Problem
An important application of the geophysical inverse theory is seismic full waveform inversion (FWI), where seismic data recorded at the surface, sea floors or in boreholes is used to estimate subsurface parameters. In FWI problem, one tries to minimize an objective function that measures the misfit between observed data and estimated data with respect to model parameters. The commonly used L2 norm objective function have multiple local minima due to the highly non-linear relationship between model parameters and data. The traditional local optimization based methods only go along the direction that reduces the objective function. It would easily stuck at one of the local minimum and fail to converge to geological meaningful solution if the starting model is far away from the global minimum. The invention overcome this issue by extensively searching for the possible solutions in the parameter space.
Solution
The invention is a starting model independent full waveform inversion (FWI) technique. It is developed to directly estimate subsurface properties from seismic data acquired on the surface or in the well bore hole without user providing specific starting models. The invention would tests various models with different features, thus sufficiently search the parameter space and not stuck at a local region. There is a slowly changing criterion that determines whether a model should be accepted or not. This changing criterion is to guarantee the convergence to the global minimum. The invention is able to find the global minimum region that provides accurate subsurface properties estimations. The inversion result is independent of starting models.
Features
The invention is able to jump across the parameter space and test models with different features. It not only allows the optimization process go along the direction that reducing the objective function but also occasionally allows the optimization process accept worse models. The gradient of the objective function also guide the search direction to help find a better model. There is a decision process that determines whether a worse model is accepted or not. Such decision process slowly changes to guarantee the optimization process finds accurate global solutions.
Benefits
The invention does not require accurate background model as the starting model. The invention is able to provide accurate subsurface properties estimation regardless of different starting models. It greatly relax the dependency of FWI on using the correct physics and seismic acquisition systems. Additionally, the invention needs minimum human intervention, which automates the subsurface properties estimation process.
Markets
Oil and gas companies, geothermal companies, geophysical services companies, medical device companies
