Machine Learning‐Assisted Microearthquake Location Workflow for Monitoring the Newberry Enhanced Geothermal System

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 3
• Main Authors: Leong, Zi Xian, Zhu, Tieyuan
• Format: Journal Article
• Language: English
• Published: United States Wiley 01.09.2024
• Subjects: enhanced geothermal system (EGS); GEOTHERMAL ENERGY; induced seismicity; locating microearthquakes; Newberry EGS; probabilistic multilayer perceptron; source location inversion
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000159

Enhanced geothermal systems (EGS) offer a sustainable energy source but face challenges in accurately locating microearthquakes induced during reservoir stimulation. Locating these microearthquakes provides reliable feedback on the stimulation progress. Current deep learning methods for locating earthquakes require extensive data sets for training, which is problematic as detected microearthquakes are often limited. To address the scarcity of training data, we propose a practical workflow using probabilistic multilayer perceptron (PMLP) which predicts microearthquake locations from cross‐correlation time lags in waveforms. Utilizing a 3D velocity model of Newberry site derived from ambient noise interferometry, we generate numerous synthetic microearthquakes and 3D acoustic waveforms for PMLP training. Accurate synthetic tests prompt us to apply the trained network to the 2012 and 2014 stimulation field waveforms. To enhance the accuracy of source localization, we carefully handpick the P‐arrival times. Predictions on the 2012 stimulation data set show major microseismic activity at depths of 0.5–1.2 km, correlating with a known casing leakage scenario. In the 2014 data set, the majority of predictions concentrate at 2.0–2.9 km depths, consistent with results obtained from conventional physics‐based inversion, and align with the presence of natural fractures from 2.0 to 2.7 km. We validate our findings by comparing the synthetic and field picks, demonstrating a satisfactory match for the first arrivals. By combining the benefits of quick inference speeds and accurate location predictions, we demonstrate the feasibility of using realistic synthetic data set to locate microseismicity for EGS monitoring. Plain Language Summary The development of enhanced geothermal systems (EGS) hinges on accurately locating induced microearthquakes during the reservoir’s stimulation process. The scarcity of microearthquake data complicates the use of traditional deep learning for this purpose. To overcome this, we employ a practical workflow, by simulating numerous synthetic data sets for training. The trained model is eventually applied to real‐world EGS microearthquake data. In this work, we create a realistic geological model of the Newberry EGS site and generate many artificial microearthquake data for deep learning training. During the application on field data from 2012 to 2014 stimulation, the computer model successfully identifies the depth and location of microearthquakes. Our results match well with what we already know about the underground structure, such as the presence of natural fractures in the rock. This study shows that our approach can effectively predict microearthquake locations even when presented with limited earthquake data for training, which is promising for monitoring and improving EGS operations in the future. Key Points We present a novel machine learning workflow to predict microearthquake locations in EGS, addressing data scarcity for training Application to Newberry EGS reveals accurate microearthquake locations, validated against known geological features Employs probabilistic multilayer perceptrons that map cross‐correlation time lags to microearthquake locations