摘要
为推进污水处理厂减污降碳协同治理,本研究聚焦污染物协同温室气体去除的实施路径,采用机器学习的方法对北京某AAO-MBR工艺污水处理厂进行水质模拟并核算其全流程温室气体排放,构建了基于NSGA-Ⅱ算法的减污降碳多目标优化模型.结果表明:①2019−2021年该污水处理厂温室气体平均排放量为854.96 t/月(以CO_(2)污染当量计),污泥处理处置和电力消耗既是主要排放源,又是后续减排的重点控制环节.②通过合理降低污泥产量和耗电量,可以在满足现有出水水质达标的状况下,进一步降低温室气体排放和运行成本.③基于污泥和曝气调整后的优化系统,在减少3.63%的温室气体排放的同时减少了4.07%的运行成本.研究显示,基于NSGA-Ⅱ算法的多目标优化模型能够优化环境目标和经济目标间的权衡,为污水处理厂的协同发展提供解决方案.
In order to promote the coordinated governance of pollution reduction and carbon reduction in wastewater treatment plants(WWTPs),this study focused on the implementation path for reducing pollutants,reducing greenhouse gases(GHG),and increasing benefit.A machine learning method was used to simulate the entire process of GHG emissions form an AAO-MBR WWTP in Beijing,and a multi-objective optimization model for carbon reduction and pollution reduction was constructed based on NSGA-Ⅱalgorithm.The results showed that:(1)The average GHG emissions of the WWTP from 2019 to 2021 were 854.96 t/month.Sludge treatment and disposal and power consumption were the main emission sources and the key control links for subsequent emission reduction.(2)By reasonably reducing sludge production and energy consumption,GHG emissions and operating costs could be further reduced while meeting the current effluent quality standards.(3)An optimized system based on sludge and aeration adjustment can reduce GHG emissions by 3.63%while reducing operating costs by 4.07%.The study showed that the multi-objective optimization model based on the NSGA-Ⅱalgorithm can balance environmental and economic objectives,providing solutions for the coordinated development of WWTPs.
作者
陈惠鑫
旷森楠
陈昕悦
李新宇
郑祥
程荣
石磊
CHEN Huixin;KUANG Sennan;CHEN Xinyue;LI Xinyu;ZHENG Xiang;CHENG Rong;SHI Lei(School of Environment&Natural Resources,Renmin University of China,Beijing 100872,China)
出处
《环境科学研究》
CAS
CSCD
北大核心
2023年第11期2148-2158,共11页
Research of Environmental Sciences
基金
中国人民大学科学研究基金项目(No.22XNH061)。
关键词
减污降碳
NSGA-Ⅱ算法
多目标优化
机器学习
pollution abatement and carbon reduction
NSGA-Ⅱalgorithm
multi-objective optimization
machine learning