Abstract: Aiming at the problems of strong nonlinearity and time-delay in the total pressure control of the hypersonic wind tunnel at AVIC Aerodynamics Research Institute, a composite control strategy integrating nonlinear autoregressive exogenous model (NARX) and sliding mode compensation was proposed to improve the flow field’s accuracy. First, a NARX model for total pressure was established based on the operational data, and the particle swarm optimization algorithm (PSO) was employed to optimize the model’s delay order and the number of hidden-layer nodes. Second, a controller combining NARX-based feedforward and sliding mode compensation was designed, utilizing the NARX model to predict system dynamics for feedforward control, while incorporating sliding-mode control to compensate for uncertainties and external disturbances. Simulation results indicate that the optimized model reduced RMSE by up to 57.69 % on the validation set and improved the coefficient of determination to 0.9959. Compared to the original PID scheme, the proposed control strategy reduced the overshoot by 5.01 percentage points in absolute terms and decreased the steady-state control error by 1.812 percentage points under specific operating conditions. Finally, through multi-condition wind tunnel tests, the method improved the total pressure’s steady-state control accuracy to 0.15%～0.3%, thereby enhancing the flow field quality of the hypersonic wind tunnel.