Temperature extraction from Brillouin sensing based on temporal convolutional networks

Abstract

The Brillouin optical time-domain sensing system has become a hot spot of research due to its ability to seamlessly monitor the temperature and strain variations in optical fibers along the line. Given its current limitations of low accuracy and inadequate real-time performance in long-distance monitoring, the Brillouin gain extraction temperature method based on temporal convolutional networks is proposed. On this basis, we established a Brillouin optical time-domain experimental system where comprehensive simulations and tests were conducted to assess the temperature extraction performance under different conditions. Besides, a comparison was made between the system and traditional methods like Lorentz fitting method and extreme learning machine method. The results have suggested that the temporal convolutional network exhibits remarkable measurement accuracy, even in scenarios with low signal-to-noise ratios and large sweep frequency steps.