2D Inversion for Structure with Angular Unconformity on Example of Troll Field

Summary Angular unconformity of bed boundaries and the oil-water contact is common for the Troll field in Norway. The depth of investigation and azimuthal sensitivity of extra-deep azimuthal resistivity (EDAR) measurements make it possible to image such complex structures. The paper describes an approach to real-time 2D inversion based on artificial neural networks (ANNs). We propose a 2D parametric model with two non -parallel boundaries suitable for scenarios with angular unconformity and pinch-out. The 2D inversion algorithm utilizes the Levenberg-Marquardt optimization method and the ANN-based solver. Training of the ANNs for the parametric model is performed using a synthetic database containing samples with the model parameters and corresponding tool responses. The inversion is performed interval by interval first with the 1D layer-cake model. If any “non-1D” behavior is observed in the data or the resulting picture, then we switch to the 2D model. On the example of one of the wells in the Troll field, we demonstrate that the described approach reconstructs the oil-water contact and the unconformable boundary with the performance fast enough for real-time.