Resilient Solar Power Infrastructure for Flood-Prone Kerala: A Machine Learning

Kerala, with its high solar energy potential and recurrent flood risks, faces challenges in establishing sustainable and resilient power infrastructure. This study presents a machine-learning-driven optimization framework for designing and deploying flood-resilient solar power systems across vulnerable, solar-rich constituencies. By integrating geospatial flood data, weather data, and infrastructure vulnerability levels, the model forecasts high-risk areas and optimally distributes solar assets to minimize infrastructure breakdown and flood damage while improving energy generation. A hybrid approach combines Random Forests for risk assessment and Genetic Algorithms for system placement and sizing. Experimental results demonstrate over 92.4% accuracy in energy distribution, outperforming state-of-the-art baseline models by 5–7%. The model achieves the lowest energy loss of 5.1%, compared with over 6.5% in traditional approaches, and improves overall rewards by 13.8%. Under simulated worst-case flood scenarios, it demonstrates 95.3% operational uptime and 93.7% flood-response stability, reducing downtime by 21.5% and energy recovery time by 4.8 seconds. Additionally, its adaptive optimizer achieves average response times below 0.14 seconds, enabling real-time responsiveness. These results highlight the model's performance, robustness, and potential for smart grid deployment in disaster-prone areas.