SVR and FCM clustering based short term load forecasting in power systems

This paper presents a new method - combined use of FCM clustering and support vector regression (SVR) for short term load forecasting in power systems. SVR has the characteristics of "small samples, high generalization ability" and satisfies structural risk minimization. Using the above advantages of SVR, the complicated nonlinear relationships between some forecasting influence factors and the forecasting load can be regressed. Meanwhile, this paper chooses training samples by fuzzy clustering according to similarity degree of the input samples in consideration of the periodic characteristic of load change, which means take the same type of the data as the learning samples for forecasting. This method guarantees the consistency of the data characteristic and enhances the history data regulation. The results of the practical applications and compares of the proposed method show that this method has higher forecasting precision and speed than the conventional artificial intelligence methods (e.g. BPNN) and single SVR method, thus can guarantee the precision and real-time requirements by electric short term load forecasting. (6 pages)