Abstract: For the engineering problem where satisfactory control of the outlet molten salt temperature of the receiver in a tower solar thermal power system is difficult to achieve with using conventional proportional integral derivative (PID) control algorithms due to large time delay and nonlinear characteristics, a predictive control method for the outlet molten salt temperature of the receiver based on adaptive robust Kalman filtering was proposed. In this method, the mismatch between the prediction model and the actual object was addressed by introducing a model error tolerance bound and adaptively updating the measurement noise covariance matrix, thus enhancing the robustness of the control system. Simulation results under large-magnitude solar radiation disturbances show that, compared with predictive control based on robust Kalman filtering, predictive control based on standard Kalman filtering, and dynamic matrix control (DMC), the overshoot of the proposed method is reduced by 2.45, 4.74, and 12.43 K, respectively, with corresponding reduction percentages of 7.37%, 12.48%, and 27.21%. Under various disturbance conditions, the proposed method is demonstrated to exhibit superior dynamic deviation suppression and faster adjustment speed, fully validating the effectiveness of the adaptive robust Kalman filtering strategy in addressing the large time delay and model mismatch problems of nonlinear objects.