基于概率神经网络的水资源可持续利用评价研究

发布时间：2018-05-16 15:00

本文选题：概率神经网络 + BP神经网络　；参考：《成都理工大学》2017年硕士论文

【摘要】：水资源可持续利用评价研究可以为未来水资源合理利用及科学管理提供可靠依据,同时对区域社会经济、生态环境及水资源的协调发展也具有重要指导意义。其研究思路通常为,采用一定的方法或模型来刻画评价指标与标准间的复杂关系,进而实现水资源可持续利用能力在时间与空间上的分类或排序。水资源可持续利用能力评价作为水资源可持续利用研究的一种主要方式,其评价的结果可划分为综合指数和等级评价两种形式,等级评价的实质是一个分类问题,就评价结果而言是将评价对象划分为不同等级。而概率神经网络(Probabilistic Neural Network,PNN)是一种可以最大限度利用先验知识来获得Bayes准则下最优解,且具有较好模式分类功能的神经网络模型。PNN在学习过程中,通过Parzen窗函数估计获得类的条件概率密度,再利用Bayes定理和基于最小风险的Bayes决策规则对样本做出分类。因此,根据PNN强大的模式分类功能,将其应用到水资源可持续利用能力评价研究中。本文以四川省成都市为研究区,以成都市水资源可持续利用能力为评价目标。从可持续角度出发,首先对其关键影响因进行分析,进而初始化了评价指标体系,并利用相关分析和粗糙集(Rough Set,RS)理论,筛选了14项具有代表性的指标,从而确定了最终的指标体系。其次,结合PNN分类原理,构建了PNN水资源可持续利用评价模型,并以2013-2015年成都市实际数据为待测样本,利用建立的评价模型计算得出评价结果。最后,为了验证PNN评价模型的合理性,与传统BP神经网络模型结果进行对比分析,从应用结果可以看出,PNN应用于水资源可持续利用能力评价中具有一定的可行性和科学性,而且取得了较好的应用效果。
[Abstract]:The evaluation of sustainable utilization of water resources can provide a reliable basis for the rational utilization and scientific management of water resources in the future, and also has an important guiding significance for the coordinated development of regional social economy, ecological environment and water resources. The research idea is usually to use a certain method or model to describe the complex relationship between the evaluation index and the standard, and then to realize the classification or ranking of the water resources sustainable utilization ability in time and space. As one of the main ways to study the sustainable utilization of water resources, the evaluation results of water resources sustainable utilization can be divided into two forms: comprehensive index and grade evaluation. The essence of grade evaluation is a classification problem. As far as the evaluation results are concerned, the evaluation objects are divided into different grades. Probabilistic Neural network (PNN) is a kind of neural network model, which can obtain the optimal solution under Bayes criterion by using priori knowledge to the maximum extent, and has a better function of pattern classification in the process of learning. The conditional probability density of classes is estimated by Parzen window function, and then the samples are classified by Bayes theorem and Bayes decision rule based on minimum risk. Therefore, according to the powerful model classification function of PNN, it is applied to the evaluation of water resources sustainable utilization ability. This paper takes Chengdu, Sichuan Province as the research area, and takes the sustainable utilization of water resources in Chengdu as the evaluation goal. From the point of view of sustainability, this paper first analyzes the key factors, and then initializes the evaluation index system, and selects 14 representative indexes by using correlation analysis and rough set RSs theory. Finally, the final index system is determined. Secondly, according to the principle of PNN classification, the evaluation model of sustainable utilization of PNN water resources is constructed, and the evaluation results are calculated by using the actual data of Chengdu from 2013-2015 as the sample to be tested. Finally, in order to verify the rationality of the PNN evaluation model, compared with the results of the traditional BP neural network model, it can be seen from the application results that the application of PNN in the evaluation of water resources sustainable utilization capacity has certain feasibility and scientific. In addition, good results have been obtained.
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TV213.9;TP183

【参考文献】