Prediction of NOx Emissions from a Coal-Fired Boiler Based on Convolutional Neural Networks with a Channel Attention Mechanism

Round 1

Reviewer 1 Report

The topic of the article is interesting and important.

Authors should consider several additions to their article.

1. It is necessary to describe in more detail the mechanisms of NOx formation and the parameters on which the NOx emission in coal / pulverized boilers depends.

2. Whether or how the elemental composition of coal affects the calculations.

3. Whether, or how, the model considers the mechanisms of the formation of nitrogen oxides.

4. In table 1, add the units for the quantities characterizing the boiler operation.

Author Response

We are very grateful to the reviewer for his constructive comments and suggestions. The manuscript has been revised as suggested by the reviewers. All changes that have been made are highlighted in red for easy comparison between the modified version and the original one. Our point-by-point response to the comments and detailed changes that have been made are summarized as follows.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

 

The work presented for review covers an important aspect for the energy sector. Importantly, it can have a measurable implementation effect. However, the text requires explanations and supplements.

Chapter 2.2 Data preparation

Verses 106-108

Please define and include in the text, what has an impact on NOx emissions in the test (in this case): e.g .:

1 1. To what extent was the boiler load considered?

2 2. To what extent has the total air flow in the case of the considered boiler loads been analysed, the total fuel flow related to the boiler load, etc. ....... see Table 1?

Dependencies (1), (2), (3), (4), (5)

Please provide the sources of the dependencies 1, 2, 3, 4, 5.

 Chapter 2.4 The brief introduction to the channel attention mechanism

Dependence (6)

Please provide the source of the dependency(6).

Verse 152

Quote: „………………a lower computational cost.”

Please provide an explanation (briefly) of the lower calculation cost in this case

 Chapter 2.5 The architecture design of the NOx prediction model

Fig.3.

Please describe and explain precisely Fig. 3. What do the next steps in the scheme mean and why are these stages important in building the model?

 Verse 173

Quote: „Three building blocks with varying parameter settings are repeated in a cascade to learn the representations.”

What parameters are included in the set of this block? What is meant by "variable set of parameters”?

Verses 174-175

How are one-dimensional vectors defined?

Verses 175-176

Based on what information does layer (FC) predict NOx emissions? 

Verse 178

What are the practical guidelines? To better understand the problem, you should specify them and briefly discuss them.

 Chapter 3.1. Training infrastructure

In my opinion, the content of this chapter should be included in chapter 2. Methods.

Chapter 3.2. Effect of the ECA module

Verse 199

Quote: “Still, we kept the parameter setting of other model components unchanged”

Please provide what are the “other model components”?

Verses 203-205

 Quote: „The overall performance of our model without ECA modules is worse than our model with ECA modules.”

How to understand the phrase „the overall performance”? Please justify why the performance of the model without ECA is worse than with the EC module?

Best regards,

Reviewer

Author Response

We are very grateful to the reviewer for his constructive comments and suggestions. The manuscript has been revised as suggested by the reviewers. All changes that have been made are highlighted in red for easy comparison between the modified version and the original one. Our point-by-point response to the comments and detailed changes that have been made are summarized as follows.

Author Response File: Author Response.docx

Reviewer 3 Report

1) In general, this paper seems to be a little bit simple, please improve it;

2) The authors should present more details about how to establish the new model.

3) The authors should give out deeper analysis of the model prediction results, or the difference with detailed data should be presented.

4) If possible, please present the key findings with data in the part of "Conclusions".

5) What is the limitations of the current study? please indicate it in the "Concluions"

Author Response

We are very grateful to the reviewer for his constructive comments and suggestions. The manuscript has been revised as suggested by the reviewers. All changes that have been made are highlighted in red for easy comparison between the modified version and the original one. Our point-by-point response to the comments and detailed changes that have been made are summarized as follows.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The manuscript can be accepted as the  authors have addressed my comments.