Forecasting Oil Production Flowrate Based on an Improved Backpropagation High-Order Neural Network with Empirical Mode Decomposition

Round 1

Reviewer 1 Report

In this paper, the authors proposed a machine learning system based on a combined empirical mode decomposition back-propagation higher-order neural network (EMD-BP-HONN) for the oilfield. The topic is relevant to the journal. The results presented well. But the specific comments are as follows:

·       The quality of Figure 4 should be improved.

·       All parameters should be described after each equation.

 

·       Since intelligent models like LSTM, …. have different parameters, how to set the related parameters of them for better performance?

Author Response

Dear Professor,

First of all, we would like to appreciate your time and effort to review our manuscript titled  Forecasting oil production flowrate based on improved back-propagation high-order neural network with empirical mode decomposition in journal of Processes. We are grateful to the reviewers for the insightful comment in our paper. We have been able to revise our manuscript as per your suggestion provided (please see the attachment of revised document).

Here is point-by-point response to your feedback and concerns

1.  Feedback :The quality of Figure 4 should be improved.
   • Response: Agreed to make it more clearer. Already revised with bolder text.
2. Feedback :   All parameters should be described after each equation.
   • Response : We agree with this and have added all parameter descriptions in equation (1), (2), (3), (16), and (22).
3. Feedback : Since intelligent models like LSTM, …. have different parameters, how to set the related parameters of them for better performance?
   • Response : Thanks for the question. As for LSTM, as mentioned in line 304-306 which the parameter being used for LSTM is number of layer and activation function of tanh, relu and sigmoid. The experiment run repeatedly with different combination of those paramaters and also different lag of input (X). The best result is taken as model result. Similar approach for ARIMA is in line of 300-302, which the parameter of p,d,q were selected from 1 to 4. For HONN, the setup mentoned in line 285-288 which comprised of different hidden neuron (2 to 10), different synaptic operation (LSO, QSO and CSO) and also the different lag of input (X).

I believe we already covered all of your valuable response and feedback. Thus, we look forward to hearing from you in due time regarding our submission.

 

Sincerely

Teguh Adji, 31 May 2022

Author Response File: Author Response.pdf

Reviewer 2 Report

In the article "Forecasting oil production flowrate based on improved back-propagation high-order neural network with empirical mode decomposition" models for oil flow forecasting for two data series were analyzed:

-data from 5 (five) wells in Cambay Basin (CB), India.

-data irom actual oilfield in Central Sumatra Basin (CS), Indonesia.

The models selected for analysis represented a statistical and machine learning approach. The choice of the model was preceded by both a rich literature review and a statistical analysis, which allowed for the selection of the correct group of methods.

As a result of the obtained results, tools for the analysis of non-stationary series were selected.

After a series of experiments, all the constructed models were compared, indicating their advantages and disadvantages. Ultimately, a model based on the hybrid approach EMD-BP-HONN and EMD-BP-MLMVN was chosen.

The article reads very well, everything is presented in a very understandable way. The Authors clearly indicated what the article brings to the literature on the subject, described the experiments carried out in great detail and presented the most important conclusions.

Author Response

Dear Professor,

First of all, I'm really appreciate your time to review our manuscript titled Forecasting oil production flowrate based on improved back-propagation high-order neural network with empirical mode decomposition to journal of Processes. After reading your review, we are grateful to fulfill your expectation on how to be qualified paper written. Your comments are really supporting our motivation to keep productive as researchers.  As my understanding, no changes required thus the original manuscript will be submitted as revision (with minor changes from other reviewer's feedback). Once again, it's my honor to get your valuable feedback. 

Thanks and regards,

Adji