Modeling and Advanced Control of an Activated Sludge Waste Water Treatment Facility
With sewage treatment plants accounting for some 60 percent of the phosphorus and nearly 35 percent of nitrogen entering the Chesapeake Bay, reducing these compounds is a primary task of the Bay's 40 percent nutrient reduction goal. Achieving large reductions, however, is expensive–cost estimates in Maryland alone range upwards of $300 million for the needed plant upgrades.
To help minimize capital costs, Thomas McAvoy and Peter Minderman are extending an initial Sea Grant-supported project to develop techniques of neural network and process modeling that will help operators of waste treatment plants optimize activated sludge processes, the biochemical reactions to remove carbon, phosphorus and nitrogen fro incoming waste.
Neural network models employ techniques of artificial intelligence, in which feedback loops take changing data into account to signal adjustments in mechanical processes, such as controlling oxygen-demand activated sludge processes. By using neural network approaches for the design of models that are specific to each waste treatment plant, this project should demonstrate techniques that can:
Thomas J. McAvoy
Peter A. Minderman, Jr.
Department of Chemical Engineering
University of Maryland