基于主成分分析和神经网络相结合的制冷剂充注量故障诊断

Refrigerant Charge Fault Diagnosis Based on Principal Component Analysis and Neural Network

空调设备的能耗在建筑总能耗中占据30%以上的比例,一旦发生故障不及时排除,很可能导致系统能耗增加、设备寿命缩短、人员体感不适甚至影响正常的工作。本文采用一种基于主元分析法(Principal Component Analysis,PCA)与基于反向算法(Back Propagation,BP)的多层前馈神经网络(BP神经网络)相结合的算法,以制冷剂充注量为例,实现了对多联机性能故障的高效诊断。首先收集多联机组实测运行数据,进行一定的数据预处理工作,然后利用PCA提取主元,最终基于BP神经网络训练建立PCA-BP模型进行联机制冷剂充注量的故障诊断。结果表明:PCA-BP神经网络能高效检测多联机制冷剂充注量的故障,较于传统BP神经网络节约了计算时长及计算空间,同时该算法也具有泛化能力,为推广到多联机其他故障的诊断奠定了基础。

The energy consumption of air-conditioning system accounts for more than 30% of the total energy consumption of buildings.If the hitch is not processed in time,it will lead to the increased energy consumptions,the shortened equipment life time and the uncomfortable feelings.In the present study,a PCA-BP（Principal Component Analysis-Back Propagation） algorithm to detect the refrigerant charge fault high effectively was proposed.At first,the raw data was pretreated.In addition,the PCA method was employed to reduce the dimension of the data.Finally,the data will be trained by BP neural network.The result shows that the proposed method has a great accuracy on the refrigerant charge fault detection.Compared to traditional BP neural network,the new method saves computing time and computing space.At the same time,this algorithm also has the generalization ability,which lays the foundation for the extension to the other diagnosis of air conditioning system.