Industrial Carbon Footprint (ICF) Calculation Approach Based on Bayesian Cross-Validation Improved Cyclic Stacking

Round 1

Reviewer 1 Report

The overall structure of the paper is appropriate, and the manuscript was written quite well. Some comments are provided to improve the quality of this manuscript. In my opinion, this manuscript should be revised before accepting in Sustainability.

1)      The authors have repeated abbreviation and full form of words many times. Please introduce the abbreviation for the first time the word appears in the text and then use the abbreviation form.

2)      In Fig. 2, it is more common using "Test" instead of "Prediction" in cross validation literature. Please replace "Prediction" with "Test" in Cross Validation box.

3)      The manuscript needs to be more concise. Please remove general content such as describing TP, FP, etc.

4)      In Section 3.3, please provide more technical details about the used machine learning algorithms. I highly recommend to consider a subsection for each algorithm. You should add one or two equations for each method. Please refer to the following papers for a short explanation of these methods:

LightGBM:

[1] “Pump Feature Construction and Electrical Energy Consumption Prediction Based on Feature Engineering and LightGBM Algorithm,” Sustainability, vol. 15, no. 1, p. 789, 2023.

XGBoost:

[2] “Modeling operational cement rotary kiln variables with explainable artificial intelligence methods – a ‘conscious lab’ development,” Part. Sci. Technol., vol. 41, no. 5, pp. 715–724, 2023, doi: 10.1080/02726351.2022.2135470.

Random Forest:

[3] “Modeling industrial hydrocyclone operational variables by SHAP-CatBoost - A ‘conscious lab’ approach,” Powder Technol., vol. 420, p. 118416, 2023, doi: 10.1016/j.powtec.2023.118416.

Extreme Learning Machine:

[4] “Short-Term Wind Power Prediction by an Extreme Learning Machine Based on an Improved Hunter--Prey Optimization Algorithm,” Sustainability, vol. 15, no. 2, p. 991, 2023.

 

5)      Please bold the best result in each Table for more reader attention.

6)      Please provide more discussion on the results.

7)      The manuscript should be proofread to ensure that there are no grammatical and typing errors.

8)      My main concern is about the experiments and comparisons. No statistical tests were performed in the paper. So, how could we determine whether the results are statistically significant?

9)      Please add a description regarding the limitations of the proposed method.

 

10)   Running time of all models should be listed and compared to each other. You should mention time complexity of the proposed method as it trains various ML algorithms.

11)   It is recommended to share a link to the source code to make the project reproducible.

 

12)   The authors should describe more on the hyperparameter tuning of the model. Please add a table and list selected hyperparameters for each model.

13)   I suggest to move Appendix A and Appendix B to the main text, as they specify the main results of the manuscript.

14)   The authors should report average and standard deviation of results for each algorithm.

The manuscript should be proofread to ensure that there are no grammatical and typing errors.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

• The manuscript has an interesting topic.
• The statistical methods used are appropriate.
• The explanations of the methods provided are detailed and in-depth.

Areas for improvement:

• The manuscript is difficult to read, the writing style might need improvement.
• The methods and explanations build up but fail to lead to strong and convincing conclusions.
• Repetitive content is present in the text.
• Some sentences are very long, particularly in the introduction.
• Abbreviations need to be reviewed and consistent throughout the document.
• Paragraph lengths vary widely, better organization and cohesion are needed.
• Would suggest moving Table 1 to an Annex for clarity.
• Would recommend combining parts of Figures 6-10 to improve their size and clarity.

Terminology is not perfect. But overall English level is good.

Author Response

Please see attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The article titled "Improving Industrial Carbon Footprint Calculation through Appliance Identification and Electricity Consumption Analysis" presents an ambitious attempt to address a crucial aspect of achieving carbon neutrality – the accuracy of Industrial Carbon Footprint (ICF) calculation. While the goal of enhancing the precision of carbon emissions estimation is commendable, a critical examination of the article reveals several strengths and weaknesses in its approach.

One of the strengths of this article is its recognition of the limitations in existing ICF calculations. It rightly points out that the estimation of direct carbon emissions in industries is often based on incomplete and outdated information provided by industries themselves. This is a valid concern, as accurate data is fundamental to any meaningful carbon footprint calculation.

The article proposes a two-pronged approach to improve ICF calculations. First, it suggests an appliance identification method that utilizes Bayesian cross-validation and a State-corrected Hidden Markov Model (SHMM) to estimate carbon emissions from appliances. Second, it considers electricity consumption as a separate source of emissions, incorporating the marginal carbon emission factor of the connected bus. These ideas represent a potentially valuable contribution to the field of carbon footprint estimation.

However, there are several notable limitations and concerns in this article:

Complexity and Feasibility: The proposed method appears complex and data-intensive, potentially making it challenging to implement in real-world industrial settings. The article should discuss the feasibility of its approach in practical applications.

Data Availability: The success of the proposed method relies heavily on the availability of comprehensive and up-to-date data on appliances and electricity consumption. In many cases, industries may not have this level of data readily accessible, which could limit the method's applicability.

