基于MCDM-BPNN的城市内涝风险评价及调蓄池选址  

Risk assessment of urban waterlogging and site selection of storage tank based on MCDM-BPNN

作　　者：郝景开 李红艳[1,2] 张峰 张翀[3] 毛立波 刘大为 HAO Jingkai;LI Hongyan;ZHANG Feng;ZHANG Chong;MAO Libo;LIU Dawei(School of Environmental Science and Engineering,Taiyuan University of Technology,Jinzhong Shanxi 030600,China;Shanxi Municipal Engineering Graduate Education Innovation Center,Jinzhong Shanxi 030600,China;Shanxi Traffic Science and Technology Research and Development Co.,Ltd.,Taiyuan Shanxi 030032,China;Science and Technology Management Department of Shanxi Dadi Environmental Investment Holding Limited Liability Company,Taiyuan Shanxi 030032,China)

机构地区：[1]太原理工大学环境科学与工程学院,山西晋中030600 [2]山西省市政工程研究生教育创新中心,山西晋中030600 [3]山西省交通科技研发有限公司,山西太原030032 [4]山西大地环境投资控股有限公司科创管理部,山西太原030032

出　　处：《中国安全科学学报》2024年第8期214-221,共8页China Safety Science Journal

基　　金：山西省自然科学研究面上项目(202203021221060);山西省研究生创新项目(2023KY254);吕梁市引进高层次科技人才重点研发项目(2021RC-1-22)。

摘　　要：为建立一套较为完善的城市内涝风险评价体系,并据此确定调蓄池位置,首先,从积水风险、超载风险和边侧进流量3个维度构建评价指标,设计一种包括改进层次分析法(IAHP)、反熵权法(AEW)和优劣解距离法(TOPSIS)的混合多准则决策框架(MCDM);然后,将IAHP-AEW-TOPSIS模型分别与IAHP-TOPSIS、AEW-TOPSIS模型对比,通过斯皮尔曼排序相关系数验证排序一致性,通过计算变异系数、相对极差和灵敏度证实IAHP-AEW-TOPSIS模型的性能;最后,结合反向传播神经网络(BPNN),建立MCDM-BPNN模型,并以山西省某一内涝易发区域为例进行验证。结果表明:积水风险对城市内涝风险评价体系的影响最为显著,所占权重为0.46,其次为超载风险,所占权重为0.36;节点位置与连接管道数量很大程度上对该节点的内涝风险产生影响,在管道汇接处或汇流面积较大处内涝出现更为频繁;IAHP-AEW-TOPSIS模型在样本判别方面具有更好的性能;在5年与10年重现期下,MCDM-BPNN模型验证集准确率分别为93.3%和100%,能够准确快速模拟和预测城市洪水;应用案例设置调蓄池后,高、中、低风险节点数量分别为7、9、30和6、19、21,内涝溢流削减效果显著。To establish a comprehensive evaluation system for urban waterlogging risk,three dimensions were selected:water accumulation risk,overload risk,and lateral inflow.This system aims to provide a reference for the optimal placement of storage tanks.Firstly,a mixed MCDM framework including the improved analytic hierarchy process(IAHP),anti-entropy weight method(AEW),and technique for order preference by similarity to ideal solution(TOPSIS)was designed.Then,the IAHP-AEW-TOPSIS model was compared with IAHP-TOPSIS and AEW-TOPSIS model respectively,and the ranking consistency was verified by Spearman ranking correlation coefficient.The performance of IAHP-AEW-TOPSIS model was confirmed by calculating variation coefficient,relative range and sensitivity.Finally,a model based on MCDM-BPNN was established and verified by a waterlogging-prone area in Shanxi Province.The results show that water accumulation risk has the most significant influence in the evaluation system of urban waterlogging risk,with the weight of 0.46,followed by the overload risk with the weight of 0.36.The location of the node and the number of connecting pipes greatly affect the risk of waterlogging of the node,and waterlogging occurs more frequently at the junction of pipes or in larger confluence areas.There was better performance exhibited by the IAHP-AEW-TOPSIS model.In the 5-year and 10-year return periods,the accuracy of MCDM-BPNN model verification set is 93.3%and 100%respectively,which can accurately and rapidly simulate and predict urban floods.After the application case is set up,the number of high,medium and low risk nodes are 7,9,30 and 6,19,21 respectively,and the effect of reducing waterlogging overflow is remarkable.

关 键 词：多准则决策框架(MCDM) 反向传播神经网络(BPNN) 城市内涝 风险评价 调蓄池 

分 类 号：X915.5[环境科学与工程—安全科学]