Talk:Open energy system models

open_eGo: open electricity Grid optimization

This project can be added when it goes live at the end of August 2016. RobbieIanMorrison (talk) 16:39, 21 July 2016 (UTC)

 Done. RobbieIanMorrison (talk) 12:05, 26 September 2016 (UTC)

The project seems to have been removed. It was clearly present at one point as indicated. In any case, the project is now known as oGoⁿ. RobbieIanMorrison (talk) 09:11, 18 October 2024 (UTC)

Split page in due course: Open energy systems data

Given that open energy system data projects are on the increase, this page could be split in due course to yield a new Open energy systems data page. RobbieIanMorrison (talk) 12:05, 26 September 2016 (UTC)

 Done. Please see Open energy system databases. RobbieIanMorrison (talk) 10:47, 4 February 2017 (UTC)

Spanish wholesale power market data

The following Spanish initiative could be a possible addition to the open data section in due course. Best wishes. RobbieIanMorrison (talk) 20:32, 5 October 2016 (UTC)

• "PriceProfor Engine: Iniciativa abierta para el modelado, el análisis y la explotación de datos de precios del Mercado mayorista eléctrico Español" [PriceProfor Engine: open initiative for modeling, analysis, and exploitation of price data from the Spanish wholesale power market] (in Spanish). Retrieved 2016-10-05.

 Not done. Shifted comment to the open energy system databases article. RobbieIanMorrison (talk) 10:56, 4 February 2017 (UTC)

Further models

Open energy system modeling projects that could be added in due course

Project	Model type	Academic publications	License details	Code published	Comments
AMIRIS	java-based agent-based modeling framework		Apache 2.0	GitLab	to add
COMANDO	python modeling framework beyond MILP		MIT	GitLab instance	to add
CREST	spreadsheet-based micro-energy system model		CC BY-ND 3.0		cannot add, the ND variant is not an open license
DDPP	spreadsheet-based energy system model	no	not specified
deeco	high-resolution energy system modeling framework		GPLv2		to add   (although project abandoned 2005)
Einstein	single-site analysis		GPLv3		to add
EnergyNumbers– Balancing	interactive electricity system model	no	GLPv3
energyRt	model from United States written in R	no	AGPLv3.0		need to confirm license status
ETHOS.FINE	Framework for Integrated Energy System Assessment		MIT	GitHub	to add
FlexTool	Power system flexibility for the energy transition		LGPLv3+	request access	to add
Friendly Sam	energy system optimization toolbox	not known	LGLPv3	GitHub	license needs to be added to codebase
GCAM	Global Change Analysis Model IAM		ECL-2.0	GCAM Core	to add
GENeSYS-MOD	global energy system model (based on OSeMOSYS)		Apache 2.0	GitLab instance	to add
GnuAE	GNU alternative household energy systems	no	GLPv2		to add
GridCal	Research oriented power system software		LGPLv3	GitHub
GridPath	Versatile power-system planning platform		Apache 2.0	GitHub	to add
Grid Sim	browser based public education model		Apache 2.0	GitHub	to add   (an initiative by google)
highRES	GAMS-based model for Great Britain and Europe		MIT	GitHub	to add
M-LED	multi-sectoral latent electricity demand platform		MIT	GitHub	to add
mat-dp	material demand projections model		MIT	GitHub	to add
MultiMod	large-scale equilibrium model of global energy markets		not specified	not yet	first top-down model?
OMEGAlpes	district scale model		Apache 2.0	GitLab
open_plan	decentralized systems with emphasis on built environment	no	AGPLv3	GitHub
OpPoDyn	research on the dynamics of networks and power grids		GPLv3	GitHub	to add
PLEXOS Open EU	based on proprietary PLEXOS software		not specified	not yet
PowSyBl	Power System Blocks	no	MPL-2.0	GitHub	to add
Rogeaulito	projected accounting model for global energy		not specified		publication indicates open license, have emailed the developers
Spine	energy system model				to add
SpineOpt	energy system model				to add
StELMOD	stochastic optimization model for European electricity markets		MIT		to add
TIMES Model Generator	framework maintained by ETSAP		GPL‑3.0‑or‑later		to add
WWS project	based at Standford University		MIT		to add
Academic publications: pending or published academic literature or the presence of secondary sources is used as a criteria for inclusion in the article. License details: projects must specify their license first (spreadsheet projects are prone to not doing so). Code published: lists projects that have indicated their intention to make their code public but have not yet done so. Blank cells indicate an affirmative.

