Short-term Load Forecasting of Microgrid Based on Wavelet Transform and Support Vector Machines

To meet the higher requirement of the load forecasting accuracy and method adaptability introduced by the construction and development of microgrid, the load curves of a typical microgrid are compared and analyzed in time and frequency domains, and stronger conclusion of load fluctuation of microgrid is obtained. Then, according to the load characteristics of microgrid, the influence of week types and meteorological factors on microgrid load is considered. Furthermore, the Discrete Wavelet Transform (DWT) is used to analyze microgrid load, and layer component is forecasted by support vector machines (SVM) model. In the end, the forecasting results are drawn by the application of decomposition formula. Research results show that the load forecast by SVM model after decomposing microgrid load curves has higher prediction accuracy by comparing with that by single SVM algorithm, which is more suitable for short-term load forecasting of microgrid.