Study of the Data Augmentation Approach for Building Energy Prediction beyond Historical Scenarios

Round 1

Reviewer 1 Report

In the manuscript entitled “Study on the data augmentation approach for building energy prediction beyond historical scenarios”, the authors propose a data-driven hybrid model method to improve the accuracy and stability of building energy consumption prediction.

If the model is based only on historical data (field measurements), it can give wrong forecasts about energy consumption when the use of the building and its systems is different from the previously recorded practice. That is why it is desirable to supplement the historical data with data obtained by a reliable simulation model, which can be used to analyze situations that may occur in the building but have not yet happened in practice.

The authors’ hybrid method is based on the use of recorded data from the case study building and additional data obtained by simulations using the physical model. In that way, more data is available and can be used to train and validate the data-driven model for the building energy consumption predictions. The authors paid particular attention to the behavior of the building's occupants and their use of the terminal units, lighting, and electronic equipment in the building to create scenarios beyond the historical experiences.

The paper must be improved by taking into account the following remarks referred to the authors.

The abstract is not clear enough and needs to be reformulated. It is necessary to emphasize more clearly that three scenarios of the use of the building and its equipment, which are not included in the historical data, were considered and to describe them briefly. 

The authors should check that all the abbreviations, symbols, and parameters are well predefined in the text or included in the nomenclature table. Since the nomenclature table is present, I suggest writing all the symbols (together with the measurement units) in that table.

The authors should avoid writing heading after heading with nothing in between. The headings should be either merged or a small paragraph should be provided in between.

The authors should check that all the references are cited and managed using the journal reference style.

Lines 43 and 44: I quote: “Meanwhile, the classical physical model method has been effectively developed due to its benefits in certain application circumstances (Bourdeau et al., 2019).”. Those “certain application circumstances” should be explained in more detail.

Line 54: A sentence is missing explaining why there is a lack of data (what is the cause of that situation).

Lines 96 and 97: The authors should name the studies that combine data augmentation and occupancy behavior data and clearly explain their own contribution in comparison to those studies. 

Line 126 and the paragraph dedicated to the data augmentation procedure: The analyzed scenarios are not clearly defined in that part of the text. The scenarios become clear only later in the text. The scenarios should be clearly defined already in step 2. Consider reorganization of the manuscript.

Line 145: The authors should take attention to write correctly the name of the commercial software used. The name of the software producer should also be written.

Line 202: I suggest using the title “Case Study” or similar.

Table 2.: The measurement units of the parameters written in the last two rows are not clear enough. The last three rows of the table should be reorganized.

Fig. 3.: The colors are too similar (if not identical). Different colors must be used.

Line 283, the section entitled “Air conditioning system parameters”: If fan coil units are used in the building, the fans in them usually have three rotational speeds that can be engaged manually or automatically. The units do not usually consume the same amount of electrical energy during the operational time. Some information about the units control strategy should be given.

It would be better to use the term "HVAC System" instead of "Air Conditioning System".

Please use the correct mathematical symbol for the multiplication operation.

Line 294: Some measurement units are missing (for the temperatures at least)

Line 348: Check the sentence. 

Fig. 8. The time scale in the charts is different from the previous cases (0-23 instead of 1-24 as in Figures 3 and 7). The time scale should be equal in all the charts. The chart for case “c” has different units at the vertical axes from the charts for the rest of the cases.

Proofreading by a native English speaker or proofreading service should be conducted to improve both language and organization quality.

 

Author Response

Please see the attachment.

 

Author Response File: Author Response.docx

Reviewer 2 Report

Building energy prediciton is a classical problem. This study proposes a data augmentation based on physical model to  improve the stability. Here are some comments:

1) How to validate the Chiller model and Pump models. Please clarify. Also provide the used model reference data. 

2) Fig. 5 shows a monthly data. Which data is used as the monthly representation? 

3) Please bring strong relevance to the scope of journal by investigating most recent literature. More literature investigation (e.g., deep reinforcement learning can be discussed and compared) should be conducted to highlight the importance and novelty of the current work. Similar conclusions should be avoided and please bring constructive conclusions further to improve readership.

4) Basically, this model is based on SVM which is a classical alghorithm. Please discuss the advantages and its limitations. 

5) Discussion should be expanded for the in-depth analysis of the results as well as comparison with relevant literature studies. Practical implications and limitations of the study should be discussed as well.

6) Check typos and grammar. 

