A Surrogate Based Optimization Approach for the Development of Uncertainty-Aware Reservoir Operational Rules: the Case of Nestos Hydrosystem

Operation of large-scale hydropower reservoirs is a complex problem that involves conflicting objectives, such as hydropower generation and water supply. Deriving optimal operational rules is a challenging task due to the non-linearity of the system dynamics and the uncertainty of future inflows and water demands. A common approach to derive optimal control policies is to couple simulation models with optimization algorithms. This paper in order to investigate the performance of a future reservoir and safely infer about its significance employs stochastic simulation, thus long synthetically generated time-series and a multi-objective version of the Parameterization-Simulation-Optimization (PSO) framework to develop uncertainty-aware operational rules. Furthermore, in order to handle the high computational effort that ensues from that coupling we investigate the potential of a surrogate-based multi-objective optimization algorithm, ParEGO. The PSO framework is deployed with WEAP21 water resources management model as simulation engine and MATLAB for the implementation of optimization algorithms. A comparison between NSGAII and ParEGO optimization algorithms is performed to assess the effectiveness of the proposed algorithm. The aforementioned comparison showed that ParEGO provides efficient approximations of the Pareto front while reducing the computational effort required. Finally, the potential benefit and the significance of the future reservoir is underlined.