Historical load curve correction for short-term load forecasting

Abstract

Short-term load forecasting (STLF) is a significant task for power system operation. The existence of bad data in historical load curve affects the precision of load forecasting result. This paper presents the second order difference method to detect the bad data, eliminate them and evaluate the real data. To decrease the effect of impulse load on the prediction result, weighted least square quadratic fitting is proposed to filter the curve. K-means clustering and support vector machine method are employed to forecast the future load. The proposed method is successfully applied to an actual power system