| Abstract [eng] |
This thesis is about investigation of adaptive pH control of fed-batch microbial cultivation process. A nonlinear MATLAB/Simulink process model was implemented, with biomass concentration, hydrogen-ion concentration and reactor volume as state variables. Using this model, a PI controller and gain-scheduling adaptation laws were implemented to enhance the control performance in time-varying operating conditions. The following controller types were compared: fixed PI controller, gain-scheduled controller with measured reactor volume and substrate feed flow, gain-scheduled controller with measured substrate feed flow and constant average reactor volume, and gain-scheduled controller with measured reactor volume and estimated substrate feed flow. Simulation results indicated that gain-scheduled controllers outperformed fixed PI controllers. The gain-scheduled controllers had lower tracking error and better disturbance rejection. The controller that used measured reactor volume and substrate feed flow had the smoothest response with the lowest overshoot, while the controller that used measured reactor volume and estimated substrate feed flow had the lowest tracking error. Simulations with Gaussian measurement noise added to the measurements demonstrated that the gain-scheduled controllers were still better at low and medium noise levels, but at the highest noise level there were no significant differences between the controllers because the response was dominated by the measurement noise. In conclusion, the results indicate that gain-scheduled PI control is an effective approach to pH control for fed-batch microbial cultivation. |
