Electricity price forecasting by linear regression and SVM

Power system has appeared as a complex interconnected network due to competitive business environment. Power producers and consumers obligate for a precise price forecasting, as this information is an important part of decision making process. Decisions, regarding optimal scheduling of generators, bidding tactics and demand side organizations are based on price forecast. In recent years, development of new approaches for short term price forecasting has attracted the interest of the researchers. Electricity as a commodity consists of some distinct features that, firstly it can't be stock piled & secondly the volatile nature of the electricity price. With these two issues, forecasting of electricity price becomes an intimidating task for system planners and designers. This paper presents a hybrid approach based on linear regression and Support Vector Machine (SVM) to forecast short term electricity price. Linear Regression patterns are developed with the help of different factors of historical data of the electricity price. Two philosophies are developed with the combination of different factors. It is observed that method of similar days is effective. Further forecasted results of the regression models are given to SVM based supervised learning model which have tuned by Particle Swarm Optimization (PSO) technique. It is observed that proposed hybrid approach shows better accuracy as compared with others.