Study On The Method Of SO₂ Concentration Measurement Based On The DOAS And Support Vector Regression

With the rapid development of modern industry and transportation, energy consumption is increasing, and the problem of the air pollution is more and more serious in recent years. It not only seriously affects people’s physical and mental health, but also greatly restricts the sustainable development of nature and society. As one of the main polluting gases, high concentration of sulfur dioxide will result in a series of environmental problems, such as smog, acid rain, and lead to significant impact and loss to people’s life. So it’s urgent and necessary to develop the gas analyzer for the detection of sulfur dioxide concentration, which can realize on- line monitoring and real-time control of sulfur dioxide emissions.Currently, gas analyzers mainly adopt the spectrum measurement technology to detect gas concentration, and UV DOAS is suitable for the detection of nitrogen oxide and sulfur dioxide, which uses molecular narrow-band absorption characteristics. With the merits of simple structure, high response speed and precision, it has been successfully applied in atmospheric environmental monitoring. The contribution of this thesis is mainly to research the prophase work of the UV gas analyzer, and makes in-depth study in the analyzer software algorithm to detect the concentration of sulfur dioxide.In this paper, there are three main parts. First, introduce the principle of gas molecules absorb and sulfur dioxide molecular spectroscopy based on the theory of absorption spectra, and illustrate the law of light absorption and the theoretical derivation process of difference spectrum technology; Second, set up the experimental platform based on the difference technology principle, briefly illustrate the process of data collection, design of filter for data preprocessing, analyze data processing in detail, and compare different ways of fitting inversion effect, finally choose the piecewise linear fitting method; Third, propose the method of SO2 concentration measurement based on manifold learning and genetic-support vector regression, introduce the method of manifold learning, the genetic algorithm optimization and support vector regression theory, and through the experimental results, show the validity and practicability of the method.In the paper, the software algorithm is designed for determination of sulfur dioxide concentration based on the difference technology principle, and the effectiveness and the deficiencies are illustrated by the experimental data. In addition, the method of SO2 concentration measurement is proposed based on manifold learning and genetic-support vector regression. It is compared to ultraviolet difference technology, which shows the advantages and disadvantages. Finally, we summarize and prospect the software algorithm of this paper, and propose practical suggestions.