神经网络模型在水质评价中的应用

发布时间：2018-04-18 23:03

本文选题：水质评价 + BP神经网络　；参考：《浙江海洋大学》2016年硕士论文

【摘要】：水质评价作为水环境保护的一项重要措施,近年来不断有研究学者投身到这个科学领域,国内外对于水质评价的研究已提出了多种方法及模型。本文例举了一些常用的传统水质评价方法并分析了这几种方法不同程度上存在的局限性,并在前人研究的基础上利用了几种基于神经网络的地表水水质评价模型对水质进行评价。本文研究了利用BP神经网络、概率神经网络和遗传算法优化的神经网络建立的三种水质模型,对同一个样本数据进行分类评价。研究的主要内容有:针对吉林敦化新甸段的水质断面的地表水水质,选取了“COD、CODMn、BOD5、石油”四项指标组成的研究样本。为了取得良好的训练效果,针对这些指标的标准再结合地表水水质标准对样本数据用随机插值的方法进行了样本的扩充。本文首先阐述了BP神经网络、概率神经网络和遗传优化算法的基本概念、原理,分析了各个神经网络的特点以及存在的局限性。通过BP模型、PNN模型对水质样本模拟得出的结果说明了神经网络水质评价的可行性。由于BP神经网络在某些问题的处理上有一定的局限性,所以结合BP网络和遗传算法各自的特点,将全局搜索能力比较强的遗传算法与BP结合构建了一种稳定、收敛快速、鲁棒性高的水质评价模型。模型结合遗传算法的忧点对BP的权重和阈值做出调整,从本文的每个实验的仿真结果可以看出改进后的BP网络不仅使迭代的次数减少,网络的收敛速度和准确率都有很明显的提升。本文的研究表明,神经网络对水质分类的评价方面在理论上是可行的,在实际应用中有待进一步的研发,本文的研究也为神经网络的发展做出了一些贡献。
[Abstract]:Water quality assessment is an important measure of water environmental protection. In recent years, more and more researchers have devoted themselves to this scientific field. Many methods and models have been put forward for water quality assessment research at home and abroad.In this paper, some traditional methods of water quality evaluation are cited and the limitations of these methods in different degrees are analyzed.On the basis of previous studies, several surface water quality evaluation models based on neural network are used to evaluate the water quality.In this paper, three kinds of water quality models based on BP neural network, probabilistic neural network and genetic algorithm optimized neural network are studied, and the same sample data are classified and evaluated.The main contents of the study are as follows: according to the surface water quality of the water quality section of Xindian section of Dunhua, Jilin Province, the study samples of "CODX CODMn BOD5, Petroleum" are selected.In order to obtain good training effect, the standard of these indexes and the surface water quality standard are used to expand the sample data by random interpolation.In this paper, the basic concepts and principles of BP neural network, probabilistic neural network and genetic optimization algorithm are introduced, and the characteristics and limitations of each neural network are analyzed.The simulation results of water quality samples based on BP model and PNN model show the feasibility of water quality evaluation based on neural network.Because BP neural network has some limitations in dealing with some problems, combining the characteristics of BP network and genetic algorithm, a stable and fast convergence is constructed by combining genetic algorithm with BP, which has strong global searching ability.Water quality evaluation model with high robustness.The model adjusts the weight and threshold of BP in combination with the worrisome points of genetic algorithm. From the simulation results of each experiment in this paper, it can be seen that the improved BP network not only reduces the number of iterations.The convergence speed and accuracy of the network are obviously improved.The research in this paper shows that the evaluation of water quality classification by neural network is feasible in theory and needs further research and development in practical application. The research in this paper has also made some contributions to the development of neural network.
【学位授予单位】：浙江海洋大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：X824;TP183

【相似文献】