Research On Fault Early Warning Method Using Data Analysis For Wind Turbine Pitch System

Pitch system is an important structural system of wind turbine, pitch system can directly affect the normal operation of safe and stable operation of the unit. And wind turbines working environment more bad areas, often appear wind, strong wind, at the same time, including sudden gust. Pitch system’s main function is to capture the wind, the maximum use of wind power, and the influence of environment on the system is more obvious. If we do not provide an effective early warning program, when the pitch system is in case of failure, we are difficult to carry out maintenance work, and also cause significant economic losses to the wind farm. So the development of effective early warning programs for typical faults of pitch system, that is not only to timely and accurate warning fault information, to avoid serious fault conditions, but also ensure the unit running economy.Now at home and abroad for large and complex systems of early warning program presented numerous research programs, including the three typical faults of pitch system, but most are based on a constant single threshold warning or a single parameter based on the working condition. Such early warning scheme has certain defects, for example, to pitch angle fault setting a single threshold, not considering the influence of the sudden wind, put forward the alarm but it did not lead to failure,to motor temperature fault set a single warning threshold value, without considering the effects of environment temperature and wind speed on the pitch motor operating,resulting in high false alarm rate. Due to all kinds of problems, not reasonable warning fault information, is not conducive to stable operation of the unit.Based on the further study of variable pitch system, on the basis of structure and typical failure causes, through the establishment of FMEA and fault tree analysis, combined with the SCADA system of history and the measured data, are put forward aiming at three typical fault early warning, so as to suppress or solve the false alarm problems. For pitch angle fault, based on the least squares fitting historical data curve, formed under the normal operation of the model, based on Euclidean distance method to determine the normal work of the single pitch angle range. For more than two pitch angle difference of processing method using data statistical analysis method, to determine the threshold of early warning, at the same time using sliding window method to calculate the fault rate of abnormal, as early warning indicators, the shape of the failure rate can be observed, determine whether it is in fault condition. This method is compared with the threshold value as the early warning index method, the effective early warning at the same time, can be observed the development trend of abnormal failure rate, reducing false alarm rate, and ensure stable operation of the unit. For pitch torque fault, proposed a method of statistical analysis based on historical data set warning threshold limit,and using residual confidence interval as warning indicators,through calculating the mean and standard deviation of the residuals to observe whether is beyond the scope of setting confidence intervals to determine whether any fault condition. The method is simple and effective, from the perspective of the mean and standard deviation, reduce the effects of abnormal datas for the analysis, at the same time it can be seen that the overall trend of the development of situation, and is advantageous for the overall analysis. For pitch motor temperature fault, puts forward the method based on interval differentiate, in order to avoid the influence of the environment temperature on pitch motor running, is based on wind speed and environmental temperature as the basis, divided into working condition of the whole is divided into m* n interval, for each interval set up early warning based on data statistical analysis, abnormal failure rate were calculated based on sliding window as alarm indicators. This method divide all conditions into different interval, overall conditions changes according to the working condition of interval,setting different warning value, compared with the single threshold warning, effectively improve the accuracy of the early warning.Through early warning scheme can effectively monitor the fault occurrence, FMEA and fault tree analysis for the cause of the problem in screening and detailed maintenance measures provided the guarantee. Based on the warning at the same time, the main reasons for failure of maintenance measures are put forward, to prevent the accident further worse, save the maintenance cost, reduce maintenance time, increase operation efficiency of the unit.