The power grid maintenance work ticket contains a lot of valuable information, and the retrieval and utilization of historical work ticket can effectively assist in the decision-making of maintenance work ticket. The traditional retrieval method is inefficient and has a large workload and cannot achieve a comprehensive and effective reference for historical work tickets. Based on this, an assistant decision-making method of power grid maintenance work ticket based on text mining is proposed. Firstly, the limitations of traditional work ticket retrieval and the features of work ticket text are analyzed and summarized. Then the text is pre-processed with power knowledge. According to the problem of non-standardization of text expression, an improved double-layer bag-of-words (BOW) model containing subject words and auxiliary words is proposed. Finally, the cosine similarity is used to achieve multivariate similarity matching between the key information of the maintenance equipment and the historical scene, and the matching results are arranged in descending order. Through the digging of implicit links between work tickets, the identification and extraction of similar and related work tickets can be realized, providing more accurate and comprehensive decision-making support for dispatchers to make working tickets., and promote the development of power grid in the direction of intellectualization. Based on the actual power grid example, the feasibility and effectiveness of the proposed method are shown.