Validation and Generalizability: The article claims superior performance in estimating device carbon emissions with less than 3% error. However, the article lacks information on the diversity and representativeness of the data used for validation, raising concerns about the generalizability of the method across different industrial settings.

Algorithm and Technique Selection: The article discusses the comparison of cross-validation techniques and machine learning algorithms but does not delve into the rationale behind the selection of specific techniques and algorithms. A more comprehensive explanation would have strengthened the article's methodology.

Environmental Impact Assessment: The article focuses on improving ICF calculations but does not address the broader issue of assessing the environmental impact of industries. A more holistic approach might consider not only carbon emissions but also other environmental factors such as water usage, waste generation, and resource consumption.

 

In conclusion, while the article offers a promising approach to enhancing ICF calculation through appliance identification and electricity consumption analysis, it falls short in addressing practical feasibility, data availability, and the broader context of environmental impact assessment. Further research and real-world testing are needed to validate the method's effectiveness and applicability in diverse industrial settings. Achieving comprehensive and accurate ICF calculation is a critical step towards carbon neutrality, but this article represents just one piece of a complex puzzle that requires careful consideration of practical constraints and broader environmental concerns.

 

 

 

 

 

 

No comments. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Dear Authors,
As it was written:

The paper elaborates on the issue of The Industrial Carbon Footprint (ICF) calculation, as a foundation to achieve carbon neutrality. By dividing a factory's carbon emissions into carbon emissions directly produced by appliances and electricity consumption carbon emissions, authors estimate the total carbon emissions of the factory. An appliance identification method is proposed based on a cyclic Stacking method improved by Bayesian cross-validation, and an appliance state correction module SHMM (State-corrected Hidden Markov Model) is added to identify the state of the appliance and then to calculate the corresponding appliance carbon emissions. Regarding the selection of artificial intelligence models and cross-validation technique required in the appliance identification method, this paper compares the effects of cross-validation techniques including Stratified K-Fold, K-Fold, Monte Carlo, etc., on machine learning algorithms such as Adaboost, XGBoost, Feed Forward Network, etc., to determine the technique and algorithms required for the final appliance identification method.
Please read my below comments and suggestions:
- Page 12 – Table 1 and Table 2 could show on separate pages, to underline the techniques 
- Page 14 - If You could reformat text, all pictures from Figure 3. Power profiles in the IAID., could be show on one page – it is better to view profiles
- Page 16 - Any comments needed after equation (28)
- Page 19 – Figure 4 could be put into properly text place relevant to description
- Page 20 – Figure 5 could be put into properly text place relevant to description
- Page 20 to Page 22 – All figures from Figure 6 to Figure 10 should be describe by text immediately before and after the figure, to better understand and underline the effects – what was made, how the correction was implemented
- Page 23 – Figure 11 could be described by text info relevant to made changes
- Page 23 - Table 4. Comparison with other State-of-art models - could be described by text info relevant to made changes
- Page 25 - Figure 12. Carbon emission results of factories – it could be good to describe some pictures from a to f for Figure 12.
- Page 25 – Conclusions - please show the reader how it was calculate - After applying the SHMM, the approach estimates device carbon emissions with an error less than 3%, demonstrating that the proposed approach can achieve comprehensive and accurate ICF calculation
- Appendix A – maybe, to better underline the calculations, it is good idea to show tables and figures as bar plot? 
I hope my suggestions will help to improve the article.

Best Regards
The Reviewer

Minor editing of English language required

Author Response

Please see attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors did not address some of my comments.

1) As the manuscript is multidisciplinary research, some of readers may not be familiar with the machine learning algorithms. So, you should provide more technical details about the used machine learning algorithms. I highly recommend to consider a subsection for each algorithm. You should add one or two equations for each method. Please refer to the following papers for a short explanation of these methods:

LightGBM:

[1] “Pump Feature Construction and Electrical Energy Consumption Prediction Based on Feature Engineering and LightGBM Algorithm,” Sustainability, vol. 15, no. 1, p. 789, 2023.

XGBoost:

[2] “Modeling operational cement rotary kiln variables with explainable artificial intelligence methods – a ‘conscious lab’ development,” Part. Sci. Technol., vol. 41, no. 5, pp. 715–724, 2023, doi: 10.1080/02726351.2022.2135470.

Random Forest:

[3] “Modeling industrial hydrocyclone operational variables by SHAP-CatBoost - A ‘conscious lab’ approach,” Powder Technol., vol. 420, p. 118416, 2023, doi: 10.1016/j.powtec.2023.118416.

Extreme Learning Machine:

[4] “Short-Term Wind Power Prediction by an Extreme Learning Machine Based on an Improved Hunter--Prey Optimization Algorithm,” Sustainability, vol. 15, no. 2, p. 991, 2023.

If you think of article length limitation, please remove less important materials, such as TP, ... explanations.

2) There are still no statistical tests performed in the revised version of manuscript. Please apply the two tailed Welch's t-test to show that the results are statistically significant.

No comment.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This reviewer's comments have been addressed. Would reccommend for publication after editing and proofreading.

Satisfactory, proofreading needed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

The authors have addressed all my comments.

No comment.