Air Conditioning Load Forecasting and Optimal Operation of Water Systems

Round 1

Reviewer 1 Report

This paper presents some interesting work on introducing load forecasting to white box control modules of air conditioning systems in hotels. Introducing SVR for load forecasting allows improved control in the study presented, increasing efficiency generally over different load ranges.

A few notes for the authors below:

While the English is generally acceptable throughout the paper (perhaps some overuse of definite articles and the occasional run-on sentence), there are quite a few punctuation and capitalization errors that should be addressed.

The second paragraph of the introduction serves as a literature review of sorts, but the connection to the paper content is not clear. Why does the cited literature lead you to this paper? Why simplify from the neural networks in those sources to an SVR in your study? What is the novelty of the work done here in contrast to the other literature studies? Much of this might be clarified in the next paragraph by changing the phrase "most of the existing studies" to "most of the studies listed above".

It seems like something is missing in Eq. 7, where the sum of multiple groups of power expressions (each with units of kW) inexplicably leads to a total consumption with units of kWh. Is there perhaps a "delta t" term missing here, or an implicit consumption of hourly discretization? If so it should be called out. It seems to me these expressions should be integrals with respect to dt which are then made discrete but are missing the delta t component.

Many symbols don't seem to be appearing correctly. This occurs in several places but is very notable just after Eq. 18.

In Figure 3 and the accompanying discussion, it's noted that there are quite a few measurements with ratio of high load forecast to true load compared to the linear fit; however, there's no discussion covering the source or impact of these out-of-mean sampling points. What are the results of neglecting these data points in your prediction mechanism? Will the chiller be significantly unprepared for some percent of operation?

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents an air conditioning load forecast and optimization for a building in China. It is essentially an application paper, where minor localized contributions are made. I believe it may be a nice reference for future works.

Specific comments:

- A better literature review could be made on building load forecast and optimization, which is quite vast. For instance, there is a very recent paper on that

Proposing a hybrid metaheuristic optimization algorithm and machine learning model for energy use forecast in non-residential buildings, Sientific Reports, 2022

and many others on specific themes of building load  forecast and optimization.

- Instead of stating that you used MATLAB to solve the optimization problem, it is more scientifically suitable to describe the method used to solve. For instance, there is no theoretical guarantee of optimal solution for problems with non-linear equality constraints: how did you manage that?

- Equation (4) seems to be misplaced, especially considering the variable description that comes in the next lines 133-135.

- Equation (7) seems to be missing period multiplication to transform power in energy.

- The average energy saving could be placed in the abstract and introduction.

- Equations must be part of the text. Something like, "... is given by

... (1)

where..."

- Put a space after periods, commas and semicolons (e.g. lines 10 and 19), put a space before opening a parenthesis (e.g. line 63).

- Use small case after semicolons (e.g. line 243).

- Use a comma as thousands separation in big numbers (e.g. line 70).

- Use italic for variables.

- Equation (20): putting x into parenthesis can misunderstood as a function, which is not true.

- There is no need to use explicit multiplication operators.

- Use superscripts for powers (e.g. lines 70 m2 and 254 R2) and subscripts for indexes (e.g. lines 117-124 N1-N16).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf