A Novel Ultra Short-Term Load Forecasting Algorithm of a Small Microgrid Based on Support Vector Regression

Based on the support vector regression (SVR), a novel ultra short-term load of forecasting algorithm is proposed in this paper. The algorithm is used to realize the online cycle prediction of the load. The historical load data is divided into three categories: working days, weekends and holidays. The similarity method is used to compute the correlation between the training dataset and the prediction load at the current time. The correlation values are sorted in descending order. The first kth load data are used for training to obtain three types of online forecasting models based on SVR. A novel particle swarm optimization (PSO) algorithm is designed to get the optimal parameters of SVR. The experiment on a commercial building of a small mirogrid system shows the designed algorithm works efficiently and stably.