基于机器学习的污水源热泵机组故障诊断模型研究

A Novel Fault-diagnosis Model for Sewage Source Heat Pump based on Machine Learning Approach

根据污水源热泵机组的状态变量和部分故障类型,提出一种基于BP神经网络的污水源热泵机组故障诊断模型。以西安市某200万m^2大型污水源热泵集中供暖系统为研究对象,采集2016年11月至2017年3月供暖期间热泵机组中蒸发器、冷凝器和压缩机的温度、压力等工况数据,利用L-M、BR、SCG训练算法分别建立故障诊断模型,并进行比较分析。研究结果表明:通过神经网络建立污水源热泵机组故障诊断模型,正常类型与故障类型的诊断结果差异显著,可以满足诊断需求。其中,L-M算法训练效果最优,迭代次数为113次,正常类型误差值均小于0.012,故障类型误差值均大于0.9。

According to the state-variable and diagnosis-type of sewage source heat pump unit,the fault-diagnosis model based on BPNN(Back Propagation Neural Network)was proposed.In this paper,a two million square meters sewage source heat pump central heating system was studied as the research case.The operation data were investigated,including temperature and pressure of evaporator,condenser and compressor,from November 2016 to March 2017.The fault-diagnosis model was respectively established and analyzed with L-M,BR and SCG algorithm.The experimental results indicated that the fault-diagnosis model based on BPNN can effectively find and distinguish the three faults of sewage source heat pump unit.The iterations of L-M can reach 113 times,and is the best one of the three different algorithms.The error values of the normal diagnosis was less than 0.012,the error values fault diagnosis was greater than0.9.