Study On Infrared Spectrum Background Deduction And Quantitative Algorithm Of SF₆ Decomposition Products Under Partial Discharge

Partial Discharge (PD) will happen in Gas Insulated Switchgear (GIS) in operation when there are insulation defects, and SF₆ insulating gases in GIS decompose into various component gases because of PD. Different insulation defects cause different relative contents of SF₆ PD decomposition components, so detection and analysis of decomposition components can be used for condition assessment and fault diagnosis of GIS. Fourier Transform Infrared Spectroscopy (FTIR) is one of the most commonly used methods to analyze SF₆ PD decomposition components now, and features high speed response, high distinguishing precision, multi-composition detection, good linear characteristic, high anti-interference ability, long service life, which makes the qualitative and quantitative analysis of SF₆ decomposition components be carried out effectively.Under the conditions of 0.1 meters and 16 meters optical length gas-cells, standard gas SO₂F₂ is infrared detected. The contrast result indicated that the infrared detection with long optical shows characters of good stability, high resolution, excellent linearity, high signal-to-noise ratio and so on. The infrared detection to the trace components can be accomplished by long optical gas-cell. The scanning number of infrared spectrometer is selected as 32 times, and the optimally resolution is set as 0.5cm-1, then the infrared detection system is adopted to obtain infrared absorption spectrums of some SF₆ decomposition components. According to spectral line profile of component gases, absorption peaks, the influence of spectrum baseline drift is analyzed.Aim to ubiquitous baseline drift, a background deduction method based on complex wavelet is presented, based on wavelet transform features. According to the characteristic that baseline usually belongs to gradual low frequency components of tested signals, this algorithm process consists of decomposing the tested signal into different frequency components, and resetting high frequency coefficients to reconstruct background signal, then subtracting background signal to get the useful signal. Four kinds of analogue signals, which are respectively mixed up with Gaussian curve, slope, complex Gaussian, Gaussian curve connected with slope backgrounds, are structured, and these backgrounds are effectively deducted by this complex wavelet method. After the measured infrared spectrums of three components SO₂F₂, SOF₂, SO2 are processed by the complex wavelet method, the baselines of these signals almost near zero coordinate horizontal axis. Experimental analysis of analog signal and actual signal reflects the availability of the background deduction method based on complex wavelet. Besides, by comparing correction effect between real wavelet and complex wavelet, the advantage of complex wavelet is proved.When Beer-Lambert law is applied for gas infrared quantitative analysis, there are some problems, such as uncertainty of absorption coefficient and fussy calculation of multi-composition detection. An infrared quantitative algorithm based on principal component regression analysis is presented, moreover, three components SO₂F₂, SOF₂, SO2 and their characteristic absorption peaks are selected as research objects. Spectra data and gas concentration are regarded as imports and exports of the algorithm, so the direct corresponding relation, between absorption spectrogram and gas concentration, is established. Concentration samples of three Characteristic components are prepared, afterwards spectra data are obtained and the infrared quantitative forecast model is established on the basis of the algorithm. Forecast results of testing samples and real data indicate that the algorithm and the forecast model have the very good applicable value.