PARTIAL INFEASIBILITY METHOD FOR CHANCE-CONSTRAINED AQUIFER MANAGEMENT

A pump-and-treat system of extraction wells is modeled to create a hyd raulic capture zone to immobilize a plume of contaminants in a confine d aquifer with uncertain hydraulic transmissivity. The stochastic appr oach is used to model the transmissivity as a random field and to gene rate random realizations of the field. The linear-response-matrix meth od is used to address a joint chance-constrained optimization problem in which constraints on velocities must be simultaneously satisfied wi th high probability. In the literature, the multiple-realization appro ach provides a robust solution but does not allow for prespecification of a desired reliability. We develop a partial-infeasibility method, in which the pumping strategy is obtained through heuristic methods th at require the designed reliability level to be satisfied for a traini ng set consisting of multiple realizations of the uncertain transmissi vity. Empirical Monte Carlo testing shows that as the size of the trai ning set increases, the actual reliability approaches the design crite rion. Thus, the method combines the intuitive appeal of the multiple-r ealization method with the ability to prescribe a desired reliability level.