Research on Hot Spot Temperature Inversion Method of Cable Intermediate Joints Based on Finite Element and Optimized BP Neural Network Method

Abstract

The cable intermediate joint plays a crucial role in power cable systems, as its temperature directly impacts insulation performance and longevity. Predicting temperature accurately poses challenges for operation and maintenance. This study introduces a model for electromagnetic-thermal coupling of a 110 kV single-core high voltage cable, enabling numerical simulation to determine temperature distributions within the cable body and middle joint. By employing a hybrid orthogonal design approach, training and test samples are generated from simulated temperature field data. Conductor current, ambient temperature, convective heat transfer coefficient, and insulation thermal conductivity coefficient of the intermediate joint are chosen as variables to compile the dataset. An inverse model-based prediction method is developed using a firefly-optimized BP neural network algorithm. Results demonstrate that the optimized model exhibits a correlation coefficient of 0.99, surpassing the prediction accuracy of traditional optimized BP neural networks. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.