Application Of Support Vector Machines In Short-term Load Forecasting Of Power System

Short-term load forecasting provides important foundation for the safety and economical operation of power system. With the fast development of modern electric power systems, the operation of power market requires high precision of short-term load forecasting for the minimal cost of power system operation. Currently there have been more studies in theory and complemented methods of load forecasting and obtained great achievement. New theory and new technology based load forecasting researches have been developed continuously. As new technology of data mining, support vector machines(SVM) have been successfully applied in pattern recognition and regression problem, et al. This paper proposes to use its advantages of non-linear processing and generating ability to accomplish short-term load forecasting of power system, so as to improve forecasting precision and executed speed. Consequently the study is significant in theory and is valuable in practice.This paper analyses the basic theories of SVM. SVM have the remarkable advantages of non-linear regression, high forecasting accuracy and small time complexity. According the non-linear relationship between the forecasting load and its influence factors, this paper proposes a short-term load forecasting model based on SVM. Compared with the forecasting method of artificial neural networks(ANN), the simulation results of the practical application show that the SVM method is much better than ANN.Because of numerous load influence factors having a great of complex characteristics, the pattern, without selecting input vectors, will lead to reduce of the precision and increase of the computering time. Therefore this paper adopts an effective fuzzy clustering analysis and process technology for the load data and combines the clustering algorithm with SVM. A new SVM method based on FCM fuzzy clusting algorithm for short-term load forecasting is first presented in this paper. Compared with the conventional SVM method, this method 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, guarantee the consistency of the data characteristic and enhance the history data regulation. The results of the...