Hydropower Portfolios Management via Markov Decision Process

This paper presents a methodology of hydropower portfolios management with hydropower producers bidding in the regional electricity market (REM) of China. Initially, the problem is constructing as two level optimization problems by time scales. At upper level with contract market, this level use the Markov decision process to establish the hydropower contracts trading. At the lower level in day-ahead market, an optimal generation management model under the risks of forecasted water inflow and electricity prices uncertainties is established to modify the upper level generation strategies. In this two models, suppose that the state-space and strategy-space of each level are non-overlapping, the performance produced by the lower level decisions will affect the upper level decisions. Duel stochastic dynamic programming is employed in the power generation process of each hydropower plants so that the maximization expected revenue can be reached. The real case studies of three Gorges cascade station participating in the middle-China REM shown that the presented methods have more satisfied hydropower management results