AI-Enabled Predictive Maintenance and Asset-Health Analytics for MV/LV Distribution Substations Using Commissioning-Test Data and Condition Monitoring

Medium Voltage and Low Voltage (MV/LV) distribution substations are critical components of modern power distribution networks, directly influencing system reliability, power quality, and operational continuity. Traditional maintenance strategies for these substations largely rely on periodic inspections and reactive fault management, which often fail to detect early stage degradation and may lead to unplanned outages and increased operational costs. This paper proposes an AI enabled predictive maintenance and asset health analytics framework that leverages historical commissioning test data combined with real time condition monitoring information. Machine learning models are employed to analyze insulation resistance, contact resistance, thermal profiles, partial discharge indicators, and operational parameters to predict asset degradation trends and remaining useful life. The proposed approach enables data driven maintenance decisions, early fault detection, and optimized asset lifecycle management. Experimental evaluation demonstrates that the framework improves fault prediction accuracy and reduces maintenance lead time compared to conventional rule based methods. The results indicate that integrating artificial intelligence with substation condition data can significantly enhance reliability, reduce downtime, and support the transition toward intelligent and resilient distribution networks.