Abstract: The biochemical treatment of coking wastewater is a complex process, so accurate regulation of chemical dosing to ensure stable effluent quality poses a great challenge. In this study, an intelligent control model based on a back propagation neural network (BPNN) was proposed to achieve precise prediction of effluent quality and chemical dosage in the biochemical treatment process of coking wastewater. The results demonstrated that the coefficient of determination (R²) of the effluent quality prediction model reached 0.891. The R² values of the chemical dosage prediction models were 0.605, 0.798, and 0.809, respectively. Data preprocessing techniques, including data cleaning and feature selection, significantly enhanced the predictive performance of the models. Moreover, the dual-output model exhibited superior predictive accuracy than the single-output models. This study provides a theoretical reference for the intelligent regulation of biochemical treatment processes and chemical dosing systems in the coking industry