The occurrence of surge in large centrifugal fans during operation can lead to a degradation of their original operational quality, resulting in substantial economic losses and even posing serious safety hazards. In this study, we propose a constant-pressure approach to approximate the surge limit line, addressing the limitations of fixed and variable surge limit methods when the flow approaches the surge limit. This control strategy effectively expands the stable operating region of large centrifugal fans, enhancing their operational efficiency. Additionally, we employ a Radial Basis Function Neural Network (RBFNN) optimized PID controller for controlling the opening of the recirculation valve. Through this approach, dynamic response and control accuracy are significantly improved. Matlab/Simulink simulations demonstrate that the RBFNN-PID controller achieves the shortest settling time, minimal overshoot, rapid response, and excellent control precision and stability.