The studying of combined power-load forecasting by error evaluation standard based on RBF network and SVM method

• Published in: 2009 Chinese Control and Decision Conference pp. 4016 - 4019
• Main Authors: Zhiping Fan, Zhong Qin, Tiansheng Hong, Yufei Zhuang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-06-2009
• Subjects: Artificial neural networks; Combined weight; Computer errors; Function approximation; Load forecasting; Mathematical model; Neural networks; Power load; Prediction methods; Predictive models; Radial basis function networks; RBF network; Support vector machines; SVM method
• ISBN: 9781424427222; 1424427223
• ISSN: 1948-9439; 1948-9447
• DOI: 10.1109/CCDC.2009.5192606

The load forecasting method usually starts from a single method, we usually improved prediction methods to get the better forecasting accuracy, but this often confined to the application of the method, combination forecasting method can achieve superiority of various methods, the forecast accuracy is higher than single forecasting method. In this paper, we used RBF neural network prediction method and support vector machine forecasting method.RBF neural network prediction method is the more popular method in recent years, it has the better generalization ability to the traditional neural network prediction method, It can effectively avoid local minima value and has a very good learning ability; SVM prediction method is transformed into one-dimensional nonlinear prediction of linear space, it has very precise calculation process and can meet the high forecast precision. Based on the combination of the two methods, not only from the Angle of artificial memory model prediction, and using the tight nonlinear model, ultimately meet the purpose of combined forecasting. The main innovation in this paper is that assess the result of every kind of prediction method by making the standards of error qualified, using the error rate to determine the weight of combination, finally, we can get the satisfactory results through an empirical analysis.