Condition Monitoring Of Wind Turbine Based On Cointegration Analysis

Wind turbine is an important equipment to realize the conversion of wind energy to electrical energy.Accurate assessment of the operation performance of wind turbines is of vital significance for the state monitoring and operation maintenance of each component of the wind turbine.Cointegration analysis method is used to analyze multiple nonstationary variables simultaneously,and the off-line model is established and verified by using the historical data SCADA wind turbine.Finally,the off-line model is transformed into an on-line monitoring program by LabVIEW software,and the on-line monitoring of wind turbine status is realized by using that program.The main research contents of the paper are as follows:1)The historical data stored in the SCADA system is used to establish the ADF stationarity test model,and the t-statistic values of model coefficients and constant terms are estimated.Several sets of non-stationary data are selected from many SCADA monitoring objects as the next analysis object,and then selected The object of the first-order difference operation is performed,and the operation result is subjected to a stationarity test,and several sets of data that meet the first-order single integer are selected as the input of the co-integration analysis.2)Use the E-G two-step method and Johansen multivariate cointegration analysis method to test the cointegration relationship of the selected data.Several groups of variables with cointegration relationship are used to establish corresponding models,and the least squares estimation method is used to estimate the cointegration vector and constant.The cointegration vector constant and the selected variables are used to establish an offline cointegration model,which is verified by SCADA historical data under normal and abnormal wind turbine conditions.The results show that the state monitoring method using only SCADA data is relatively sensitive to electrical appliance failures.3)Select high-frequency vibration data as an input for cointegration analysis.First,the singular value difference spectrum method is used to denoise and reconstruct the original vibration signal.Item input establishes a cointegration model and verifies.The results show that the sensitivity of the proposed method to mechanical failure is improved by adding vibration data.4)In the online phase,the offline model is written as an executable real-time monitoring program: through observation of multiple wind farm SCADA systems,it is found that the monitoring objects of the SCADA systems of different wind farms are slightly different,in which wind speed and output power are detected by all wind farms In view of the universality of the program,the E-G two variable offline model was transformed into an online monitoring program.Use formula functions,handle attributes and other functions to transform the offline state monitoring model into an online monitoring program,and use multiple wind turbine SCADA data to verify the online program.The results show that the developed program can effectively identify the abnormal operation status of wind turbines,and has compatibility and good applicability to different wind turbine SCADA data.