Research Article
A Convolutional Neural Network Approach for Stratigraphic Interface Detection
@INPROCEEDINGS{10.1007/978-3-030-67514-1_42, author={Jingya Zhang and Guojun Li and Juan Zhang and Guanwen Zhang and Wei Zhou}, title={A Convolutional Neural Network Approach for Stratigraphic Interface Detection}, proceedings={IoT as a Service. 6th EAI International Conference, IoTaaS 2020, Xi’an, China, November 19--20, 2020, Proceedings}, proceedings_a={IOTAAS}, year={2021}, month={1}, keywords={Stratigraphic interface detetion Geophysical well logging Convolutional Neural Network}, doi={10.1007/978-3-030-67514-1_42} }- Jingya Zhang
Guojun Li
Juan Zhang
Guanwen Zhang
Wei Zhou
Year: 2021
A Convolutional Neural Network Approach for Stratigraphic Interface Detection
IOTAAS
Springer
DOI: 10.1007/978-3-030-67514-1_42
Abstract
Making full use of the remaining oil-gas resources is a practical and feasible way to solve the problem of oil shortage. And the most important step of this method is the accurate detection of stratigraphic interface. In this paper, we propose a Convolutional Neural Network (CNN) approach for stratigraphic interface detection from the geophysical well logging data. Our proposed approach can automatically extract representative features from well logs data. It can reduce errors caused by human factors such as manual randomness and lack of experience. First of all, we normalize the data in the form of a single point of the well logging data and convert data points to 2D segment. We then feed segments into the CNN model for training. Secondly, we predict the formation of the well logging data using the trained model. Finally, we introduce a post-processing method to perform the stratigraphic interface detection. The experimental results demonstrate the proposed approach is able to achieve(89.69\%)of the average accuracy of stratigraphic interface detection. Moreover, the relative error between the predicted boundary points with the ground-truth is only(1\%), which indicates the proposed approach satisfy the real-world application requirements.
