Heating Load Prediction for Heating Systems Based on Support Vector Regression with Cross Validation

Support Vector Regression( SVR) has achieved some results in the research of heating load forecasting. However,the fitting degree accuracy and generalization performance of the SVR models depend on the selection of its parameters,optimization process is difficult to find,with adequate prior information. In view of this situation,the thought of Cross Validation( CV) was put forward to carry on the grid division for several important parameters( the penalty factor C and RBF kernel function parameter γ),which could search the best parameter in the training set automatically,so as to obtain the optimal model of regression prediction for the test set. In the experimental study of a heat source data,the results illustrate that this method can quickly establish prediction model,and effectively predict the heating load,with a high fitting degree and strong generalization ability.