基于人工神经网络的分流制系统雨水泵站雨天排放污染成因研究

魏卿; 陈咏琪; 谢一凡; 林晶莹; 尹海龙

doi:10.13205/j.hjgc.202312006

基于人工神经网络的分流制系统雨水泵站雨天排放污染成因研究

doi: 10.13205/j.hjgc.202312006

魏卿^1,2,
陈咏琪^1,2,
谢一凡³,
林晶莹^1,2,
尹海龙^1,2, ,

1. 同济大学环境科学与工程学院, 上海 200092;
2. 同济大学污染控制与资源化研究国家重点实验室, 上海 200092;
3. 清华大学环境学院, 北京 100084

基金项目:

国家重点研发计划课题(2021YFC3200703)

国家自然科学基金(521701013)

详细信息

作者简介:
魏卿(1996-),男,博士研究生,主要研究方向为城市排水系统智慧诊断和优化调控。qingw@tongji.edu.cn

通讯作者:
尹海龙 (1976-),男,教授,主要研究方向为城镇智慧水务和水环境治理。yinhailong@tongji.edu.cn

计量
- 文章访问数: 53
- HTML全文浏览量: 2
- PDF下载量: 9
- 被引次数: 0
出版历程
- 收稿日期: 2023-11-15
- 网络出版日期: 2024-03-08

POLLUTION MECHANISM OF WET-WEATHER PUMPING DISCHARGE IN SEPARATED STORMWATER DRAINAGE SYSTEMS BASED ON ARTIFICIAL NEURAL NETWORK

WEI Qing^1,2,
CHEN Yongqi^1,2,
XIE Yifan³,
LIN Jingying^1,2,
YIN Hailong^{1,2
, ,}

1. College of Environmental Science and Engineering, Tongji University, Shanghai 200092, China;
2. State Key Laboratory of Pollution Control and Resource Utilization Research, Tongji University, Shanghai 200092, China;
3. School of the Environment, Tsinghua University, Beijing 100084, China

摘要

摘要: 分流制系统雨水泵站排放造成雨天河道水质恶化,阐明其排放污染成因是削减雨天污染和改善河道水质的基础。针对上海市中心城区两个分流制系统,分别采用反向传播神经网络和径向基神经网络建立降雨、管网运行等11个参数和雨水泵站雨天排放水质的非线性响应关系,探究雨天排放污染的主要影响因素。结果表明:与反向传播神经网络相比,径向基神经网络具有较高模拟效果,COD、NH₃-N和SS的平均绝对误差、均方根误差、平均百分比误差分别下降了15.6%~31.9%、12.3%~18.3%和12.6%~53.9%,决定系数提高了3.1%~5.4%。基于径向基神经网络对输入参数进行重要性分析,确定了5个优先参数,分别为距离上次开泵时间、开泵水位、峰值降雨量、前期不降雨天数、停泵水位。应开展雨污混接调查和实施雨污分流改造,从根本上减少旱天污水在管道沉积造成的雨天污染"零存整取"和污染效应放大。此外,通过优化雨水泵停泵水位,也可以削减雨天排放浓度。
- 分流制系统 /
- 雨水泵站 /
- 人工神经网络 /
- 雨天污染排放 /
- 雨污混接
Abstract: Wet-weather discharges from stormwater pumping stations in separate stormwater drainage systems contribute to the deterioration of river water quality. Understanding the pollution mechanism of pumping discharge is the key to alleviating wet-weather pollution and improving water quality. To identify the dominant factors for controlling wet-weather discharge pollution, this study employed back propagation neural network (BPNN) and radial basis neural network (RBNN) models to establish the nonlinear relationships between 11 input parameters (i.e., rainfall and stormwater network operation) and the wet-weather discharge pollution from two pumping stations in separate stormwater drainage systems in downtown Shanghai. The findings indicated that RBNN performs better in simulating wet-weather pumping discharge pollution than BPNN. For water quality indicators of COD, NH₃-N, and SS, the RBNN achieved a higher performance of 15.6% to 31.9%, 12.3%~18.3%, and 12.6%~53.9%, respectively than BPNN, measuring by average absolute error, root mean square error, and average percentage error, respectively. Coefficients of determination were improved by 3.1%~5.4% when comparing RBNN with BPNN. Moreover, the importance ranking of input parameters using RBNN identified five key parameters, including time duration between two consecutive pump operation events, wet-well water level for starting the pumps, peak rainfall intensity, antecedent dry-weather periods, and wet-well level for stopping the pumps. Actions should be taken to investigate the sewage sources illicitly discharging into storm drains and then correct these illicit discharges. These actions can eliminate the sewage accommodation and the corresponding accumulated pollutant deposits within the storm pipes on dry-weather days, which are inappropriately flushed into the receiving waters during the operation of wet-weather storm pumps. Therefore, the wet-weather discharge pollution would be obviously alleviated. Optimizing the in-pipe water level to stop stormwater pumps, can also aid in reducing the wet-weather discharge pollution.
- separated storm and sewer system /
- stormwater pumping station /
- artificial neural network /
- wet-weather pollutants discharge /
- illicit sewage connection into storm pipes

