A hybrid prediction model for heating load of buildings within residential communities considering occupancy rates

• Analyzed the impact of occupancy rates from inter-household heat transfer and building zoning perspectives. • Proposed a new method for quantifying occupancy rates. • Introduced occupancy rates as input parameters in the prediction model. • Enhanced prediction accuracy with OGGWO-GRU integration. • Improved the predictive accuracy for different buildings within residential communities. Accurate load prediction for each building is crucial for the optimized control of heating networks in residential areas. However, current residential heating load feature selection mainly focuses on environmental parameters and building-related factors, neglecting the influence of occupancy rates on different residential buildings, which leads to lower prediction accuracy. This study aims to obtain accurate short-term heating load predictions for different residential buildings within residential communities. The impact of occupancy rates on heating load is analyzed from the perspectives of inter-apartment heat transfer and building zoning, and a novel method for quantifying occupancy rates is proposed. Based on this, an OGGWO-GRU hybrid model is introduced, which uses the Local Opposition-Learning Golden-Sine Grey Wolf Optimization (OGGWO) algorithm to adjust hyperparameters, enhancing the model’s prediction accuracy. Finally, real heating load data is employed to validate the superiority of the proposed hybrid model. The results show that after introducing quantified occupancy rates, the Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE) of the hybrid prediction model decreased by approximately 7.92% and 25.99%, respectively. Compared to other prediction models, the proposed hybrid model achieves minimum reductions of 11.47% in MAPE and 8.42% in RMSE.