A Cyber-Physical Residential Energy Management System via Virtualized Packets
, Hafiz Majid Hussain 2, Sohail Wahid 2 and Arun Narayanan 2Round 1
Reviewer 1 Report
The paper proposed a cyber-physical system for energy management for residential houses using virtualized energy packets. The system includes three types of flexible loads an energy server and energy router and aims to minimize the energy import in such a system.
The main strength of the paper is that the management of energy in this paper is not based on the price of the energy but on direct requests that impose a "deadline" when the service must be complete. The computational burden is also important.
The main weakness of the paper is that the results are based only on simulation results.
Are the load profiles used in this paper realistic? Are there any databases that can supply these type of data. Can somehow the load profiles be validated.
Is there any energy management used in the paper? The energy management is not well described. It is not clear what algorithm is used to serve the loads. An algorithm flowchart would make things more clear.
Can the energy management system be implemented (even on small scale) to have some real-life validation of the simulated results?
The organisation of section 3 Power scheduling is a little ambiguous and hard to follow. In my opinion it should be organised better.
There is a typo on page 8 row 252 "the the level of reliability...".
All the figures must have captions on x and y axes otherwise it is difficult to understand what is represented in the graphs and figures.
Fig. 8 can not be understood.
The simulation methods or simulation programs are not presented. What programs were used for simulation?
Author Response
To Reviewer #1
First, we would like to thank the reviewer to provide his/her valuable feedback, which included many important aspects that needed to be improved in the previous version of the manuscript. We expect that our answers and changes can clarify the paper’s contributions.
In this document, italic fonts are the Reviewer’s enumerated comments, followed by our answers and actions related to it. In the manuscript, we highlight the changes with red fonts; references and pages in this document are related to the new version of the paper.
######
General comment: The paper proposed a cyber-physical system for energy management for residential houses using virtualized energy packets. The system includes three types of flexible loads an energy server and energy router and aims to minimize the energy import in such a system.
The main strength of the paper is that the management of energy in this paper is not based on the price of the energy but on direct requests that impose a "deadline" when the service must be complete. The computational burden is also important.
The main weakness of the paper is that the results are based only on simulation results.
Authors’ response: Thanks for the appreciation of the paper. We fully agree that our main strength is to provide an effective residential management looking at the “appliances’ needs” and the “available energy” without being mediated by price. While we agree that having these results solely based on simulations is a weakness, we plan to consider in the future both queueing theory as in [8] and multi-agent control as in [32].
Authors’ change: We included in our intention of extending the paper beyond simulations in the conclusions section (page 15, lines 481-483).
######
Concern 1: Are the load profiles used in this paper realistic? Are there any databases that can supply these type of data. Can somehow the load profiles be validated.
Authors’ response: This is an important aspect that deserves a better clarification from our side. We indeed do not consider a load profile that incorporates all appliances of the household (i.e., whole aggregate consumption). This is a key point of our contribution that was indeed unclear. What we are considering are only specific flexible loads and a “pool of energy resources”, which is exclusively used to supply these loads. This management is done “virtually” so we call this as cyber-physical system. Our goal is to allocate the requests of those loads based on the proposed algorithm considering the individual requests and the overall state of the “energy pool”. Therefore, we consider that the other loads that build up the actual household load profile are not included in this analysis. On the other hand, the appliances used in the paper are based on available data from real appliances (EV and Sauna from specifications, dishwasher from REDD dataset).
In any case, we agree that including the full profile is of greater importance and this is an extension we are already work on. However, including a more diversified load profile including all appliances would make harder the assessment of the proposed solution. In the present paper, our focus was to create a model to analyse the specific performance of the proposed management algorithm under very controlled conditions (in some research fields, they call this as “toy-models”).
Authors’ change: We clarified the approach of the paper following our answer above. See page 4 (lines 134-142).
######
Concern 2: Is there any energy management used in the paper? The energy management is not well described. It is not clear what algorithm is used to serve the loads. An algorithm flowchart would make things more clear.
Authors’ response: This is a good point. Yes, there is an energy management that is “rule-based”, as presented in Section 2 and 3. The management is related to the flexible loads and it is done at the energy server, which allocates the resources related to the “pool of available energy”; see answer of comment 1. However, our presentation about this was poor and we fully agree that a flowchart would greatly improve the understanding of the proposed packetized management scheme.
Authors’ change: We have prepared a flowchart (see below and in page 5) to illustrate the packetized energy management run at the server.
