Research On Identification Method Of Outliers In Operating Data Of Wind Turbine

The accuracy of wind farm wind speed and power prediction is determined by various factors,and the effectiveness and reliability of wind power operation data affect the practical application effect of the prediction method.In the actual operation process,due to the wind power limitation,sensor failure,meter error and the influence of environmental factors such as dirt and ice on the wind turbine blades,it is difficult to avoid abnormal data in wind speed data collected from wind farms.If it is directly used as the original input data of wind power prediction model and the basic data of wind power impact on the system,it will affect the accuracy of prediction and the reliability of analysis results.In this thesis,the wind farm is taken as the research object,and the abnormal data in the wind turbine operation data is effectively identified,which provides effective data support for the safe and reliable operation of wind power grid connection.The measured wind speed and power data obtained by the supervisory control and data acquisition system(SCADA)is an important basis for assessing the economic and technical level of wind farms.The measured data collected from the wind farm usually contains abnormal data points characterized by missing,accumulated,and over-limit data.These abnormal data mainly come from faults or disturbances in various aspects such as data acquisition,measurement and storage,as well as fan shutdowns caused by unplanned maintenance units and abandonment of wind power.The historical wind speed data collected by wind farms exhibits non-stationary changes with time.To ensure the accuracy and effectiveness of wind speed data,the method for identifying abnormal wind speed data based on combined auto regression integrated moving average(ARIMA),wavelet decomposition(WD)and hidden Markov model(HMM)algorithm is proposed.The ARIMA model is used to fit the sampling time series,and the fitting residuals reflecting the abnormal wind speed data are obtained.Then the fitting residuals are deal with wavelet-decomposed,and the double random process of HMM algorithm is used to describe the wind speed anomalies.Through the analysis and calculation of measured wind speed data of Jiuquan wind farm,the results show that the proposed algorithm is effective and feasible for the identification of outliers in non-stationary wind speed time series with large data volume,which can provide reference for improving wind power power prediction accuracy and optimizing wind farm operation.A large number of abnormal points in the measured wind speed-power curve will affect the analysis of the operating characteristics of wind turbines,and there is a certain correlation between wind speed and power.A method for identifying abnormal data of wind turbine based on Copula theory is proposed.The Copula function is used to establish the probabilistic power curve of the correlation between wind speed and power.The corresponding anomaly data recognition model is obtained by combining three types of anomaly data features.According to the measured data of Jiuquan Wind Farm,the effectiveness of the method is verified.