Predictive Maintenance Framework for Solar Inverters and Smart Grid Assets Using Edge AI and Advanced Fault Analytics

Solar photovoltaic systems are increasingly connected to smart grids, making equipment reliability a major concern. Failures in solar inverters and grid connected components reduce energy output and increase operational cost. Most existing predictive maintenance studies focus either on PV systems or on smart grid assets separately and rely mainly on centralized cloud processing. This paper proposes a unified predictive maintenance framework that integrates solar inverter and smart grid monitoring within an edge-based architecture. Electrical and thermal signals are processed locally, where time and frequency domain features are extracted and analyzed using a CNN–LSTM model for real time fault classification. A health index model is applied to estimate remaining useful life for condition-based maintenance planning. Experimental results show 96.8% classification accuracy and a reduction in inference latency from 85 ms in cloud-based processing to 18 ms at the edge. The proposed framework reduces communication load, supports faster decision making, and improves operational stability in distributed renewable energy systems.