A Two-Step Energy Management Method Guided by Day-Ahead Quantile Solar Forecasts: Cross-Impacts on Four Services for Smart-Buildings

Round 1

Reviewer 1 Report

It is a very interesting research which is also very important for renewable energy implementation in smart buildings. The four discussed services are also useful in the energy management. However, the flowing points are need to revise. Generally, it is better to revise the English in a more scientific way. For example, some of part it write like “it seems”, “probably”, which should not be used in papers. The title of 3.4 should not be a question. The abstract should revised as the structure of challenges and issues, proposed method and main result.

Author Response

All your comments were considered pertinent for the improvement of the paper. The corresponding corrections proposed are detailed as follows:

English corrections:

Non-scientific english expressions were corrected, as suggested by the reviewer.
"It seems" has been replaced in lines: 461,470,472,491,606,615,622
"maybe" has been replaced in lines: 307,398
"might" has been replaced in lines: 467,474,507,903

Title of section 3.4 should not be a question.It has been changed by:
'Impact in performance of targeting one service over the non-target services':
line 499

Abstract revision:
Some changes in structure of challenges and issues, proposed method and main result
are suggested by reviewer. While trying to keep the abstract in a reasonable extension
we proposed the following corrections:

Clarifications added in line 112:
"trhought the use of quantile forecasts"
"namely: Energy cost, carbon footprint, grid peak power
and grid commitment are the four services evaluated,
being the latter a novel service proposed in the domain of microgrids"

Line 114 rephrased:

"The opposing nature among some of the target services as well as the different requirements
in terms of forecasts when optimizing for different services and seasons of the year,
two of the main conclusions of this work. This fact highlights the usefulness of the quantile
forecasting approach in an energy management system, as a tool to deal with the intrinsic
uncertainty of PV power production."

Reviewer 2 Report

Section 1 & Ref list

Reference list could be expanded considering the MG relevant related topics and trends (Artificial Intelligene, BESS optimization, Local Energy markets, Blockchain technology application, etc.)

Please look at (for example):

Multiobjective intelligent energy management for a microgrid

Automatic generation control of microgrid using artificial intelligence techniques

Optimal management algorithm for battery energy storage system included in an islanded Micro-Grid

Artificial intelligence framework for smart city microgrids: State of the art, challenges, and opportunities

Local energy markets: Paving the path toward fully transactive energy systems

Blockchain applications in microgrids an overview of current projects and concepts

[...]

***********

Section 2:

The real case study is interesting.
Please consider a mix of NWP and AI as potential technique (see comments above)

The forecasted method is essential a persistence model? You should motivate this choice among potential other hybrid solution.
Line 209: forecasts are never "perfect" and PF methods are not convinncing for this reviewer. I'm not sure it is valuable a
comparison not taking into consideration other relevant methods.

Service 4 is very interesting. However it is not feasible in many countries. Please give more details about tariffs in the
context of your case study (France, isn't it?) and try to generalize for a gobal/general value.

Figure 4: please double check flow chart (it seems Pbatt is always negative).

 

***********

Section 3:

Please double check grammar (E.g. "This results show"..)

Section 3.1-3.2: your interesting questions in line 409 (what does it mean "a proper forecasting method"? What characteristics define it? [..])
seem not well addressed as promised in the following section. The explaination is not clear and too many strategies/prameters mixed together.
Too many acronyms are confusing for the reader.
Then new indicators appear showing a dependence on the sub-period/sesonal time!! (no atincipated, thus the logic flux is again confusing).

Maybe a final table which summarize just rlevant outcomes can help to understnad the main concept.
It seems that an initial trial design planning was missing.

****

The conclusion section is too long (honestly the paper also with alle the appendices material.

Final major concern: no indcation about time of the algorithm and EMS response are given in the manuscript. This should be a relevant issue
in this specific context.

However the generla methodology and proposed framework for EMS application in in MG (group of buildings?) are valuable for this reviewer who suggest
a general revision to make more effective the proposed approach.

Author Response

Section 1 & Ref list

Remark: Reference list could be expanded considering the MG relevant related topics and trends
(Artificial Intelligene, BESS optimization, Local Energy markets, Blockchain technology
application, etc.).

