Machine Learning‐Based Health Index Prediction for XLPE Power Cables: Enhancing Reliability and Cost‐Effective Maintenance

ABSTRACT The reliability of power cables is essential for ensuring efficient and uninterrupted power distribution. Insulation degradation caused by thermal aging, moisture ingress, partial discharge (PD), and natural corrosion significantly impacts cable performance and lifespan. Accurate Health Index (HI) estimation is crucial for maintenance planning, failure prevention, and asset management. However, traditional HI assessment methods, which rely on scoring and weighting techniques, often suffer from data availability and subjective evaluations, leading to reduced accuracy. In contrast, this study introduces advanced machine learning (ML)‐based models for predicting the HI of Cross‐Linked Polyethylene cables using both classification and regression models via Python. Real‐time key diagnostic parameters, including Insulation Resistance, Tan Delta, Loading Conditions, and PD activity, were incorporated to enhance model accuracy. To address missing data challenges, this study employs data imputation strategies, ensuring reliable HI predictions even with incomplete data sets. On the basis of the predictive accuracy, a cost‐saving analysis is conducted to quantify the financial benefits of ML‐driven predictive maintenance, highlighting its potential to optimize maintenance costs. Among the eight regression and classification models, Random Forest Regression achieved the highest accuracy of 99.01%. At the same time, Random Forest Classification reached up to 97.2%, demonstrating the effectiveness of ML in HI assessment. The results emphasize that data‐driven HI assessment enhances predictive decision‐making for power cable systems, offering a superior alternative to conventional approaches.