Low-frequency distributed acoustic sensing (LF-DAS) exploits the optical phase shift of Rayleigh backscatter in fiber optic cables to obtain distributed measurements of changes in strain and temperature. Fiber optic cables are often installed for multistage hydraulic fracture diagnostics in horizontal wells. LF- DAS in an untreated well provides far-field strain measurements while offset wells are hydraulically fractured. Such a configuration is called cross-well LF-DAS sensing. Cross-well LF-DAS measurements have proven useful to diagnose frac hits, fracture azimuth, planarity, cluster efficiency, fracture propagation rates, and the dynamic distance to the fracture front. In contrast, in-well LF-DAS is conducted on the actively fractured well. Due to cool fracture fluid being injected at high injection rates, the strain component of the LF-DAS response is largely obscured by temperature changes. In permanent fiber optic cable installations, distributed temperature sensing (DTS) is often conducted simultaneously with LF- DAS. An opportunity exists to decouple the temperature and strain components of LF-DAS sensors to observe strain changes on in-well LF-DAS.
The LF-DAS response is modeled as linearly dependent on strain and temperature changes. Theoretical LF-DAS temperature and strain sensitivity coefficients are derived based on changes to the index of refraction and length of the fiber. Using the DTS measurements, temperature changes are computed, smoothed, filtered, and compared to the LF-DAS response. Cross-plots of the in-well LF-DAS measurements and temperature changes from DTS measurements far from the actively fractured region are used to validate the theoretical sensitivity coefficients. Uncertainty in the temperature component of the LF-DAS response is quantified. The difficulty in corresponding the different spatial and temporal resolutions of the DTS and LF-DAS measurements is overcome by comparing the responses over a moving temporal and spatial window. If the LF-DAS response at the center of the window agrees with the DTS response within uncertainty, the measurement is filtered out. After filtering, the remaining non-zero in-well LF-DAS measurements are due to changes in strain. The data is then visualized in waterfall plots. The results indicate that the theoretical and observed strain and temperature coefficients agree within 10%. After the temperature component of the in-well LF-DAS response is extracted, the remaining non-zero measurements are located primarily within the actively treated region. Locations with peaks in the strain response are interpreted to indicate fracture initiation points. These fracture initiation points are compared with in-well DTS and high frequency DAS noise measurements across multiple stages to better understand fracture initiation along the horizontal well.