Real-Time 3D Inversion of Ultra-Deep Resistivity Logging-While-Drilling Data

Abstract

Abstract Ultradeep resistivity logging-while-drilling (LWD) is now a routine service for real-time well landing,geosteering, and reservoir and fluid contact evaluation. Progressing beyond layered earth inversions to three-dimensional (3D) inversions helps improve real-time decisions to deliver better well placement, completion, and production. To this end, the first real-time 3D inversion of ultradeep resistivity LWD data is realized by exploiting the fact that the sensitive volume of a given transmitter-receiver pair is far smaller than the total logging volume. This implies that the global mesh can be decoupled into multiple independent, localized inversion and modeling meshes that are tractable for the efficient solution of the forward and inverse problems in real time using moderate computer resources. The authors' implementation is based on a 3D finite-volume method discretized on locally refined octree meshes. It uses the regularized Gauss-Newton method for minimizing the objective function for data subsets on local inversion meshes, which iteratively update the global mesh. Nonlinear Kalman filtering is applied using prior information on each local inversion mesh from the updated global mesh to introduce new observations optimally. A model study and a case study of trilateral well placement in a mature reservoir in the Norwegian Continental Shelf demonstrate the efficacy of the method. Run times on modest computer resources enable the first real-time 3D inversion of ultradeep resistivity LWD data.