Electric Price Forecast using Interbreed Approach of Linear Regression and SVM

Abstract

Electricity price forecasting is a hypercritical issue due to the involvement of consumers and producers in electricity markets. Price forecasting plays an important role in planning and managing economic operations related with the electrical power (bidding, trading) and other decisions related with load shedding and generation rescheduling. It is also useful for optimization in electrical energy trade. This paper explores an interbreed technique based on Support Vector Machine (SVM) and linear regression to predict the day ahead electricity price using historical data as a raw insert. Different 27 linear regression models are formed to create initial framework for forecasting engine. Comparison of the performance of different forecasting engines is carried out on the basis of error indices namely Mean Square Error (MSE), Sum Square Error (SSE) and other conventional error indices. A detailed explanation of linear regression system based model is presented and simulation results exhibit that the proposed learning method is able to forecast electricity price in an effective manner.