R\ The introduction was classed using subtitles as suggested by the reviewer in his
example, to improve the clarity of this section. Titles included:
"energy management of microgrids" (line 161)
"Resources scheduling and forecasts" (line 172)
Paragraph of line 175 deleted
"that constitutes one of the main contributions of this
work regarding the uncertainty challenge" added to line 193

However, due to the nature of the article (not a review article) and the already long extension of
the article (as pointed out by you in the last part of your remarks), we prefered to keep the reference list as short
as possible sticking only to the very relevant subjects/areas directly related to the work performed.

Section 2:

The real case study is interesting.
Please consider a mix of NWP and AI as potential technique (see comments above)
R/: AI is definitely one of the paths to explore in the future (including reinforcement learning),
in order to benchmark the proposals made on this paper and to produce forecasts of production and
consumption (i.e. net demand).

The forecasted method is essential a persistence model? You should motivate this choice among potential other hybrid solution.
R/ No, the Persistence model is only a benchmark method (typically used with that purpose within the
meteorological community). The proposed quantile forecasting method, obtained from the analogs ensembles
method, is a probabilistic forecasting method. The adapted analogs ensembles probabilistic forecasting method
has been also developed by this author and is subject of another publication that is currently under review.

Line 209: forecasts are never "perfect" and PF methods are not convinncing for this reviewer. I'm not sure it is valuable a
comparison not taking into consideration other relevant methods.

R/ Actually we acknowledge the fact that foreacsts are never perfect as you mention. However, we used
Perfect forecasts (PF) just as the best (theoretical) benchmark method one could ever have. This set the highest reference line
of comparison, even though unachievable, give a good tool to evaluate performance of forecasting methods.
But we also compare the results obtained with other benchmark forecasting methods, such as persistence or the commercially
available NWP forecasting method from meteo-France (e.g. see table 3, table 6).

Service 4 is very interesting. However it is not feasible in many countries. Please give more details about tariffs in the
context of your case study (France, isn't it?) and try to generalize for a gobal/general value.
R/Fig. 3 added to clarify the tariffs in our case study.
"Countries or electricity markets where the contracted power is charged using a similar fee schedule
as the one presented in figure \ref{fig:TempoTariff}, could benefit from this service." added in line 302.

Figure 4: please double check flow chart (it seems Pbatt is always negative).
R/ We actually use a generator convention and Pbatt can be positive or negative. This is stated in the central
diamond box of figure 4 (now figure 5), where the action is taken based on the sign of the power of
the battery (negative to the left, positive to the right).

Section 3:

Please double check grammar (E.g. "This results show"..)
R/checked and corrected

Section 3.1-3.2: your interesting questions in line 409 (what does it mean "a proper forecasting method"?
What characteristics define it? [..]) seem not well addressed as promised in the following section.
The explaination is not clear and too many strategies/prameters mixed together.
R/ We agree with your remark, the following explanation was added in line 480 to clarify the answer to the question:
"...Therefore, at the view of this results and acknowledging the fact that a forecasting error will be always present,
a proper forecasting method could be defined as such that, due to its persistent under or over estimative nature,
allows the EMS to produce optimal performance in a particular service by providing certainty on the sign of the
imbalance between production and consumption."

Too many acronyms are confusing for the reader.
R/ There are some acronymas that are basic and should not be removed, namely those related to the services provided:
GC (grid commitment), EC (energy cost), GPP (grid peak power) and CO2 (CO2 emmissions). Those are directly related with
the strategies meant to minimize them (e.g. GCmin, ECmin, CO2min et GPPmin).
There are other basic and recurrent acronyms that are considered pertinent, such as: EMS (energy management system),
MG (microgrids), NWP (numerical weather prediction), AnEn (analogs ensembles), among others. However, it is true that
some other acronyms might be unnecessary and make the reading more confusing, then they have been replaced by their
full definition, namely:

DERs (distributed energy resources) line 415
PF (Perfect forecast) line 437
PE (persistence forecast) line 444,
through-output-enegy (TOE) line 505,
BAL (balancing) line 552,
carbon footoprint line 660,

Then new indicators appear showing a dependence on the sub-period/sesonal time!!
(no atincipated, thus the logic flux is again confusing). Maybe a final table which summarize just
rlevant outcomes can help to understnad the main concept. It seems that an initial trial design planning was missing.
R/ We agree that this section has a lot of results and its comprehension might become a bit difficult. Following your advice,
we included figure 8, where the logical design flow of the section is presented, which includes the main research questions and
references to where the answers to those questions can be found.

