Work Package 3 provides hardware sensor fusion and data integration, specifically using machine learning techniques and processes. Significant cost-savings and overall efficiency can be achieved through sensor fusion onto single vehicle platforms. The multiple datasets from these sensing platforms, as well as other sources, need to be combined to create high-fidelity, probabilistic maps of undocumented well locations. These data can range from historical records and production records located in various state databases to data collected in the field by the other work packages that need to be synthesized and analyzed. Challenges associated with this task are to harmonize information across disparate data sources, and to develop algorithms that can pinpoint well locations and their characteristics based on the limited data that are available.
Develop sensor fusion hardware methodologies to simplify collection of multiple data types. Develop computational techniques to combine distinct data streams using machine learning and other statistical algorithms to optimize identification and characterization of UOWs.
- Short term impact
- Create reference remote sensing platforms for collecting multi-modal data to aid in the detection, identification, and characterization of wells that can address the scale of the challenge.
- Develop machine learning methods, processes, and tools to leverage a wide data set and provide intuitive results to allow for efficient localization of UOWs.
- Long term impact
- Provide sensing systems and designs to allow interested parties to effectively collect data to drive well identification and characterization.
- Deploy algorithms, frameworks, and tools to allow interested parties to identity, localize and characterize UOWs and document their findings for the community.