Annual electricity consumption forecasting with least squares support vector machines

In the electricity market, electricity consumption reflects the electric power usage of the whole society. As an important part of load forecasting, annual electricity consumption forecasting plays an important role for all the market participants, especially for the decision-makers establishing bidding strategies of generation companies and the planners of market investors. In this paper, factor selection on the inputs of electricity consumption forecasting is first performed to find proper features from the data in terms of their statistical information. After that, the regression LS-SVM is generalized for the forecasting modeling in this work that is built using the data after factor analysis. In this way, electricity consumption forecasting is accomplished, which not only overcomes the over-fitting of traditional forecasting methods, but also quickens the computation velocity of standard SVM by converting the quadratic optimization problem into the computation of the linear equations. In the end, the case studies are carried out to test the proposed model, which indicates that the built model is an accurate and quick method to predict the electricity consumption.