The conclusion section is too long (honestly the paper also with alle the appendices material).
R/ Conclusions where rephrased and shortened, trying to make them more concise.

Final major concern: no indication about time of the algorithm and EMS response are given in the manuscript.
This should be a relevant issue in this specific context.
R/ Reference to computation times added in line 346: " The computation time of the GA (~10-15 minutes) is a
disadvantage compared to the fast convergence of the NLP (less than one minute)"
and in line 363:
"The BAL module, being a rule-based method, requires very few time (< 10s) to find a solution."
and in line 161:
"An advantage of this method is its short computation time (<30s with a two-year database), compared to the several-hours
required by commercial probabilistic forecasting methods, such as the one produced by the European centre for medium range
weather forecasts (ECMWF)"

Reviewer 3 Report

In this paper the authors present a novel two-stage energy management strategy in the context of microgrids, which combines optimization-based and rule based approaches. The proposed approach also relies on a novel forecasting scheme aimed at reducing uncertainty of PV power production. The authors show that the proposed approach outperforms some reference strategies in terms of different performance metrics. Furthermore, the authors discuss the trade-off of the considered objectives and establish requirements of forecasting strategies for different service optimization, as well as different seasons of the year.

In my opinion, the timeliness of the covered topic is good and the scientific soundness of the proposed research is appreciable. Nevertheless, this paper present some flaws and can be improved. In particular, my main concern is related to the exposition and manuscript organization. Concepts and definitions are scattered all over the text without a logic structure and this makes the manuscript hard to read.
Here are my main comments:

1) The introductory section lacks of a summary of the main research contributions. I suggest to include a paragraph (or a subsection) that highlights, in a clear and concise way (eg, bullet list), the novelty of the proposed approach and how it differs from state-of-the art solutions (e.g., the use of quantile forecasting, etc.)

2) The acronym TSO (line 77) should be defined.

3) In subsection 2.1 ("The Analogs Ensembles method") the authors introduce the concept of quantile forecasting (QE) and they state that "it is used as a novel forecasting method ...". The authors must include a more detailed description of this method or at least a brief descriprion corroborated by a significant reference.

Furthermore, In line 161, the notation AE _ {\ tau = 0.1} suddenly appears without being defined before. Which is the meaning of this notation?

The same notation is partially explained in line 417 (Results section). Why? I believe that this part (from line 417-426) should be included in the description of the QE technique. Has AnEn the same meaning of AE?

4) Line 218: Benchmark forecasting method. Is this a subsection?

5) What is table PVB_ {max ^ -} in Table 2? Does not seem to be defined before. Pleas try to better clarify the notation.

6) The acronym TOE is used in line 498 before being defined. A definition of TOE is only provided in Appendix A. Even here, TOE is used multiple times before being defined in line 781.

7) In my opinion, the conclusion section is too verbose. It this form, it appears as a redundant discussion on obtained results. The conclusions section should concisely report the most important results and the main message of the paper. It should clearly summarize the importance and the novelty of the proposed research with emphasis on obtained results.

Author Response

1) The introductory section lacks of a summary of the main research contributions. I suggest to include a paragraph
(or a subsection) that highlights, in a clear and concise way (eg, bullet list),
the novelty of the proposed approach and how it differs from state-of-the art solutions
(e.g., the use of quantile forecasting, etc.)
R/ We agree with the remark, then we included a bullet list with the main contributions of the work, as suggested (lines 123-139)

2) The acronym TSO (line 77) should be defined.
R/ It is actually not used anywhere else in the text so it was substituted by its full name

3) In subsection 2.1 ("The Analogs Ensembles method") the authors introduce the concept of quantile forecasting
(QE) and they state that "it is used as a novel forecasting method ...". The authors must include a more detailed
description of this method or at least a brief descriprion corroborated by a significant reference.
R/ It is considered that the basic explanation of the analogs ensembles and quantile forecasting
method is presented (lines 156 - 187). The references to the articles are also presented in line 160 ([20][21])
The comparison criteria, which is the core of the Analogs Ensembles method, is presented in equation 1, as
well as a graphical scheme of the workflow of the method is presented in figure 2. Therefore, with the aim of
respecting also the recommendations of other reviewers, who have pointed out the the article was already
very long, we decided not to go deeper in the explanations of this method.

