Office buildings are significant contributors to energy consumption and carbon emissions due to high occupancy density and prolonged operation. To balance decarbonization with indoor environmental quality, this study proposes a simulation-driven multi-strategy optimization framework for a three-story office building in Jinan. This study integrates EnergyPlus 23.2, jEPlus+EA 2.3.2, and the NSGA-II algorithm to co-optimize building performance. We evaluate the synergistic effects of roof photovoltaic coverage ratio, night ventilation turn-on temperature difference, and HVAC control strategies on carbon emissions and thermal comfort, while ensuring that CO₂ concentrations remain within health thresholds. The results indicate that the night ventilation temperature turn-on temperature difference is the most influential parameter. It yields standardized regression coefficients (SRCs) of 0.7456 for carbon emissions and 0.5325 for thermal discomfort. The Pareto-optimal solution achieves a carbon footprint of approximately 477 tCO₂, with only 8.8% indoor discomfort hours. This framework provides a robust, practical approach for the low-carbon and healthy operation of office buildings.

Heating load is jointly affected by meteorological conditions, system operating states, and historical evolution, exhibiting nonlinear, time-varying, and locally fluctuating characteristics. To improve short-term forecasting accuracy, this study proposes a particle swarm optimization (PSO)-based Trend–convolutional neural network (CNN)–bidirectional long short-term memory (BiLSTM)–Attention model. The model constructs trend-enhanced features from meteorological variables, operating parameters, temporal periodicity, historical lags, rolling statistics, differenced features, and exponentially weighted moving averages. CNN is used to extract local temporal features, BiLSTM captures bidirectional temporal dependencies, and Attention identifies key time steps. PSO further optimizes key hyperparameters. Two datasets are constructed from hourly heating-season operating data, and the proposed model is compared with BiLSTM, CNN-BiLSTM, CNN-BiLSTM-Attention, and Trend-CNN-BiLSTM-Attention models. The proposed model achieves the best performance on both datasets. For Dataset 1, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R²) are 0.628 GJ, 0.928 GJ, 9.308%, and 0.851, respectively; for Dataset 2, they are 0.3987 GJ, 0.5954 GJ, 10.32%, and 0.8709. These results indicate that trend-enhanced features and PSO improve forecasting performance and can support heating system operation scheduling.