基于改进小波神经网络的水质评价建模研究

发布时间：2018-07-02 07:48

本文选题：小波神经网络 + 水质评价　；参考：《江西理工大学》2015年硕士论文

【摘要】：水资源是一种不可替代的资源。近年来,国内外一直重视水资源的保护和治理工作。然而随着科技和产业的发展,水资源的问题还是一直制约着社会和生态环境的发展。同时传统的水质评价方法面对水环境问题的复杂性和非线性缺少高效的处理效率,因此,提高水资源的保护措施刻不容缓。人工神经网络(ANN)的发展为水质研究带来了新的方向,目前国内外已经有不少关于基于人工神经网络的水质方面的研究。本文根据前人对人工神经网络和水质评价的研究,深入研究小波神经网络的理论、结构和算法后,尝试采用小波神经网络(Wavelet Neural Network,WNN)应用于水质评价研究。论文主要研究包括以下几个方面:1.鉴于传统水质评价方法存在一定的局限性,利用小波神经网络收敛速度快、泛化能力好、精度高和良好非线性处理能力,提出采用小波神经网络用于水质评价建模,把评价结果和传统评价实验结果进行对比,证明该想法的可行性。2.由于传统小波神经网络算法,存在收敛速度慢等缺点,因此引入自适应学习和动量因子,加快网络学习速度,提高网络的学习能力。3.由于传统小波神经网络算法易陷入局部极小,将遗传算法(Genetic Algorithm,GA)引入小波神经网络中,虽然遗传算法具有良好的自适应学习能力和全局搜索能力,但其收敛速度慢,因此将一种改进的遗传算法-自适应遗传算法(Adaptive Genetic Algorithm,AGA)应用于小波神经网络的优化研究。在遗传算法的基础上引入自适应调整参数,加快收敛速度,提高算法的性能;在创建水质评价模型时,先采用自适应遗传算法优化WNN的初始权值、阈值、伸缩和平移参数,然后将选择好的参数作为改进WNN的初始参数值,该方法结合了AGA算法的全局搜索能力以及自适应动量梯度下降法的局部搜索能力,经过仿真结果比较研究,证明该理论的可实现性。4.分别对传统WNN算法、改进WNN算法和AGA算法建立基于小波神经网络的水质评价模型,进行仿真实验,对实验结果进行对比分析。研究结果表明:采用AGA算法的小波神经网络模型比其他方法有较大的提高,该方法可以对水环境的评价有较高的准确性和有效性。最后,创建基于WNN的水质评价图形用户界面(GUI),方便用户的使用。
[Abstract]:Water resource is an irreplaceable resource. In recent years, domestic and international attention has been attached to the protection and management of water resources. However, with the development of science and technology and industry, the problem of water resources still restricts the development of society and ecological environment. At the same time, the traditional water quality assessment methods face the complexity of water environmental problems and lack of efficient treatment efficiency. Therefore, it is urgent to improve the protection measures of water resources. The development of artificial neural network (Ann) has brought a new direction to water quality research. At present, there have been a lot of research on water quality based on artificial neural network at home and abroad. Based on the previous researches on artificial neural network and water quality evaluation, the theory, structure and algorithm of wavelet neural network are deeply studied in this paper, and then wavelet neural network (WNN) is applied to water quality evaluation. The main research includes the following aspects: 1. In view of the limitations of traditional water quality assessment methods, wavelet neural network is used to model water quality evaluation by using wavelet neural network, which has the advantages of fast convergence, good generalization ability, high precision and good nonlinear processing ability. Compare the evaluation results with the traditional experimental results to prove the feasibility of the idea. 2. 2. Because the traditional wavelet neural network algorithm has some shortcomings such as slow convergence speed, so the adaptive learning and momentum factor are introduced to accelerate the learning speed of the network and improve the learning ability of the network. Because traditional wavelet neural network algorithm is easy to fall into local minima, genetic algorithm (GA) is introduced into wavelet neural network. Although genetic algorithm has good adaptive learning ability and global searching ability, its convergence speed is slow. Therefore, an improved genetic algorithm (Adaptive genetic algorithm) is applied to the optimization of wavelet neural networks. On the basis of genetic algorithm, adaptive adjustment parameters are introduced to accelerate the convergence speed and improve the performance of the algorithm, and the adaptive genetic algorithm is used to optimize the initial weight, threshold, scaling and translation parameters of WNN when the water quality evaluation model is created. Then the parameters are selected as the initial parameters of the improved WNN. The method combines the global search ability of the AGA algorithm and the local search ability of the adaptive momentum gradient descent method. The simulation results are compared and studied. The realizability of this theory is proved. 4. The traditional WNN algorithm, the improved WNN algorithm and the AGA algorithm are used to establish the water quality evaluation model based on the wavelet neural network, and the simulation experiments are carried out, and the experimental results are compared and analyzed. The results show that the wavelet neural network model based on AGA algorithm is more accurate and effective than other methods. Finally, the graphical user interface (GUI) for water quality evaluation based on WNN is created to facilitate the use of users.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：X824;TP183

【参考文献】