Model-Free Predictive Control for Nonlinear Processes with Variable Delay

This paper presents a Model-Free Predictive Control (MFPC) strategy for regulating a nonlinear mixing tank process with time-varying delay. Unlike traditional Model Predictive Control (MPC), MFPC eliminates the need for an explicit system model by leveraging an ultra-local model and real-time optimization. The approach dynamically compensates for disturbances, delays, and setpoint changes while adhering to operational constraints. Simulation results demonstrate that MFPC outperforms conventional PID and intelligent PID (iPID) controllers, achieving superior performance metrics such as reduced overshoot, smoother control actions, and lower RMSE, IAE, and ITAE values. Notably, MFPC avoids the oscillatory behavior and actuator stress observed in PID-based methods, proving its effectiveness for nonlinear processes with variable delay.