Fuzzy neural very-short-term load forecasting based on chaotic dynamics reconstruction

This paper presents an improved fuzzy neural system (FNS) for electric very-short-term load forecasting problem based on chaotic dynamics reconstruction technique. The Grassberger–Procaccia algorithm and least squares regression method are applied to obtain the value of correlation dimension for estimation of the model order. Based on this order, an appropriately structured FNS model is designed for the prediction of electric load. In order to reduce the practical influences of the computation error on correlation dimension estimation, a dimension switching detector is devised to enhance the prediction performance of the FNS. Satisfactory experimental results are obtained for 15 min ahead forecasting by using actual load data of Shandong Heze Electric Utility, China. To have a comparison with the proposed approach, similar experiments using conventional artificial neural network (ANN) are also performed.