Research On Key Technology Of Methane Detection System Based On TDLAS

As a kind of clean energy,natural gas is widely used,but there are also transportation risks and use risks.Methane is the main component of natural gas.Detection of methane leakage can prevent disastrous accidents caused by natural gas leakage.Tunable diode laser absorption spectroscopy(TDLAS)is widely used due to its advantages of high resolution,fast response,accurate measurement and non-contact.However,the ambient light interference introduced by open light louie will affect the measurement results.Aiming at the problem that the analog denoising algorithm has weak frequency selection ability in processing strong noise signals and low signal-to-noise ratio,a denoising algorithm based on Visu Shrink wavelet threshold is proposed.The threshold width is calculated by noise variance,and the unstable discontinuity caused by hard threshold jump discontinuity is optimized by Visu Shrink function.The signal-to-noise ratio,mean-square error and waveform similarity coefficient parameters are calculated experimentally,and the meansquare error coefficient,signal-to-noise ratio and normalized correlation coefficient of Visu Shrink wavelet threshold denoising algorithm are 0.01370,28.4 and 0.999,respectively.Compared with the fourth order Butterworth filter,it shows superior performance in waveform recovery ability,smoothness and calculation speed.It can effectively improve the detection accuracy and speed of methane sensor.The source of error in feature extraction of actual measurement results is studied.In view of the accuracy error of discrete sampling peak-seeking algorithm,a feature extraction algorithm combined with Gaussian polynomial fitting peak-seeking algorithm is proposed.Gauss-polynomial is used to fit the discrete points at the peak end of the maximum wave,and the position of peak points is determined by derivation.In the process of obtaining the minimum mean square error by Gaussian fitting,the dependence on the acquisition accuracy is reduced.When the sampling density is below 400 points per second,the sampling accuracy error will reach one-fifth.A practical long-optical TDLAS methane detection device was built,and the gas distribution system was used to obtain array methane gas samples within the concentration range of 0-10000 ppm.The influence of Visu Shrink wavelet threshold denoising algorithm and traditional denoising algorithm on sample analysis effect was compared.The data show that the linear correlation of Visu Shrink wavelet threshold denoising algorithm is improved from 0.91 to 0.94,and the overall error accuracy is reduced from 7% to 3% by error analysis,which proves that the algorithm is actually effective in improving the accuracy of TDLAS detection system.