Author Response

Please see the attachment

 

Author Response File: Author Response.docx

Reviewer 3 Report

The article concerns the possibility of increasing the accuracy of forecasting energy consumption in public buildings by taking into account scenarios regarding the possible operating conditions of these buildings. The presented case study illustrates the practical use. The article was mostly written intelligibly, the purpose of the work was clearly presented, I consider the composition of the text to be correct. Nevertheless, I have a few comments.

In my opinion, the article can be published in the journal Buildings after taking into account the suggestions outlined below.

 

1. Nomenclature

RMSE - no description. Repeated below.

 

2. Research methodology. Was any other software used in the analysis besides EnergyPlus? If so, to what extent?

 

3. Table 1. What is “handbook”?

 

4. Lines 160-161. „Due to many possible combinations of the three factors, this study analyzes three levels of the three parameters compared to typical operating conditions, which are 0.3, 0.6, and 0.9.”. Please explain what these values mean.

 

5. Is equation (4) well edited? I have doubts.

 

6. Lines 207-208. „The structure uses a variety of energy-saving and renewable energy (photovoltaic) technologies, which is designed to be energy-efficient.”. Please describe it in more detail. Does this affect the calculations presented later?

 

7. Table 2. What is “window-wall ratio ... east/south/west/north”?

 

8. Table 2. External wall U=0.44, External window U=2.4. It's very poor. My guess is that a building located in southern China does not need to be heated. Is it true? Please write it clearly. In a colder climate zone, heating would be a key factor in terms of energy consumption.

 

9. Figure 3. The red and blue lines in the graphs are the same. It is impossible to distinguish what parameters they represent.

 

10. Figure 3 and the next. What is "Utilization Rate"? Please describe it in more detail in the text.

 

11. Lines 304-305. „The deviation of energy simulation results is smaller than the required error when compared to the ASHARE/FEMP standard ...”. Citation necessary.

Please explain in more detail what the values in Table 5 and Figure 6 mean.

 

12. Figure 9. What is “time ID”.

 

13. In the discussion and conclusions, attention can be drawn to the possible practical use of the method. Based on the possible scenarios and the associated energy consumption calculations, can appropriate recommendations be made on the optimal exploitation of the building?

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 4 Report

The paper addresses the important and actual issue of energy consumption prediction methods and their influence on the optimization of the air-conditioning system operation. as well as the energy-saving diagnosis. The authors propose a data-driven hybrid model based on the expansion of physical simulation data to improve the accuracy and stability of building energy consumption prediction. Overall the paper is interesting and well written. The methodology is sufficiently described and the results are very well presented. However, the references are limited and the authors analysed mainly the Chinese sources, while it would be really beneficial to check the full state of the knowledge. Additionally, the conclusions are not fully clear. I recommend to re-write it with the focus on the benefits of the newly proposed hybrid method, while avoiding the dry list of bullet points in this part. There are several typos and grammar errors so the paper would benefit from the final revision of the English language. Having said this, I truly recommend this article for publishing after minor revisions.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have modified their manuscript to a considerable extent. In doing so, they respected the reviewers' advices, although they were not always completely consistent.

Some shortcomings of the manuscript still remain apparent.

The abstract needs to be revised. The sentence starting in line 14 should be a continuation of the previous sentence, or the previous sentence should be reworded. Further in the abstract some working steps are listed, but without a proper introduction. Furthermore, to someone who only reads the abstract, the considered scenarios for the prediction model verification process remain unclear, as well as the sentence in lines 16 and 17 and the values written in it.

Considering the nomenclature section, the list of abbreviations and the list of symbols should be separated under different section names (abbreviations/symbols).

The abbreviation for the American Society of Heating, Refrigerating and Air-Conditioning Engineers is ASHRAE, not ASHERAE.

Line 34: The very first sentence in the Introduction section must be deleted.

Fig. 3. It is still impossible to differentiate the curves labeled „actual weekday“, „actual weekend“ and „actual holiday“ (all three curves have the same color) and the curves labeled „default weekday“ and „default weekend“ (both curves have the same color). Please use different colors for every curve.

I suggest not mixing the terms “air conditioner” and “fan coil unit” in the manuscript. It would be better always to use the term “fan coil unit”.

Fig.5 and 6. Check the charts labels and figure captions. It seems there are some mistakes there. Please keep the same x-axes scale.

Proofreading by a native English speaker or proofreading service should be conducted to improve both the language and still quality.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Additional Note for Point 2: Fig. 5 shows a monthly data. What kind of data is used as the monthly representation? (monthly mean? or others) Please add explanation. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx