Heating load prediction based on support vector regression machine with parameters optimized by genetic algorithm

In order to further improve the prediction accuracy of the heating load, analyses the influence of parameters on support vector regression machine (SVR), puts forward a model based on the SVR optimized by genetic algorithm for the heating load forecasting. The method takes advantage of cross validation in aspect of model performance evaluation and selection, combined with the ability of global optimization of genetic algorithm, realizes automatic selection of optimal parameters, and obtains the best model to forecast the heating load. In an experimental study on a heat source data, compares with other algorithms, the results show that the average absolute value of the relative error of the method is 4.33%, 10.77% lower than that of traditional SVR machine and 5.28% lower than that of wavelet neural network.