Change log

• the following models could be added in due course — RobbieIanMorrison (talk) 17:18, 2 December 2016 (UTC)
• updated list — RobbieIanMorrison (talk) 07:31, 9 December 2016 (UTC)
• transfered to table format and updated again — RobbieIanMorrison (talk) 10:14, 20 December 2016 (UTC)
• updated WWS project entry — RobbieIanMorrison (talk) 00:12, 21 December 2016 (UTC)
• added Einstein project — RobbieIanMorrison (talk) 10:51, 22 December 2016 (UTC)
• added EnergyPLAN — RobbieIanMorrison (talk) 13:44, 22 December 2016 (UTC)
• updated Einstein license details — RobbieIanMorrison (talk) 15:57, 22 December 2016 (UTC)
• removed EnergyPLAN as it is a binary only distribution — RobbieIanMorrison (talk) 23:12, 22 December 2016 (UTC)
• added the Open Government License to the UKTM entry — RobbieIanMorrison (talk) 10:54, 4 February 2017 (UTC)
• added entry for GnuAE project — RobbieIanMorrison (talk) 12:27, 6 February 2017 (UTC)
• updated entry for PLEXOS, see this newsgroup posting — RobbieIanMorrison (talk) 10:48, 6 March 2017 (UTC)
• ONSSET entry removed as it is now included in the article — RobbieIanMorrison (talk) 10:51, 13 March 2017 (UTC)
• added deeco — RobbieIanMorrison (talk) 11:00, 13 March 2017 (UTC)
• added CREST (CREST integrated PV electricity demand model 1.0) model / Loughborough University — RobbieIanMorrison (talk) 23:03, 14 March 2017 (UTC)
• added Friendly Sam from RISE, Sweden — RobbieIanMorrison (talk) 15:17, 25 April 2017 (UTC)
• added Grid Sim from Google scientists  — RobbieIanMorrison (talk) 17:28, 3 October 2017 (UTC)
• added GridCal  — RobbieIanMorrison (talk) 10:41, 7 January 2021 (UTC)
• added Spine  — RobbieIanMorrison (talk) 12:56, 27 January 2021 (UTC)
• added FINE  — RobbieIanMorrison (talk) 13:36, 19 May 2021 (UTC)
• added AnyMOD.jl  — RobbieIanMorrison (talk) 13:40, 12 August 2021 (UTC)
• added open_plan  — RobbieIanMorrison (talk) 09:04, 8 September 2021 (UTC)
• added GridPath — RobbieIanMorrison (talk) 11:42, 11 October 2021 (UTC)
• added COMANMD  — RobbieIanMorrison (talk) 05:44, 3 November 2021 (UTC)
• added PowSyBl from Linux Foundation  — RobbieIanMorrison (talk) 21:15, 16 November 2021 (UTC)
• added AMIRIS (Agent-based Market Model for the Investigation of Renewable and Integrated Energy Systems) — RobbieIanMorrison (talk) 14:35, 26 January 2022 (UTC)
• added OMEGAlpes  — RobbieIanMorrison (talk)
• removed UKTM or UK TIMES model: still not released and I asked at a webinar last month and clearly not imminent — RobbieIanMorrison (talk) 13:47, 6 April 2022 (UTC)
• added M-LED  — RobbieIanMorrison (talk) 14:50, 30 April 2022 (UTC)
• added GENeSYS-MOD  — RobbieIanMorrison (talk) 07:35, 3 June 2022 (UTC)
• added FlexTool  — RobbieIanMorrison (talk) 16:17, 10 June 2022 (UTC)
• Added SpineOpt  — RobbieIanMorrison (talk) 09:59, 28 July 2022 (UTC)
• removed AnyMOD.jl, now live — RobbieIanMorrison (talk) 06:31, 29 August 2022 (UTC)
• added mat-dp model that provides LCA for energy systems analysis  — RobbieIanMorrison (talk) 06:07, 7 September 2022 (UTC)
• added CGAM integrated assessment model  RobbieIanMorrison (talk) 22:37, 4 October 2022 (UTC)
• added TIMES Model Generator  RobbieIanMorrison (talk) 10:04, 21 February 2023 (UTC)
• added highRES  RobbieIanMorrison (talk) 09:07, 26 August 2024 (UTC)
• removed RAMP (now on main page) (retrospective note) RobbieIanMorrison (talk) 09:11, 26 August 2024 (UTC)
• added OpPoDyn, here is a pubication on the PowerDynamics.jl component  RobbieIanMorrison (talk) 07:11, 31 August 2024 (UTC)
• added ETHOS.FINE (although seems related to FINE)  RobbieIanMorrison (talk) 09:28, 3 February 2025 (UTC)
• removed FINE after talking to developers, now named ETHOS.FINE RobbieIanMorrison (talk) 10:56, 3 February 2025 (UTC)

