Design of a generalized predictive controller for a biological wastewater treatment plant M. Sadeghassadi, C. J. B. Macnab and D. Westwick ABSTRACT This paper presents a generalized predictive control (GPC) technique to regulate the activated sludge process found in a bioreactor used in wastewater treatment. The control strategy can track dissolved oxygen setpoint changes quickly, adapting to the system uncertainties and disturbances. Tests occur on an Activated Sludge Model No. 1 benchmark of an activated sludge process. A T filter added to the GPC framework results in an effective control strategy in the presence of coloured measurement noise. This work also suggests how a constraint on the measured variable can be added as a penalty term to the GPC framework which leads to improved control of the dissolved oxygen concentration in the presence of dynamic input disturbance. M. Sadeghassadi (corresponding author) C. J. B. Macnab D. Westwick University of Calgary, 2500 University Dr NW Calgary, AB T2N 1N4, Canada E-mail: msadegha@ucalgary.ca Key words | Benchmark Simulation Model No. 1, biological wastewater treatment process, model predictive control, recursive least squares algorithm INTRODUCTION The control of a biological wastewater treatment process provides a challenging task, due to the nonlinear dynamics, high number of parameters and uncertainties, and multiple operating points. The treatment of wastewater consists of three major parts: removing the nitrogen, removing the phosphorus, and removing the organic substrate. This paper concerns the removal of organic substrate, which is accomplished by regulating the level of biomass in the acti- vated sludge. Oxidation provides the energy that is required for growth and division of the biomass. Therefore, the key idea in biological wastewater treatment becomes the control of the dissolved oxygen concentration. Researchers often use the Benchmark Simulation Model No. 1 (BSM1) for a model. This is a well-established model published by the International Water Association (IWA), Alex et al. (), Copp () and Jeppsson & Pons (). In this paper we use the benchmark parameters (stoi- chiometric parameters, kinetic parameters and mass balances) for the tanks as directly given from benchmark manual description without any change. This gives a fair comparison between the results of our new control strat- egies and the proportional-integral (PI) controller results published by IWA. This model consists of a MATLAB model of five tanks in series. Three aerobic tanks follow two anoxic tanks. In the fifth tank a control system regulates dissolved oxygen by manipulating the oxygen coefficient. Some researchers have used the model predictive control (MPC) method for control of the BSM1, including Holenda et al. (), Zhe et al. (), Francisco et al. (), Han et al. () and Yang et al. (). The strategies of dynamic matrix control, quadratic dynamic matrix control and non- linear MPC with feed-forward compensation appear in Shen et al. (). The closed-loop performance in the pres- ence of large disturbances, both with and without feedforward compensation, has been compared. As expected, the control performance improves with feedfor- ward compensation. Some researchers have used fuzzy/neural network models with MPC. For example, the method in Yang et al. () utilized a fuzzy model of the activated sludge waste- water treatment process; a fuzzy model made a suitable basis for designing a nonlinear model predictive controller. A fuzzy c-means algorithm clustered the space of input vari- ables. Then, the least squares method identified the consequent parameters. Finally, a solution of a convex optimization problem resulted in the control signal. In Han et al. (), a self-organizing radial basis function neural network MPC method regulated the BSM1 to main- tain the oxygen concentration. A nonlinear constrained 1986 © IWA Publishing 2016 Water Science & Technology | 73.8 | 2016 doi: 10.2166/wst.2016.050 Downloaded from https://iwaponline.com/wst/article-pdf/73/8/1986/462653/wst073081986.pdf by guest on 19 July 2020