← Back to Sessions

WITHDRAWN DAS-recorded Microseismic Monitoring in Geothermal Field Stimulation With Waveform Imaging and Deep Learning

Description: 

WITHDRAWN Enhanced Geothermal Systems (EGS) have the potential to provide clean energy on a large scale by amplifying the permeability of basement rocks through fracture stimulation. Accurate mapping of the stimulated fracture network is essential to optimize EGS efficiency. Microseismic events generated by the rupture of fractures provide the most reliable information for this purpose. The advent of distributed acoustic sensing (DAS) using fiber optic cables enables high-resolution recording of microseismic signals with improved signal-to-noise ratios. This advancement enhances the ability to map the finer details of fracture networks, necessitating the development of new algorithms to leverage the data provided by DAS technology. In this study, I propose a methodology to detect and locate microseismic events recorded by three DAS fibers installed in wells at a geothermal project near the FORGE site. The wells and corresponding fibers have varying geometries: horizontal, vertical, and deviated.

The proposed methodology combines waveform seismic imaging with deep learning to detect and locate seismic events. Sequential time-lapse images of potential microseismic sources are generated, and a deep convolutional neural network is trained to classify these images as containing a source or not. If a source is detected, another neural network determines its location and timing. This approach offers higher sensitivity by amplifying the seismic signal relative to background noise during imaging. Additionally, DAS-specific noise is attenuated, resulting in an improved signal-to-noise ratio and enhanced detectability. The imaging process produces point-like representations of sources at the time of microseismic triggering, facilitating accurate estimation of source location and time. The results demonstrate that the trained neural networks efficiently interpret these images, delivering reliable and precise source estimates. Overall, the methodology proves to be both robust and successful, highlighting its potential for advancing microseismic event detection and location in EGS applications.

Session: Seismology for the Energy Transition [Poster]

Type: Poster

Date: 4/16/2025

Presentation Time: 08:00 AM (local time)

Presenting Author: Julio

Student Presenter: Yes

Invited Presentation: 

Poster Number: 129

Authors