Prediction of public building energy consumption by using artificial fish-swarm algorithm

In order to overcome the shortage of traditional BP neural network method in the prediction of public building energy consumption, this paper proposes a neural network prediction model based on time series self-correlation analysis. Firstly, we determined the dimension of the input variables based on the energy consumption of building standards of self-correlation analysis, then combined with artificial fish-swarm algorithm, which has the advantage of higher optimization speed and easy to jump out of the extreme, to optimize initial weights and threshold value of the BP neural network, and established the energy consumption prediction model, finally, we used the model to predict a month´s energy consumption values of a university building in Xi´an. The results show that compared with the traditional BP neural network model, this model has faster convergence speed and prediction accuracy is in plus or minus one, and the prediction error decreases with the increase of the number of iterations.