A Method Of Wind Turbine Fault Diagnosis Based On SCADA Data

With the increasing emphasis on the environment and the shortage of traditional energy sources,the development of new energy sources has become more and more rapid in recent years.More and more electric power generation in China is supplied clean energy like wind power.Wind power plants are generally located in areas with harsh environmental conditions.In addition,the characteristics of the equipment itself are high in operation and maintenance costs.Therefore,appropriate fault detection methods are needed to detect faults early to reduce operation and maintenance costs.Currently,wind turbines are commonly equipped with monitoring and data acquisition(Supervisory Control and Data Acquisition,SCADA)systems.By applying a machine learning method,a large amount of data collected for the SCADA system can be used for condition monitoring and fault detection.This paper first introduces the basic components of the fan and the SCADA system,and explains some common faults of the wind turbine.Then it introduces the application of machine learning technology in wind turbine fault diagnosis,including data preprocessing,feature extraction,common modeling methods,and model verification.After that,the complete fault prediction process used in this paper is described,and the significance and effect of each step are explained in combination with the experiment.Finally,several components of the wind turbine are tested and applied to the relevant statistics of the entire wind farm.The main innovations of this paper are as follows: Based on the general modelbased condition monitoring and fault detection method,the random forest is used to find the best input features to avoid more time-saving and reasonable than the artificial selection of attributes(RF-DNN)the optimization method and appropriate features for input and output data There are few researchers who choose to have a great influence on the detection results.The method proposed in this paper uses monitoring and data acquisition data as data points,and uses wavelet analysis signals to perform noise reduction processing to achieve better input effects and utilize A recursive least squares(RLS)filter reduces the false alarm rate.By applying these methods,a more accurate output can be obtained,which greatly reduces the false alarm rate.Experiments on several components of the wind turbine were carried out on the actual SCADA data,and the effectiveness of the method was verified and applied to the entire wind field.