Improving the wastewater treatment plant performance through model predictive control strategies

Using the Benchmark Simulation Model No. 1 as virtual plant, the development of model based control strategies for an activated sludge process was addressed in this work. The dynamic matrix control algorithm was used to obtain the optimal control of ammonia and nitrate concentration by using dissolved oxygen concentrations in the bioreactor, and internal recycle flow rate as manipulated variables. The main goal of the proposed control strategies was the minimization of aeration and pumping energy consumption by guaranteeing good nitrogen removal efficiency. In order to mimic a more realistic situation, process model identification was carried out considering time varying inputs. Recurrent neural network were used to describe the required input-output relationships. Results showed that ammonia and nitrogen removal was enhanced even in the coldest season, with a reduction of energy consumption if compared with BSM1 default control strategy.