References

Likely AI generated material

Hi - I added the AI generated tag to this article, as the edits by RobbieIanMorrison display several indications of LLM use and thus need extra review for accuracy/hallucinations, sourcing issues, editorializing/non-neutral tone and synthesis, and the like. Gnomingstuff (talk) 22:09, 6 October 2025 (UTC)

@Gnomingstuff: could you expand on that? @RobbieIanMorrison: did you use LLMs to help write some of the stuff in this article?

Let's see if we can work towards removing the tag at some point. —Femke 🐦 (talk) 17:32, 28 November 2025 (UTC)

Absolutely not. I think that tag is not warranted is this case. I have never used Google search to help write text. I have never used ChatGPT at all, ever. I might occasionally run a sentence or two through DeepL Write more recently to see if the text could be better framed. Perhaps someone could give an example of the "offending" material? That would really help me to comment specifically. I should add that I have worked in this field for 30 years. And I have read all the sources I have cited. Interesting times, I guess! Best, RobbieIanMorrison (talk) 18:05, 28 November 2025 (UTC)

And can we compare timestamps? I would guess that most of that article was written prior to LLMs being available. RobbieIanMorrison (talk) 18:07, 28 November 2025 (UTC)

DeepL is a large language model tool. Therefore I have restored the tag, as it is warranted on the grounds of AI use literally being confirmed. "Interesting times" indeed.

The primary visible problem is AI editorializing -- asking AI to "better frame" something is turning over the article prose to synthesis and original research, resulting in such AI-characteristic text like highlighting the need to reduce coal consumption, highlighted the need to coordinate infrastructure policies and import strategies. Which carries several issues:

• It's the exact same kind of superficial editorializing that AI produces for virtually any topic (see WP:AISIGNS, "Superficial analyses" and "AI vocabulary" sections)
• It is opinion stated in wikivoice; your idea of a "need" may not be my idea of a "need," which is probably not the person reading this's idea of a "need" either. The people holding the opinion, in this case, are the researchers of the study cited. However...
• AI-generated text like this tends either to be not backed up by the source at all, or an overly close paraphrase of the source material. The second statement is the latter, as it is almost verbatim from the study cited. And that's just one clause; to "work toward removing the tag," any text written, "better framed," or otherwise touched by LLM tools needs to be similarly scrutinized.

And no, this kind of text was not present in versions prior to widespread AI availability. Gnomingstuff (talk) 14:02, 3 December 2025 (UTC)

Fellow Wikipedia editors: can we gain some perspective please? Firstly:

• LLMs can be used in a number of ways, ranging from improving grammar to generating entire texts from a single prompt
• my use of DeepL.Write, if indeed deployed when contributing to this article, would have been limited solely as an aid to single sentence formulation

Can someone point to Wikipedia policy that relates to this use‑case of sentence-by-sentence formulation assistance under fine‑grain human agency?

More specifically, relative to this article:

• by my estimate, I could have utilized DeepL.Write for the GENeSYS and RAMP sections (but I have no record, of course, of having done that)
• by my estimate, I could have utilized DeepL.Write for five or more one sentence edits (the exact number is hard to gauge)
• any such usage by me would have been restricted to single sentences under careful control
• the change log shows that 84% percent of article commits predate November 2022, a date used by German academics to mark the public uptake of LLMs
• I am responding directly because I have authored circa 90% of this article

I want to repeat that I have never used AI to generate text, as normally understood, when editing Wikipedia. Never.

The example in the previous reply regarding the term "highlighting" is bizarre in my view — with each usage backed up by a separate peer‑reviewed article. Those sentences are also separated in the text and do not run together as implied earlier. Both cited publications use the term "highlight*" multiple times (counts: Burandt et al (1 December 2019): 6, Hainsch et al (15 January 2022): 11). The example quoted is indeed the kind of framing that predated LLMs. And, yes, it is acceptable and probably desirable to use phrasings from the original sources when editing articles (without violating copyright of course).

I find the the use of assertions presented as certainties and the reverse burden‑of‑proof evident in this discussion somewhat problematic.

Finally, I did not confirm my use of DeepL.Write when editing this article. That is a misrepresentation that should be corrected (@Gnomingstuff: please note). @Femke: FYI. Best, RobbieIanMorrison (talk) 16:56, 3 December 2025 (UTC)

by my estimate, I could have utilized DeepL.Write for five or more one sentence edits (the exact number is hard to gauge)

I want to repeat that I have never used AI to generate text, as normally understood, when editing Wikipedia. Never.

So which is it, then? Gnomingstuff (talk) 17:05, 3 December 2025 (UTC)

There's a difference between generating completely new text (which has a high hallucination rate), and copyediting (which has a low hallucination rate, and hallucinations are easier to spot as you know what the source says). I do the second sometimes, even though I have no patience for the former. —Femke 🐦 (talk) 17:10, 3 December 2025 (UTC)

There's a difference, but the above comment consists of hundreds of words Gish galloping through minor nitpicks (no shit, those two quotations that don't syntactically fit together are in fact two separate quotations?) while carefully talking around what was and was not done.

As far as the specific issue with the stuff I mentioned (which are only two examples out of who knows how many): If someone states an opinion, that opinion must be attributed to the source, not stated in wikivoice. This is policy, per WP:NPOV: Usually, articles will contain information about the significant opinions that have been expressed about their subjects. However, these opinions should not be stated in Wikipedia's voice. Rather, they should be attributed in the text to particular sources, or where justified, described as widespread views, etc. For example, an article should not state that genocide is an evil action but may state that genocide has been described by John So-and-so as the epitome of human evil. It just so happens that AI violates WP:NPOV in this specific way all the time. Gnomingstuff (talk) 17:16, 3 December 2025 (UTC)

Some finesse is necessary. "As normally understood" refers to generative AI — where a single prompt can produce an entire research report. DeepL.Write is a service that improves grammar and framing under human agency. Very different use cases. I could not have included (quoting from the AI template) "hallucinated information, copyright violations, claims not verified in cited sources, original research, or fictitious references". That is simply not possible under my potential usage   and I stress "potential" usage. Therefore the AI tag should be removed because it cannot apply. RobbieIanMorrison (talk) 17:22, 3 December 2025 (UTC)

As someone who does copyedit with AIs, hallucinations do happen in this use case still, even though they tend to be more subtle. AIs occasionally change the meaning of sentences when they copyedit. —Femke 🐦 (talk) 17:24, 3 December 2025 (UTC)

