A multi-step joint noise reduction method of the distributed acoustic sensor data for flow rate monitoring

Abstract

Distributed acoustic sensor (DAS) based on phase-sensitive optical time domain reflectometry (Ф-OTDR) has become a novel and effective acoustic perception tool in various fields. However, when applied to the fluid monitoring, the target acoustic signal obtained by DAS becomes blurred due to the weak amplitude of the Rayleigh backscattering (RBS) signal itself, as well as the interference caused by the working environment. Here, we propose a multi-step joint (MSJ) noise reduction method to suppress the noise of the raw DAS data in flow rate monitoring effectively, which contains four steps, i.e., moving average (MA), wavelet packet transform (WPT), bandpass filtering (BPF), and envelope extraction (EE). To evaluate the noise suppression effect of the DAS data, we build a novel criterion, the distance series cumulative signal-to-noise ratio (dCSNR). Compared to the raw data, the dCSNRs of the denoised data at different flow rates are improved by more than 15 dB. In addition, the R² value of the fitted curve between flow rate and relative acoustic energy (RAE) using the denoised data by the MSJ method is improved from 0.9086 to 0.9635 compared to that using the raw DAS data. The established flow rate prediction model with the noise-suppressed DAS data also demonstrates a mean absolute percentage error (MAPE) as low as 8.81 % and the average relative uncertainty is 4.1 %. We believe the efforts here, including the DAS data denoising method (i.e., MSJ method) and the DAS data quality assessment criterion (i.e., dCSNR), will advance the DAS application in various fields, especially the oil and gas industry.