| 96 | 0 | 153 |
| 下载次数 | 被引频次 | 阅读次数 |
我国洪涝灾害的发生具有频率高、影响范围广、损失大的特点。因此，快速准确地预测洪涝灾害受灾人口对制定科学合理的应急响应策略至关重要。建立了基于灰狼优化算法(Grey Wolf Optimizer, GWO)的反向传播(Back Propagation, BP)神经网络的预测模型，该模型在综合考虑致灾因子、孕灾环境、承灾体、防灾减灾能力和灾害严重度的基础上，选取10个影响洪涝灾害受灾人口的指标，构建了洪涝灾害受灾人口预测模型。并以历年洪涝灾害的数据作为试验数据，对神经网络模型进行训练和测试。结果表明，GWO-BP神经网络相比BP模型和粒子群算法的BP模型平均预测准确率分别提高了65.68%和63.27%,证明该模型具有更好的预测准确性和可靠性，可以为应急策略的制定提供可靠的数据支持。
Abstract：Flood disasters in China are characterized by high frequency, extensive spatial coverage, and severe socioeconomic impacts. Accurately predicting the affected populations is crucial for optimizing disaster preparedness, emergency response, and resource allocation. To address this critical need, this study presents a hybrid machine learning model that combines the Grey Wolf Optimizer(GWO) with a Backpropagation(BP) neural network to effectively predict flood-affected populations. This model systematically incorporates 10 key indicators across four core dimensions to account for the complexity of flood impacts. These dimensions include disaster-causing factors, disaster-forming environments, disaster-bearing entities, disaster prevention and mitigation capabilities, and disaster severity. A significant advantage of the GWO-BP model is its ability to overcome the inherent limitations of standalone BP networks, such as slow convergence speed and susceptibility to local optima. The GWO algorithm adaptively optimizes the initial weights and thresholds of the BP network, enhancing its learning efficiency and prediction stability. Experimental validation was conducted using a comprehensive dataset of 72 major flood events in China from 2015 to 2020. The data was divided into a training set(6/7 of the total data) and a test set(1/7 of the total data) to ensure rigorous evaluation. The results indicated that the GWO-BP model achieved a mean prediction accuracy of 83.8%, significantly outpacing the standalone BP model(66.5%) and the Particle Swarm Optimization(PSO)-BP hybrid model(65.9%). Quantitative performance metrics further confirmed its superiority, with a Root Mean Square Error(RMSE) of 12.597 5 and a Mean Absolute Percentage Error(MAPE) of 16.2%. Compared to the BP and PSO-BP models, these figures represent substantial improvements of 65.68% and 63.27% in RMSE and MAPE, respectively. A case study focusing on the catastrophic 2021 Henan flood—one of the most destructive flood events in recent Chinese history—further validated the model's robustness and practical applicability, achieving an outstanding prediction accuracy of 99.48% for the actual affected population. Collectively, these findings confirm that the GWO-BP model effectively captures the nonlinear relationships between flood-related indicators and affected populations, providing a reliable tool for evidence-based emergency management. Future research will focus on expanding the dataset to include more flood events and diverse geographic regions, integrating additional optimization algorithms to enhance the model's performance further, and incorporating real-time meteorological and hydrological data to enable dynamic, short-term predictions of flood impacts.
[1] United Nations Office for Disaster Risk Reduction.Global assessment report on disaster risk reduction 2022:our world at risk:transforming governance for a resilient future[R].Geneva:United Nations,2022.
[2] United Nations Office for Disaster Risk Reduction.Global assessment report on disaster risk reduction 2021:special report on drought[R].Geneva:United Nations,2021.
[3] United Nations Office for Disaster Risk Reduction.Global assessment report on disaster risk reduction 2023:mapping resilience for the sustainable development goals[R].Geneva:United Nations,2023.
[4] Guo Jingdong,Jiao Haiyan.Quick prediction of seismic casualties using fuzzy case-based reasoning technology[J].Journal of Intelligent & Fuzzy Systems,2023,44(2):1743-1753.
[5] Zhu Xiaoxin,Sun Baiqing.Final and dynamic mortality prediction from earthquakes in China[J].International Journal of Emergency Management,2017,13(2):131-149.
