Office building air conditioning load forecasting method based on wavelet decomposition and support vector machine

Proposes an air conditioning load forecasting modeling method based on wavelet decomposing and support vector machine (SVM). Decomposes strongly stochastic and nonlinear air conditioning load series into different frequency sub-series. Each of sub-series is modeled separately using support vector regression. This method largely avoids the fluctuation of SVM forecasting accuracy caused by the incompleteness of training data. Simulation results show comparing with that of the expected error percentage is reduced by 33.6% comparing with that of the simple SVM method, and the accuracy is improved significantly.