Neural Net Modeling and Control of a Municipal Waste Water Process

One municipal facility is beginning to consider the benefits of using model predictive control as a means of improving product quality and reducing energy costs. To date, the initial steps of this project have been completed. The first step was to upgrade the basic control and data acquisition systems. The second step was to collect experimental data in order to build a process model. The third step was to build this model; a dynamic nonlinear finite impulse response model was constructed using the neural network partial least squares algorithm. This model has been used to analyze the steady state behavior of the plant and this analysis has helped identify an improved strategy which lowers annual operating costs. The implementation of these ideas awaits the completion of a process retrofit. After this expansion, the modified process will be remodeled, and the suggested control strategy will be experimentally verified.