Use of genetic algorithms in bounded search for design of biological nitrification/denitrification waste treatment systems

This thesis presents an innovative approach to the design of biologically-mediated wastewater treatment systems. The design of a single-sludge, nitrification/denitrification system is used in the development of this approach. The bioreactor system is composed of a plug-flow deoxygenation reactor, an anoxic continuously-stirred tank reactor (CSTR) reactor, and an aerobic CSTR operated in series. Various components and operational settings are considered as design elements. The design space is formed by upper and lower bounds for each design element, accompanied with an interval size by which the range between the upper and lower bounds is divided. International Association on Water Quality biosystem Model 1 is used to predict concentration profiles of eleven wastewater partition components. Design alternatives are evaluated under static and dynamic influent conditions and with variability in kinetic and control parameters. This degree of testing was included to enhance the robustness of solutions developed. Design points are evaluated by their ability to achieve design goals. A genetic algorithm directs the search into regions of the design space where alternatives of increasing performance may be located. Test problems are analyzed for maximization of flow through an existing system, and for minimization of reactor volume for a fixed flow. There is evidence of improvement of the design points as the algorithm proceeds and that the tool provides a method for generating high valued alternative solutions. Solution movement is toward limiting values for key operational considerations.