A decomposition approach for stochastic dynamic programming models of reservoir networks

Stochastic dynamic programming is often used to model reservoir control problems. However, in most practical cases, these models are computationally intractable due to the high dimension of the state and action spaces. This paper presents an approach which decomposes the problem into low-dimensional subproblems, each concentrating on one reservoir in the network. The approach works well for certain types of reservoir network under the objective of maximizing expected value. The paper examines how the approach can be adapted for different networks and objectives which take more account of risk.