Fault Diagnosis Of Wind Power System Based On Recurrent Neural Network

With the continuous expansion of the scale of wind power system,wind power enterprises put forward higher requirements for the production and operation of wind power equipment and equipment management and control.In order to improve the rapidity and reliability of fault diagnosis of wind power system,combined with data-driven technology and deep learning algorithm,complex non-linear fault diagnosis problems are discussed.The new solution is of great significance in guaranteeing the personal safety of maintenance and technical personnel,reducing or even avoiding unnecessary economic losses.Deep learning is essentially a mechanism for disposing and processing input and output signals.With the development of large data and cloud computing technology in recent years,the problem of low training efficiency has been solved by using greatly increased computing power.By expanding training data sets,the risk of network over-fitting has been greatly reduced.Comparing with traditional methods,the results of complex deep learning network model in various fields of classification and regression problems are significantly improved.By learning different deep learning algorithms,this paper proposes a multi-fault diagnosis method for wind power generation system based on recurrent neural network(RNN-MFD).According to the data of actual wind speed,440,000 pieces of normal operation and fault data of wind system are obtained by system modeling.A recurrent neural network algorithm is established,which takes 30 characteristic parameters such as wind speed,rotor speed,generator speed and power generation as input and 10 different types of fault labels of wind as output.Classification prediction model and intelligent learning of the specific rules formed in the wind system sample data through the model;continuous training,optimization and testing of the model,at the same time,adding Gauss interference matrix in the training and testing set to test the robustness of the model,accuracy and loss results verify the feasibility of the algorithm;By comparing the four evaluation criteria with the other five related algorithms,such as deep confidence network,which have higher recognition in deep learning,such as Precision,Recall,MDR and F1-Measure,the results show that the algorithm model can effectively solve the multi-classification problem of wind power system fault diagnosis.