 , xi2. . . b id ) in thedimensional space. The best previous position of the i-thparticle is recorded and represented as:Pbesti = (Pbesti 1, Pbest i ,2 ,..., Pbest id )
The index of best particle among all of the particles in the
group is gbest d .The velocity for particle i is represented as vi= (v i 1, v .2,...,v. d ).The modified velocity and position of
each particle can be calculated using the current velocity andthe distance from Pbest id to gbest d as shown in the following
formulas [17]:
v[] = v[] + c1 * rand() * (pbest[] - present[]) + c2 *
rand() * (gbest[]-present[]) (11)
present[]=persent[]+v[] (12)
Where
v[] is the particle velocity, persent[] is thecurrent particle (solution). pbest [] and gbest[] are defined asstated before. rand () is a random number between (0,1). c1, c2are learning factors. Usually c1 = c2 = 2.
The evolution procedure of PSO Algorithms is shown in Fig.8. Producing initial populations is the first step of PSO. Thepopulation is composed of the chromosomes that are realcodes. The corresponding evaluation of populations calledthe "fitness function". It is the performance index of apopulation. The fitness value is bigger, and the performance isbetter.
After the fitness function has been calculated, the fitnessvalue and the number of the generation determine whether ornot the evolution procedure is stopped (Maximum iterationnumber reached?). In the following, calculate the Pbest of each particle and Gbest of population (the best movement of all particles). The update the velocity, position, gbest and pbestof particles give a new best position (best chromosome in ourproposition).
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 5/7
5
Fig. 8 Evaluation procedure of PSO Algorithm
For the data set considered the fuzzy inference system hasbeen optimized for six parameters (maxima and minima of each of the input fuzzy variable EL, ET, EH), considereing 49particles. Hence each particle is a six dimensional one.Theinitial values of the fuzzy inference system are obtained byusing the May and June data. These values are incorporatedinto the fuzzy inference system to obtain the forecast errors of June month.
The particle swarm optimization function accepts thetraining data i.e. May and June 99, and the objective is toreduce the RMS MAPE error of the forecast days(June 99)using May data. The RMS MAPE is taken as the fitnessfunction and the particle swarm optimizer function is run for100 iterations(by then the RMS MAPE) is more or less fixed.After every iteration the optimizer updates the latest particleposition using the optimizer equations based on the PBest andGbest of the previous iteration if the fitness function value isbetter than the previous one. The parameters thus obtained
after the PSO optimization are the final fuzzy input parametersand these are used to forecast the load of the testing data seti.e. July month.
V. SIMULATION RESULTS
The performance of the method for the short term loadforecast is tested by using the 7 months data, from January toJuly of a particular data set used. The method has beensimulated using the fuzzy logic toolbox available inMATLAB. Load forecasting is done for the month of July.
The fuzzy inference system is designed for the computersimulation. First, fuzzy parameters are s based on the existingtraining data set, as described in the above section. Hence,the data of the month June has been used for the selectionof similar days. The number of similar days used for theforecasting is five.
The initial parameters of the fuzzy membership functionsare determined through the simulation of the load curveforecasting in the previous month to the forecast day. Theparameters of the membership functions for the input andoutput variables for the next-day load curve forecasting for the
month of June are as follows:TABLE1
PARAMETERS OF MEMBERSHIP FUNCTIONS OF INPUTVARIABLES
TABLE2PARAMETERS OF MEMBERSHIP FUNCTIONS OF OUTPUT
VARIABLES (b1,b2,b3) (-0.3,-0.25,-0.2)(b3,b4,b5) (-0.2,-0.15,-0.10)(b4,b5,b6) (-0.15,-0.10,-0.05)
(b5,b6,b7) (-0.10,-0.05,0)(b6,b7,b8) (-0.05,0,0.05)(b7,b8,b9) (0,0.05,0.1)(b8,b9,b10) (0.05,0.1,0.15)(b9,b10,b11) (0.1,0.15,0.2)(b10,b11,b12) (0.15,0.2,0.25)(b11,b12,b13) (0.2,0.25,0.3)
Second, the fuzzy parameters are optimized using PSOalgorithm only to find the optimal range of the membershipfunctions (FIS with PSO algorithms). Last, the optimizedfuzzy inference system on the PSO algorithm is used to
forecast the load of the month of July which is the testingdata set. After the optimization process, the optimal valuesof a1, a2, a3, a4, a5 and a6 in FIS with PSO algorithm arecalculated and are given in the Table 3.
