Kohonen neural network and wavelet transform based approach to short-term load forecasting

This paper presents Kohonen neural network and wavelet transform (WT) based technique for short-time load forecasting of power systems. Firstly, the historical seasonal load data are classified into four patterns using Kohonen neural network and then Daubechies D2, D4 and D10 WTs are adopted in order to forecast the hourly load. The wavelet coefficients associated with certain frequency and time localisation are adjusted using the conventional multiple regression (MR) method and the components are reconstructed to predict the final loads through the five-scale synthesis technique. The outcome of the study clearly indicates that the proposed composite model of Kohonen neural network and WT approach can be used as an attractive and effective means for short-term load forecasting. (C) 2002 Elsevier Science B.V. All rights reserved.