Abstract: The existing identification methods of electrical equipment defects rely on single modal data or single type features for identification, and the adaptability to thermal and physical composite defects of electrical equipment in photovoltaic stations is insufficient, resulting in low overall identification accuracy. Therefore, an intelligent defect recognition method for electrical equipment in photovoltaic stations based on infrared and visible light image registration is proposed. Firstly, feature points are extracted and false matching pairs are eliminated. The algorithm achieves pixel level registration of bimodal images through affine transformation. Next, construct a Laplacian pyramid and complete image fusion according to the rule of "priority preservation of high-frequency details and weighted fusion of low-frequency contours". Further, by using adaptive morphological segmentation to lock the device area, temperature and structural features are extracted layer by layer, and a complete defect feature vector is constructed. Finally, based on the collaborative mechanism of "rule judgment model classification", combined with the preset defect rule library and MobileNetV2 network, precise defect recognition and classification are achieved. The experimental results show that the mAP of this method for identifying thermal defects, physical defects, and composite defects reaches 94.2%, 92.8%, and 90.5%, respectively, with an overall mAP of 92.5%, which is superior to existing comparative methods and achieves accurate and efficient identification of electrical equipment defects.