Performance analysis of moving average filtering used in TDLAS carbon dioxide sensing system

Reducing carbon emissions is a global priority due to human impact on atmospheric pollution and the greenhouse effect. Achieving carbon peak and neutrality requires real-time monitoring of CO2 concentrations. However, developing high-sensitivity, portable, and anti-jamming gas detection solutions is challenging. Among spectroscopic techniques, Tunable Diode Laser Absorption Spectroscopy (TDLAS) is highly sensitive for detecting CO2 concentrations. This paper elaborates on the principles of TDLAS for detecting CO2 concentrations and proposes a noise reduction algorithm to meet diverse environmental requirements. Simulations were performed using software to simulate CO2 absorption spectra at approximately 1.57866535μm under high-intensity noise (0.1mW - 1mW). Based on this simulation, we applied the Wavelength Modulation Spectroscopy (WMS) technique to calculate the ratio of the output differential signal's second harmonic intensity to the first harmonic S2f/1f(T) and output power to reduce light intensity influence and improve concentration inversion linearity. The weighted convolutional moving average filtering was utilized to optimize WMS denoising, utilizing weight transfer to make the process more precise and reliable. After analyzing various window functions, it was concluded that a window length of 9 would be the most optimal. The algorithm improved the signal-to-noise ratio (SNR) by 22.435% under these conditions. When the noise level increased fourfold from the original signal, the algorithm enhanced the SNR by 59.514%, enabling reliable CO2 monitoring even under challenging conditions.