Sure, DeepL.Write can change the meaning. But I check what I use of course. And that is exactly what I am claiming (given that I did use DeepL.Write for the cited example). Best RobbieIanMorrison (talk) 17:32, 3 December 2025 (UTC)

Policy Wikipedia:LLMCOMM "does not apply to using LLMs to refine the expression of one's authentic ideas". That is exactly the case here in the context of possibly a very few edits or perhaps no edits. I am going to remove the AI tag in 24 hours time, unless I hear further arguments to the contrary. Best RobbieIanMorrison (talk) 18:25, 3 December 2025 (UTC)

Three days or so have passed and I am just about to remove that AI tag. Stepping back, the process I experienced as a longtime Wikipedia editor, as documented above, has left a bad taste in my mouth. Wikipedia needs to improve these processes for identifying and rectifying the inappropriate use of generative AI, which I don't doubt is a problem. The apparent current "shoot first and ask questions later" AI policy seems at odds with the normal ethos of Wikipedia. In any case, I found the preceding discussion quite difficult. Best, RobbieIanMorrison (talk) 11:15, 7 December 2025 (UTC)

 Done The certain=yes qualifier on the AI-generated template was especially unwarranted.

Again, as a longstanding editor, I would really like to be believed when I make considered statements. Best, RobbieIanMorrison (talk) 11:23, 7 December 2025 (UTC)

I do believe you, given that you have danced right up to the edge of saying you used DeepL Write -- oh, excuse me, I'm sorry, "may have used it," "my potential usage," etc. Gnomingstuff (talk) 19:57, 25 April 2026 (UTC)

@Gnomingstuff. Thanks. I am going to initiate a Wikipedia disputes process in one week from now unless other relevant information surfaces. Your assessments are not remotely legitimate or acceptable in my view. Kind regards, RobbieIanMorrison (talk) 10:26, 26 April 2026 (UTC)

Why wait a week, RobbieIanMorrison? The tag seems pretty legitimate to me, there's plenty more of WP:SUPERFICIAL in the article than the two examples that have been pointed out here, and I have hard time reading this conversation as anything other than you having used DeepL to edit the article. --Gurkubondinn (talk) 12:40, 26 April 2026 (UTC)

@Femke Can you offer some concrete suggestion? As far as I can tell the tag is in place because @RobbieIanMorrison said might occasionally run a sentence or two through DeepL Write more recently to see if the text could be better framed. However, in my opinion such a use matches the exception listed Wikipedia:Writing_articles_with_large_language_models "Editors are permitted to use LLMs to suggest basic copyedits to their own writing, and to incorporate some of them after human review, provided the LLM does not introduce content of its own." Two editors have made extensive reviews of the content on this page, so the encyclopedia has suffered no harm from this use of AI. Nevertheless there seems to be no way to remove the tag. Johnjbarton (talk) 18:10, 23 June 2026 (UTC)

Hi @Johnjbarton. I see you've been tagging some sentences as failed verification. Were these new additions. AI hallucinations are a common cause of failed verification, but this in general failed verification is quite a common thing. What was your sense. We have the original author, me and you leaning against placing a tag, and Gurkubondinn and Gnomingstuff in favour. That means we'll need to get good arguments to move away from the status quo. In solidarity, —Femke (talk) 🐦 20:34, 23 June 2026 (UTC)

Yes, I marked failed verification for some content from this edit, https://en.wikipedia.org/w/index.php?title=Open_energy_system_models&diff=prev&oldid=1252016268. As I discussed below it also had some issues with non-encyclopedic content. However, no one has confessed to using AI to generate that addition. Since the content was sourced, I have no basis for reverting it. I could fix the content, but I don't see the point. As far as I can tell, once AI has been declared, no fixes will undo it. Hence my query. Johnjbarton (talk) 21:07, 23 June 2026 (UTC)

