Learning-based hierarchical control of water reservoir systems

Abstract: The optimal control of a water reservoir systems represents a challenging problem, due to uncertain hydrologic inputs and the need to adapt to changing environment and varying control objectives. In this work, we propose a real-time learning-based control strategy based on a hierarchical predictive control architecture. Two control loops are implemented: the inner loop is aimed to make the overall dynamics similar to an assigned linear through data-driven control design, then the outer economic model-predictive controller compensates for model mismatches, enforces suitable constraints, and boosts the tracking performance. The effectiveness of the proposed approach as compared to traditional dynamic programming strategies is illustrated on an accurate simulator of the Hoa Binh reservoir in Vietnam. Results show that the proposed approach performs better than the one based on stochastic dynamic programming.