主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2026, Vol. 34 ›› Issue (1): 268-281.doi: 10.16381/j.cnki.issn1003-207x.2023.2025cstr: 32146.14.j.cnki.issn1003-207x.2023.2025

• • 上一篇    下一篇

基于风险特征溯源的城市暴雨级联事件风险评估模型构建

刘昭阁1,4(), 李向阳2, 朱晓寒3   

  1. 1.厦门大学公共事务学院,福建 厦门 361005
    2.哈尔滨工业大学经济与管理学院,黑龙江 哈尔滨 150001
    3.武汉东湖新技术开发区管委会,湖北 武汉 430075
    4.厦门大学应急管理研究中心,福建 厦门 361005
  • 收稿日期:2023-12-04 修回日期:2024-04-05 出版日期:2026-01-25 发布日期:2026-01-29
  • 通讯作者: 刘昭阁 E-mail:zhaogeliu@xmu.edu.cn
  • 基金资助:
    国家自然科学基金青年项目(72404232);国家自然科学基金重大研究计划项目(91746207);国家社会科学基金重点项目(23AZD072);福建省自然科学基金青年创新项目(2023J05011)

Risk Assessment Model Construction of Urban Rainstorm Cascading Events Based on Risk Feature Tracing

Zhaoge Liu1,4(), Xiangyang Li2, Xiaohan Zhu3   

  1. 1.School of Public Affairs,Xiamen University,Xiamen 361005,China
    2.School of Management,Harbin Institute of Technology,Harbin 150001,China
    3.Administrative Committee of Wuhan East Lake High-tech Development Zone,Wuhan 430075,China
    4.Center for Emergency Management Research,Xiamen University,Xiamen 361005,China
  • Received:2023-12-04 Revised:2024-04-05 Online:2026-01-25 Published:2026-01-29
  • Contact: Zhaoge Liu E-mail:zhaogeliu@xmu.edu.cn

摘要:

城市暴雨极易诱发基础设施损毁、居民受困等大范围、小粒度、差异化的级联事件(urban rainstorm cascading events, URCE),灾害预防亟需全面精准地评估不同类型的URCE风险,但由于样本数据风险特征的不足,模型效果受到限制。本文考虑历史案例对URCE风险特征的全面描述,提出一种基于风险特征溯源的风险评估模型构建方法。该方法遵循由“果”寻“因”基本思路,以风险的情境描述为衔接,利用案例推理从历史案例中抽取风险特征,再采用机器学习方法学习风险特征与URCE风险类之间的关联关系,由此完成风险评估模型自适应构建。对武汉市6类典型URCE风险评估的用例结果表明:所提方法有助于通过基于案例的风险特征溯源解决风险评估模型构建的特征不完备问题,由此提升风险评估精度,在风险间区分性和小样本建模方面亦具有较好效果;与传统综合评价、灾损曲线等方法相比,所提方法在小粒度、复杂事件风险耦合场景下的风险评估模型构建中优势显著,更适应复杂风险的精准化管理目标。

关键词: 城市暴雨, 暴雨级联事件, 风险评估模型构建, 风险特征溯源, 案例推理

Abstract:

Urban rainstorm disasters can easily induce large-scale, small-size, differentiated cascading events (urban rainstorm cascading events, URCE) such as damage to infrastructure and residents being trapped, etc. Disaster prevention urgently needs comprehensive and accurate assessment of the risks of different types of URCEs, but it faces the model performance constraints caused by sample data risk feature incompleteness. The comprehensive description of URCE risk features in historical cases is considered, and a risk assessment model construction method is proposed based on tracing back risk features. This method follows the basic idea of seeking “causes” from “effects”, uses case-based reasoning to trace risk features from historical cases with the connection of risk scenario descriptions, and then uses machine learning methods to learn the association between risk features and URCE risk categories, thus completing the adaptive construction of risk assessment models. The case study results of risk assessment for 6 typical URCE risks in Wuhan show that: the proposed method helps to solve the problem of incomplete features in constructing risk assessment models by tracing risk features based on cases, thus improving the accuracy of risk assessment. It also has good effects in distinguishing risks and small sample modeling. Compared with traditional comprehensive evaluation, damage curve and other methods, the proposed method has significant advantages in constructing small-grained, complex event coupled risk assessment models, which better meets the goal of accurate management of complex risks.

Key words: urban rainstorms, rainstorm cascading events, risk assessment model construction, risk features tracing, case-based reasoning

中图分类号: