DESC0024854
Project Grant
Overview
Grant Description
Informer (Induced Seismicity Transformer)
Awardee
Funding Goals
INFORMER (INDUCED SEISMICITY TRANSFORMER)
Grant Program (CFDA)
Awarding Agency
Funding Agency
Place of Performance
San Mateo,
California
94402-2516
United States
Geographic Scope
Single Zip Code
Related Opportunity
Stottler Henke Associates was awarded
Project Grant DESC0024854
worth $199,950
from the Office of Science in February 2024 with work to be completed primarily in San Mateo California United States.
The grant
has a duration of 10 months and
was awarded through assistance program 81.049 Office of Science Financial Assistance Program.
The Project Grant was awarded through grant opportunity FY 2024 Phase I Release 1.
SBIR Details
Research Type
SBIR Phase I
Title
11b InFormer (Induced Seismicity Transformer)
Abstract
Induced seismic activity is a category of seismic activity that is caused by human activity, which includes hydraulic fracturing, deep and shallow salt water disposal, and carbon sequestration, among others. In the past decade, induced seismic activity has increased in the United States, especially in the central US, causing continued danger and disruption to local communities as well as to commercial operations. While it is known that changes in pore pressure and/or poroelastic stress from injection or extraction of fluids is the primary trigger mechanism for induced seismicity, the broader causes for these changes are complex and multi-dimensional; meanwhile, current regulatory attempts are broad and reactive in nature and thus expose local communities to continued risk. Machine learning techniques like transformers have demonstrated the ability to learn complex, multivariate relationships that have been traditionally difficult to study. The proposed InFormer system will use an approach inspired by Spacetimeformer, a transformer-based multivariate time series forecasting technique which will be adapted to analyze the complex spatial and temporal relationships between well injection and production activity and induced seismicity. InFormer aims to find latent connections between variables; the tool will be useful to regulators and industry to evaluate existing and proposed operations on their seismic hazard before they occur. In the proposed effort, key seismological features from an existing set of injection data will be cleaned and engineered. This dataset will then be used to train a prototype machine learning pipeline based on the SpacetimeFormer architecture. The end-to-end pipeline will be evaluated on both in-distribution data and out-of-distribution data, i.e., data at the same geographical location as the training data and data at a different geographical location. This partition will demonstrate the efficacy of the proposed methods as well as their generalizability and domain adaptation capabilitiesÅthe evaluation results will be used to produce visualizations and reports representative of the full Phase II system. The proposed solution will help inform Government agencies how to best regulate injections (primarily saltwater disposal), which will help reduce environmental and societal impacts related to induced seismicity. InFormer will also help private companies evaluate new injection sites and potential risks associated with the locations. This will provide a method for prospectors to target construction in lower- impact areas and inform buyers of expected future site instabilities.
Topic Code
C57-11b
Solicitation Number
DE-FOA-0003110
Status
(Complete)
Last Modified 3/11/24
Period of Performance
2/12/24
Start Date
12/11/24
End Date
Funding Split
$199.9K
Federal Obligation
$0.0
Non-Federal Obligation
$199.9K
Total Obligated
Activity Timeline
Additional Detail
Award ID FAIN
DESC0024854
SAI Number
None
Award ID URI
SAI EXEMPT
Awardee Classifications
Small Business
Awarding Office
892430 SC CHICAGO SERVICE CENTER
Funding Office
892401 SCIENCE
Awardee UEI
FFEAH6Z5CK27
Awardee CAGE
0K501
Performance District
CA-15
Senators
Dianne Feinstein
Alejandro Padilla
Alejandro Padilla
Modified: 3/11/24