【摘要】 针对基于数据驱动的制冷系统故障诊断模型缺乏可解释性的问题，提出一种融合XGBoost与SHAP的制冷系统故障预测及其特征分析方法。首先，基于XGBoost模型对制冷系统制冷剂泄漏、冷凝器风机故障及蒸发器风机故障分别进行预测，其次，结合SHAP解释方法对三种故障进行特征分析，结果表明：可以仅凭压缩机进气压力判断系统是否发生冷凝器风机故障及蒸发器风机故障，而制冷剂泄漏则需要多个特征进行判断，主要是压缩机排气压力、制冷量、膨胀阀出口压力。更多还原

【Abstract】 Aiming at the problem that the fault diagnosis model of refrigeration system based on data-driven is lack of interpretability, a fault prediction and feature analysis method of refrigeration system based on XGBoost and SHAP was proposed. Firstly, based on the XGBoost model, the refrigerant leakage, condenser fan fault and evaporator fan fault of the refrigeration system were predicted respectively. Secondly, combined with the SHAP interpretation method, the characteristics of the three faults were analyzed. The results show that the compressor inlet pressure can be used to judge whether the system has condenser fan fault and evaporator fan fault, while the refrigerant leakage requires multiple characteristics to judge, mainly the compressor exhaust pressure, cooling capacity, and expansion valve outlet pressure.更多还原