权重的选择是各种洪水灾害评估模型中的一个关键而又难以确定的问题。根据现有主观赋权法和客观赋权法的特点,在差分进化(DE)算法的基础上,引入决策者偏好过滤掉不满足条件的个体,通过优化考虑决策者偏好的模糊聚类迭代模型获得洪灾样本的指标权重向量,结合洪灾损失样本特征值矩阵得出各样本的灾情综合评价值;然后依据灾情综合评价值与聚类矩阵求出各等级的特征值,并依此自动识别聚类矩阵中各行的类别属性;最后依据识别出的类别属性和各样本的灾情综合评价值,对洪灾样本在不同的决策者偏好下进行等级划分和灾情排序。以2013年四川省和1996年新疆的洪灾样本为例进行仿真试验,实现了不同决策者偏好下的洪灾等级评估,并为水利部门选择偏好类型提供了参考性建议。

Assigning appropriate weight to different indexes is a key also a difficult problem to various flood disaster evaluation models. According to the characteristics of the present subjective and objective assigning weight methods, decision maker’s preferences is introduced into the differential evolution (DE) algorithm to filter out those individuals which dissatisfy the preferences, and the indexes’ weight vectors of flood samples can be obtained by optimizing the fuzzy clustering iterative model which considers the decision maker’s preferences. Furthermore, with the characteristic value matrix of the floods samples, the comprehensive evaluation value of each flood disaster is obtained. According to the comprehensive evaluation value of each flood disaster and the cluster matrix, the eigenvalue of each degree that is followed by identifying the degree of each row in the cluster matrix automatically, can be obtained. Finally, all the flood samples under different decision maker’s preferences are assessed and sorted based on the identified degree of each row and the comprehensive evaluation value of each flood disaster. Simulation test on two flood samples, namely in Sichuan occurred in 2013 and Xinjiang in 1996, reveals the results of flood rating in different decision maker’s preferences, and provides reference for water conservancy department on choosing preference type.