Multiobjective stochastic programming with recourses for real-time flood water conservation of a multireservoir system under uncertain forecasts

• Published in: Journal of hydrology (Amsterdam) 2020-11, Vol.590, p.125513, Article 125513
• Main Authors: Xu, Bin, Zhong, Ping-an, Lu, Qingwen, Zhu, Feilin, Huang, Xin, Ma, Yufei, Fu, Jisi
• Format: Article
• Language: English
• Subjects: Flood water conservation; Forecast uncertainty; Multiobjective optimization; Reservoir operation; Stochastic programming with recourses
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2020.125513

•Proposed uncertain correlated scenarios generation method by copula function.•Developed a multiobjective stochastic programming model for real-time flood conservation.•Explored risk-robust noninferior strategies by resolving conflicts of multi-risks. Flood water conservation realized through real-time multireservoir operations is effective in mitigating water scarcity. Owing to the influence of real-time inflow forecast uncertainty, determining an informed operation plan necessitates resolution of the conflict between upstream flood risk, downstream flood risk, and water scarcity risk. This study developed a multiobjective stochastic programming with recourses (MOSP) model to seek robust risk-averse plans under multiple risks. In the proposed approach, inflow forecast errors are modeled and sampled as spatially and temporally correlated stochastic processes using a copula function. The multiobjective stochastic programming is solved using the epsilon constraint method based on explicit formulation of multiple risk objectives with discretized inflow scenarios, and a rolling horizon with recourses for real-time operational updating is modeled to address the dynamic decision-making characteristics. The proposed methodologies were applied to the multireservoir system of the Pi River Basin (China), and comparisons with model results of a traditional coupled simulation and a deterministic single objective optimization model were conducted. Results indicated that capturing spatial and temporal dependencies of forecast uncertainty provides informative forecast support. Moreover, noninferior solutions obtained from MOSP dominate the deterministic optimal solution, which could either conserve 4.86 million m3 (12%) more water without increase in flood risk, or reduce the upstream flood risk by 3.1% without increase in either water scarcity or downstream flood risk during a typical receding flood. The proposed methodology provides a dynamic decision-making modeling tool that could reduce overall risk and determine a compromise for flood water conservation under uncertainty.