Application of indoor temperature prediction based on SVM and BPNN

Aiming at the problems for predicting the building indoor temperature so as to set up a reasonable indoor environment, the support vector machine (SVM) model and back propagation neural network (BPNN) model of the indoor temperature prediction were established in this paper. The LibSVM toolbox and neural network toolbox were respectively used to predict the indoor temperature in this paper. The sample data was trained in the two models, the output of the two models is the target predicted value. In final, the predicted value and actual value were compared in this paper. The experimental results shown that the prediction error of the SVM model were less than the prediction error of the BPNN model. The experimental results also indicated that the SVM model has the better prediction accuracy, the most importantly, it proved that the application of the SVM predicting method in the building indoor temperature prediction is really effective. The SVM predicting method can be also promoted in the other field of prediction.