Abstract: In view of time-varying parameters of the plant model and the large time delay of AGC (automatic gauge control) system for tandem cold rolling, an adaptive Smith generalized predictive control algorithm (ASGPC) is proposed. In this algorithm, a Smith predictor is introduced to eliminate the negative effect of time delay. It uses laser velometer to measure the time delay of AGC system and uses fading memory recursive least square algorithm to identify the plant parameters and then modifies the model of the non-time-delay part of the Smith predictor. The PID controller in traditional Smith predictive algorithm is replaced with the adaptive GPC controller designed based on the identified model parameters. Simulation results show that it still has good control performance and robust stability even though there is model mismatching and disturbances.