中央空调系统机理模型在长期运行过程中，由于管道结垢、设备老化等实际因素影响，关键物理参数会发生漂移，导致模型精度逐渐下降，严重影响基于模型的优化控制效果。为确保模型的长期可靠性和控制策略的有效性，亟需对所构建的物理模型进行深入的性能分析，并基于实际运行数据对关键参数进行有效辨识。本文以构建的中央空调系统机理模型为基础，通过敏感性分析提取对系统性能影响显著的关键参数集，建立了以冷却水进出水温度、EER等目标变量的误差总和最小为目标的优化函数，将参数辨识问题转化为优化问题，采用近似优化方法进行求解，并基于实际工程数据对辨识方法进行验证。结果表明：该方法可显著降低各目标变量的误差。以EER为例，辨识前的MAE为1.93，RMSE为1.98，MAPE高达61.4%；经辨识后，单目标辨识下MAE降至0.008，RMSE降至0.004，MAPE降低至0.29%，多目标辨识下MAE降至0.28，RMSE降至0.51，MAPE降低至9.74%。验证结果表明，该方法在提高模型精度和工程适用性方面具有良好效果。

In the long-term operation of a central air conditioning system's mechanistic model, key physical parameters drift due to practical factors such as pipe scaling and equipment aging, leading to a gradual decline in model accuracy and severely impacting model-based optimization control effectiveness. To ensure the long-term reliability of the model and the effectiveness of control strategies, it is imperative to conduct in-depth performance analysis of the constructed physical model and effectively identify key parameters based on actual operational data.  This study builds upon the established mechanistic model of the central air conditioning system. Through sensitivity analysis, a critical parameter set significantly influencing system performance is extracted. An optimization function is established, aiming to minimize the total error of target variables such as cooling water inlet and outlet temperatures and EER. The parameter identification problem is transformed into an optimization problem, solved using approximate optimization methods, and validated with real engineering data.  Results demonstrate that this method significantly reduces errors in target variables. Taking EER as an example, the MAE before identification was 1.93, RMSE was 1.98, and MAPE reached as high as 61.4%. After identification, under single-target identification, MAE dropped to 0.008, RMSE decreased to 0.004, and MAPE fell to 0.29%. Under multi-target identification, MAE was reduced to 0.28, RMSE to 0.51, and MAPE to 9.74%. The validation results confirm the method's effectiveness in improving model accuracy and engineering applicability.