Estimation of Wastewater Treatment Plant State For Model Predictive Control of N-premoval at Medium Time Scale

Abstract

State estimates are needed for optimising model predictive control of nitrogen and phosphorous removal in a wastewater treatment plant due to limited state measurements available. The MPC optimiser to implement an information feedback from the plant uses the estimates. Parameters of a grey box model used by MPC for the output prediction purposes need to be updated as well. The state estimates are then used, as pseudo measurements of the states by the parameter estimation algorithm. Otherwise the joint state and parameter estimation does not provide needed accuracy due to limited measurement programme. The paper applies extended Kaiman filter to the plant model that is based on ASM2d model of the biological reactor. The estimator is tested by simulation on a benchmark producing encouraging results.