Energy Consumption Prediction for HVAC Systems in Public Buildings Based on SSA-CNN-LSTM Neural Network

Abstract Accurate predicting dynamic energy consumption of heating, ventilation, and air conditioning (HVAC) systems for public buildings in severe cold regions is critical for coordinated dispatch of heat pump units and optimization of district boiler plants. To address the nonlinear, non-stationary, and multi-scale characteristics of HVAC electricity load in public buildings, a hybrid neural network model is constructed based on the Sparrow Search Algorithm (SSA)-Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) for HVAC energy consumption forecasting. This model combines the spatial feature extraction capability of CNN and the temporal modeling capability of LSTM, while employing SSA to optimize hyperparameters to improve prediction accuracy and robustness. The prediction results indicate that the SSA-CNN-LSTM model can effectively capture the spatiotemporal characteristics of HVAC energy consumption, with MAE, RMSE, and MAPE values of 30.44, 38.05, and 0.065, respectively. This approach provides a reference for intelligent management and control of HVAC system.