We are interested in the management of a cascade of hydroelectric dams. This problem can be solved by the straightforward Dynamic Programming (DP) approach when the cascade is small. For problems involving a large number of dams, one may use decomposition approaches to overcome the curse of dimensionality. We implement the Dual Approximate Dynamic Programming (DADP) method, that is, an approximated decomposition scheme based on price coordination. We first present exhaustive results on a three dams problem, and then apply the DADP method to real-life examples which are intractable using the standard DP method.
Mots clés : Dynamic Programming, Decomposition methods, DADP, Multistage stochastic optimization, Energy management