Abstract: The overheating defect of disconnectors is the most common and potentially dangerous type of defect in electrical equipment. This paper analyzes the heating mechanism of disconnectors. Based on the fundamental principles of the Particle Swarm Optimization-Support Vector Machine algorithm, a predictive model for connector overheating defects using multiple characteristic parameters is proposed. By training the model using a training set and predicting temperatures for defects in the test set, the root mean square error between predicted and actual values was 0.7638 for the training set and 1.5509 for the test set. The predictive accuracy of this model demonstrates its effectiveness in predicting connector overheating defects, providing a theoretical foundation for such predictions.