oa Artificial Intelligence Based DAS Microseismic Source Mechanism Inversion

Artificial intelligence technologies perform feature extraction by building deep models and training large-scale datasets, which can improve the ability to deal with complex problems. Distributed Acoustic Sensing (DAS) is a new data acquisition technology, which can significantly improve the real-time monitoring capability of reservoir stimulation. In hydraulic fracturing microseismic monitoring, high-pressure fluid injecting into shale reservoirs will change reservoir stress and generate microseismic. The study of the microseismic focal mechanism induced by the fractures can reveal the state of reservoir stress and effectively optimize the design of fracturing, and realize efficient development and safe production.

In hydraulic fracturing DAS microseismic monitoring, fiber is deployed along horizontal well to obtain monitoring data of the entire well section, high acquisition density and spatial continuity. However, the current single well DAS monitoring makes the coverage along the vertical direction of the optical fiber smaller, which makes the analysis of the microseismic source mechanism more uncertain. We propose the network for high-resolution and low-uncertainty microseismic focal mechanism inversion for the DAS microseismic data.

In the method, we create the network of the focal mechanism inversion using 4 repeated blocks and 3 fully connected (Fc) layers; then create the training dataset using synthetic DAS events and real background noise; finally train and validate the network using the training dataset. The results show that the network can obtain high-resolution and low-uncertainty source mechanism for both low- and high-SNR datasets.