A bilevel modeling framework to analyze the institutional gap between research and operation practices – Case on the Three Gorges Reservoir pre-impoundment problem

Abstract

Dealing with institutional settings involving multiple stakeholders is essential for improving the applicability of reservoir operation optimization models. However, a centralized decision maker (CDM) is conventionally assumed to represent multiple players with perfect cooperation and information exchange for modeling the decision-making process with hierarchical structure and non-cooperative interactions in real problems. In this paper, we define the deviation of operation performance between the assumed centralized and the practical hierarchical decision-making procedures as an institutional gap (IG). To further derive the causes of the IG and evaluate the influence of the CDM’s preference on the IG, a comparative analysis between bilevel programming and multi-objective programming (MOP) is conducted for the reservoir pre-impoundment problem (RPP). Compared to optimizing two conflicting objectives in MOP, our proposed bilevel optimization framework models the flood control rule maker and reservoir refill rule maker as minimizing the potential flood risk and maximizing the hydropower production benefit respectively with a hierarchical institutional structure. Our case study of the Three Gorges Reservoir (TGR) in China indicates that the centralized optimization model without considering the hierarchical interaction between flood control and reservoir refill decision-makers would lead to an IG in terms of underestimation of flood risk. We further find that the magnitude of the IG is highly related to the selection of the CDM’s preference, and the largest IG is obtained when the flood control or hydropower generation objective is solely considered. Particularly, we specify the range of preference setting which results in the domination of centralized model solutions as compared to the bilevel solution, although the solutions within this range are not realistic for practical operation. Our findings provide an insight into the cause of the IG as well as the effects of preference on the IG, facilitating future improvement of institutional structure design and preference setting for practical reservoir operation optimizations.