Research on the Application of the Wavelet Neural Network Model in Peak Load Forecasting Considering of the Climate Factors

According to the characteristics of the peak-load, a new approach to power system peak-load forecasting is proposed based on wavelet transform and artificial neural network (ANN). As we all know, the different meteorological type will cause the difference of modes of load, thus it will increase the training time of the neural network and influence the precision of predicting notably. For this reason, based on the analysis of the relationship between the climatic factor and load data, this paper adopts a new method of clustering analysis by self-study membership to choose the historical data as the samples and train them firstly, which has the most same meteorological characteristic as the predicted day, so the training time of the neural network can be shortened; then choose the Morlet wavelet to establish a wavelet neural network secondly, which means to adopt the nonlinear wavelet base to replace the commonly used nonlinear Sigmoid function to predict the historical data array; and to reach the global best approximation effect through an expansion and contraction factor and a translation factor. The WNN combines the time-frequency localization characteristic of wavelet and its self-learning ability, so it helps to overcome the defects of ANN such as the difficulty of rationally determining the network structure and the existence of partial optimal points. The practical example shows that this method possesses high forecasting accuracy and better adaptability than traditional forecasting methods.