New approaches to screening infrastructure investments in multi-reservoir systems– Evaluating proposed dams in Ethiopia and Kenya

Abstract:

Two approaches have been used traditionally to screen infrastructure investments in multi-reservoir systems: scenario analysis of a few simulated designs and deterministic optimization, sometimes using hydro-economic models or screening optimization models. Simulation models realistically represent proposed water systems and can easily include multiple performance metrics; however each prospective system operating rules need to be formulated and simulated for each proposed design (time consuming. Optimization models have been used to overcome this burden. Screening optimization models use integer or non-linear programming and can be challenging to apply to large and/or multi-objective systems. Hydro-economic models that use deterministic (implicit stochastic) optimization must be modified to examine each different plan and they cannot always reproduce realistic or politically acceptable system operations.

In this presentation we demonstrate the application of a new screening approach to multi-reservoir systems where operating rules and new assets (dams) are simultaneously optimized in a multi-criteria context. Results are not least cost investment plans that satisfy reliability or other engineering constraints, but rather Pareto-optimal sets of asset portfolios that work well under historical and/or future scenarios. This is achieved by using stakeholder-built simulation models linked to multi-criteria search algorithms (e.g. many objective evolutionary algorithms, MOEA). Typical output is demonstrated through two case-studies on the Tana and Blue Nile rivers where operating rules and reservoir assets are efficiently screened together considering stakeholder-defined metrics. The focus on the Tana system is how reservoir operating rules and new irrigation schemes should be co-managed to limit ecological damages. On the Nile system, we identify Blue Nile river reservoir capacities that least negatively impact downstream Nile nations. Limitations and new directions of research in the application of search-based system planning approaches are surveyed. Current limitations include the need to integrate investment scheduling (in addition to suggesting optimal asset bundles) and adaptive planning requirements real stakeholders have into the planning problem formulation.