An Intelligent Inversion Method for Azimuth Electromagnetic Logging While Drilling Measurements
Zhengming Kang,Yi Zhang,Haojie Qin,Wei Gan,Gang Chen
DOI: https://doi.org/10.1109/access.2023.3298972
IF: 3.9
2023-08-05
IEEE Access
Abstract:Azimuth electromagnetic (EM) Logging While Drilling (LWD) tools play an important role in geological steering and reservoir evaluation. Its measuring response is amplitude ratio (ATT) and phase difference (PS), which can't directly reflect formation information. To obtain direct formation information such as resistivity and formation boundary, accurate and efficient inversion method is essential. However, the existing inversion methods (i.e., iterative method) have some problems, such as slow computation speed, dependence on initial value selection and easy to be trapped by local minimum. Therefore, this paper proposes an intelligent inversion method for azimuthal EM LWD measurements based on the U-net deep learning network framework. Firstly, an efficient analytical solution is used to generate amounts of EM LWD data. Those samples are divided into training and test sets in a 9:1 ratio, which are used for training and testing of the network, respectively. Then, the network parameters are constantly adjusted during training to ensure its inversion performance. Finally, the trained network is utilized to invert the samples of the test sets, and the inversion results are compared with the forward formation model. The study's results demonstrate that the network is capable of efficiently and precisely inverting both isotropic and anisotropic formations, with a single sample being inverted less than 0.05 seconds. Its noise layer parameters can be improved to successfully invert noisy data, leading to good robustness. In addition, the network has good applicability for the inversion of complex formations. These consequences highlight the significant potential of this method in azimuthal EM LWD inversion applications.
computer science, information systems,telecommunications,engineering, electrical & electronic