Control of Nitrate Flow in Pre-Denitrification Systems using Long-Range Identification for Predictive Control

This paper is concerned with long-range adaptive predictive control of the biological nitrogen removal process in a pre-denitrification activated sludge plant. A control-oriented identification procedure is used for modelling a multi-step ahead (long-range) predictor using a reduced model of the nitrate concentration at the plant anoxic zone, in the range required by the predictive controller. The optimal structure of the reduced model and its parameters are identified by an advanced regression orthogonal estimator, based on the Modified Gram-Schmidt algorithm. The proposed controller controls the internal nitrate recirculation flow in order to maintain a desired nitrate concentration set-point at the anoxic zone. Simulation results, using a benchmark model, compare the performance of the proposed adaptive predictive controller against the performance of a well tuned PI controller, showing that the proposed controller presents much better external pro- cess disturbances attenuation and equivalent set-point tracking ability.