Forecasting building energy consumption based on hybrid PSO-ANN prediction model

As a popular data driven method, Artificial Neural Networks (ANNs) are widely applied in building energy prediction field. In this paper, a hybrid prediction approach that combines Particle Swarm Optimization (PSO) and ANN is presented. Before the prediction model applied, the principal component analysis (PCA) is used for the selection of the input variables, which helps to reduce the input dimension and simplify the model structure. To improve the prediction accuracy, PSO is used to adjust the ANN model's weight and threshold values. The performance of the proposed hybrid model is investigated using the data set of the Energy Prediction Shootout I contest, and the results indicate that PSO-ANN have better performance than regular ANN in term of prediction accuracy. In addition, another kind of hybrid prediction model which combines Genetic Algorithm (GA) and ANN is also proposed. Performance comparison shows that PSO-ANN has the same accuracy level with GA-ANN, and has simpler structure which is more suitable for online prediction tasks.