Primal and dual stochastic dynamic programming in long term hydrothermal scheduling

This work presents a comparative study between primal and dual stochastic dynamic programming in the hydrothermal scheduling problem. The comparison is made by simulation using historical inflows records of Brazilian hydroelectric system, to the specific case of systems comprising a single hydro plant. The stochastic variable of the system is modeled by a lag-one parametric autoregressive model. In the case of the dual approach, a parametric autoregressive model of superior order is also considered. Results have demonstrated that the performance of the primal and dual stochastic dynamic programming is not similar.