The increasing quantity of air pollutants generated by automobiles can cause significant harm in relatively enclosed indoor environments. Studying the distribution of pollutants under different conditions in underground parking garages is of great significance for improving indoor air quality and reducing casualties in the event of a fire. This article presents a geometric model of an underground parking garage based on PHOENICS modeling. The related results of CO concentration distribution and fire temperature distribution under ventilation and fire conditions are obtained. Based on the CO concentration and velocity distribution as well as the temperature distribution during a fire, reasonable suggestions are proposed to improve indoor air quality and reduce casualty rates in fire incidents. The results show that under ventilation conditions, adjusting the position of the induced ventilation fan can maintain CO concentrations below 30 ppm in partitions one to three and below 37 ppm in partitions four to six. The temperature of smoke gases remained below 50 °C during the evacuation time, and only a small area exhibited CO levels exceeding 2000 ppm. The existing ventilation exhaust system provides effective fire protection, as it minimally affects personnel evacuation due to the relatively lower smoke temperature.

Addressing the deficiencies of data-driven heating, ventilation, and air conditioning (HVAC) models in multi-objective joint prediction, long-horizon prediction accuracy, and uncertainty quantification, this study proposes a heterogeneous deep ensemble framework. Through co-simulation technology, a multi-operating-condition operational dataset for a typical variable air volume (VAV)–Reheat system was constructed. By adopting an operating-condition-level partitioning strategy, the input features were strictly decoupled into historical system states, internal loads, and external control commands. On this basis, the framework integrates the advantages of Long Short-Term Memory (LSTM) networks, Gated Recurrent Units (GRU), and Temporal Convolutional Networks (TCN). Utilizing an inverse-variance weighting mechanism, it achieves the synchronous and accurate prediction of 11-dimensional targets, encompassing total system power, five-zone temperatures, and five-zone CO₂ concentrations. Simultaneously, the prediction divergence among the ensemble models is utilized to quantify epistemic uncertainty, and a variance scanning mechanism is introduced to calibrate the prediction confidence intervals. The results indicate that within a 6 h prediction horizon, the absolute errors of the model in power, temperature, and CO₂ are reduced by 80.17%, 63–72%, and 75–80.62%, respectively, compared to the physical baseline. On the independent test set, the actual coverage rate of the calibrated prediction intervals approaches 90%, exhibiting an adaptive enveloping variation characterized by tightening during steady states and widening during transients. The proposed heterogeneous ensemble framework demonstrates superior prediction accuracy and long-horizon prediction stability, providing reliable model support for multi-objective optimal operation and robust control of buildings.