Mid-Term Load Forecasting: Level Suitably of Wavelet and Neural Network based on Factor Selection

This paper proposes the level suitably of a wavelet transform and a neural network method that are very significant technique for a load forecasting. An accurate forecast of the load demand is an essential activity for fuel reserve planning in a power system. Level adjustment of wavelet is demonstrated in this research based on mid-term load demand forecasting in Thailand. The factor correlating with approximate and detail of each level of wavelet is chosen by using the correlated value between factors and components. All of Features input selected are used to the neural network for training and forecasting. The forecasted results show that “two-level” has good prediction properties compared to other level of wavelet transform clearly.