TinyML powered UAV inspection for solar panel monitoring with optimized path planning

Renewable energy is a vital element of sustainable power solutions, with solar panels playing a crucial role. Performance and maintenance of solar infrastructure are increasingly essential as the adoption of solar energy grows worldwide with government incentives and large-scale installations. However, traditional UAV-based inspection methods are limited by their adaptability, computational efficiency, and real-time processing ability making the process labor-intensive and very time-consuming. To overcome these problems, a new UAV-powered framework that combines lightweight TinyML algorithms that are tuned for path planning was proposed for cheap and efficient solar panel monitoring and damage detection. A* and RRT* based conventional path planning algorithms do not adapt dynamically to environmental changes, producing suboptimal trajectories, while consuming more energy. Like traditional detection models, they also require significant computational resources and are not feasible for real-time UAV operation or large-scale deployment. The proposed framework uses A* for initial path planning and iteratively improves the result using Simulated Annealing to achieve good energy efficiency and obstacle avoidance. The damage classification model can be predicted on a TinyML platform and classified into damaged and undamaged panels. Real-time geospatial data integration enables the extraction of actionable insights and enhances the adaptability of the framework during dynamic inspection scenarios. Experimental evaluations show that energy consumption is reduced by 24%, and spatial coverage is improved by 31% compared to the conventional techniques. The TinyML classification model for damage detection gives an F1 score of 96.8% while remaining computationally minimal and suitable for UAV deployment. In that sense, this framework presents a transformative solution and it bridges the gap between the UAV deployment challenge and precision solar panel monitoring in the real world.