Research On Wind Turbine Condition Monitoring Based On Data Mining

Wind power generation is a renewable energy which has a big development potential. It has been paid attention worldwide in recent years. Condition monitoring of wind turbine large equipment become an important component of wind power research field. This paper researches on wind turbine condition monitoring based on data mining. The research is divided into two directions:1、the relation between faults;2、the relation between faults and parameters.The research content is as follows:1、 Analysis wind turbine fault data, research association rules and improve the algorithm. This paper took blade angle asymmetry fault as the research object, used improved Apriori algorithm to analysis the alarm data which appeared after pitch fault. Combined with the running mechanism of pitch system, this paper found some close relations between faults. These rules can act as a reference for operators to improve their work efficiency.2、Too much dimensions of parameters will produce’dimension disaster’ when looking for the hidden relationship between faults and parameters. Therefore this paper studied feature selection algorithm, built ReliefF feature selection model, then reduced the wind turbine parameters’dimension, extracted8characteristic parameters form47parameters to eliminate redundant information and reduce the dimension of feature vector. These work laid a solid foundation for classification work.3、In order to analysis the relations between faults and parameters, this paper studied the theoretical knowledge and basic steps of classification algorithm. And we built BP neural network classification model to monitor wind turbine condition with the parameters which have been selected in part2. Combined with the real data, we found that BPNN can judge whether the wind turbine is running good or fault happened properly, and achieve the goal of wind turbine condition monitoring.