Discrimination Between Reservoir Models in Well-Test Analysis

Anraku, T.; Horne, Roland N.

doi:10.2118/26426-pa

Cited by 14 publications

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The pressure signals for HO-IA, HO-NF, and HO-PS reservoirs have been simulated using the mathematical models available in the literature 34 , 43 . Stehfest’s algorithm has been considered for numerical inversion of the Laplace transformation 44 .…”

Section: Methodsmentioning

confidence: 99%

Metaheuristic algorithm integrated neural networks for well-test analyses of petroleum reservoirs

Pandey

Aggarwal

Nath

et al. 2022

Sci Rep

View full text Add to dashboard Cite

In recent years, well-test research has witnessed several works to automate reservoir model identification and characterization using computer-assisted models. Since the reservoir model identification is a classification problem, while its characterization is a regression-based task, their simultaneous accomplishment is always challenging. This work combines genetic algorithm optimization and artificial neural networks to identify and characterize homogeneous reservoir systems from well-testing data automatically. A total of eight prediction models, including two classifiers and six regressors, have been trained. The simulated well-test pressure derivatives with varying noise percentages comprise the training samples. The feature selection and hyperparameter tuning have been performed carefully using the genetic algorithm to enhance the prediction accuracy. The models were validated using nine simulated and one real-field test case. The optimized classifier identifies all the reservoir models with a classification accuracy higher than 79%. In addition, the statistical analysis approves that the optimized regressors accurately perform the reservoir characterization with mean relative errors of lower than 4.5%. The minimized manual interference reduces human bias, and the models have significant noise tolerance for practical applications.

show abstract

Section: Methodsmentioning

confidence: 99%