Aiming at the problem that the equipment maintenance data in the power network are scattered, the knowledge structure is different from each other, and the single source information is difficult to accurately characterize the overall condition of the equipment, it is necessary to evaluate the state of the power grid equipment in order to reduce the requirements of reducing power outage or even no power outage in the power grid maintenance. In this paper, a state evaluation method of distribution network equipment based on multi-source heterogeneous information fusion is proposed. Firstly, the multi-source heterogeneous information is processed, the structured data is transformed into a recursive graph corresponding to the unstructured data by using variational modal decomposition and Hilbert transform, and the unstructured data and structured data are unified into the same dimension, Then the unified multi-source data are input into the improved convolutional neural network for training and feature vector extraction. The obtained feature vectors are spliced and fused to realize equipment state perception. Finally, an example of switchgear is used to verify the correctness of the proposed method. The results show that the distribution network equipment status evaluation method based on multi-source heterogeneous information fusion can well retain the original input information, complement each other between multi-source heterogeneous information, have higher fusion diagnosis accuracy, and can realize more accurate evaluation of equipment status.

Multi source heterogeneous information; Improved convolutional neural network; Eigenvector; Splicing and fusion; State evaluation