Well log data generation and imputation using sequence based generative adversarial networks

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 31, 2025

Well log analysis is significant for hydrocarbon exploration, providing detailed insights into subsurface geological formations. However, gaps and inaccuracies in well data, often due to equipment limitations, operational challenges, harsh conditions, can introduce uncertainties reservoir evaluation. Addressing these challenges requires effective methods both synthetic data generation precise imputation of missing ensuring completeness reliability. This study introduces a novel framework utilizing sequence-based generative adversarial networks (GANs) specifically designed imputation. The integrates two distinct GAN models: time series (TSGAN) generating sequence (SeqGAN) imputing data. Both models were tested on dataset from the North Sea, Netherlands region. For task, input comprises logs with values output corresponding imputed logs; complete real that mimic statistical properties original All measurements are normalized 0-1 range using min-max scaling, error metrics reported units. Different sections 5, 10, 50 points used. Experimental results demonstrate this approach achieves superior accuracy filling compared other deep learning spatial analysis. method yielded [Formula: see text] 0.92, 0.86, 0.57, mean absolute percentage (MAPE) 8.320, 0.005, 166.6, (MAE) 0.012, 0.002, 0.03, respectively. MAE, 0.35, MRLE 0.01. These set new benchmark integrity utility geosciences, particularly

Language: Английский

Impact of data size on ML-based prediction of shear and compressional slowness

Carbonates and Evaporites, Journal Year: 2025, Volume and Issue: 40(2)

Published: April 20, 2025

Language: Английский