Research On State Identification Of Wind Turbines Based On Relief And NSET Algorithm

Wind energy is a kind of green new energy with large reserves,mature development technology and high utilization value.The most typical way to use wind energy is wind power generation.The traditional method for the condition monitoring of wind turbines is mainly the fixed threshold.Owing to the influence of many factors on the condition of the output of the wind turbines,this method often has some problems such as inaccurate identification of the wind turbines in practical application.At the same time,the existing optimization algorithm ignores the screening of training samples,so the accuracy improvement is not obvious.Therefore,this paper is to improve the identification accuracy of wind turbine status as the object of research study,the main contents are as follows:First of all,the wind speed parameters in the wind turbine monitoring parameters are prone to deviations,so the identification criterion for the state parameters of the wind turbine is established.A new method of wind speed correction based on power coefficient(Cp)empirical formula is proposed.The wind power curve is normalized according to the measured ambient air density.The results show that the standardized wind power curve is more in line with the actual operation of wind turbine.Secondly,aiming at the monitoring data preprocessing,the research on the identification and cleaning of abnormal state parameters was carried out.A standard deviation gain method based on three-dimensional data(wind speed-rotor speed-power)is proposed.Using the model proposed in this paper,data preprocessing of multiple wind turbines in wind farms shows that the method can quickly and accurately clean out abnormal points and preliminary judge the normal and abnormal state of wind turbines.Thirdly,the model of state feature quantity mining for wind turbines based on Relief algorithm is studied.Weights are used to measure the correlation between parameters and categories.Euler distance is selected to eliminate the redundancy of sample data to construct the index system of health status identification of wind turbines.The case study shows that the state feature quantity mining model established in this paper can effectively recognize the importance of state parameters of different units.Finally,the NSET multi-input and multi-output nonlinear state assessment model is used to predict and evaluate the key parameters of wind turbines.Through the sliding window residual analysis,the alarm thresholds of different status levels are obtained,which realizes the real-time identification of the wind turbine state deterioration degree and predicts the warning fault in advance.Meanwhile,by analyzing the residual value of specific parameters,the main source of fault can be identified.The case study shows that the model of wind turbine status identification established in this paper can identify the abnormal state of the wind turbine in advance and effectively warn the fault.