A hybrid model of CNN-BiLSTM and XGBoost for HVAC systems energy consumption prediction

Predicting energy consumption in Heating, Ventilating and Air Conditioning(HVAC) systems is of great significance for optimizing energy consumption, as well as for energy saving and emission reduction in modern buildings. However, using a single energy consumption prediction method has inherent limitations. In this paper, we propose a new hybrid model of CNN-BiLSTM and XGBoost for predicting energy consumption in HVAC systems. The new model extracts time series features of the data through a CNN network, and simultaneously mines forward and backward information using a BiLSTM network. Then, the information extracted by the two networks is integrated to obtain a pre-training model. Finally, the model is optimized by XGBoost to obtain the final prediction result. Comparing the proposed new model with the prediction results of CNN, LSTM, BiLSTM, CNN-LSTM and CNN-BiLSTM models, it can be concluded that the proposed model has higher accuracy and lower error value. Therefore, the hybrid model shows promising potential for predicting energy consumption in HVAC systems.