基于机器学习水轮机汽蚀自动监测研究  

Research on Automatic Fault Monitoring of Electrical Equipment in Hydropower Station Based on Machine Learning

作　　者：刘苏程 LIU Sucheng(Huizhou Baipenzhu reservoir engineering authority,Huidong 516341,China)

机构地区：[1]惠州市白盆珠水库工程管理局,广东惠州516341

出　　处：《广东水利水电》2022年第8期70-73,88,共5页Guangdong Water Resources and Hydropower

摘　　要：水轮机汽蚀的形成是由于旋转叶片、尖锐曲线或湍流引起局部压降而形成的,是一种潜在的极具破坏性的复杂现象,会导致机械表面形成凹坑和侵蚀涡轮机材料,从而缩短水轮机转轮的寿命。研究结合了多种机器学习方法,先采用加速度计以及声发射传感器对转轮进行监测,之后利用均方根(RMS)和平方根(MD)振幅计算方法来提取输入数据特征值,最后采用随机森林方法(RF)、决策树(DT)、人工神经网络(ANN)、支持向量机(SVM)和logistic回归5种机器学习算法,对RMS和MD数据进行了对比分析,得到了最佳监测精度,研究成果可为相关工程提供参考。The formation of turbine cavitation is due to the local pressure drop caused by rotating blades,sharp curves or turbulence.It is a potentially destructive and complex phenomenon,which will lead to the formation of pits on the mechanical surface and the erosion of turbine materials,so as to shorten the service life of turbine runner.This study combines a variety of machine learning methods.Firstly,the accelerometer and acoustic emission sensor are used to monitor the runner,and then the root mean square(RMS)and root mean square(MD)amplitude calculation methods are used to extract the eigenvalues of the input data.Finally,the RMS and MD data are compared and analyzed by using five machine learning algorithms:random forest method(RF),decision tree(DT),artificial neural network(ANN),support vector machine(SVM)and logistic regression,The best monitoring accuracy is obtained,and the research results can provide reference for related projects.

关 键 词：机器学习 水电站 水轮机汽蚀 故障监测 

分 类 号：TK73[交通运输工程—轮机工程]