Estimation and detection of rotor mass imbalance of wind turbines using convolutional neural network

Rotor imbalance in wind turbines takes a negative problem. Particularly for large wind turbines, rotor mass imbalance could have a severe impact because of the large size of the rotor. In this study, we investigated the impact of rotor mass imbalance on the performance of wind turbine. This study led to the proposal of a new framework for the estimation and detection of rotor mass imbalance. The signals of different roto imbalances are collected to test and verify the impact of rotor mass imbalance. Firstly, a method based on optimized maximum correlated kurtosis deconvolution is proposed for the identification of rotor imbalance, the intrinsic mode functions (IMFs) of nacelle vibration is used as the object variable, obtained through variational mode decomposition. Secondly, a convolutional neural network with a new structure is used to estimate the magnitude of the rotor mass imbalance. The first layer of the structure is widened for improved feature extraction, moreover, training enhancement strategies are applied to the model to improve its robustness to wind. The acceleration of the nacelle is accepted as input. This structure offers good accuracy and robustness. Thirdly, the position of the rotor mass imbalance is achieved by matching rotor azimuth and the most sensitive IMF. Finally, the performance of the framework is tested with the help of a high-fidelity simulation environment with different noise and various turbulence intensities. The results demonstrate the satisfactory performance of the proposed framework.