######
Concern 3: Can the energy management system be implemented (even on small scale) to have some real-life validation of the simulated results?
Authors’ response: We believe that this is the most important point here. Although our example is a simplified model as previously discussed, we are very sure that the proposed management can be already implemented based on a variation of packetized management proposed by [36] and commercialized by their authors in the start-up company Packetized Energy (https://packetizedenergy.com/). Their approach considers a similar “request-server” relation but focusing on “real-time conditions” from both individual appliances and the grid perspective, using a solution based on local randomization of the frequency that the requests are sent (like random access problems in sensor networks [48]). Although this solution is effective - already commercialized - in smaller scales, our proposed packetized energy management offers a scalable by design looking that incorporate longer time horizons (minutes to hours) via a proactive energy server based on priorities. This could be integrated in the existing market solutions based on Exclusive Groups [41] or even be the basis of a novel model of governance based on commons [50], where electricity is shared among the community. All in all, our ambitious plan is to use the Packetized Energy Management proposed here as the basis of a future Energy Internet [37].
Authors’ change: We extended the discussions including more explicitly the discussions presented above. See pare 14 (mainly in lines 433-441)
######
Concern 4: The organisation of section 3 Power scheduling is a little ambiguous and hard to follow. In my opinion it should be organised better.
Authors’ response: We agree with it following the indications provided by the previous comments. We rearranged the Sections 2, 3 and 4 so they have a more logical organization.
Authors’ change: We have (i) explained the rationale of the section 3 organization and change the name from power allocation to “Numerical results”, (ii) added the flow diagram to explain how the packetized energy management works at the energy server in a new subsection 2.3, as described in Concern 2, and (iii) converted section 4 in a subsection of section 3. See page 9 (lines 287-300), page 10 (lines 320-325) and page 12 (line 375-386).
######
Concern 5: There is a typo on page 8 row 252 "the the level of reliability...".
Authors’ response & change: Corrected.
######
Concern 6: All the figures must have captions on x and y axes otherwise it is difficult to understand what is represented in the graphs and figures.
Authors’ response & change: Corrected.
######
Concern 7: Fig. 8 can not be understood.
Authors’ response: We agree. Figures 7 and 8 illustrate the energy consumption and the “variation” of the EV change levels based on the allocation rules. Figure 8 shows that, when a “soft limit” is considered (i.e., charge the EV regardless of the “available pool of resources”) to satisfy the appliance request, then there are consumption peaks, while no variation in the end battery levels. In Figure 7, the soft limit guarantees that the limit is never passed, but this leads to a variation of the battery levels at the end. In both cases, we showed the impact of different allocation rules.
Authors’ change: We have better explained the plots following the new structure of Section 3, merged with previous Section 4. See page 12 (line 375-386).
######
Concern 8: The simulation methods or simulation programs are not presented. What programs were used for simulation?
Authors’ response and change: This is a very good point. We simulate the scenario using pyhton. We have developed the code ourselves.
######
Reviewer 2 Report
This paper proposes a residential energy management system based on energy packet.
The rule-based approach seems advantageous especially for the implementation.
The authors also present the numerical verification in several different setups.
However, in order to fully understand/evaluate the significance of this paper, there are four concerns that should be addressed.
1. The authors argue that the proposed method can guarantee "a given Quality of Service (QoS)" (Line 383).
However, I could not understand what the "QoS" exactly is and which result indicates its fulfillment.
In fact, the authors also argue that it is not possible to achieve a reasonable QoS and the peak reduction (Lines 363-364).
The authors are recommended to give a clear definition of QoS, and to explain how it can be evaluated based on the results.
2. The time length of a unit packet is set at 10 minutes.
The authors denote the charging process may consist of multiple unit packets.
However, in Fig.3, once the charging starts, the status continues to be "charging" until it becomes full.
The authors are suggested to add some more explanation on the figure and the result to show what is actually happening in the charging process.
3. Description in Lines 7-8 and Lines 115-119 sound somehow contradictory.
If the authors are to argue the advantage over the existing approaches, there should be more detailed discussion on the comparison in the concluding section or elsewhere.
4. Some minor comments:
- It will be very helpful for readers if the authors put labels (e.g. names and units) of the x and y axes in the figures of the results, or at least add such descriptions in their captions.
- Algorithm 1 is not referred to in the body text.
- The appearance of the number "4" in x-axis of Fig.5 seems to be collapsed.
- Some typos, e.g. Line 350: kWh to kW, Line 145: An to A, Line 29: Gelambe to Gelenbe.