HTML全文

参考文献(14)

[1]	ZGHEIB S, MOILLERON R, CHEBBO G. Priority pollutants in urban stormwater: part 1-Case of separate storm sewers[J]. Water Research, 2012, 46(20): 6683-6692.
[2]	徐祖信, 徐晋, 金伟, 等. 我国城市黑臭水体治理面临的挑战与机遇[J]. 给水排水, 2019, 55(3): 1-5 , 77.
[3]	ELLIS J B, BUTLER D. Surface water sewer misconnections in England and Wales: pollution sources and impacts[J]. Science of the Total Environment, 2015, 526: 98-109.
[4]	LI Y, ZHOU Y, WANG H, et al. Characterization and sources apportionment of overflow pollution in urban separate stormwater systems inappropriately connected with sewage[J]. Journal of Environmental Management, 2022, 303: 114231.
[5]	THÉVENOT D R, MOILLERON R, LESTEL L, et al. Critical budget of metal sources and pathways in the Seine River basin (1994—2003) for Cd, Cr, Cu, Hg, Ni, Pb and Zn[J]. Science of the Total Environment, 2007, 375(1/2/3): 180-203.
[6]	李田, 戴梅红, 张伟, 等. 水泵强制排水系统合流制溢流的污染源解析[J]. 同济大学学报(自然科学版), 2013, 41(10): 1513-1518, 1525.
[7]	WU J, WANG Z. A hybrid model for water quality prediction based on an artificial neural network, wavelet transform, and long short-term memory[J]. Water, 2022, 14(4): 610.
[8]	SALEH A, ALI H. Wastewater pollutants modeling using artificial neural networks[J]. Journal of Ecological Engineering, 2021, 22(7): 35-45.
[9]	赵梦圆,王建龙,黄涛,等.北京市雨水径流中颗粒物沉降特性[J].环境工程,2019,37(2):67-72.
[10]	王新民, 张超超. 基于深度学习的旧金山湾水质预测[J]. 吉林大学学报(地球科学版), 2021, 51(1): 222-230.
[11]	张秀菊, 王柳林, 李秀平, 等. 基于BP神经网络的潇河流域水质预测[J]. 水资源与水工程学报, 2021, 32(5): 19-26.
[12]	孙跃扬,武利,郭楠,等.具有在线自组织功能的RBF网络的COD预测[J/OL].控制工程:1-7.https://doi.org/10.14107/j.cnki.kzgc.20230237. 2023-12-17.
[13]	XIE Y, CHEN Y, LIAN Q, et al. Enhancing real-time prediction of effluent water quality of wastewater treatment plant based on improved feedforward neural network coupled with optimization algorithm[J]. Water, 2022, 14(7): 1053.
[14]	ZHANG Q, LI Z, SNOWLING S, et al. Predictive models for wastewater flow forecasting based on time series analysis and artificial neural network[J]. Water Science and Technology, 2019, 80(2): 243-253.