Predicting hourly electricity consumption of chillers in subway stations: A comparison of support vector machine and different artificial neural networks

Abstract

It is essential to predicting the electricity consumption of air-conditioning systems. In this study, three models, namely, backpropagation neural network (BPNN), cascade correlation neural network (CCNN), and support vector regression (SVR), were proposed for the prediction of the electricity consumption of chillers in subway stations. For demonstration, historical data from a subway station in Guangzhou were used for training and testing. The simulation results showed that the proposed models could predict the electricity consumption of chillers effectively. In comparison with BPNN, CCNN and SVR showed higher accuracy. For the refrigeration of public spaces (PSs), weak electricity regions (WERs), and comprehensive management regions (CMRs), CCNN exhibited the minimum mean absolute percentage error (MAPE) and root-mean-square error (RMSE) for the test samples, the values of which were 1.56% and 3.15, respectively, followed by those of SVR and BPNN. For the refrigeration of strong electricity regions (SERs), PSs, WERs, and CMRs, SVR presented the smallest error, and its MAPE and RMSE for the test samples were 2.01% and 6.09, respectively, followed by those of CCNN and BPNN. Through the comprehensive consideration of input characteristics, the comparison of different prediction models and the distinction of different operating conditions, the prediction accuracy was improved effectively. Based on the forecasting, the operational parameters of the air-conditioning system can be optimized in advance, which is of great significance for energy saving in subway stations.