TABLE3PSO OPTIMIZED FUZZY PARAMETERS
(a1,a2) (a3,a4) (a5,a6)(-17274,15766) (-57.69, 35.55) (-7.25, 24.27)
TABLE4PARAMETERS OF PSO ALGORITHM
Population Size 49Number of Iteration 100
w max 0.6
w . 0. 1c 1 = c 2 1.5
The forecasted results of 4 representative days of the Julymonth in a week are presented. These days represents fourcategories of classified days of week in the presentmethodology namely Saturday, Sunday, Monday, and Tuesday.
The humidity variations for the days to be forecasted i.e.23 rd - 27 th July are as shown in Fig. 9.
(a1,a2) (a3,a4) (a5,a6)(-1000,1000) (-20,20) (-20,20)
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 6/7
6
Fig. 9 Humidity curve between 23 rd July-27 th July
0 5 10 15 20 251.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2x 10
4 Load curves of July 24
Hours
L o a d ( M W )
ActualForecasted
Fig. 10 Load curves of July 24
0 5 10 15 20 251.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8x 10
4 Load curves of July 25
ActualForecasted
Fig. 11 Load curves of July 25
0 5 10 15 20 251.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2x 10
4 Load curves of July 26
ActualForecasted
Fig. 12 Load curves of July 26
0 5 10 15 20 251.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1x 10
4 Load curves of July 27
ActualForecasted
Fig. 13 Load curves of July 27
The forecast results deviation from the actual values arerepresented in the form of MAPE. Mean Absolute PercentageError (MAPE) is defined as:
1
1100
i i N A F
i
i A
P P MAPE
N P
(13)
PA, P F are the actual and forecast values of the load. N is thenumber of the hours of the day i.e. N = 1, 2, . .24 With theproposed method the MAPE error for the considered days, forwhich forecasted results are shown in Fig. 9-12, is calculatedand these are as follows in Table 5:
TABLE5MAPE OF FORECASTED RESULTS
Day MAPE Error 24 July (Saturday) 2.357425 July (Sunday) 1.985526 July (Monday) 1.740427 July (Tuesday) 1.391
In the present work, a novel PSO optimized fuzzy logic system forshort term load forecasting is developed to forecast the load for nextday. In the fuzzy inference system, the optimization of membershipfunctions was required, as it reduces the MAPE and so the inputparameters limits are adjusted in optimal values so as to give aMAPE error of low values. The computer MATLABsimulation demonstrates that the fuzzy inference system associatedto the PSO algorithm approach became very strong, giving goodresults and possessing good robustness and the results will improvein quality with the use of historical data of more number of years.
VI. CONCLUSION
In this paper an optimal fuzzy logic is designed using PSOalgorithm. The fuzzy inference system with PSO algorithmis better than the traditional fuzzy inference systemwithout PSO algorithms as we observed during oursimulation study and proposed system provides betterimprovement and good robustness. The system takes intoaccount the effect of humidity as well as temperature on
load and a fuzzy logic based method is used to correct thesimilar day load curves of the forecast day to obtain the loadforecast. Also, a new Euclidean norm with weight factors isproposed, which is used for the selection of similar days.Fuzzy logic is used to evaluate the correction factor of theselected similar days to the forecast day using the informationof the previous forecast day and the previous similar days.
To verify the forecasting ability of the proposedmethodology, we did simulation to do load forecasting for themonth of July in a data set of 7 months and results forfour representative days of a week in the month of July are
given. The results obtained from the simulation show thatthe proposed forecasting methodology, which proposes the useof weather variables i.e. temperature as well as humidity, givesload forecasting results with reasonable accuracy. Theproposed methodology would certainly give improved resultswhen it is implemented on historical data set of many years,
8/6/2019 5_amit
http://slidepdf.com/reader/full/5amit 7/7
7
which we will present in our next research paper. Authorshope that the proposed methodology will be helpful in usingmore weather variables, which will certainly be better thanusing only temperature as the weather variable that affects theload, in short term load forecasting and will further propagateresearch for short term load forecasting using new optimization
techniques like PSO.
VII. REFERENCES
[I] G. Gross and F. D. Galiana, "Short-Term LoadForecasting,"Proceedings of IEEE, vol. 75, no. 12, pp.1558 - 1573, Dec. 1987.
[2] A. D. Papalexopoulos and T. C. Hesterberg,"A regression-based approach to short-term load forecasting,"IEEE Trans. Power Syst., vol. 5, no. 4, pp. 1535-1550, 1990.[3]T. Haida and S. Muto, "Regression based peak load
forecasting using a transformation technique," IEEE Trans.
