A Combination Approach of the Numerical Simulation and Data-Driven Analysis for the Impacts of Refracturing Layout and Time on Shale Gas Production

Round 1

Reviewer 1 Report

Dear Editor

I would thank you for the invitation to review the manuscript entitled “ A combination approach of numerical simulation and data-driven analysis for the impacts of refracturing layout and time on shale gas production”, The authors propose a method which somewhat innovative that combines numerical simulation and artificial neural network (ANN) analysis to quickly check out the effects of refracturing layout and time on shale gas production, using 500 simulation results to train an ANN (20 hidden neurons) to predict the cumulative shale gas production with an adequate accuracy (90%). Although the suggestions and comments appended below in order to ameliorate the Ms, in my opinion, the study is worth of publishing in first international rank journals. Minor revisions are suggested.

Major Comments:

1.      Introduction:

· Line [65, 66], why the current refracturing design does not consider the effect of fracture creep, although it’s almost based on the assumption of an elastic reservoir? (Explain more).

2.      Methods:

· 50000 is considered as a sufficient dataset to conduct this study, why there is no statistical analysis of it (Descriptive stats, Correlation, P-value)

· Some of the formulas need more explanation: eqs [1~3], eqs [5~6], eq [22], eqs [41~42]

3.      Results:

· There was informed that more than 20 hidden layer neurons in Fig 17 are more accurate, why only 20, 40 is not better?

4.      Discussion:

· as mentioned before depending on the statistical analysis the true findings can be explained (significance)

·So, based on these finding, how do we proceed with this method, in different fields?

· Do you feel the current data results (yours) can be generalized in the different shale gas production (different than the Sichuan Basin)? Please elaborate and discuss.

Minor Comments:

·  Data collection should be clear

· The conclusion should contain a description of the limitation of the study areas for future research

 

 

 

 

Author Response

Please the attachement of 'Responses to Reviewer 1'.

Author Response File: Author Response.pdf

Reviewer 2 Report

1，The language of the manuscript needs to be further improved

2，The content of the abstract needs to be further revised to highlight the key research content of the article, which needs to be further simplified

3，In the introduction, the comparison of relevant research achievements between China and other countries in the world should be added to highlight the importance of the research content of this manuscript

4，Relevant explanations of Figure 6 should be added to make the reader more clear about the meaning expressed by the author

5，Figure 21 should unify the number fonts in the figure

6，The 712-715 lines in the manuscript should be in a uniform format. The format of the current version is incorrect

7，The overall format and graphic format of the manuscript need to be further improved and modified

Author Response

Please the attachement of 'Responses to Reviewer 2'.

Author Response File: Author Response.pdf

Reviewer 3 Report

1. The authors developed a mixed approach combining both the conventional numerical simulations and deep learning methods. It is impressing that the authors not only applied deep learning in shale gas studies, but also incorporated the trained model into practical investigations. My major concern is the general adaptivity of the trained model. In another word, the authors should clearly state whether the research founding can be extensively applied in general scenarios.

2. The unit of the Langmuir pressure constant in Table 2 seems not correct. 

3. Only one single layer is used in the neural network, and the result seems still acceptable. However, the authors are still encouraged to show the tuning of the network hyperparameters to check the performance, or at least analyze on the possible impacts. The authors can refer to "https://doi.org/10.1016/j.petrol.2020.107886" for further information on how to show the effects of network hyperparameters. 

Author Response

Please the attachement of 'Responses to Reviewer 3'.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The author's attitude in revising the article is serious, which conforms to the quality of a scientific researcher. I recommend that the paper be accepted and published