Dear Marco Cinelli , Peter Burgherr , Miłosz Kadzinski , Roman Słowinski

I have read your paper

Proper and improper uses of MCDA methods in energy systems analysis

My comments

1- In the abstract you say “lead Decision Makers in shaping the energy systems of the future”

Sorry, but I disagree. As far as I know no MCDM method, except Linear Programming (LP), can handle it. Why?

Because analysis of energy systems involves aspects like contamination, installation costs, variable demand, factor load, replacement time, construction time, new technologies, peak loads, global warming, sun radiation, maintenance costs, final disposal cost (fundamental in nuclear systems and wind blades), material reuse, wild life impact, simultaneous use of wind and PV (photovoltaic cells) installations, wind losses in parallel wind installations, etc., and fundamentally, all of them interconnected, something that is alien to all MCDM, expect LP.

In addition, it is necessary to consider dates as far as 2035, 2050 and 2075, and the energy installations that must last for decades, as well at the advent of hydrogen cells plants (until now, only to 1 MW) in portable units (Ballard, Canada), as well as nuclear fusion plants (D-T), working with hydrogen and lithium (INTER project, in South France).

We also need to consider plants operated by ocean waves movement, some of them already working, as in Brazil.

2- In my opinion weights must not be used. How do you rate the relative importance or utility of wind turbines and PVs? On what basis?

3- Page 1 you say “The energy systems analysis community has shown considerable advancements in the last decades in terms of approaches and strategies for assessing these systems”

Sorry, I cannot support this assertion. What they did is to treat problems with different menthols, considering in each one a total independency of alternatives and not taken into account that the whole energy system is a SYSTEM, and as such, it must be considered holistically, and in addition, they are using invented subjective weights to quantify criteria.

Where is the advance since the 80s?

4- In page 1 “the fit between the types of DMPs and the suitability of the used MCDA methods has been rarely verified”

This is undoubtedly true. It can’t be verified because none of them give information on the extent or range in which each objective is satisfied. Then don’t even demonstrate that the problem is feasible, and they assume that it is, without any proof.

Your box 2.4 in Figure 1 is fundamental. The procedure normally followed at present is that if some characteristic of a problem can’t be modelled, it is easier just to ignore it. For instance, how are at present considered the construction of energy plants for say in 2035? That is, how a precedence or continuity of actual plants is taken into account? Will they continue in that future or will be shut down?

How energy decarbonization is assured? Replaced by what, and on what reasons?

How do they consider that PV installations cannot work the whole day, but that a solar thermal plant may continue working long hours after the sunset?

It appears that the trend is to concentrate in mathematics rather than in reality.

5- In page 2 “The MCDA-MSS is freely accessible software that provides a structured and traceable path for the identification of the MCDA methods most suitable to a specific DMP, applicable to both new (prospective) and published (retrospective) case studies. The upgrade consisted in formulating the guidelines to be respected when choosing an MCDA method”

Do you see? The main problem which is MODELING REALITY, is not even mentioned

6- In page 8 you refer about guidelines for DM and they are indeed useful to select a method.

Don’t you think that the best and unique guideline could be ‘Select the MCDM method that best models all characteristics of a problem’?

7- In page 8 “In case the answer is “Yes” to the question “Does the MCDA method support this(these) missed feature(s)?”, it implies that the MCDA method chosen by the authors of the publication supports the missed feature(s), which leads to outcome 4 in Fig. 1. This implies that the developers of the MCDA-MSS can update its database, re-run the software and obtain outcome 1”

In my opinion, this must be the first stepin the selection of a MCDM method. Once you know the alternatives, and all the characteristics of the problem, expressed by the responsible of e ach area, the DM can define the corresponding criteria and be in a position to see which method fits them the best. Then, going back to the responsibles of each area and consult them about the whole scheme and getting their approval. Normally this people are very knowledgeable in their respective areas and completely ignorant, and not interested in the methods to be used, let alone in their mathematics.

8- In page 8 “Importance coefficients weights express the intrinsic importance of each criterion, meaning their voting power”

Subjective criteria weights are trade-offs that don’t have any voting power, because they are not related to the alternatives they should evaluate. Different is the case of objective criteria weights from entropy, CRITIC or standard deviation.

9- In page 9 “The formulation of every DMP requires selecting the type of decision recommendation the stakeholders would like to receive”

This makes a lot of sense but is difficult in practice. A production engineer may need a costly equipment that the financial officer does not want to buy. The only solution is a compromise solution where each part cedes something, but for that, these two stakeholders must be linked, something that does not happen in MCDM methods.

10- In my opinion, there are three expressions that should be banned in MCDM. They are subjective weights , preferences, and pair-wise comparison. They do not have room in real-life scenarios, unless they refer to trivial and personal scenarios like selecting a movie, choosing a mode to travel, or picking a restaurant for dinner.

We have to work with facts, with reasoned assumptions, with research and with information, not based on intuitions, pair-wise comparisons or using a MCDM method because it is popular’

Of course, there are subjective issues, that can be described by surveys and by using a simple scale for criteria considered independently.

Participation and opinion of the DM is absolutely necessary, utterly important and essential.

But there must be a symbiosis between the mathematics of a method’s result and the DM. For this reason, it is not advisable that the DM modifies, according to his/her idea or opinion the initial data.

Once the result is achieved by any method, not using wrights and preferences, the DM is the person to analyze the result, make whatever changes in the initial matrix according to his judgement, (not preferences), and even reject the result. He may decide selecting the second-best alternative instead of the best, according to his interpretation of result and sensitivity analysis, analyzing aspects that can vary and change the result, etc.

11- In page 11 “The research we propose in this paper shows an interesting link with what Robyn Dawes did in the late 1970s’, when he studied the proper and improper strategies to develop (linear) psychological models”

Not in my opinion, psychology is not related to MCDM

The article puts emphasis in weights. Naturally, not all criteria have the same importance and this is a fact that must be considered. As an example, Linear Programming does not use weights, however, the significance of each criterion is analyzed and modified at each iteration (and there could be thousands in a complex problem), and is an essential part in the method.

12- In page 12 “This research points out the need to stress that MCDA methods have their own implementation checklists, which can vary notably from one method to another’

Possibly this is true, but what happened when there is no method that is able to manage the problem?

The reason? Its inability to model all characteristics of the problem.

What I can assure you because I proved it, is that a method based on Linear Programming may address the very difficult problem of energy analysis.

13- In page 11 “This paper presents the first systematic assessment of the suitability of using MCDA methods in a representative set of real-world case studies concerning energy systems analysis”

I am afraid that this assertion in inexact. There is a paper published by Springer as a pre-print months ago, that address extensively this subject. Its title is:

Transition to renewable energy – An attempt to model the mix of existing and future generation technologies for 2035 and 2050

These are my comments. I will be more than glad to discuss this problem in depth with my colleagues of Poland, either in RG or using my email

[email protected]

Hope these comments may be useful

Nolberto Munier