Hybrid Floating Photovoltaic-Hydropower (HFPVH) systems offer a promising solution for climate-resilient energy generation by integrating hydropower and solar energy to enhance grid reliability. Such systems aim to utilize the water reservoir surface’s natural cooling ability to enhance solar panel efficiency. This article presents research that is part of a larger project aiming to develop a robust optimization framework that combines hydrological modeling, machine learning-based forecasting, and adaptive decision-making to optimize HFPVH system operations under uncertain climate conditions. In this paper, we introduce a novel LSTM architecture, with a dynamic outlier filter, improving the forecasting accuracy of hydrological inflows and energy generation patterns under uncertainty. This proposed LSTM aims to offer flexibility in simulating scenarios under uncertainty, providing a more comprehensive understanding of potential system failures, while enabling enhanced monitoring for more informed decision-making and improved performance tracking over time.