基于改进递归神经网络算法的风电机组故障诊断研究OA

The fault diagnosis of wind turbine based on classical recurrent neural network uses single value data,lacking the consideration of system uncertainty,which affects the accuracy of diagnosis.Therefore,an improved recurrent neural network algorithm based wind turbine fault diagnosis scheme is proposed.By introducing hierarchical k-means clustering and interval value data technology,the sensitivity of the fault diagnosis algorithm to system parameter changes and external uncertain inter-ference is reduced,so that it can maintain satisfactory robustness and diagnosis accuracy in a long operation process.The sys-tem model of the wind turbine is analyzed,and the hierarchical k-means clustering and interval value data technology are intro-duced to improve the classical recurrent neural network algorithm.The unit fault data of an offshore wind power generation ar-ea in Guangdong is used for verification.The proposed wind turbine fault diagnosis program can achieve a diagnostic accuracy of greater than 98％.It has higher diagnostic accuracy than classical recurrent neural network algorithm and other neural net-work algorithm of the same type.