Research On Fault Early Warning For Wind Turbine Based On Neural Network

The bad operating environment of wind turbine leads to the frequent failure of wind turbine,which has a negative impact on the safe and stable operation of wind turbine,and then poses a serious challenge to the development of wind power generation.Therefore,fault early warning for wind turbine is so important to large scale development of wind power generation.Due to the variability and unpredictability of wind turbine operating condition,the research on fault early warning for wind turbine has gradually become an important research direction in wind power.In recent years,more and more advanced intelligent algorithms are successfully applied in various fields with the development of artificial intelligence technology.Therefore,this paper adopts the neural network algorithm on the study of fault early warning for wind turbine.Based on the modeling of wind turbine state parameter,this paper presents a method of fault early warning for wind turbine combined information entropy with neural network.Firstly,locality preserving projections is used to extract the feature from wind turbine state parameters based on the high dimensional and nonlinear characteristics of wind turbine state parameters.Then,a neural network prediction model of the target state parameter is established.Finally,the information entropy method is used to analyze the residual trend of the prediction model,the early failure of wind turbine is detected.In order to realize the fault early warning of wind turbine,the premise is to establish a prediction model of the target state parameter with good stability and high prediction accuracy.In this paper,the concept of locality preserving projections and extreme learning machine is introduced.However,due to the random selection of hidden layer parameters in extreme learning machine,the prediction model is not stable.Therefore,the kernel function is used to improve the hidden layer of extreme learning machine.The simulation results show that the LPP feature extraction method reduces the difficulty of modeling and prediction,and improves the prediction accuracy.Moreover,extreme learning machine among the neural network algorithm has more advantages than the traditional BP neural network in learning speed and generalization performance.Compared with extreme learning machine,the prediction model of kernel extreme learning machine has a certain improvement in the stability and prediction accuracy.However,the residual of the target state parameter prediction model based on the neural network varies largely and violently.Combined with the information entropy method,this paper presents a method of fault early warning for wind turbinebased on information entropy and neural network.The information entropy method combined with the concept of entropy can quantify the intensity of data changes.The simulation results show that this method can detect the early faults of wind turbine.