Title: Logging curve reconstructions based on MLP multilayer perceptive neural network

Authors: Li Ming; Rui Deng; Chengquan Gao; Rigong Zhang; Xiaojun He; Jie Chen; Tao Zhou; Zhixian Sun

Addresses: Yangtze University, No. 111 Daxue Road, Caidian District, Wuhan, Hubei Province, China; PetroChina Qinghai Oilfield Company, No. 9 Kunlun Middle Road, Qili Town, Dunhuang City, Gansu Province, China ' Yangtze University, No. 111 Daxue Road, Caidian District, Wuhan, Hubei Province, China ' PetroChina Tuha Oilfield Company, No. 1 Chuangye Road, Yizhou District, Hami, Xinjiang, China ' PetroChina Tuha Oilfield Company, No. 1 Chuangye Road, Yizhou District, Hami, Xinjiang, China ' PetroChina Qinghai Oilfield Company, No. 9 Kunlun Middle Road, Qili Town, Dunhuang City, Gansu Province, China ' PetroChina Tuha Oilfield Company, No. 1 Chuangye Road, Yizhou District, Hami, Xinjiang, China ' PetroChina Qinghai Oilfield Company, No. 9 Kunlun Middle Road, Qili Town, Dunhuang City, Gansu Province, China ' Qingdao Real Estate Registration Center, No. 9-11, Wuxia Road, Shinan District, Qingdao, Shandong Province, China

Abstract: As an important component of the petroleum industry, logging curves play an important role in lithology identification, reservoir evaluation, and geological structure analysis. However, instrument measurement or drilling problems often result in distorted or discontinuous logging curves in certain profiles. Retesting is costly and challenging to operate. Traditional linear fitting and statistical analysis are insufficient for the analysis and evaluation of ultrafine reservoirs. To address this issue, MLP neural network (multi-layer perception) technology has established a curve prediction model based on training data. The nonlinear factors are introduced through the activation function, the optimal solution is found by using the optimisation methods of loss function and MBGD (small batch gradient descent method), and the best neural network model is obtained by optimising the number of hidden layers and nodes of the neural network. Compare the predicted curve with the original measurement curve for quality control to obtain the best result in constructing curve weights. The results show that MLP neural network technology can provide reliable technical support for accurately predicting logging curves and enhancing reservoir analysis and evaluation. [Received: March 8, 2023; Accepted: May 5, 2023]

Keywords: logging curve reconstruction; MLP multilayer perceptual neural network; optimal solution; prediction model; curve missing.

DOI: 10.1504/IJOGCT.2023.133540

International Journal of Oil, Gas and Coal Technology, 2023 Vol.34 No.1, pp.25 - 41

Received: 07 Mar 2023
Accepted: 05 May 2023
Published online: 19 Sep 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article