Research on support vector machine model and its application on long-term distribution load forecasting

Abstract

In power market, the accuracy of long term load forecasting has significant effect on power distribution system planning. In this paper, a new methodology for long term distribution load forecasting in the framework of support vector machine with input vector selection was presented. The strength of this technique lies in its satisfactory generalization capability since it adopts structure risk minimization principle and can get better study effect with fewer samples. As usual, the long-term load influential factors such as social ones, economic ones and past load data should be considered first; while in this paper, we introduced the concept of past load data time-series into SVM model since the past load demand could affect and imply the future load demand. Based on the past load data series, SVM time-series forecasting model is constructed. Forecasting results for the period 1998-2004 indicate that the proposed model structure appears to offer improved prediction accuracy. (5 pages)