[6] Mishra V,Anderson E R,Edwards S,et al.Improving tornado casualty predictions in the US with population exposure data and a modified social vulnerability index[J].International Journal of Disaster Risk Reduction,2023,87:103588.
[7] Ayenew W A,Kebede H A.GIS and remote sensing based flood risk assessment and mapping:The case of Dikala Watershed in Kobo Woreda Amhara Region,Ethiopia[J].Environmental and Sustainability Indicators,2023,18:100243.
[8] Zhu Zijiang,Zhang Yong.Flood disaster risk assessment based on random forest algorithm[J].Neural Computing and Applications,2021,34(5):3443-3455.
[9] 单九生，徐星生，樊建勇，等.基于GIS的BP神经网络洪涝灾害评估模型研究[J].江西农业大学学报，2009,31(4):777-780.Shan Jiusheng,Xu Xingsheng,Fan Jianyong,et al.A study on flood disaster assessment model of BP neural network based on GIS[J].Acta Agriculturae Universitatis Jiangxiensis,2009,31(4):777-780.
[10] Wang Xiaoyan,Jiang Fangzhou,Zhong Lei,et al.Intelligent post-disaster networking by exploiting crowd big data[J].IEEE Network,2020,34(4):49-55.
[11] 刘晓平，叶子超，刘宇航，等.长短期记忆网络在隧道火灾移动通风预测中的应用研究[J].安全与环境学报，2024,24(5):3210-3220.Liu Xiaoping,Ye Zichao,Liu Yuhang,et al.Application of long short-term memory network in prediction of mobile ventilation during tunnel fire[J].Journal of Safety and Environment,2024,24(5):3210-3220.
[12] Opella J M A,Hernandez A A.A study on flood risk assessment using support vector machine and convolutional neural network:a conceptual framework[C]//Proceedings of the 2019 IEEE 15th International Colloquium on Signal Processing & Its Applications.Piscataway:Institute of Electrical and Electronics Engineers,2019.
[13] Mirjalili S,Mirjalili S M,Lewis A.Grey wolf optimizer[J].Advances in Engineering Software,2014,69(5):46-61.
[14] Mosavi M R,Khishe M,Ghamgosar A.Classification of sonar data set using neural network trained by gray wolf optimization[J].Neural Network World,2016,26(4):393-415.
[15] 詹伟，程春鑫.基于GWO-BP的震后过渡安置阶段应急物资需求预测[J].中国安全科学学报，2024,34(10):17-23.Zhan Wei,Cheng Chunxin.GWO-BP-based forecasting of emergency material demand in post-earthquake transitional resettlement phase[J].China Safety Science Journal,2024,34(10):17-23.
[16] Li Boyi,Gong Adu,Zeng Tingting,et al.A zoning earthquake casualty prediction model based on machine learning[J].Remote Sensing,2021,14(1):30.
[17] Yuan Shuai,Wang Guizhi,Chen Jibo,et al.Assessing the forecasting of comprehensive loss incurred by typhoons:a combined PCA and BP neural network model[J].Journal on Artificial Intelligence,2019,1(2):69-88.
[18] Kia B M,Pirasteh S,Pradhan B,et al.An artificial neural network model for flood simulation using GIS:Johor River Basin,Malaysia[J].Environmental Earth Sciences,2012,67(1):251-264.
[19] Hassanin M F,Shoeb A M,Hassaniyen A E.Grey wolf optimizer-based back-propagation neural network algorithm[C]//Proceedings of the 2016 12th International Computer Engineering Conference (ICENCO).Piscataway:Institute of Electrical and Electronics Engineers,2016.
[20] Qie Zairong,Rong Lili.An integrated relative risk assessment model for urban disaster loss in view of disaster system theory[J].Natural Hazards,2017,88(1):165-190.
[21] Hecht-Nielsen R.Theory of the backpropagation neural network[M]//Wechsler H.Neural networks for perception.Amsterdam:Elsevier,1992:65-93.
基本信息:
DOI：10.13637/j.issn.1009-6094.2025.0963
中图分类号:TP183;TV87
引用信息:
[1]关文玲,王然,岑浩栋,等.基于GWO-BP神经网络的洪涝灾害受灾人口预测研究[J].安全与环境学报,2026,26(07):2773-2781.DOI:10.13637/j.issn.1009-6094.2025.0963.
2026-07-15
2026-07-15