A Two-Layer Optimization Strategy For An Efficient Day-Ahead Generating Scheduling In Multi-Unit, Head-Dependent Hydro Plants

This paper proposes a two-layer mixed-integer linear programming (MILP) strategy to efficiently solve the day-ahead generation scheduling (DAGS) problem in large multi-unit, head-dependent hydro plants subject to environmental and grid connection constraints. The DAGS problem in such cases is computationally challenging due to the nonlinear coupling between head, turbined outflow, and generator efficiency, multiple identical generating units (GUs), and time-coupled environmental constraints. The proposed two-layer approach addresses these challenges by decoupling the problem into two sequential MILPs. In the first layer, the hydro production function (HPF) is approximated by a piecewise-concave model to determine reservoir level trajectories while mitigating symmetry issues among identical GUs. In the second layer, given the reservoir levels from the first stage, a sequence of one-step MILPs refines the dispatch using a more accurate HPF representation. The approach is tested on the Brazilian Jirau hydropower plant (50 units, 3750 MW), revealing that the proposed strategy yields generation schedules equivalent to those of a full nonconvex MILP formulation while drastically reducing computation time. These results demonstrate the potential of hierarchical MILP decomposition to improve the scalability and realism of large-scale hydro day-ahead scheduling models.