Furthermore, In line 161, the notation AE _ {\ tau = 0.1} suddenly appears without being defined before.
Which is the meaning of this notation?
R\You are right in your remark, the paragraph has been rephrased to:
"...different quantiles are obtained with probability levels ranging from 10\% to 90\% ($AnEn_{\tau=0.1}$
to $AnEn_{\tau=0.9}$) which are used as deterministic forecasts. The notation $AnEn_{\tau}$ refers to a
quantile with probability level $\tau$ obtained from an ensemble produced with the Analogs Ensembles method"

The same notation is partially explained in line 417 (Results section). Why? I believe that this part
(from line 417-426) should be included in the description of the QE technique.
R/ The notation is included there because it is the notation used in the tables (from table 2 to 6) and
in figure 7. It is just the notation used to present the results that use quantiles as forecasts that's why
they it was included there. However, as remarked above the notation is now also introduced in line 161.

Has AnEn the same meaning of AE?
R/Yes, it was corrected in line 161

4) Line 218: Benchmark forecasting method. Is this a subsection?
R\Yes, numered label was added

5) What is table PVB_ {max ^ -} in Table 2? Does not seem to be defined before.
Pleas try to better clarify the notation.
R/ It is actually explained in section 2.4.3, line 399

6) The acronym TOE is used in line 498 before being defined.
A definition of TOE is only provided in Appendix A. Even here,
TOE is used multiple times before being defined in line 781.
R\Explanation added in line 538

7) In my opinion, the conclusion section is too verbose. It this form,
it appears as a redundant discussion on obtained results.
The conclusions section should concisely report the most important results and
the main message of the paper. It should clearly summarize the importance and the
novelty of the proposed research with emphasis on obtained results.
R\It was also pointed out by other reviewer. We agreed and try to rephrase in a more concise
manner this section

Round 2

Reviewer 2 Report

The authors have addressed most of the reviewer remarks. However a further minor step is necessary before final publication.

 

Comments:

 

Lines 16-18: please rephrase.

Figure 3: could you please use English language?

line 394: unexpected bold font should be double checked.

Table 1: Quadratic Programming never cited in the text: could you please motivate why used for GPP?

Table 2 strategy, strategy: repetition

717-719: please rephrase for sake of clarity.

Conclusion section is still too long and should be compressed. You can consider to create a previous section
with "Discussion of results".

Reference list: should be revised

(e.g [20] Conference proceedings, journal or what?? [30] "distribuitor"??)

Concern: you cannot say you it is not possible to expand the reference list after 31 pages article.

Remark again: Reference list could be completed considering the MG relevant related topics and trends
(Artificial Intelligene, BESS optimization, Local Energy markets, Blockchain technology
application, and last emerging topic in recent literature).

 

Author Response

First of all, thank you for the follow up, the detailed revision of the article and the constructive comments you have made. This will definitely help to the quality of the publication.

Lines 16-18: please rephrase.

R/ It has been rephrased to:

"...The fact of how optimizing for some services during the scheduling (i.e. grid commitment) can be highly detrimental for the performance of the non-targeted services, is an interesting finding of this work. The differences regarding the optimal forecasting eccentricity (i.e the forecasting quantile) required when optimizing for the different services and seasons of the year is also considered an important conclusion of the study..."

Figure 3: could you please use English language?

R/ Changed to english

line 394: unexpected bold font should be double checked.

R/ Bold eliminated

Table 1: Quadratic Programming never cited in the text: could you please motivate why used for GPP?

R/ It is actually justified in section 2.4.1. Non-linear programming (NLP) has been changed to quadratic programming (QP) as we were actually referring to the same algorithm.

Table 2 strategy, strategy: repetition

R/ Corrected

717-719: please rephrase for sake of clarity.

R/ Rephrased to:

"...It is important to remark the importance of taking into account the embedded CO\textsubscript{2} emissions of the battery and PV panels when assessing the advantages of implementing an energy management strategy with a view to reduce CO\textsubscript{2} emissions...."

Conclusion section is still too long and should be compressed. You can consider to create a previous section with "Discussion of results".

R/ Conclusions have been compressed to less than one page

Reference list: should be revised

(e.g [20] Conference proceedings, journal or what?? [30] "distribuitor"??)

R/Reference list revised and corrected

Remark again: Reference list could be completed considering the MG relevant related topics and trends
(Artificial Intelligene, BESS optimization, Local Energy markets, Blockchain technology application, and last emerging topic in recent literature).

R/ Suggested topics added in lines 71-83, 120-126