Dinghan Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Dinghan Wang has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
Andrea Thompson
2 shared publicationsA. Thompson
1 shared publication- Elnara Rustamzade
The University of Texas at Austin
1 shared publication Hongjie Xiong
University Lands, Houston, TX, USA
1 shared publicationJ.T. Foster
1 shared publicationJohn Foster
1 shared publication
Latest Publications
- Open Access
- 14 September 2026
- Published by Elsevier in Applied Computing and Geosciences
- Vol. 32
Bayesian updating with machine learning and geostatistics for posterior predictions of unconventional reservoir performance
- 14 September 2026
- Published by Elsevier in Applied Computing and Geosciences
- Vol. 32
Open AccessBayesian updating with machine learning and geostatistics for posterior predictions of unconventional reservoir performance
- Open Access
- 15 May 2026
- Published by Elsevier in SSRN Electronic Journal
Bayesian Updating with Multivariate Machine Learning-based Prior and Spatial Geostatistics-based Likelihood for Posterior Predictions of Unconventional Reservoir Performance
Accurate forecasting of hydrocarbon production in unconventional reservoirs requires integrating spatial heterogeneity with well-scale geological and engineering data. Traditional machine learning regression approaches often neglect spatial context, including spatial continuity, spatial trends and data locations, resulting in potential overestimation of uncertainty. To address this limitation, we introduce a Bayesian updating workflow with a multivariate machine learning model as a prior model and a spatial geostatistical model of well production as a likelihood model. The workflow is applied after normal score transformation of all features in Gaussian space to provide complete local uncertainty models from geostatistical kriging estimates and variances and to facilitate the use of closed-form analytical Bayesian updating. Our proposed workflow is demonstrated with 9 geological and 6 completion secondary features to forecast cumulative one-year oil production as the primary feature for an unconventional reservoir. A multivariate machine learning-based prior model is first calculated using local ridge regression with hyperparameter tuning to relate the secondary features to production. The likelihood model is calculated by ordinary kriging based on an additive deterministic trend and residual estimation. Bayesian updating with K-fold cross-validation is then applied to tune the workflow, demonstrating a reduction of the high uncertainty of the prior model, yielding a posterior distribution that is both more accurate and precise than the machine learning or geostatistical distributions on their own.
- 15 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessBayesian Updating with Multivariate Machine Learning-based Prior and Spatial Geostatistics-based Likelihood for Posterior Predictions of Unconventional Reservoir Performance
Accurate forecasting of hydrocarbon production in unconventional reservoirs requires integrating spatial heterogeneity with well-scale geological and engineering data. Traditional machine learning regression approaches often neglect spatial context, including spatial continuity, spatial trends and data locations, resulting in potential overestimation of uncertainty. To address this limitation, we introduce a Bayesian updating workflow with a multivariate machine learning model as a prior model and a spatial geostatistical model of well production as a likelihood model. The workflow is applied after normal score transformation of all features in Gaussian space to provide complete local uncertainty models from geostatistical kriging estimates and variances and to facilitate the use of closed-form analytical Bayesian updating. Our proposed workflow is demonstrated with 9 geological and 6 completion secondary features to forecast cumulative one-year oil production as the primary feature for an unconventional reservoir. A multivariate machine learning-based prior model is first calculated using local ridge regression with hyperparameter tuning to relate the secondary features to production. The likelihood model is calculated by ordinary kriging based on an additive deterministic trend and residual estimation. Bayesian updating with K-fold cross-validation is then applied to tune the workflow, demonstrating a reduction of the high uncertainty of the prior model, yielding a posterior distribution that is both more accurate and precise than the machine learning or geostatistical distributions on their own.
- Open Access
- 19 March 2026
- Published by SAGE Publications in Energy Exploration & Exploitation
- Vol. 44 (4) , 1806-1829
A comprehensive machine learning workflow for classifying production profiles in unconventional reservoirs
Accurate determination and prediction of production profile types in unconventional reservoirs is a critical task for optimizing petroleum exploration and development. While machine learning (ML) methods have been increasingly applied in petroleum engineering, a comprehensive and generalizable workflow for classifying and predicting production profiles remains lacking. To support unconventional reservoir decision-making, we propose an ML workflow integrating geological, completion, and production data. The workflow begins with univariate and multivariate analyses to explore data relationships, followed by feature selection using multiple ranking methods: maximum relevance–minimum redundancy, variance inflation factor, partial correlation coefficient, recursive feature elimination, and model-based random forest importance. This yields six key features selected from the initial dataset. K -Means clustering applied to a dataset of 189 Permian Basin wells identifies two distinct production patterns (high and low), with normalized oil production histories showing clear separation. Multidimensional scaling (MDS) is used to reduce feature dimensionality to two components while preserving pairwise sample distances, enabling improved visualization and interpretability. A random forest classification model trained on these MDS components achieves a higher test accuracy (97.5%) and stability (lower variance) compared to a reference model using the original high-dimensional features (95.0%) and the model using two principal components of principal component analysis (90.0%). This workflow facilitates the automated classification of production profiles and provides a data-driven foundation for optimizing completion designs in unconventional reservoir development.
- 19 March 2026
- Published by SAGE Publications in Energy Exploration & Exploitation
- Vol. 44 (4) , 1806-1829
Open AccessA comprehensive machine learning workflow for classifying production profiles in unconventional reservoirs
Accurate determination and prediction of production profile types in unconventional reservoirs is a critical task for optimizing petroleum exploration and development. While machine learning (ML) methods have been increasingly applied in petroleum engineering, a comprehensive and generalizable workflow for classifying and predicting production profiles remains lacking. To support unconventional reservoir decision-making, we propose an ML workflow integrating geological, completion, and production data. The workflow begins with univariate and multivariate analyses to explore data relationships, followed by feature selection using multiple ranking methods: maximum relevance–minimum redundancy, variance inflation factor, partial correlation coefficient, recursive feature elimination, and model-based random forest importance. This yields six key features selected from the initial dataset. K -Means clustering applied to a dataset of 189 Permian Basin wells identifies two distinct production patterns (high and low), with normalized oil production histories showing clear separation. Multidimensional scaling (MDS) is used to reduce feature dimensionality to two components while preserving pairwise sample distances, enabling improved visualization and interpretability. A random forest classification model trained on these MDS components achieves a higher test accuracy (97.5%) and stability (lower variance) compared to a reference model using the original high-dimensional features (95.0%) and the model using two principal components of principal component analysis (90.0%). This workflow facilitates the automated classification of production profiles and provides a data-driven foundation for optimizing completion designs in unconventional reservoir development.