Subspace-based optimal control of N-removal activated sludge plants

In this paper, a multivariable feedforward/statefeedback Linear-Quadratic (LQ) + (1) controller is designed and used to control the NOx - N concentration (nitrate-nitrogen plus nitrite-nitrogen) in both anoxic and aerobic zones of a pre-denitrifying activated sludge process. The controller design is based on a discrete state-space model of the process, and it incorporates integral error and anti-reset windup. The state-space model was obtained using a subspace-based identification technique, from input/output data. The "artificial" states generated by the subspace algorithm are estimated on-line through a Kalman filter (KF) and coupled with the LQ-controller. This technique is evaluated through simulations using the ASWWTP-USP benchmark and the results show the potential efficiency of the proposed approach.