Application of Stochastic Dual Dynamic Programming Using Markovian Stochastic Modeling for Formulating Long-term Operation Policies of Hydropower Systems

In the last decades, various methods have been proposed for stochastic operation optimization of reservoirs out pf which one is Stochastic Dual Dynamic Programming (SDDP). Reviewing the literature shows in most previous studies, inflow uncertainties have been mostly modeled by auto-regressive AR(1) or seasonal aut-regressive PAR(2) models when using SDDP. Another approach toward inflow uncertainty modeling by SDDP uses Markov Chain which has received less attention. In this study, MSDDP model has been developed which uses Markov chain concept. Furthermore, applicability of SDDP in developing long-term operation policies has not been tested before. In this study, after presenting an algorithm for using SDDP and MSDDP for formulating long-term operation policies, their performances have been compared in weekly operation optimizaition of Karoun multi-reservoir hydropower system. The benefits estimated for operation of the system based on MSDDP policies has shown 35 percent improvement compared with SDDP results. The results have shown absolute superiority of MSDDP in formulating long-term operation policies. Furthermore, operation policies obtained from SDDP have shown highest compatibility with the best possible operation policy when there is no uncertainty (deterministic conditions).