Heating load prediction for heating systems based on support vector regression and wavelet packet

Based on an analysis of the factors causing changes in the load of heat-supply network, preprocesses the data of the heating load, and decomposes the load sequences into different scales through the wavelet packet transform. Develops respective support vector regression predicting models for these sub-sequences. After reconstructing sequences, obtains the predicting results. Simulation results show that the method is superior in predicting accuracy to the traditional BP neural network and the support vector regression method without wavelet packet decomposition.