With the continuous expansion of the construction scale of power grid projects, the turnover and backlog of materials are constantly increasing. In addition, the maintenance and emergency repair activities of distribution networks in various regions are also related to many factors such as natural aging and disasters, and their material needs exhibit seasonality and periodicity. To better match materials and reduce the management costs and disposal risks of power grid companies, this paper uses the ReliefF-PCA method to extract features from inventory materials. Then, hierarchical clustering is performed on inventory materials, and a pattern recognition method based on the BPF neural network is used to achieve multi-dimensional automatic classification; Using the multi-source data analysis method based on the SCA to extract the association rules between inventory matching and backlog inventory materials, and corresponding inventory material circuit management references are provided. From case analysis, it can be proven that this model can achieve multi-dimensional automatic classification of inventory materials, break through the data barriers between protocol inventory matching and material management business, improve the processing efficiency of backlog inventory materials, and provide data support for the refined management of power grid inventory materials.