I'm now only seeing how this discussion continued. Disregard the above. I'll do a bit of a counting exercise and propose a way forward later this week, which might be asking for input from the AI cleanup folks, who should be able to give better arguments on whether anything untoward happened here. I feel people are getting so tired of AI slop that it might be difficult to examine the type of text that AI models take their cues from (academic text). In solidarity, —Femke (talk) 🐦 20:42, 23 June 2026 (UTC)

Looking through the diff I found the following significant changes to text:

1. The content of the section on sci2grid: https://en.wikipedia.org/w/index.php?diff=1337953115 Temp user. Looks like AI to my untrained eye.
2. The content of the section on Remix: https://en.wikipedia.org/w/index.php?title=Open_energy_system_models&diff=prev&oldid=1252016268 One time editor. I tagged this content as inaccurate when I verified the sources.
3. Section on RAMP: https://en.wikipedia.org/w/index.php?title=Open_energy_system_models&diff=prev&oldid=1230074946

Of these only #3 was authored by @RobbieIanMorrison.

I propose to revert the first two and delete the last paragraph of the RAMP section to complete the AI cleanup. Johnjbarton (talk) 23:26, 28 May 2026 (UTC)

Regarding #2. I know Berit Müller (talk · contribs) personally. I've not got the inclination to look into the Remix material myself. But this will not be a malicious edit. RobbieIanMorrison (talk) 23:15, 29 May 2026 (UTC)

No one has claimed the edit was "malicious". I added it to my list to be comprehensive in the additions during the time frame of possible AI additions. Was this content prepared by AI? I can't tell. I've seen all kinds of nonsense entered by humans, but evidently the style of AI generated content sets some people off.

Independent of any AI issues, the content struck me not completely inline with the one source in the middle section and "advertising". A couple of examples:

• "REMix offers several key features that make it a robust tool for energy system modeling."
  • "Robust" is marketing speak, not a technical evaluation.
• "A notable feature is its custom accounting capability, provided through the indicator module, which enables flexible definitions of what contributes to the objective functions."
  • Encyclopedias provide facts, this is an ad.

I would attribute this to enthusiasm and lack of experience, not malice. Johnjbarton (talk) 01:48, 30 May 2026 (UTC)

I will email Berit Müller (presuming I have the right person) this morning and ask her for some background. I have a couple of thoughts on what might have happened but am not going to speculate here. Best, RobbieIanMorrison (talk) 09:42, 30 May 2026 (UTC)

Berit tells me she wrote the article together with people at the German Aerospace Center. She will try and fix it next week. Best, RobbieIanMorrison (talk) 15:57, 31 May 2026 (UTC)

I rewrote the REMix section. The sources given may have illustrated the characteristics of REMix claimed in the source, if you understood REMix. But generally these sources said little about the characteristics of REMix: they are applications papers. In my opinion, most readers are better served by hints about the possible applications than abstract properties of the framework. Thus my rewrite focused on the applications as reported in the sources.

This completes the items of clean-up I identified earlier. Johnjbarton (talk) 21:28, 30 June 2026 (UTC)

Many thanks. I've just let the folk at the German Aerospace Center know. They could well make further edits and improvements. But I myself will not be contributing as a regular editor to this section. Best, RobbieIanMorrison (talk) 07:36, 1 July 2026 (UTC)

Folks: can we WP::CUTITOUT? Editors use all sorts of writing aids: dictionaries, thesauruses, Strunk & White, Fowler, and more. People composing offline have grammar checkers, spelling chekcers, Apple writing tools, Grammarly, Deepl translators. The current policy is intended to avoid using AIs to compose entire articles or extended passages, especially in the characteristic style of GPT circa 2025. It is not a purity test certifying that every word is untainted and unprocessed. If there is superficial material here, edit that. MarkBernstein (talk) 14:10, 11 May 2026 (UTC)

As the predominant editor of this article, I want to underscore two claims:

My use of DeepL.Write was extremely minor, if at all: I have never used an LLM like ChatGPT or Claude. I have a paid account with DeepL and use Write to improve my writing. I usually only add a couple of sentences to the input dialog and normally do not paste back into Wikipedia but instead take note of the suggested changes and enter them by hand. In all honestly, I may have never used DeepL.Write while editing the article in question. That type of usage easily falls within Wikipedia policy guidelines on AI editing assistance by my reading.

