为解决电站辅机运行工况多变、结构复杂以及故障频发等问题，提出了一种基于核主元分析（KPCA）和门控循环单元（GRU）神经网络的电站辅机故障预警方法。通过核主元分析法提取电站辅机设备故障征兆参数，进行原始数据的约简。采用GRU神经网络进行电站辅机设备故障预警模型的建立。以神华福建某电厂HP843/Dyn中速磨煤机为例进行故障预警模型的训练、测试以及验证，该方法可以有效且提前发现中速磨煤机故障征兆。

In order to solve the problems of variable operating conditions,complex structure and frequent failures of auxiliary equipment in power station,this paper proposes a power station auxiliary equipment fault warning method based on kernel principal component analysis (KPCA) and gated recurrent unit (GRU) neural network.The kernel principal component analysis method is used to extract the failure symptom parameters of the auxiliary equipment in the power station,and the original data is reduced.The GRU neural network is used to establish the fault early warning model of the auxiliary equipment in the power station.The HP843/Dyn medium speed coal mill of a certain power plant of Shenhua Fujian Company is taken as an example to train,test and verify the failure warning model.The results show that the method can effectively detect the failure symptoms of the mediumspeed coal mill in advance.

TM621.7

上海市"科技创新行动计划"地方院校能力建设专项项目（19020500700）