基于AHP-BP模型的文山州水资源可持续利用评价分析

基于层次分析法和BP神经网络的基本原理和方法,结合区域实际,利用层次分析法构建了符合丰水地区水资源可持续利用指标体系和评价标准,从水资源条件、水资源开发利用效率、生态环境状况、水资源合理配置和水资源管理能力五个方面提出50个评价指标,运用双隐层BP神经网络,建立AHP-BP水资源可持续利用评价模型,对文山州不同规划水平年水资源可持续利用进行综合评价。结果表明:①不同规划水平年各评价区域水资源可持续利用评价为2～3级,即处于可持续与基本可持续之间,反映了文山州现状及中、长期水资源可持续利用状况,符合区域发展实际。②AHP-BP评价模型克服了层次分析法判断矩阵构造主观性强和一致性不易检验等缺点,满足客观评价要求,且双隐层BP神经网络具有比单隐层网络学习时间短,参数收敛迅速,自适应能力强等优点。

Evaluation of sustainable use of water resources in Wenshan based on AHP-BP model

Based on the basic principle and method and BP neural network AHP, Combined with the actual construction area, the paper built thea index system and evaluation standard of sustainable utilization of water resources in accordance with bundant water area by using the analytic hierarchy process, brought forward 50evaluation indexes from five aspects of water resources condition, development and utilization efficiency of water resources, state of ecological environment, rational allocation of water resources and water resources management ability.Using double hidden layer BP neural network, it established the sustainable utilization evaluation model of AHP-BP water resources and comprehensively evaluated the state of sustainable utilization of water resources of Wenshan state in different level years. The results show that the evaluation grade of sustainable utilization of regional water resources in different planning level years is 2~ 3, witch is between sustainable and basic sustainable and reflects the state of Wenshan and the status of long-term water resources sustainable utilization that is accordance with the actual of regional development.The result shows that the model of AHP-BP evaluation and the evaluation method established in the study is reasonable and feasible. The AHP-BP evaluation model overcomes the shortcomings of the subjectivity of AHP judgment matrix and the consistency being not easy to test and can meet the requirements of objective evaluation.The two-hidden layer BP neural network has the advantages of short learning time ,quick converge speed of parameters and strong adaptive ability etc compared with single hidden layer network.