Early Fire-Warning Strategy for High Voltage Cables Based on Pyrolysis Characteristics of PE Outer Sheath at Different Temperatures

With the widespread application of high-voltage power cables in urban power grids, the fire risk associated with these cables poses an increasing threat to the safe operation of power systems. Pyrolysis gases emitted from the cable’s outer sheath serve as critical indicators of early-stage thermal anomalies, acting as diagnostic signals rather than direct ignition sources. To investigate the gas generation characteristics for fault diagnosis and to develop a robust early warning method, this study conducted pyrolysis experiments on polyethylene (PE) outer sheaths across a temperature range of 90–250 °C. An optimized Random Forest (RF) early-warning model was established, underpinned by multi-dimensional feature engineering and the Synthetic Minority Over-sampling Technique (SMOTE) to address data imbalance. Experimental results reveal that the pyrolysis products primarily consist of alkanes, alkenes, alkynes, CO, and CO2, exhibiting distinct non-linear trends with increasing temperature. Based on these characteristics, the cable condition was categorized into three warning levels: Attention, Alert, and Alarm. The proposed RF-based model achieved a mean stratified 5-fold cross-validation accuracy of 92.28%, with an independent test accuracy of 93.75%, demonstrating superior stability and generalization capability compared with Support Vector Machine (SVM) and Logistic Regression (LR) classifiers. This study provides a reliable, data-driven strategy for identifying incipient thermal faults and preventing fire accidents in high-voltage power cables.