Investigation on Deep Learning-Based Temperature Prediction Model for Cable Core of Duct Cable

Accurate prediction of cable core temperature is of utmost importance in the effective operation of power transmission systems, as prolonged overheating of power cables can result in thermal breakdown of insulation and subsequently impact the stability of the power grid. However, measuring cable core temperatures is costly due to the limitations of cable installation environment, while machine learning-based prediction methods exhibit limited accuracy. To address this issue, this paper proposes a deep learning-based cable core temperature prediction model that integrates bidirectional long short-term memory (BiLSTM), bidirectional gate-controlled recurrent unit (BiGRU), and simple attention mechanism (SimAM). First, BiLSTM is utilized to capture the correlation within fault information to reduce information loss. Next, BiGRU is employed for model learning to enhance convergence speed while preventing overfitting. Finally, the utilization of SimAM serves to enhance the capacity of model to concentrate on crucial information, thereby further enhancing its predictive performance and generalization ability. The results show that mean absolute error (MAE) and mean square error (MSE) of the proposed model are 0.1332 °C and 0.0319 °C², respectively.