Optimal operation of soil aquifer treatment systems under uncertainty

The objective of this research is to develop a new methodology for SAT (Soil Aquifer Treatment) operation under uncertainties. The optimal operation of SAT systems is formulated as a discrete-time optimal control problem and solved by interfacing the SALQR (Successive Approximation Linear Quadratic Regulator) optimizer with the MSTS (Multiphase Subsurface Transport Simulator) simulation model. Both deterministic and stochastic formulations have been solved. A chance constrained formulation of the optimization is utilized to account for the uncertainties of water content as a result of the variation of parameter uncertainties. The Monte Carlo method is used to explicitly deal with the influence of the spatial variability of aquifer properties on the SAT operation management. This approach enables one to quantify the uncertainties in the SAT system and automatically account for them in the decision-making process through the SAT management model. The methodology provides a valuable tool for the decision maker to quantitatively analyze and determine SAT operation under uncertainties. The research are divided into two phases.; Phase I was the development of a deterministic SAT optimal operation model by interfacing SALQR (Successive Approximation Linear Quadratic Regulator) with the MSTS model (Multiphase Subsurface Transport Simulator). The major objective was to develop a computer program for determining the optimal SAT operation schedule for the purpose of maximizing infiltration. In this model, both water content and infiltration volume are expressed as the function of loading schedule through the modified MSTS unsaturated flow model. The model is solved by an SALQR optimizer. In this step some numerical experiments and tests of the MSTS model are performed.; Phase II was the development of an SAT optimal management model under uncertainties by utilization of some stochastic optimization techniques. In order to quantify the uncertainties in the SAT operation management model the stochastic formulations using chance-constrained and Monte Carlo methods were used respectively to explicitly quantify parameter uncertainties and spatial variability of aquifer properties. Through the use of first-order analysis of uncertainty of the simulator, MSTS model, the uncertainties of the unsaturated hydraulic conductivity and water content due to the uncertainties of the simulation parameters were evaluated. The turning band method has been used to generate two-dimensional hydraulic conductivity fields for the input of the Monte Carlo model. The results from the Monte Carlo model were used to characterize the probability distribution of the optimal loading schedule and then these data were used to analyze the effects of heterogeneity of the soil on the operation of the SAT process.