Computational improvement for dynamic programming models by diagnosing infeasible storage combinations

A dynamic programming (DP) optimization model is developed for long term planning of multiple-reservoir operations. To overcome the well-known dimensionality problem associated with such a model, a heuristic approach is used to narrow the needed search algorithm within the state space of the DP model. This method can recognize many infeasible transitions from the initial to the final state of the DP stages. By diagnosing these infeasible transitions in advance and removing them from further computations, significant improvement in computational load is achieved so that the computer time for solving the model is reduced more than 50 times for the reservoir system under study. This methodology is applied to a four-reservoir system located in Iran.