A Physical Model-Based Data-Driven Approach to Overcome Data Scarcity and Predict Building Energy Consumption
Round 1
Reviewer 1 Report
This study shows a hybrid method using physics-model and data-driven models, which overcome the limited datasets and low model performance. The topic is very interesting and the validation is peformed based on the real field data. My review comments and questions are as below:
1) Clarify the intended meanings of four cases constructed.
2) I think the MAE is better than MBE, especailly for the absolute MBE in Figure 8.
3) Please discuss more about the results and the exceptionin Case 2.
4) What is the reanson why the result of A1 in Case 1 is lower than that of Case 3. This is because Case 1 has much more datsets in the training.
5) Explan which case (or cases) show the issue of data scarcity.
Author Response
Please see the attached file
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Interesting study to predict the energy consumption of the building. The conventional data-driven method has been combined with physical model forming a hybrid method.
The structure of the paper was built well. The findings of the paper show there is an improvement in prediction when the hybrid model is adapted however, the error is still high. I believe some points need to be addressed:
-More detail is required especially about the physical model.
- There is a huge error in predicting the summer operation. Does the reason for this really the window opening? if it is, the physical model should already cover it because the room set temperature has been given already.
- to make the work more attractive, the summer operation may be removed from the paper ( but the data have been collected already, and the data can be still useful). or instead, more explanation-discussion could be added.
-Please add some quantitative findings to the Abstract.
-Please provide a list for all abbreviations
Author Response
Please see the attached file
Author Response File: Author Response.pdf
