Application of Empirical Wavelet Transform in Power Load Forecasting

According to the characteristics of electric load, the influence of date, temperature, weather and other conditions on the power load is analyzed. In order to solve this problem, a combined prediction method based on empirical wavelet transform and SimpleMKL is adopted. First, the EWT method is used to decompose the original power load data, and then the various modal component signals generated after the decomposition are combined with the SimpleMKL method, and the different prediction models are constructed. Then the prediction results are superimposed to get the final prediction results. This paper forecasts the power load data of a certain area, and then compares the prediction results with the results of BP, RBF neural network algorithm, support vector machine, Kernel extreme Learning Machine, SimpleMKL algorithm. The results show that the EWT-SimpleMKL based method has higher prediction accuracy and better generalization performance and reliability.