The AI-generated warning template contains claims that simply cannot be correct. There is zero chance that my edits contain "hallucinated information, copyright violations, claims not verified in cited sources, original research, or fictitious references". That statement is therefore highly misleading to Wikipedia readers. Best, RobbieIanMorrison (talk) 14:48, 11 May 2026 (UTC)

WP:LLM says that "Editors are permitted to use LLMs to suggest basic copyedits to their own writing, and to incorporate some of them after human review, provided the LLM does not introduce content of its own." @Gnomingstuff: might we instead discuss and improve some section you consider requires improvement? MarkBernstein (talk) 16:07, 11 May 2026 (UTC)

Please don't ping me. Gnomingstuff (talk) 18:23, 11 May 2026 (UTC)

Why don't you turn off notifications for mentions from your settings instead of repeatedly making this request? Is there some problem with that solution? Jähmefyysikko (talk) 13:02, 30 May 2026 (UTC)

Arbitrary Break

I have inspected every available edit by @RobbieIanMorrison from December 2022 (when Deepl Write was introduced) to the present, and have spot-checked a number of online references. RobbieIanMorrison’s claim that they don’t remember whether they used Deepl.Write or not is, in my opinion, extremely plausible; most edits introduce summary information about research computing systems culled from papers and conference presentations, and most are extremely dull though doubtless useful to (checks notes) energy systems model enthusiasts. I might question whether all this is WP:DUE, but I see no reason to be concerned about AI usage here and no indication that the systems in question are not properly represented. I observe, however, that the first sentence of the article is not all that we could wish; might we have some attention there, and to other gnomish stuff? MarkBernstein (talk) 16:32, 11 May 2026 (UTC)

There are examples above given of WP:AISIGNS in the text. As the tag says, please do not remove it while the AI-generated text remains. Feel free to clean it up though. --Gurkubondinn 16:36, 11 May 2026 (UTC)

Please pick a specific sign in a specific passage. We can discuss that. I’ve reread the section above, and I’m unsure what examples are given above. MarkBernstein (talk) 17:18, 11 May 2026 (UTC)

Just read the section above. The example there remains in the text. It is just an example and a sympton of the problem, not the actual problem (so just removing that sentence does not fix the problem). The point is that there is AI-generated text in the article. The meaning of the tag is "there is AI-generated text in this article", so the tag is correctly applied and should remain. Recognising this stuff just comes down to practice (I have more than I ever wanted to have), so if you want to learn more about what AI-generated text looks like, and what problems (for Wikipedia) that causes, then WP:AISIGNS is a pretty good breakdown. It has become pretty dense at this point, but it is a very good resource of our current understanding. --Gurkubondinn 17:49, 11 May 2026 (UTC)

WP:AISIGNS covers AI chatbots. I have never knowingly used a chatbot in my life. If people would like an affidavit on this, I am happy to visit an intellectual property lawyer in Berlin and then prepare and sign one to this effect. RobbieIanMorrison (talk) 18:20, 11 May 2026 (UTC)

My paid account with DeepL dates from 21 July 2023. RobbieIanMorrison (talk) 18:24, 11 May 2026 (UTC)

I believe you that you aren't using AI chatbots and that your DeepL account was from 2023. (The timeline of the edits and when they started displaying LLM indicators matches up exactly with when DeepL Pro started being LLM-based.) But WP:NOLLM covers AI-generated or AI-assisted text as a whole.

