Variable-speed Constant-frequency Wind Turbine Image Recognition Algorithm Based on mRMR-GRU
In order to accurately recognize the operating state of variable-speed constant-frequency wind tur-bines,a variable-speed constant-frequency wind turbine image recognition algorithm based on mRMR-GRU was proposed.By analyzing the mechanical dynamic structure of variable-speed and constant-frequency WTGs,the important control parts of the overall variable-speed and constant-frequency WTGs operation were obtained;based on the Internet of Things(IoT)technology,the integrated intelligent terminal sensing e-quipment collected the variable-speed and constant-frequency WTGs imaged in a specific structure;analy-zing the specific structure of WTGs generated the original images of the variable-speed and constant-frequen-cy WTGs;based on the histogram equalization method,the image of the variable-speed and constant-fre-quency WTGs was enhanced processing,the output of high-resolution acquisition of images;based on the mRMR algorithm,the definition of the mutual information of feature variables,to obtain the most representa-tive and relevant feature subset;fusion of GRU units and mRMR values to generate mRMR-GRU algorithm,optimize the image recognition network structure,the output of the wind turbine image state recognition re-sults.The experimental results show that:the designed algorithm has a MAPE value of 4.46,an RMSE val-ue of 3.57,and the highest value of image recognition accuracy is 0.98,which is enough to effectively regu-late the feature variables of the image,effectively identify the abnormalities that exist in the image of varia-ble-speed and constant-frequency WTGs,and it has obvious image pixel-point equalization effect and image enhancement recognition effect.
NETL
NSTL
万方数据
