Data-driven anti-windup compensator synthesis for unknown linear systems with time delay

Delay and actuator saturation are inevitable in practical control engineering, which may lead to system performance degradation, or even divergence of controlled variables. Combination of Smith predictor (SP) and model recovery anti-windup method is an effective solution to anti-windup synthesis for linear delayed systems, but model-dependent property brings obstacles for the application and promotion. To overcome such a difficulty, a few-shot data-driven approach has been proposed in this paper, and operating data are collected to simultaneously tune feedback controller parameters and estimate internal model applied in SP. Subsequently, the anti-windup compensator is constructed with the estimated model and compensator gain is calculated by using the Lyapunov equation. Finally, the effectiveness of the proposed method is verified through simulations that implemented three typical processes in engineering.