Application Of Hybrid Forecasting Methods In Prediction Of Building Energy

Establishing forecasting model of building energy was the basis of building energy conservation work, and also was the prerequisite of effective management to buildings. The building energy was a non-linear, dynamic and indeterminacy system. The traditional way is difficult to predict building energy. So hybrid methods were proposed to forecast the building energy consumption in this paper. Specific research work was-divided into three chapters:The prediction accuracy of traditional prediction models were so low that they could not eliminate data’s redundancy in affecting factors of building energy. In order to improve the prediction accuracy of building energy consumption, a prediction combined PCA with RBF neural network was proposed. Principal components obtained by PCA were used as inputs of RBF neural network; the results showed that the improved PCA-RBF model can effectively improve the accuracy of building energy consumption prediction.Aimed at the problems existed in the building energy systems such as noise in field data and multiple correlations of variables, an improved prediction algorithm based on weighted LSSVM and Kernel Principle Component Analysis (KPCA) is proposed. Firstly, KPCA is used to eliminate the noisy of data, lower the multiple correlations and reduce the dimensions of the input samples. Then each training sample’s penalty with different weighted coefficient was assigned according to the importance of each training sample. At the end, particle swarm algorithm (PSO) was used to optimize the punishment factor and kernel parameter, then a model with good generalization was established. Results showed that the KPCA-WLSSVM model had a better prediction.In order to improve the accuracy of college building energy consumption, a mehthod made of gray theory and RBF neural network was proposed. This method combined the advantages of gray model and RBF neural network, and is used to predict the energy consumption, the results showed that the combination approach can be more effective use of useful information to improve model prediction accuracy.Studies showed that the combination forecasting method proposed in this paper had better prediction results than single method while used to predict the energy consumption, and is important to improve the efficiency of integrated design and the level of management.