Detection Method Of Abnormal Value In Operating Data Of Wind Turbine

With the steady development of wind power science and technology in China,the demand for wind power generation and grid operation is becoming higher and higher.The operation data of wind power of wind speed fluctuation and intermittence effectively and accurately predict become the basis of wind power operation,the high quality of the wind turbine operation data is necessary,the wind power generation performance evaluation of wind power forecast work.With the acquisition of wind turbine operation data contains a large number of outliers,the running data of low quality will lead to further deepen the application of information and data so as to reduce the misjudgment of the wind power prediction accuracy,even adversely affect the operation of the power grid dispatching.Therefore,the identification and correction of outliers in operational data is a necessary prerequisite for predictive modeling analysis.In order to ensure the safety and economic operation of wind power integration,the following contents are included in this paper.Firstly,the causes of abnormal values in the operation data of wind turbines and the distribution characteristics of abnormal values in wind power scatter plots are analyzed,and the classified multiple model is used to detect outliers.According to the characteristics of abandoned wind data,but the fluctuation of wind speed value recognition continuous detection model in a certain range of power fluctuation;by using statistical method to identify a robust four smaller proportion and isolated existence of outlier outlier data,to eliminate the abnormal data caused by sensor error;using the idea of data mining,selection fuzzy C means algorithm identification data scattered cluster of deviation points in the graph,the electromagnetic interference is eliminated in the process of transmission and storage in the computer terminal due to failure of the operation data of pollution.Then,from the analysis of the correlation between wind speed time data,select the Adaboost-BP network with the least squares support vector machine to build the optimal combination model obtain future time prediction value,residual measurement value and predictive value of the anomaly information mining.For the general residuals,the residual values are normal distribution,and the posterior residual logarithm is used to detect the residual residual information in the residual sequence,and then the location of the outliers is determined.In order to avoid the detection error caused by detection threshold,an adaptive detection method is proposed,which can adaptively recognize outliers according to the characteristics of the detected residual sequence.Finally,in order to ensure the continuity and utilization of the operation data of the wind turbines,the three spline interpolation method is used to correct the outliers.Measured operation data from Gansu Jiuquan wind farm and wind farms were low Pu outlier detection and correction.Prediction evaluation algorithm for outlier detection accuracy using RBF neural network,the prediction results show that after pretreatment to improve the accuracy of prediction of 10% or so,and the adaptive detection method compared to the Bayesian posteriori detection method is better than the processing efficiency and accuracy,the accuracy can be improved by about 15%.The evaluation results show that this method can effectively eliminate abnormal values,and has some practical value for wind speed and power prediction of wind farms,and further improves the prediction accuracy of short-term wind speed and power.