基于贝叶斯网络的暴雨-地质、暴雨-洪涝灾害链推理模型

inference model of the Rainstorm-Geology and Rainstorm- Flood Disaster Chain Based on Bayesian Network

摘要:

为解决现有山区暴雨-地质、市区暴雨-洪涝灾害模型在链式反应推理、精度上的不足，本文分析其演化规律，结合贝叶斯复杂网络理论，构建灾害链推理模型。通过国内灾害历史数据和相关文献丰富灾害节点及状态,并据此构建灾害链拓扑结构，应用EM算法求得条件概率表，基于因果推理实现次生灾害和基础设施损毁等级的预测，并以湖南省宁远县滑坡洪灾灾害进行实例验证。结果表明：实例的预测结果与灾害实际情况吻合较好，Brier 检验B 值小于0.6，所构建的暴雨-地质、暴雨-洪涝灾害链推理模型具有可行性，可为国家减灾委等部门提供及时断链减灾的参考。

Abstract:

In order to solve the deficiencies of the chain reaction reasoning and accuracy in the existing mountainous storm-geology, urban storm-flood disaster models, this paper analyzes its evolutional process and combines the Bayesian complex network theory to construct a disaster chain reasoning model. Through domestic disaster historical data and related literature, disaster nodes and statuses are enriched and then construct a disaster chain topology based on that. The conditional probability table was obtained by applying the EM algorithm. Using the causal reasoning, the secondary disasters and infrastructure damage levels was predicted finally and the data on landslide flood disaster in Ningyuan County, Hunan Province was tested as a example. The results show that the prediction results are in good agreement with the actual situation of the example. The Brier test B value is less than 0.6. The constructed storm-geology, storm-flood disaster chain reasoning model is feasible, and it can provide timely reference for the National Disaster Reduction Center of China to mitigate disaster.