Novel discoveries are going to prove essential for attaining a significant percentage of oil output rather than solely depending on existing fields to keep up with the energy demanding of the United States. As one of the critical sectors of Exploration and Production, seismic surveys are used to produce detailed images of local geology to determine the location and size of possible oil and gas reservoirs. Despite this essential role and over 50 years of evolution, the technology is far from optimal and is still rapidly improving. Full waveform inversion (FWI) is a recent exciting technique for the seismic industry because of its ability to deliver very detailed subsurface imaging. In order to obtain a stable FWI result, low-frequency data play a very important role. However, due to the limitation of acquisition hardware and the extremely high cost, low-frequency data are usually not available. In this project, we proposed a novel computational approach to synthesize the low- frequency data through a deep learning network, which is trained to map the high- frequency data to the low-frequency space. In addition, an unsupervised transfer learning technique will be implemented to ensure the adaptiveness of the mapping function for different field surveys. Phase I will demonstrate the feasibility of the proposed approach by verifying the performance improvement of the full waveform inversion over both benchmark synthetic data and field data after using low-frequency data synthesized by the deep learning network. This proposed approach, if proven to be feasible, will close the gap between the full waveform inversion requirements and the data sets available. This project will lead to a large-scale full waveform inversion based commercial product that provides the highest level of subsurface velocity model accuracy in the industry. The product would significantly improve the performance of the oilfield exploration and production, and be of great interests to other important industrial applications such as enhanced geothermal systems, CO2 sequestration, and subsurface waste disposal and environmental remediation.
