Nonlinear metabolic model-based control of ethanol in Escherichia Coli

A 34 state cybernetic model consisting of 15 reaction is developed for the anaerobic growth of E. coli. The cybernetic model is used in the model predictive control (MPC) framework for set point tracking of ethanol concentrations in a continuous mode of operation of the bioreactor by manipulating the dilution rate. A linear MPC algorithm that includes a Kalman estimator with partial state measurements is developed. Two nonlinear MPC algorithms are also developed and both use the extended Kalman estimator for state update. The first algorithm uses an instantaneous linear model for prediction and optimization whereas the second algorithm uses the nonlinear model which significantly increases the computational load. The performance of these algorithms is tested in a highly nonlinear operating regime and in the presence of plant-model mismatch, input and measurement noise and variable constraints. Results show that performance of the nonlinear MPC based algorithms is superior to that of the linear MPC based algorithm. There are insignificant differences in the performance of the two nonlinear MPC algorithms thus enabling higher performance achievement without unreasonable computational loads.