Author(s): Duan Chen; Xinlong Deng; Yufang Ni

Linked Author(s): Duan Chen

Keywords: Reservoir Operation; Inflow Ensemble Forecast; Robust Optimization

Abstract: Ensemble Streamflow Forecasting (ESF) has significantly enhanced the flexibility of streamflow predictions for reservoir operations compared to single streamflow forecasts. However, the uncertainty inherent in ESF remains underexplored, particularly regarding its quantification, propagation, and control within the context of reservoir optimization. This study proposes an innovative optimization model that integrates the concept of Robust Optimization (RO) with the Karhunen-Loeve (KL) expansion method to address these challenges. In contrast to traditional stochastic optimization, which typically assumes predefined probability distributions, the proposed model utilizes an uncertain set to characterize the ESF uncertainty. The KL expansion method is employed to quantify the uncertainty, preserving the statistical structure of the ESF through a combination of eigenvalues and eigenfunctions. This uncertainty quantification is then incorporated into a multi-objective optimization model, where the uncertainty set is updated at each iteration. The model seeks to determine the optimal reservoir outflow that maximizes the operational objective while minimizing variance across multiple potential inflow scenarios. The effectiveness of the proposed framework is demonstrated through a case study of reservoir operations in the Liuxihe Basin, China, and is compared to a benchmark Monte Carlo simulation. The results indicate that the proposed model provides a viable and computationally efficient approach to optimizing reservoir operations under ESF uncertainty, offering both a straightforward framework and high computational tractability.

DOI: https://doi.org/10.64697/978-90-835589-7-4_41WC-P2113-cd

Year: 2025