基于改进的神经网络方法的风暴潮灾害经济损失预测

Improved neural network-based economic loss prediction of storm surge disaster

风暴潮灾害一直以来对中国东南沿海地区的社会经济发展具有较为严重的负面影响,是对中国造成危害最为严重的海洋灾害之一,建立一个准确有效的损失评估模型进行风暴潮灾害损失预测,对风暴潮灾害的预防具有重要的意义。本文在现有研究的基础上收集了2000—2018年中国东南沿海的琼、粤、闽、浙等省份记录较为完整的风暴潮灾害相关数据,在综合考虑危险性、承灾体脆弱性、孕灾环境和防灾减灾能力的基础上,建立起更为完整的风暴潮灾害损失的指标体系。相较于单一的BP神经网络,本文在借鉴机器学习相关理论的基础上搭建了差分进化灰狼算法(DEGWO)优化的BP神经网络,对样本进行训练和仿真测试。结果表明,通过DEGWO算法优化后的模型误差更小,数据的拟合程度更高,对比而言,提高了风暴潮灾害损失预测的精确性,能够为风暴潮灾害损失预测的研究提供新的思路,同时也为风暴潮灾害的防灾减灾管理提供了指导。

Storm surge disasters have serious negative impact on the social and economic development of China’s southeast coastal areas and are one of the most serious marine disasters in China.Therefore,it is highly important to establish an accurate and effective loss assessment model for storm surge disaster loss prediction,which is crucial for the prevention and management of storm surge disasters.Based on existing research,this study collects rela-tively complete storm surge disaster-related data of Hainan,Guangdong,Fujian,and Zhejiang provinces on the southeast coast of China from 2000 to 2018.It establishes a complete indicator system for storm surge disaster losses based on comprehensive consideration of risk,the vulnerability of disaster bearers,pregnant environment,and disaster prevention and mitigation capabilities.Compared with a single back propagation(BP)neural network,this study constructs a BP neural network optimized by the differential evolutionary gray wolf algorithm(DEGWO)based on machine learning-related theories and trains and simulates the samples.The results show that the proposed network model demonstrates a smaller error and a higher fit of the data than single BP neural network model,thus improving the accuracy of storm surge disaster loss prediction.These results can provide new insights for the study of storm surge disaster loss prediction and guidance for storm surge disaster prevention and mitigation management.