The paper introduces a simulation/optimization procedure for the assessment and the selection of infrastructure alternatives in a complex water resources system, i.e. in a multisource (reservoirs) multipurpose bulk water supply scheme. An infrastucture alternative is here a vector X of n decision variables describing the candidate expansions/new plants/water transfers etc. Each parameter may take on a discrete number of values, with its own investment cost attached. The procedure uses genetic algorithms for the search of the optimal vector X through operators mimicking the mechanisms of natural selection. For each X, the value of the objective function (O.F.) is assessed via a simulation model. Simulation is necessary as the O.F. contains, besides investment costs, also incremental operation costs and benefits that depend on the incremental water amounts which the alternative can provide. The simulation model transforms a thirty-year hydrologic input at daily/monthly scale in water allocations, accounting for the usual nonnegativity constraints and using some simple, sytem-specific rules aimed at reducing spills and at sharing water deficits among demand centres. Different O.Fs and constraints have been tested, such as incremental financial cost/benefit minimization under various maximum water deficit constraints scenarios or cost/benefit mimization including scarcity costs. This latter approach has the advantage of implicitly allowing for the magnitude of deficits, but requires the assessment of deficit-scarcity cost relationships. The application of the procedure to a water resources system in south-western Sicily shows that the model is able to converge to results that are consistent with the planning options expressed by the selected O.Fs.
Keywords: genetic algorithms, infrastructure optimization, loss functions, water resources systems