Goto

Layer | Fill | Outline |
---|

Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|

Fill | Stroke |
---|---|

We carry out the inversion of marine controlled-source electromagnetic data using real coded genetic algorithm to estimate the isotropic resistivity. Unlike linearized inversion methods, genetic algorithms belonging to class of stochastic methods are not limited by the requirement of the good starting models. The objective function to be optimized contains data misfit and model roughness. The regularization weight is used as a temperature like annealing parameter. This inversion is cast into a Bayesian framework where the prior distribution of the model parameters is combined with the physics of the forward problem to estimate the aposteriori probability density function in the model space. The probability distribution derived with this approach can be used to quantify the uncertainty in the estimation of vertical resistivity profile. We apply our inversion scheme on three synthetic data sets generated from horizontally stratified earth models. For all cases, our inversion estimated the resistivity to a reasonable accuracy. The results obtained from this inversion can serve as starting models for linearized/higher dimensional inversion.

Presentation Date: Monday, October 15, 2018

Start Time: 1:50:00 PM

Location: Poster Station 13

Presentation Type: Poster

annual meeting, Artificial Intelligence, conductivity, CSEM data, electromagnetic data, evolutionary algorithm, frequency-domain marine controlled-source electromagnetic data, genetic algorithm, geophysics, inversion, iteration, machine learning, Mallick, marine controlled-source electromagnetic data, model parameter, model space, optimization problem, Reservoir Characterization, resistive layer, resistivity, seg international exposition, synthetic data, Upstream Oil & Gas, well log data

SPE Disciplines:

Technology:

We developed a multi-objective optimization strategy for inverting marine controlled source electromagnetic data The second category of inverse methods called the using a fast-non-dominated sorting genetic algorithm. Here, probabilistic methods explore the posterior, which is the we propose a robust stochastic global search method which probability of a certain earth model given the observed data considers the objective as a two-component vector and (Trainor-Guitton and Hoversten, 2011). Some examples simultaneously minimizes both components: data misfit and where stochastic methods have been applied include the model roughness. By providing an estimate of the entire set inversion of resistivity sounding data using simulated of the Pareto-optimal solutions, the method allows a better annealing (SA) (Sen et.

SEG-2019-3215410

Artificial Intelligence, constable, CSEM data, data misfit, evolutionary algorithm, genetic algorithm, geophysics, horizontal resistivity, inversion, MacGregor, machine learning, marine controlled source electromagnetic data, model roughness, multi-objective inversion, multi-objective optimization, nsga ii, optimization problem, Reservoir Characterization, reservoir layer, resistivity, seg international exposition, Upstream Oil & Gas, vertical resistivity

SPE Disciplines:

Technology:

We carry out the two-dimensional inversion of marine controlled-source electromagnetic data from the SEG advance modeling program using MARE2DEM Software.We applied this inversion on three survey lines from the given data set to image the salt body and delineate thin hydrocarbon reservoirs that are present near the salt flanks.The inversion was unconstrained and did not use any a priori information about the salt body from the seismic imaging or nearby well logs. Despite the complex 3D structure of thesalt model, our inverted results agree well with the truemodel demonstrating the robustness of the method in imaging the reservoirs and their lateral extents without any prior information.

Presentation Date: Tuesday, October 16, 2018

Start Time: 9:20:00 AM

Location: Poster Station 15

Presentation Type: Poster

annual meeting, application, CSEM data, electromagnetic data, exploration, frequency-domain marine-controlled-source electromagnetic data, geophysics, grid, hydrocarbon reservoir, Imaging, information, inversion, line 50, Reservoir Characterization, salt body, SEAM model, seg international exposition, seg seam model, survey line, Upstream Oil & Gas

Oilfield Places:

- Oceania > Australia > Western Australia > North West Shelf > Carnarvon Basin > Scarborough Field (0.99)
- Oceania > Papua New Guinea > Gulf (0.95)

**Summary**

In an inversion for the subsurface conductivity distribution using frequency-domain Controlled-Source Electromagnetic data, various amounts of horizontal components may be included. We investigate which combination of components are best suited to invert for a vertical transverse isotropic (VTI) subsurface. We do this by probing the solutionspace using a genetic algorithm. We found, by studying a simple horizontally layered medium, that if only electric data are used, either the horizontal or the vertical conductivity of a layer can be estimated properly, but not both. Including the crossline electric field does not add additional information. In contrast, including the two horizontal magnetic components along with the two horizontal electric components allows to retrieve a better estimate of some of the VTI parameters. For an isotropic subsurface, the electric field is sufficient to invert for the subsurface conductivity.

Artificial Intelligence, conductivity, conductivity distribution, electric field, em inversion problem, evolutionary algorithm, genetic algorithm, geophysics, global minimum, information, inversion problem, Invert, machine learning, ocean bottom, Reservoir Characterization, seg denver 2014, sensitivity, solution space, solutionspace, subsurface, Upstream Oil & Gas

SPE Disciplines:

Technology: Information Technology > Artificial Intelligence > Machine Learning > Evolutionary Systems (0.65)

Alumbaugh, David (Chevron Energy Technology Company) | Hoversten, Michael (Chevron Energy Technology Company) | Stefani, Joseph (Chevron Energy Technology Company) | Thacher, Catherine (Chevron Energy Technology Company)

**Summary**

A modeling study has been undertaken to investigate the resolution and accuracy of marine magnetotelluric (MT) and controlled source electromagnetic (CSEM) data to image the base of basalt and sediments in an environment representative of the North Atlantic Margin. The three dimensional (3D) model was constructed using regional well logs, seismic and geologic data to create a cube of porosity and saturation that is 32 km long, 16 km wide and 5 km deep. The porosity and water saturations were then converted to resistivity using well log-derived relationships in the sediments above the basalt, three different volcanic units, sub-volcanic sediments, and flat lying basement. Transverse anisotropy was incorporated in most of the units by making the vertical resistivity more resistive than the horizontal, while the flow basalts were made tridiagonally anisotropic to account for lower resistivities along strike caused by faulting, and higher vertical resistivities caused by the flow structures. Simulated MT and CSEM data were computed using a 3D finite difference algorithm. A two-dimensional (2D) pixilated inversion algorithm was used to produce images of the synthetic MT and CSEM data along lines that were roughly perpendicular to the regional strike of the basalt structure. The images show that the MT data by themselves lack resolution to warrant use alone, but help to constrain the inversion process when jointly inverted with CSEM data. Constraints in the form of a discontinuity at the top of the basalt further enhance image resolution.

algorithm, boundary, CSEM data, data misfit, discontinuity, em imaging, hyaloclastite, Imaging, inversion, investigate em imaging, iteration, log analysis, misfit, MT data, Reservoir Characterization, resistivity, resolution, sediment, seg houston 2013, structural unit, sub-basalt structure, Upstream Oil & Gas, well logging

SPE Disciplines:

Thank you!