TL;DR:
- Chengping Chai leads a groundbreaking research project at ORNL to enhance seismic data processing for geothermal and carbon capture projects.
- The project combines machine learning techniques and advanced algorithms to improve accuracy and speed.
- Seismic data helps understand subsurface fractures, ensuring safe and effective project implementation.
- A machine learning-enhanced seismic monitoring workflow is being developed for the rapid processing of seismic event catalogs.
- Data from two sites, Oklahoma and EGS Collab, is used to validate the effectiveness of the approach.
- Deep learning models significantly reduce processing time and improve seismic event locations.
- The technique aids in identifying fractures and optimizing operations in geothermal projects.
- The preliminary results show a potential 99.9% reduction in processing time.
Main AI News:
A groundbreaking research endeavor led by Chengping Chai, a distinguished geophysicist affiliated with the Department of Energy’s esteemed Oak Ridge National Laboratory (ORNL), is poised to revolutionize the processing of seismic data. By harnessing the power of cutting-edge machine learning techniques and advanced seismic data processing algorithms, Chai aims to accelerate the development of faster and more accurate methodologies. These advancements hold immense potential for mitigating risks in geothermal and carbon capture projects, paving the way for enhanced safety and operational efficiency.
Collaborating closely with ORNL’s esteemed scientist, Monica Maceira, Chai’s tireless pursuit of seismic excellence builds upon Maceira’s extensive experience in managing the laboratory’s seismology portfolio. Their shared vision encompasses utilizing seismology as a powerful tool to unravel the intricacies of subsurface fractures, offering invaluable insights for both geothermal and carbon capture initiatives. By gaining a comprehensive understanding of the location and movement of these fractures, stakeholders can make informed decisions that optimize the effectiveness and safety of their projects.
Chai emphasizes the importance of leveraging seismic data analysis tools at various stages of these ventures, from initial site selection and borehole drilling to ongoing monitoring during the injection of water or carbon dioxide. Such continuous monitoring enables the assessment of fluid movement patterns and fracture evolution. Post-project evaluation ensures that fractures remain aligned with the desired objectives or identifies any deviations that warrant attention.
At the heart of this ambitious project lies the development of an innovative seismic monitoring workflow fortified by machine learning capabilities. This cutting-edge workflow facilitates the rapid and automatic processing of seismic event catalogs, empowering researchers to analyze both natural and induced seismicity across diverse spatial scales, ranging from 10 meters to 100 kilometers. To enrich the workflow’s efficacy, data was meticulously collected from two distinct sites: the Oklahoma region (10-kilometer scale) and the EGS Collab site (10-meter scale).
In the Oklahoma scenario, a staggering number of approximately 235,000 three-component body wave seismograms were diligently procured. Manual picking to measure signal arrival times for this extensive dataset would have been an arduous and time-consuming task. However, the advent of deep learning models has ushered in a remarkable transformation, reducing the phase picking process to a mere 38 minutes—a staggering improvement compared to the over 100 days typically required by human analysts.
Furthermore, the integration of deep learning has yielded superior seismic event locations. Notably, the deep-learning-derived picks have exhibited tighter linear trends in comparison to manual picks. Additionally, the results obtained through deep learning showcase a remarkable degree of alignment with the moment tensor solutions derived from the highly regarded Saint Louis University earthquake catalog.
Monica Maceira highlights that activities involving subsurface extraction or injection may induce changes in stress levels, leading to seismicity. A pertinent example can be found in Oklahoma, where the introduction of fracking coincided with an upsurge in earthquake occurrences. To safeguard operations and maintain optimal safety standards, the implementation of a system capable of swiftly and accurately processing high-resolution seismic data is paramount. Such a system empowers operators to make real-time decisions that can adapt and optimize their processes on the fly. Geothermal projects, in particular, stand to gain immensely from accurate seismic models, as they assist in identifying permeable channels—facilitating the flow of fluids and maximizing operational outcomes.
Although Chai’s technique is still undergoing refinement based on the study’s findings, preliminary results obtained from the deep learning approach are exceptionally promising, showcasing the potential for a remarkable 99.9% reduction in processing time. This significant time-saving breakthrough holds the key to accelerating research efforts and operational decision-making, ultimately advancing the efficiency and safety of geothermal and carbon capture projects.
Conclusion:
ORNL’s seismic research project, spearheaded by Chengping Chai, represents a significant advancement in the field of geothermal and carbon capture projects. The integration of machine learning techniques and advanced algorithms not only accelerates seismic data processing but also enhances accuracy, leading to safer and more effective project implementation. This breakthrough has the potential to revolutionize the market, enabling stakeholders to make data-driven decisions, optimize operations, and pave the way for a more sustainable energy landscape. The industry can expect improved efficiency, reduced risks, and enhanced performance as a result of these pioneering research efforts.
