Abstract: The reliable operation of high-voltage oil-filled cable terminals is a prerequisite for the stable operation of cable lines, but the traditional diagnosis model for oil-filled cable terminal faults has problems such as low efficiency and poor reliability. In order to accurately judge oil-filled cable terminal faults, this paper proposes a fault diagnosis method for oil filled cable terminals based on the maximum mutual information coefficient (MIC) and the improved Archimedes optimization algorithm (IAOA) to optimize the deep trust network (DBN). Firstly, the MIC theory is used to reduce the dimensionality of the dissolved gas concentration in the silicone oil filling agent for cable terminals and perform feature extraction.Secondly, the optimal feature quantity is taken as the input of the DBN network model, and in view of the difficulty in selecting the hyperparameter of the DBN network, the IAOA is proposed to optimize the hyperparameter of the DBN network model. It is easy for the AOA algorithm to fall into local optimization and weak search ability, thus a variety of improvement strategies are introduced to optimize the optimization performance of the AOA method and improve the optimization ability of the AOA. Finally, the feasibility of the model was verified by constructing experimental platform for simulation of oil filled cable terminal faults, collecting fault sample data of oil filled cable terminals, and creating category sample labels. The example verification shows that the oil filled cable fault diagnosis method proposed in this paper can be adopted to effectively complete fault diagnosis, with an accuracy of 98.33% in the test set. Compared with traditional fault diagnosis models, the proposed method has good stability and high recognition accuracy, which can provide a theoretical basis for guaranteeing the reliable operation of high-voltage oil filled cable terminals.