Study On Condition Monitoring Model Of Wind Turbine Based On Hybrid Input Fuzzy Neural Network

In pursuit of environmental protection,the use of wind energy for power generation is a good renewable energy generation method.It has matured technology and good commercialization characteristics and has received worldwide attention.In recent years,the state has vigorously advocated wind power generation.China's wind power installed capacity has grown rapidly,and the proportion of large/ultra-large wind turbines has increased year by year.As the proportion of wind turbines increases year by year,how to formulate the operation and maintenance decision plan of wind turbines is the key to ensure the reliable operation of wind turbines.Therefore,it is necessary to conduct research on the state monitoring and health status of the wind turbine system.In this thesis,using the wind turbine operating data and fault data obtained by FAST system simulation,the state monitoring and health research of wind turbine system based on FAST system data is carried out.Its research content mainly includes: Briefly describe the basic principle of wind turbines and introduce the characteristics and working principle of doubly-fed wind turbines and introduce the operating characteristics and working conditions of wind turbines.The basic theory of FAST system and the model simulation of wind turbine by FAST system are introduced,and the state parameter data of wind turbine is obtained.1)In this thesis,the basic principle of wind turbine is briefly introduced,and the characteristics and working principle of the doubly fed wind turbine are introduced.The operating characteristics and working conditions of the wind turbine are introduced.The basic theory of FAST system and the model simulation of wind turbine by FAST system are introduced,and the state parameter data of wind turbine is obtained.2)For the problem of dealing with categorical variables,this thesis proposes to solve this problem by using hybrid input fuzzy neural network.The model structure,model structure identification and training algorithm of hybrid input fuzzy neural network are introduced.The numerical simulation comparison experiment is carried out.By comparing the T-S fuzzy neural network and the hybrid fuzzy neural network input model,it is proved that the hybrid input fuzzy neural network model has a good effect in dealing with categorical variables.3)In the past,the monitoring of multiple parameters of wind power generators was usually achieved by establishing multiple state monitoring models,which easily caused problems such as multiple models and high false positive rate.Considering the different operating conditions of the wind turbine system,this thesis proposes a state parameter model based on the hybrid input fuzzy neural network for the normal operation of the wind turbine.The model can introduce categorical variables into the model and solve the problem that the state parameters of different operating conditions of the wind turbine system have different mapping relationships.At the same time,it solves the problems of many previous models.4)The purpose of condition monitoring and health status of the wind turbine is to determine whether the wind power generator is in an abnormal state.In this thesis,the multivariate Gaussian function model is used to identify the wind turbines.The residual Gaussian function is used to establish the residual model of the wind turbine normal state parameter model.The threshold is selected by the method of gradient descent,and whether the wind turbine is in an abnormal state is determined by whether the threshold is exceeded.