基于模糊神经网络的台风灾害损失模型及应用

Prediction Model of Typhoon Disaster and Its Application Based on Fuzzy Neural Network

利用海南省1992-2011年台风灾害损失数据,采用主成分分析法确定了台风灾害损失的致灾因子及灾情指标因子。借助广义动态模糊神经网络(GD-FNN)在学习过程中逐渐形成模糊规则的优点,设计了一种模糊神经网络模型并应用于台风灾害损失的预测预警中,定量地研究了台风灾害致灾因子与灾情指标因子之间的规律。训练结果表明实际样品值与训练结果拟合较平滑,受台风影响较大的局部降雨量峰值增强了间接经济损失预测值。将建立的预测模型应用201119"尼格"台风的灾害损失预测,实验结果表明该模型能较好地预测台风灾害中倒塌房屋、受灾农作物面积、受灾人口及间接经济损失。

In this paper the typhoon disaster loss data from 1992 to 2011 in H ainan Province are used for statistical analysis and hazard and disaster indicators factor are determined by means of principal component analysis. With the GI〉FNN gradually formed in the process of learning the advantages of fuzzy rules, a fuzzy neural network model is designed and applied to the typhoon disaster loss forecast, and the law of the typhoon disaster hazard factor with disaster indicators factors is studied using Quantitative method. The training results show that the actual data value relatively fits the results but local rainfall peak enhances the predictive value of indirect economic losses. The prediction model is applied to the forecast of ＂Nalgae＂ Typhoon disaster loss, with the experimental result that the model has a better ability to predict typhoon disaster loss.