Partial Discharge Fault Prediction of Tunnel XLPE Cable Based on Deep Learning

Partial discharge can induce internal insulation degradation in tunnel XLPE cables, generate tree-like discharge channels, accelerate material aging, increase breakdown risk, and threaten the reliability of tunnel power supply. This paper proposes an LSTM-based fault prediction method for partial discharge in tunnel XLPE cables. First, through MATLAB and Simulink simulations as well as on-site measurements, a dataset containing time-domain pulses, discharge magnitude, and discharge phase was constructed. Subsequently, an LSTM-based partial discharge prediction model was developed, and model training methods were explored in conjunction with the characteristics of discharge data. Experiments based on the constructed dataset demonstrated that the proposed method can exploit the advantages of multi-source monitoring data, accurately identify degradation trends, overcome the limitation of traditional methods in insufficiently capturing early-stage partial discharge features, achieve condition assessment and fault early warning, and ensure the safe operation of cables.