Multi-objective short-term hydropower generation operation for cascade reservoirs and stochastic decision making under multiple uncertainties

Abstract

In real-world, the multi-objective short-term hydropower generation operation (STHGO) faces changing environment and challenges derived from inherent uncertainties in streamflow forecast, multi-objective optimization and decision-making processes. For decision makers (DMs), these multiple uncertainties always bring harmful effect and risk on optimal STHGO decision making. Traditionally, studies concerning to STHGO are usually based on deterministic streamflow forecast, optimization and decision making without considering nonnegligible uncertain information. Therefore, in this work, a unified framework is developed for solving multi-objective STHGO under multiple uncertainties and quantifying risk information propagated between each process. First, the modified martingale model of forecast evolution (MMMFE) is used to disclose forecast uncertainty and improvements evolution and generate simulated reservoir inflow scenarios based on historical streamflow. Then, improved multi-objective particle swarm optimization (IMOPSO) is developed to obtain optimal Pareto solutions. Pareto front uncertainty is analyzed by importing synthetic inflow scenarios and risk information is quantified through risk criteria. Meanwhile, novel stochastic multicriteria acceptability analysis (SMAA) model coupling with TOPSIS and grey correlation analysis (GCA) is proposed to solve stochastic decision making for STHGO under uncertain environment. The uncertainty propagated from statistically distributed Pareto non-dominated solutions to stochastic decision making is quantified by risk of errors in decision making process (REDMP). Finally, we demonstrate proposed methodology with case study in Qingjiang cascade hydropower stations. Simulation results effectively reveal risk transmission process and quantify risk information. The effects caused by various levels of inflow forecast uncertainty are disclosed on stochastic decision making for STHGO. Comparison analysis results demonstrate the superiority of novel SMAA model and verify it an efficient tool for showing more convincible solution and measurable risk information to DMs.