[1]胡金财,丁涛,施昱,等.基于卷积网络分析和定量氮去除的智能加药研究[J].中国给水排水,2026,42(7):1-7.
　HUJincai,DINGTao,SHIYu,et al.Intelligent Carbon Source Addition Control Based on Convolutional Neural Network Analysis and Quantitative Nitrate Removal[J].China Water & Wastewater,2026,42(7):1-7.
点击复制

基于卷积网络分析和定量氮去除的智能加药研究

中国给水排水[ISSN:1000-4062/CN:12-1073/TU] 卷: 第42卷 期数: 2026年第7期 页码: 1-7 栏目: 出版日期: 2026-04-01
Title:
Intelligent Carbon Source Addition Control Based on Convolutional Neural Network Analysis and Quantitative Nitrate Removal
作者:
胡金财1， 丁涛1， 施昱1， 支丽玲1， 刘雅菲1， 陈诺1， 郑凯凯2， 谢海强3， 沈知章3， 李激3
（1.无锡市水务集团有限公司，江苏 无锡 214031；2.无锡普汇环保科技有限公司，江苏 无锡 214125；3.江南大学 环境与生态学院，江苏 无锡 214122）
Author(s):
HU　Jincai1， DING　Tao1， SHI　Yu1， ZHI　Liling1， LIU　Yafei1， CHEN　Nuo1， ZHENG　Kaikai2， XIE　Haiqiang3， SHEN　Zhizhang3， LI　Ji3
（1. Wuxi Water Group Co. Ltd.， Wuxi 214031， China； 2. Wuxi Puhui Environmental Protection Technology Co. Ltd.， Wuxi 214125， China； 3. School of Environment and Ecology， Jiangnan University， Wuxi 214122， China）
关键词:
碳源投加智能控制;  卷积网络;  氮负荷;  前馈补偿;  反馈调节
Keywords:
intelligent carbon source addition control;  convolutional neural network;  nitrate loading;  feedforward compensation;  feedback regulation
摘要:
针对人工经验型碳源投加导致的浪费问题，设计开发了碳源投加智能控制系统，考察了碳源投加量与氮负荷变化及脱氮效果的关系。结果表明，改进型时间卷积网络能够有效捕捉变量间以及时间维度上的依赖关系，筛选出了以出水总氮、水温和过程硝态氮为主的关键因素，以此为框架构建了耦合“前馈补偿+定量核算+反馈调节”等过程的智能控制系统。在碳源投加智能控制系统应用后，达到了碳源投加泵流量与氮负荷呈正相关的目的，在出水总氮平均削减量同比提高15.7%、出水总氮浓度控制在7 mg/L以下概率达到96.7%的情况下，碳源平均投加量从218.4 g/m3下降至142.2 g/m3，降幅为34.9%。该碳源投加智能控制系统综合了进水、过程参数和出水水质，考虑了脱氮过程的各个阶段，克服了单一过程调控难以应对水质变化的难题，显著降低了运行药耗。
Abstract:
Addressing the waste carbon source addition issue caused by manual and empirical control, an intelligent control system had been designed and developed. Research had been conducted on the relationship between carbon source addition quantity and changes in nitrogen loading, denitrification efficiency. The results indicated that the improved convolution network could effectively capture the dependencies between variables and on the time dimension, and identified the main influencing factors as total nitrogen of effluent, water temperature, and process nitrate. Based on this framework, an intelligent control system centered around “feedforward compensation + quantitative accounting + feedback regulation” was constructed. After application of the intelligent control system, the objective of achieving a positive correlation between the flow rate of the carbon source pump and nitrogen loading was met. With a increase of 15.7% in the average reduction of total nitrogen in effluent and a probability of 96.7% for the total nitrogen in effluent to be controlled below 7 mg/L, the average carbon source dosage decreased from 218.4 g/m3 to 142.2 g/m3, with a decrease of 34.9%. The intelligent carbon source addition system established in this study integrates the influent, process parameters, and effluent, comprehensively consideres the various stages of denitrification process, and overcomes the difficulty of single process regulation in dealing with water quality changes,which significantly reduces carbon source consumption.
更新日期/Last Update: 2026-04-01