← Back to Gallery

Supervised and Unsupervised Machine Learning Applications for Induced Seismic Data Analysis at Illinois Basin Decatur Project Site

Quantifying in-situ subsurface conditions and understanding of slip mechanisms along the faults are critical to reducing risks of induced seismicity and improving subsurface energy activities. In this work, we present a novel integration of both supervised and unsupervised machine learning methods to process and characterize microseismic data obtained at the Illinois Basin - Decatur Project (IBDP) site where a CO₂ injection process lasted three years (2011-2014) as a field demonstration project. For detection and phase picking of microseismic (MS) data, a set of preprocessing and data augmentation techniques were used to feed waveform time-frequency information to a convolutional neural network (CNN) to accurately detect true events and estimate p and s wave arrivals (>98% accuracy with testing data). After training CNN-based detection model with data from MS cluster #2 at the IBDP, the CNN model was retrained with another cluster #4 to evaluate transfer learning approach. In addition, the original PhaseNet model that was developed based on conventional seismic data was retrained to MS events to accurately obtain p and s arrival times. In both cases we achieved higher true event detection rate compared to the original catalog (manual picks). We discuss the advantages of each method in terms of their ability to detect and/or phase pick. Second, an unsupervised machine learning using the Nonnegative Matrix Factorization and the Hidden Markov Model was used to construct a time dependent probabilistic architecture. The resulting spatio-temporal patterns are taken as fingerprints of spatio-temporal waveform characteristics related to changes in pore pressure and stress caused by CO₂ injection. This study will improve characterizing seismic waveforms by machine learning approaches and the detection of low-magnitude seismic events leading to the discovery of hidden fault/fracture systems. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

Session: Imaging, Monitoring and Induced Seismicity: Applications to Energy and Storage

Type: Oral

Room: Grand C

Date: 4/20/2022

Presentation Time: 04:45 PM Pacific

Presenting Author: Hongkyu Yoon

Student Presenter: No

Additional Authors