Evaluation of machine learning methods for formation lithology identification: A comparison of tuning processes and model performances

Xie, Yunxin; Zhu, Chenyang; Zhou, Wen; Li, Zhongdong; Li, Xuan; Tu, Mei

doi:10.1016/j.petrol.2017.10.028

Cited by 230 publications

(70 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Our previous research shows that the ensemble methods are applicable to identify lithology classes given the number of logging data is limited. e gradient tree boosting model performs better in classifying the four sandstone classes than the SVM classifier, neural network classifier, and random forest classifier [10]. In this case study, the average precision scores of GTB classifier on the sandstone classes in DGF and HGF are 76.1% and 83.%, respectively, while the average precision scores of the optimized stacked boosting model are 77.5% and 84.2%, respectively.…”

Section: Results and Analysismentioning

confidence: 76%

“…Boosting approach is one of the ensemble methods. Moreover, our previous results showed that boosting method has a better capability in distinguishing sandstone classes compared with other classifiers [10]. Based on our previous work, this study conducts a meticulous evaluation over three boosting models, namely, AdaBoost, Gradient Tree Boosting (GTB), and eXtreme Gradient Boosting (XGBoost) with 5-fold cross validation.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Towards Optimization of Boosting Models for Formation Lithology Identification

Xie

Zhu

et al. 2019

Mathematical Problems in Engineering

Self Cite

View full text Add to dashboard Cite

Lithology identification is an indispensable part in geological research and petroleum engineering study. In recent years, several mathematical approaches have been used to improve the accuracy of lithology classification. Based on our earlier work that assessed machine learning models on formation lithology classification, we optimize the boosting approaches to improve the classification ability of our boosting models with the data collected from the Daniudi gas field and Hangjinqi gas field. Three boosting models, namely, AdaBoost, Gradient Tree Boosting, and eXtreme Gradient Boosting, are evaluated with 5-fold cross validation. Regularization is applied to the Gradient Tree Boosting and eXtreme Gradient Boosting to avoid overfitting. After adapting the hyperparameter tuning approach on each boosting model to optimize the parameter set, we use stacking to combine the three optimized models to improve the classification accuracy. Results suggest that the optimized stacked boosting model has better performance concerning the evaluation matrix such as precision, recall, and f1 score compared with the single optimized boosting model. Confusion matrix also shows that the stacked model has better performance in distinguishing sandstone classes.

show abstract

Section: Results and Analysismentioning

confidence: 76%

Section: Introductionmentioning

confidence: 99%

Towards Optimization of Boosting Models for Formation Lithology Identification

Xie

Zhu

et al. 2019

Mathematical Problems in Engineering

Self Cite

View full text Add to dashboard Cite

show abstract

“…Advanced statistical models have been introduced to automate the task of facies identification. These include methods such as non-parametric regression, factor analysis, principal component analysis, classification trees, clustering and techniques based on machine learning and artificial intelligence [9][10][11][12]. The electrofacies and the lithofacies are similar in attempting to identify and group rocks based on large-scale geologic and petrophysical features as shown by the log responses.…”

Section: Introductionmentioning

confidence: 99%

Application of Multivariate Statistical Methods and Artificial Neural Network for Facies Analysis from Well Logs Data: an Example of Miocene Deposits

Puskarczyk

2020

Energies

View full text Add to dashboard Cite

The main purpose of the study is a detailed interpretation of the facies and relate these to the results of standard well logs interpretation. Different methods were used: firstly, multivariate statistical methods, like principal components analysis, cluster analysis and discriminant analysis; and secondly, the artificial neural network, to identify and discriminate the facies from well log data. Determination of electrofacies was done in two ways: firstly, analysis was performed for two wells separately, secondly, the neural network learned and trained on data from the W-1 well was applied to the second well W-2 and a prediction of the facies distribution in this well was made. In both wells, located in the area of the Carpathian Foredeep, thin-layered sandstone-claystone formations were found and gas saturated depth intervals were identified. Based on statistical analyses, there were recognized presence of thin layers intersecting layers of much greater thickness (especially in W-2 well), e.g., section consisting mainly of claystone and sandstone formations with poor reservoir parameters (Group B) is divided with thin layers of sandstone and claystone with good reservoir parameters (Group C). The highest probability of occurrence of hydrocarbons exists in thin-layered intervals in facies C.

show abstract

“…Therefore, it is required to automate the procedure of reservoir characterization and, at this point, computer technologies has shown suitable to lithology identification [2,3,4,5]. These computer technologies assist the geologists to avoid the unnecessary data analysis work and improve the lithology identification accuracy [6]. As a result, geologists can build better quantitative evaluation models of different rock properties, which can also improve overall evaluation.…”

Section: Introductionmentioning

confidence: 99%

“…Cross plots and Principal Component Analysis were used to lithology characterization and mineralogy description from geochem-ical logging tool data [10]. In [6], five machine learning methods were employed to classify the formation lithology identification using well log data samples. Horrocks et al [2] explores different machine learning algorithms and architectures for classifying lithologies using wireline data for coal exploration.…”

Section: Introductionmentioning

confidence: 99%

Extreme Learning Machine combined with a Differential Evolution algorithm for lithology identification

Saporetti

Duarte

Fonseca

et al. 2018

RITA

View full text Add to dashboard Cite

Lithology identification, obtained through the analysis of several geophysical properties, has an important role in the process of characterization of oil reservoirs. The identification can be accomplished by direct and indirect methods, but these methods are not always feasible because of the cost or imprecision of the results generated. Consequently, there is a need to automate the procedure of reservoir characterization and, in this context, computational intelligence techniques appear as an alternative to lithology identification. However, to acquire proper performance, usually some parameters should be adjusted and this can become a hard task depending on the complexity of the underlying problem. This paper aims to apply an Extreme Learning Machine (ELM) adjusted with a Differential Evolution (DE) to classify data from the South Provence Basin, using a previously published paper as a baseline reference. The paper contributions include the use of an evolutionary algorithm as a tool for search on the hyperparameters of the ELM. In addition, an activation function recently proposed in the literature is implemented and tested. The computational approach developed here has the potential to assist in petrographic data classification and helps to improve the process of reservoir characterization and the production development planning.

show abstract

Evaluation of machine learning methods for formation lithology identification: A comparison of tuning processes and model performances

Cited by 230 publications

References 21 publications

Towards Optimization of Boosting Models for Formation Lithology Identification

Towards Optimization of Boosting Models for Formation Lithology Identification

Application of Multivariate Statistical Methods and Artificial Neural Network for Facies Analysis from Well Logs Data: an Example of Miocene Deposits

Extreme Learning Machine combined with a Differential Evolution algorithm for lithology identification

Contact Info

Product

Resources

About