Author Response
To Reviewer #2
First, we would like to thank the reviewer to provide his/her valuable feedback, which included many important aspects that needed to be improved in the previous version of the manuscript. We expect that our answers and changes can clarify the paper’s contributions.
In this document, italic fonts are the Reviewer’s enumerated comments, followed by our answers and actions related to it. In the manuscript, we highlight the changes with red fonts; references and pages in this document are related to the new version of the paper.
######
General comment: This paper proposes a residential energy management system based on energy packet. The rule-based approach seems advantageous especially for the implementation. The authors also present the numerical verification in several different setups. However, in order to fully understand/evaluate the significance of this paper, there are four concerns that should be addressed.
Authors’ response & changes: We appreciate the feedback given the Reviewer. We tried to improve the weak points of the previous version of the paper following the proposed comments, which certainly help to clarify the present contribution. The key changes are presented next, when we answer the specific concerns.
######
Concern 1: The authors argue that the proposed method can guarantee "a given Quality of Service (QoS)" (Line 383). However, I could not understand what the "QoS" exactly is and which result indicates its fulfillment.
In fact, the authors also argue that it is not possible to achieve a reasonable QoS and the peak reduction (Lines 363-364). The authors are recommended to give a clear definition of QoS, and to explain how it can be evaluated based on the results.
Authors’ response: That is an important question and we had indeed provided a poor explanation about it. The Quality of Service (QoS) refers to the appliances’ requests. For example, if the appliance is an EV that needs to be fully changed at 7am, but my allocation scheme only provided 80% of charging at 7am, therefore the QoS was compromised.
Authors’ change: We have reformulated the previous sections 3 and 4 (now the part related QoS is subsection 3.3). The new text incorporates an improved explanation of what we meant by QoS. See page 12 (mainly lines 394-396)
######
Concern 2: The time length of a unit packet is set at 10 minutes. The authors denote the charging process may consist of multiple unit packets. However, in Fig.3, once the charging starts, the status continues to be "charging" until it becomes full. The authors are suggested to add some more explanation on the figure and the result to show what is actually happening in the charging process.
Authors’ response: The reviewer is correct. There are two points here: (i) Fig. 4 (previously Fig. 3) only considers EVs so there is no priorities related to different flexible appliances involved, and (ii) the priority rule among the EV (earliest deadline first) that regulates the charging by fully charging the EV that “leaves” earlier. In this specific scenario, there is no charging interruption, which for example is not the case in the example presented in Figure 5. Our previous version was not clear in this regard and, as noted by the Reviewer, our example may be misleading.
Authors’ change: We included a new paragraph explaining this point in page 10 (lines 320-325).
######
Concern 3: Description in Lines 7-8 and Lines 115-119 sound somehow contradictory. If the authors are to argue the advantage over the existing approaches, there should be more detailed discussion on the comparison in the concluding section or elsewhere.
Authors’ response: The Reviewer is very correct, and we appreciate to point this textual contradiction. The point in lines 115-119 of our previous version is that we cannot fairly compare the existing approaches (i.e., centralized utility-optimizers, or decentralized price-driven-optimizers) with the proposed solution because we follow a different method of allocation that does not consider any utility function or price. Our proposed approach is a rule-based one that looks at an aggregated energy available to be shared between the flexible loads and allocates it accordingly. The proposed approach has the capability of shaping loads and directly defining individual priorities with “if-then” structures, without relying on, for example, meta-heuristics or stochastic optimization (computationally complex). This is our claim in lines 7-8 from the previous version. In this sense, we can “qualitatively” compare the advantages of our proposed solution, but we cannot “fairly quantify” it due to the way the problem is approached.
Authors’ change: We rewrote both parts based on the above to fix this apparent contradiction. See page 1 (lines 7, 8 and 11) and page 3 (lines 117, and 120-122).
######
Concern 4: Some minor comments:
It will be very helpful for readers if the authors put labels (e.g. names and units) of the x and y axes in the figures of the results, or at least add such descriptions in their captions. Algorithm 1 is not referred to in the body text. The appearance of the number "4" in x-axis of Fig.5 seems to be collapsed. Some typos, e.g. Line 350: kWh to kW, Line 145: An to A, Line 29: Gelambe to Gelenbe.
Authors’ response and changes: Corrected.
######
Round 2
Reviewer 1 Report
I have no further comments to your paper
Reviewer 2 Report
The authors addressed all my concerns. I recommend this paper to be published.
