基于BP神经网络的电氧化-电絮凝处理重金属络合废水的实时控制研究

The fluctuant inflow of complex heavy metals wastewater (CHMWs) brings challenges to its stable treatment by electrooxidization-electroflocculation (EO-EC). The real-time control strategy, capable of optimizing the treatment process in real time, is an effective and feasible way to solve these challenges. However, it is difficult to apply in practice due to the limitation of soft sensors of complex and heavy metal ions and the complexity of coupling process of EO and EC. Therefore, this project will aim at building a real-time control system based on BP neural network to improve the performance of EO-EC on CHMWs treatment. Firstly, we will investigate the mechanism of the decomplexation and heavy metal removal and coupling EO and EC. Secondly, based on this, the surrogate parameters will be selected and then used to construct the soft sensors by BP neural network (BPNN). Finally, we will construct a high performance of real-time control system based on BP neural network (R>0.98, RMSE<0.1) and provide the theoretical basis and methods on how to build a real-time control system in the future.

重金属络合废水的水质水量波动等不确定因素给电氧化-电絮凝（EO-EC）运行的稳定性带来了挑战。实时控制具有实时优化调控工艺的能力，是解决该不确定性问题的经济有效手段。然而络合物和重金属软测量模型的局限性以及EO-EC过程耦合的复杂性，导致构建的实时控制系统的经济稳定性差而难以实际应用。为此，本项目探究破络及重金属去除的途径与机理以及EO-EC耦合机制，在此基础上分别选取络合物和重金属传感器的替代参数，借助人工智能领域的BP神经网络分别构建络合物和重金属的软测量模型，最后将其嵌入EO-EC工艺的实时控制过程，搭建高性能的基于BP神经网络的实时控制系统（R>0.98，RMSE<0.1），以期显著改善EO-EC工艺深度处理重金属络合废水的经济性和稳定性，同时为今后构建同类型废水处理工艺的实时控制系统提供相关的理论及方法依据。

重金属络合废水的水质水量波动等不确定因素给电氧化-电絮凝（EO-EC）运行的稳定性带来了挑战。实时控制具有实时优化调控工艺的能力，是解决该不确定性问题的经济有效手段。然而络合物和重金属软测量模型的局限性以及EO-EC过程耦合的复杂性，导致构建的实时控制系统的经济稳定性差而难以实际应用。本项目开展了基于RuO2-IrO2/Ti 和Fe0电氧化-电絮凝对含铊络合废水的去除机理及耦合机制研究，在此基础上开展了含铊实际废水的间接参数筛选与优化研究，最后构建与验证了基于BP和DNN神经网络的电氧化-电絮凝实时控制模型，搭建了高精度的DNN神经网络的实时控制仿真系统，显著改善了EO-EC工艺深度处理重金属络合废水的经济性和稳定性，且为今后构建重金属污染土壤修复废液处理的实时控制系统提供相关的理论及方法依据。本项目在Water Research，Chemical Engineering Journal，Journal of Hazardous Materials等期刊上发表标注论文8篇，授权发明专利2项，本项目成果获湖南省科技进步二等奖1项。