电化学还原硝酸盐过程关键在于该废水处理过程中参数的有效控制。基于硝态氮电化学还原的测试数据和各参数间的相关性，得出与出水效果密切相关的四因素，即反应时间、初始浓度、初始pH和电流密度，采用BP神经网络算法建立了电化学法还原硝态氮的预测模型，并验证了模型的准确性。结果表明，4-7-1型BP神经网络网络构型最优，模型预测的去除效果与实测值相吻合，R²为0.9095。利用BP神经网络模型对参数调控，可以优化电化学处理过程：对电流密度进行阶段性调控，在相同处理量下可降低15%的能耗；在水质波动情况下进行电流密度控制，在相同处理时间内可保证出水达标。该研究结果可以为智能控制电化学去除硝态氮的过程提供参考。

Achieving effective control of parameters in the process of nitrate wastewater treatment is critical to electrochemical water treatment. The powerful nonlinear mapping ability, self-adaptation and self-learning ability of neural network technology can optimize the electrochemical processing. However, there are few researches in this direction. Hence, based on the test data of the electrochemical reduction of nitrate, an electrochemical prediction model was established by using the BP neural network algorithm. Considering the correlation of various parameters in the electrochemical process, the reaction time, initial nitrate nitrogen concentration, pH and current density were determined as the input layer of the BP neural network for model establishment. Results showed that the optimal network configuration of 4-7-1 was achieved by optimizing the hyperparameters of hidden layers number, and the numbers of neurons and epochs. The predicted value of nitrate nitrogen concentration was consistent with the measured value, and the R² value of 0.9095 was obtained. Meanwhile, the model predicts the effects of initial concentration, pH and current density on the removal efficiency of nitrate nitrogen. In the weak alkaline environment, the stability and reliability of nitrate electroreduction were higher than those in acidic and alkaline environments, and the predicted value of nitrate nitrogen is highly correlated to the true value (R²=0.9908). The initial concentration was negatively correlated to the removal rate, while the current density was positively correlated. Finally, the neural network model was used to control the electrochemical nitrate reduction process. Energy consumption tests were designed by optimizing current density, and 15% reduction energy consumption was obtained within the same processing time and processing efficiency. Also, through the prediction model, the effluent quality can be guaranteed by timely adjusting the parameter in the case of sudden water quality changes. The research results can provide a reference for the intelligent control in the electrochemical removal of nitrate. At the same time, combining the understanding of the electrochemical treatment system and artificial intelligence technology, several ideas are proposed for the application of artificial intelligence technology in the field of electrochemical water treatment.