Large Wind Turbine Fault Prediction Technology Research

As a kind of clean energy, wind power is the key to promote the electric power reform. However, with the rapid growth of wind turbine installed capacity, wind power generation benefit is far lower than expected. The main reason is that frequent wind turbine faults, which reduce the utilization rate of wind power and increase the operation maintenance cost. In order to ensure safe and reliable operation, wind turbine fault prediction technology causes many scholars' attention both at home and abroad.This paper made it clear that what are research background and significance of wind turbine fault prediction technology, and expounded its research status in detailly. In order to further understand research subject, research methods and technical difficulty, fault tree and fault prediction technology for wind turbine were analyzed in detail. It introduced several common wind turbine condition monitoring models, which lays the theoretical foundation for the following modeling. The paper put forward the improvement measures for the existing wind turbine fault prediction technologies. Firstly, putting the wind turbines as a whole, this paper presented a new method for wind turbine abnormal recognition or early failure prediction. On the one hand, the least squares piecewise linear fitting was used for wind power scatter to improve the fitting precision; On the other hand, the sliding vertical comparison plan was proposed for power curve, in order to avoid the influence of interference factors. Secondly, in view of the specific types of fault, the paper put forward a improved fault prediction method based on least square support vector machine, namely, multi-parameter fault prediction strategy with the weight, which avoids the existing methods' disadvantages that single parameter prediction accuracy is lower and multi- parameter do not consider the weight of parameters.The proposed method is verified by the field test data. Through analysis and comparison with the existing methods, it is concluded that this method is more accurate and practical than the existing methods.