Given the significant interference and the variable model parameters encountered in pH value control in the sewage treatment reaction process, this study capitalizes on the independence between the set value response and the interference response of the internal model control to proposed a pH optimization control strategy, integrating internal model control and a neural network inverse model. By incorporating a low-pass filter into the system and using the RBF neural network for online identification of the inverse model of the controlled object, the robustness and anti-interference capability of pH value control in the sewage treatment are improved. This approach effectively addresses the challenge of varying model parameters in the neutralization reaction pH value control process. MATLAB simulation results show that compared with conventional PID control and neural internal model control strategies without a filter, the proposed optimal control strategy reduces overshoot by up to 17.4% and shortens the adjustment time by up to 113.6 s. These improvements effectively improve the system’s robustness and anti-interference capabilities. Engineering applications validate the effectiveness of the proposed strategy, ensuring pH value control deviation within ±0.2. Consequently, the control accuracy and system stability are significantly improved.