Feasibility Improved Stochastic Dynamic Programming for Optimization of Reservoir Operation

Stochastic Dynamic Programming (SDP) is a major method for optimizing reservoir operation. Handling non-linear, non-convex and non-differentiable objective functions and constraints are some advantages of SDP. Besides the mentioned advantages, this method suffers drawbacks like infeasibility. Infeasibility occurs in SDP naturally because of discretization process and random combination of state variable values. The main idea in this paper is to mitigate the infeasibility drawback in SDP by properly adjusting the reservoir volume interval indices. These indices are determined based on the continuity equation and meeting the monthly minimum release so that some infeasible policies are transformed to feasible ones. Simulation of reservoir operation for 60-year planning period indicates an improvement up to 30% in objective function by employing the proposed technique in a case study.