Condition Monitoring Of Windturbine Pitch System Based Onperformance Curve And RBF Fuzzyneural Network

High frequency failure of key components of wind turbines is one of the reasons hindering the development of wind power industry.Improving the reliability of the wind turbine and reducing the operation and maintenance cost are the most effective solutions.As an important part of wind turbines,the pitch system long downtime caused by high frequency of faults,which accounts for a large proportion of the operation and maintenance costs of wind turbines.Therefore,this paper chooses to study the method of condition monitoring of pitch system,finds the abnormal conditions of the pitch system in advance,and takes maintenance measures in time to prevent more serious damage to the pitch system.This paper studies the condition monitoring of the pitch system as follows:Firstly,the operating state of the wind turbine is divided into seven categories according to the wind speed.The range of wind speed,power,impeller speed and pitch angle in each operating state is divided in detail,and draw the two-dimensional curve of the pitch system under ideal conditions to illustrate the trend of the operating parameters in each condition.The scatter plots of the six performance curves of the pitch system are plotted by SCADA data,and the characteristics of the six curves are analyzed.Two different methods are proposed to monitor the condition of the pitch system.Secondly,a condition monitoring model of pitch system based on performance curve is proposed.When the actual data point deviates from the theoretical power-impeller speed(P-N)and pitch angle-impeller speed(PA-N)curves,it indicates that the pitch system has a fault.In order to obtain the standard P-N and PA-N curves under actual wind turbine operation,this paper uses the 3? criteria to eliminate the anomaly points.Then uses the local weighted linear regression method and the constant function to fit the standard P-N and PA-N curves.The thresholds for the distance of the actual data points to the two standard curves are set according to the technical parameters of the wind turbine in different wind speed ranges.Thirdly,a condition monitoring model of pitch system is established based on RBF fuzzy neural network.Corresponding fuzzy rules are obtained by analyzing the abnormal point distribution areas in the power-wind speed(P-V),impeller speed-wind speed(N-V),generator torque-wind speed(T-V),and pitch angle-wind speed(PA-V)scatter plots.FourRBF fuzzy neural network condition monitoring models are established by using the obtained fuzzy rules.The SCADA historical data with fault labels are used as training data to train the four models separately.The four models are aggregated to form a multiple monitoring model for the pitch system.Finally,by using the fault data of the wind turbine of a wind farm in Hebei Province,the proposed two condition monitoring models of the pitch system are verified separately.The results show that the proposed model can detect the abnormal condition of the pitch system well and is earlier than the alarm of the SCADA system.