基于紫外-可见光谱法水质 COD检测方法与建模研究

发布时间：2019-05-24 08:12

【摘要】：近年来,随着我国经济飞速发展和城市化脚步加快,随之而来的水污染问题日趋严重,此问题已经成为我国乃至全世界水资源面临的最严重问题之一。化学需氧量(COD)作为评价水体污染程度的重要指标,可以表征水体有机物的浓度。紫外-可见光谱法检测COD无二次污染、周期短、可实现在线检测,是一种绿色检测技术。本文针对紫外-可见光谱法水质COD检测方法和建模开展了如下研究工作:1.光谱系统与光谱采集为了检测水质COD值,本文开发了一种紫外-可见光谱水质COD检测系统,并进行实验室邻苯二甲酸氢钾标准溶液制备与检测,采集紫外-可见吸光度光谱数据。2.光谱数据预处理技术研究针对原始光谱受到大量噪声影响的问题,本文需要一种去噪过程中尽可能少丢失真实信息的方法进行原始光谱去噪,小波分析可以满足要求。本文采用小波函数db8,对原始光谱进行5层小波分解,然后利用软阈值方式进行量化处理,重构后的水质COD光谱曲线十分光滑,去噪效果显著。小波去噪后依然存在光谱信息冗余和多重共线性问题,采用主成分分析法对光谱数据进行降维处理,有效去除冗余信息,保留有用特征信息,提高了机器学习效率。3.水质COD检测预测模型研究由于紫外-可见光光谱数据与水质COD值的关系具有复杂的非线性,无法采用传统的机理建模方法。建立基于BP神经网络的水质COD预测模型,可以有效的预测水质COD值。为了提高预测精度,采用改进的鲸鱼优化算法优化BP神经网络参数,建立基于鲸鱼优化算法BP神经网络的水质COD预测模型,预测结果表明,该模型的预测精度更高,可以应用于水质COD检测的预测。4.优化算法改进针对基本鲸鱼优化算法收敛速度慢、收敛精度低的缺陷,提出改进的鲸鱼优化算法(MWOA),MWOA主要研究了种群初始化机制和非线性自适应权重策略。仿真结果表明,改进的算法能在寻优过程中保持初始种群多样性,具有更好的收敛速度和收敛精度。
[Abstract]:In recent years, with the rapid development of economy and the acceleration of urbanization in China, the problem of water pollution is becoming more and more serious, which has become one of the most serious problems faced by water resources in China and even in the world. Chemical oxygen demand (COD), as an important index to evaluate the pollution degree of water body, can characterize the concentration of organic matter in water body. UV-vis spectroscopy is a green detection technology because of its no secondary pollution, short period and on-line detection. In this paper, the following research work has been carried out on the COD detection method and modeling of water quality by UV-vis spectroscopy: 1. In order to detect the COD value of water quality, a UV-vis spectral water quality COD detection system was developed in this paper, and the standard solution of potassium hydrogen phthalate was prepared and detected in laboratory. Collection of UV-vis absorbance spectral data. 2. In order to solve the problem that the original spectrum is affected by a lot of noise, this paper needs a method to Denoise the original spectrum with as little real information as possible in the process of denoising, and wavelet analysis can meet the requirements. In this paper, the wavelet function db8, is used to decompose the original spectrum by 5 layers of wavelet, and then the soft threshold method is used to quantify the original spectrum. The reconstructed COD spectral curve of water quality is very smooth and the denoising effect is remarkable. After wavelet denoising, there are still spectral information redundancy and multiple collinearity problems. Principal component analysis (PCA) is used to reduce the dimension of spectral data, effectively remove redundant information, retain useful feature information, and improve the efficiency of machine learning. Study on the prediction model of COD detection and prediction of water quality because of the complex nonlinear relationship between UV-vis spectral data and COD value of water quality, the traditional mechanism modeling method can not be used. The COD prediction model of water quality based on BP neural network can effectively predict the COD value of water quality. In order to improve the prediction accuracy, the improved whale optimization algorithm is used to optimize the parameters of BP neural network, and a water quality COD prediction model based on whale optimization algorithm BP neural network is established. The prediction results show that the prediction accuracy of the model is higher. It can be applied to the prediction of COD detection of water quality. 4. Aiming at the defects of slow convergence speed and low convergence accuracy of the basic whale optimization algorithm, an improved whale optimization algorithm (MWOA), MWOA is proposed, which mainly studies the population initialization mechanism and nonlinear adaptive weight strategy. The simulation results show that the improved algorithm can maintain the initial population diversity in the optimization process, and has better convergence speed and accuracy.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X832

【参考文献】