摘要
本文提出了一种基于共形预测的冷水机组鲁棒故障诊断方法。采用ASHRAE RP-1043项目数据,分析了不同数据驱动模型在七种故障场景下分别采用随机划分和变工况故障诊断两种测试方法的诊断效果。结果表明:该方法可以有效缓解故障诊断模型在新工况点下诊断能力下降的问题,所提出的故障诊断方法相较于传统机器学习模型的准确率、召回率平均最大提升幅度可分别达到17.35%和20.53%。
A robust fault diagnosis method for chillers based on conformal prediction is proposed in this paper.Using the data from ASHRAE RP-1043 project,the diagnostic performance of different data-driven models is analyzed across seven fault scenarios,employing two testing methods:random partition and variable operating condition test.The results indicate that the proposed method effectively mitigates the decline in diagnostic performance of fault diagnosis models under new operating conditions.Compared with traditional machine learning models,the proposed fault diagnosis method achieves an average maximum improvement in accuracy and recall of 17.35%and 20.53%,respectively.
作者
刘卓轩
曹承
张天乐
张哲铭
朱旭
晋欣桥
杜志敏
LIU Zhuoxuan;CAO Cheng;ZHANG Tianle;ZHANG Zheming;ZHU Xu;JIN Xinqiao;DU Zhimin(Institute of Refrigeration and Cryogenics,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai Marine Equipment Research Institute,Shanghai 200031,China)
出处
《制冷技术》
2024年第4期7-15,共9页
Chinese Journal of Refrigeration Technology
基金
国家自然科学基金(No.52276011)。
关键词
冷水机组
共形预测
机器学习
故障诊断
Chiller
Conformal prediction
Machine learning
Fault diagnosis