Power Syst., vol. 9, no. 4, pp. 1788-1794, 1994.[4] S. Rahman and O. Hazim, "A generalized knowledge-based short term load-forecasting technique," IEEE Trans.Power Syst., vol. 8, no. 2, pp. 508-514, 1993.[5]S. J. Huang and K. R. Shih, "Short-term load forecastingvia ARMA model identification including nongaussianprocess considerations," IEEE Trans. Power Syst., vol. 18, no.2, pp. 673-679, 2003.[6]H.Wu and C. Lu, "A data mining approach for spatialmodeling in small area load forecast," IEEE Trans. PowerSyst., vol. 17, no. 2, pp. 516-521, 2003.[7]S. Rahman and G. Shrestha, "A priority vector based
technique for load forecasting," IEEE Trans. Power Syst.,vol. 6, no. 4, pp. 1459-1464,1993.[8]H. S. Hippert, C. E. Pedreira, and R. C. Souza, "Neuralnetworks for short-term load forecasting: A review andevaluation," IEEE Trans.Power Syst., vol. 16, no. 1, pp. 44-55, 2001.[9]C. N. Lu and S. Vemuri, "Neural network based
short term load forecasting," IEEE Trans. Power Syst., vol. 8,no. 1, pp. 336-342, 1993.[10] T. Senjyu, Uezato. T, Higa. P, "Future Load CurveShaping based on similarity using Fuzzy LogicApproach," IEE Proceedings of Generation,Transmission, Distribution, vol. 145, no. 4, pp. 375-380,
1998.[II] S. Rahman and R. Bhatnagar, "An expert system basedalgorithm forshort term load forecast," IEEE Trans. Power Syst., vol.3, no. 2, pp.392-399, 1988.
[12] T. Senjyu, Mandal. P, Uezato. K, Funabashi. T, "NextDay Load Curve Forecasting using Hybrid CorrectionMethod," IEEE Trans. Power Syst., vol. 20, no. 1, pp. 102-109, 2005.
[13] P. A. Mastorocostas, J. B. Theocharis, and A. G.Bakirtzis, "Fuzzy modeling for short term loadforecasting using the orthogonal least squares method,"IEEE Trans. Power Syst., vol. 14, no. 1, pp. 29-36, 1999.
[14] M. Chow and H. Tram, "Application of fuzzy logictechnology for spatial load forecasting," IEEE Trans.Power Syst., vol. 12, no. 3, pp. 1360-1366, 1997.
[15] J. Kennedy, R. Eberhart: Particle Swarm Optimization,Proc. IEEE Int. Conf. on Neural Network, Vol. 4, 1995,pp. 1942 - 1948.
[16] H. Yoshida, K. Kawata, Y. Fukuyama, S. Takayama, Y.Nakanishi: A Particle Swarm Optimization for ReactivePower and Voltage Control Considering Voltage
Security Assessment, IEEE Trans. on Power Systems,Vol. 15, No. 4, Nov. 2000, pp. 1232 - 1239.[17] Chang C.S., Fu W. "Area load-frequency control using
fuzzy gain scheduling of PI controllers", Electric Power system Research, Vol. 42, 1997, pp. 145-52.
VIII. BIOGRAPHIES
Amit Jain graduated from KNIT,India in Electrical Engineering. Hecompleted his master’s and Ph.D.from Indian Institute of Technology,New Delhi, India.He was working in Alstom on thepower SCADA systems. He wasworking in Korea in 2002 as a Post-
doctoral researcher in the Brain Korea 21 project team of Chungbuk National University. He was Post DoctoralFellow of the Japan Society for the Promotion of Science(JSPS) at Tohoku University, Sendai, Japan. He also workedas a Post Doctoral Research Associate at Tohoku University,Sendai, Japan. Currently he is Head of Power SystemsResearch Center at IIIT, Hyderabad, India. His fields of research interest are power system real time monitoring and
control, artificial intelligence applications, power systemeconomics and electricity markets, renewable energy,reliabili ty analysis, GIS applications, parallel processing andnanotechnology.
M Babita Jain is PhD student at theInternational Institute of Information Technology, Hyderabad,India since 2006. The areas of interest include IT Applications
to Power Systems, Load forecasting and Multi Agent Systems.She has actively involved in establishing IEEE student branchand Women in engineering affinity group activities.
E. Srinivas is a Master's Student in PowerSystem esearch Center,International Institute of Information
Technology, Hyderabad, India. Hereceived his B. Tech degree from SriIndu College of Engineering andTechnology, Ibrahimpatan. His areas of interest includeApplications of computational intelligence techniques to power systemsand o pera tions plann ing o f power systems.