Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

ELECTRIC LOAD FORECASTING BASED ON SUPPORT VECTOR MACHINE OPTIMIZED BY EVOLUTIONARY ALGORITHMS

Polyakhov N.D. 1 Prikhodko I.A. 1 Van Efen 1
1 Saint-Petersburg State Electrotechnical “LETI”
Prediction of power consumption is the basis for the reliable operation of the power system. Changing electrical load is a stationary stochastic process depending on many factors. To handle the raw observational data is using support vector machines to reduce the dimension of the training set. To predict power consumption method is chosen support vector machines. Advantage of this method is that the parameters of the regression model are based on quadratic programming problem having a unique solution. Optimizing parameters of kernal functions implemented based on genetic algorithm and particle swarm optimization algorithm. Research on the simulation confirm the efficiency of the proposed approach. Forecast error decreased compared with a forecast based on statistical models in 2 times.