Model-free adaptive predictive control for cell culture: Decoupling environmental dynamics and compensating disturbances

Bioreactor control systems, particularly the precise regulation systems of dissolved oxygen and pH, are facing challenges of coupled environmental variables, dynamic parameter variations, and external disturbances, To address such issue，this study proposes a dynamic disturbance-compensation model-free adaptive predictive control method. Firstly, a dynamic linearization model is established at each operating point utilizing only I/O data, then the method is formulated,which virtually equivalent to an ideal I/O controller via adaptive, time-varying control gains. Additionally, online adaptive learning mechanisms are designed to adjust the control gains based on the error differential characteristics of DO and pH. Lyapunov analysis guarantees uniformly ultimately bounded tracking error. Simulation and experimental results validate the method's superior disturbance rejection, significantly enhancing the stability and controllability of bioreactor systems.