Research on dynamic evaluation model of health state of distribution transformer based on multimodal data fusion

Traditional power system equipment operation status diagnosis methods often rely on single-modal data, which are faced with challenges such as incomplete information, susceptibility to noise interference, and complex working environment, resulting in insufficient diagnostic accuracy, robustness, and environmental adaptability. In order to solve this problem, this paperproposes a dynamic evaluation model of distribution transformer state of health based on multi-modal data fusion. The model combines multi-modal data such as power equipment alarm text, infrared image and vibration audio, and constructs the BC-AT-LSTM model for feature extraction and fusion, aiming to achieve accurate diagnosis of equipment operation status. At the same time, in view of the ambiguity and randomness in the health state assessment of distribution transformers, the method combining cloud model and D-S evidence theory was used to evaluate the membership degree of each index, and the health status was evaluated through the modified probability distribution. Experimental results show that compared with the single-modal data model, the diagnostic accuracy of multi-modal data fusion is significantly improved, which verifies the feasibility and effectiveness of the method in the health assessment of distribution transformers. The research results provide an important basis for the dynamic health state assessment and condition maintenance of distribution transformers.