Development of visual prediction model for shale gas wells production based on screening main controlling factors

Niu, Wente; Lu, Jialiang; Sun, Yize; Guo, Wei; Liu, Yuyang; Mu, Ying

doi:10.1016/j.energy.2022.123812

Cited by 19 publications

(5 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Try to find the appropriate hidden layer and output layer corresponding to the linear coefficient matrix W , bias vector b , so that all the training sample input calculated output as much as possible equal to or close to the sample label. In order to avoid the overfitting of neural networks, regularization processing techniques, including L1 & L2 regularization and dropout means, are usually added to the network to enhance the generalization ability of the model. , …”

Section: Methodsmentioning

confidence: 99%

“…In order to avoid the overfitting of neural networks, regularization processing techniques, including L1 & L2 regularization and dropout means, are usually added to the network to enhance the generalization ability of the model. 31,54 Figure 2 shows the architecture of a typical DNN model. The model consists of an input layer, two hidden layers, and an output layer.…”

Section: Deep Neural Networkmentioning

confidence: 99%

“…However, the poor quality of field data often makes the prediction model derived from these technologies not very effective . Existing research shows that data-driven modeling methods provide an alternative to many problems in the oil and gas industry, and in some cases, better solutions can be provided. , For the production and EUR prediction of shale gas wells, a number of data-driven techniques and machine learning methods are available. − All these successful applications have proved the applicability of data-driven methods in this field.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Toward Production Forecasting for Shale Gas Wells Using Transfer Learning

et al. 2023

Self Cite

View full text Add to dashboard Cite

Accurate prediction of shale gas well production and estimated ultimate recovery (EUR) is always a difficult and hot spot in shale gas development. In particular, the production and EUR prediction of shale gas wells in new production blocks are faced with the lack of field gas well data and the difficulty of model development. In view of the above problems, this study proposes a new deep transfer learning strategy, which uses transfer component analysis (TCA) and deep neural network (DNN) to achieve shale gas well production and EUR prediction across formations/blocks. The feature extractor based on TCA can narrow the input feature distribution of the source and the target domains. The neural network model can be used to establish a domainadaptive transfer learning model without the prediction performance degradation caused by distribution offset. Validity and accuracy of the model were analyzed using gas well data from Weiyuan and Luzhou blocks in Sichuan Basin, China. The results appear that the reasonable application of TCA can greatly improve the prediction performance of shale gas well transfer learning model. For data sets of the same size, compared with the transfer learning model developed by classical machine learning algorithms, the proposed neural network-based transfer learning model can significantly improve the accuracy of production prediction across formations/ blocks. In addition, the proposed model can also be extended to other types of oil and gas production prediction tasks cross formations/blocks.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Deep Neural Networkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Toward Production Forecasting for Shale Gas Wells Using Transfer Learning

et al. 2023

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the study of this paper, a large number of factors affecting EUR were selected, 31 in total, from which important factors need to be selected and their number determined. Commonly used methods, such as grey correlation analysis and the distance correlation coefficient, can analyze the correlation or distance correlation coefficient of each factor with EUR, but they cannot accurately give the number of important influencing factors, and they need to be selected by human beings, which increases the uncertainty of the inputs to the model [24,25].The RF-RFE algorithm is very suitable for solving the problem of feature selection, especially in the case of a large number of features and uncertainty about which ones are the most important, and it is able to give the number of important influencing factors. Recursive feature elimination (RFE) is a feature selection algorithm that ranks feature variables [26].…”

Section: Rf-rfe Algorithmmentioning

confidence: 99%

Machine Learning-Based Research for Predicting Shale Gas Well Production

Qi,

Li,

et al. 2024

Symmetry

View full text Add to dashboard Cite

The estimated ultimate recovery (EUR) of a single well must be predicted to achieve scale-effective shale gas extraction. Accurately forecasting EUR is difficult due to the impact of various geological, engineering, and production factors. Based on data from 200 wells in the Weiyuan block, this paper used Pearson correlation and mutual information to eliminate the factors with a high correlation among the 31 EUR influencing factors. The RF-RFE algorithm was then used to identify the six most important factors controlling the EUR of shale gas wells. XGBoost, RF, SVM, and MLR models were built and trained with the six dominating factors screened as features and EUR as labels. In this process, the model parameters were optimized, and finally the prediction accuracies of the models were compared. The results showed that the thickness of a high-quality reservoir was the dominating factor in geology; the high-quality reservoir length drilled, the fracturing fluid volume, the proppant volume, and the fluid volume per length were the dominating factors in engineering; and the 360−day flowback rate was the dominating factor in production. Compared to the SVM and MLR models, the XG Boost and the RF models based on integration better predicted EUR. The XGBoost model had a correlation coefficient of 0.9 between predicted and observed values, and its standard deviation was closest to the observed values’ standard deviation, making it the best model for EUR prediction among the four types of models. Identifying the dominating factors of shale gas single-well EUR can provide significant guidance for development practice, and using the optimized XGBoost model to forecast the shale gas single-well EUR provides a novel idea for predicting shale gas well production.

show abstract

“…In the petroleum industry, a vast amount of exploration, exploration, development, and transportation data related to oil and gas has been accumulated. Leveraging this data, numerous cutting-edge machine learning techniques techniques have been effectively implemented employed in carbonate gas well production prediction, including support vector machines [6,7], gradient boosting decision trees [8,9], random forests [10,11], and deep neural networks [12].…”

Section: Introductionmentioning

confidence: 99%

Fusion attention mechanism-based bidirectional long short term memory for carbonate gas well productivity prediction

Ni,

Xu,

Liu

et al. 2023

International Conference on Algorithms, High Performance Computing, and Artificial Intelligence (AHPCAI 2023)

View full text Add to dashboard Cite

Development of visual prediction model for shale gas wells production based on screening main controlling factors

Cited by 19 publications

References 25 publications

Toward Production Forecasting for Shale Gas Wells Using Transfer Learning

Toward Production Forecasting for Shale Gas Wells Using Transfer Learning

Machine Learning-Based Research for Predicting Shale Gas Well Production

Fusion attention mechanism-based bidirectional long short term memory for carbonate gas well productivity prediction

Contact Info

Product

Resources

About