Application of improved artificial neural networks in short-term power load forecasting

Power load forecasting is a key element for power system management and planning. However, it has been proven to be a hard task due to various unstable factors. This paper presents a forecasting methodology based on this particular type of neural network. The scope of this study presents a solution for short-term load forecasting based on a three-stage model which starts with pattern recognition via self-organizing map (SOM), a clustering of the previous partition by K-means algorithm, and finally demand forecasting for each cluster with back propagation neural network (BPNN) improved by additional momentum and variable learning rate methods. The effectiveness of SOM-K-BPNN model has been verified by the final simulation which shows that the proposed model outperforms the BPNN model with default parameters and Grey System GM (1, 1); therefore, empirical results show that the proposed SOM-K-BPNN model is feasible and can fulfil the shortterm load forecasting requirements of China. (C) 2015 AIP Publishing LLC.