AISIGNS is an essay, not a guideline, and if it gives the impression that it's only about chatbots then that's a problem with the essay that should probably be fixed. Gnomingstuff (talk) 17:48, 23 June 2026 (UTC)

@Gurkubondinn: I do research in an AI-adjacent area, and I have some experience as an editor. I have read the discussion (twice). I have read the article (alas). It can be difficult not to be superficial when trying to write a one or two sentence overview of a large technical system, even if you are perfectly human. Repeating myself once again: please cite a specific passage which, in your opinion, contains WP:AISIGNS, so we can improve it. Iterate until happiness breaks out! Profit! @RobbieIanMorrison: Let’s try to lower the temperature, OK, just as an experiment. I know it's frustrating. Thanks! MarkBernstein (talk) 19:08, 11 May 2026 (UTC)

The only person who knows exactly what is AI-generated and what isn't, is the editor that inserted the text. I'm not sure what temperature problems you may or may not have, but I have none. I've told you to simply read the discussion, because this is much more complex than just pointing at a sentence or two, because those individual sentences are not the problem. If you fix those sentences (for example the one that has been given as an example here), then how have improved something, but you have not fixed the problem itself (there would still be AI-generated text in the article). You are still free to improve it and start the cleanup. There is an example directly above here, but it is just one example. It can be fixed (and probably should), but the problem would still remain overall. I don't think that it is productive to talk too much about this one example, because that gives the impression that this sentence (or any other specific sentence) is the problem, when that is not the case. So that's why I hope that you can understand why "just point me to a passage" is not a very helpful question, nor would it be very helpful if I pointed you at some passage. As you say, this is a large and complicated problem, so it is not possible to just distill it down to "give me a paragraph that I can edit so I can remove the tag", because that tag isn't there because of a discrete number of sentences or paragraphs, it is there because someone used an AI to generate text that they pasted in. They have been less than forthcoming about this, which is very unfortunate, because this could have been cleared up right when it was detected if they had just told us what they let an AI write and which AI they used. Instead they have been saying that they "may have" and "if I did, then I might have" and etc (which you can read all here above). So here we are, and the article will have to remain tagged until this has been figured out one way or another. The goal here isn't "removing the tag", the goal is "cleaning up AI generated text" and "disclose to the readers that there is a good possibility of the article they are reading contains AI-generated text". So while that is the case, the tag is serving a purpose, and will remain. --Gurkubondinn 19:35, 11 May 2026 (UTC)

You’re a skilled and experienced Wikipedian. Take a step back. AGF. I’m asking you what edits you would make to remove AI generated text. Is your suggestion to remove all edits by @RobbieIanMorrison? To roll back the article to its state in December 2022? To replace all em-dashes? What do you want to do? (Are you confident that RobbieIanMorrison is the sole source of text you consider AI-generated?) MarkBernstein (talk) 19:53, 11 May 2026 (UTC)

I have made no such suggestion. There very well may have been other editors adding AI-generated text to this article. Believe it or not, it happens constantly and at a much higher rate than we stand a chance to clean up. If you notice anyone, then please flag it (the WP:AINB noticeboard is where you can ask for opinions or help with cleanup). Ideally, I would have liked seeing the editor drop the "may have"/"if I did"/"I might possibly have"/"if I may have then maybe I did" qualifiers and simply come clean, but seeing as they have attempted to open an ArbCom case, I am not expecting much cooperation. Maybe the best case scenario is someone else deciding to rewrite those parts of the article? Note that there is no harm in an article being tagged long term, it does not remove any content and it does not nominate the article for deletion of any kind. It is just reader disclosure and a maintenance tag. It serves the purpose of informing readers and editors of the well-founded belief that this article most likely contains AI-generated text. The best way to avoid this tag appearing on an article is to not create this problem, by not using any AI-tools (such as DeepL, Grammarly or ChatGPT) for Wikipedia. --Gurkubondinn 20:06, 11 May 2026 (UTC)

To note that most people don't regard DeepL.Write and similar writing aids as LLMs. I certainly didn't. So please allow some latitude while the terminologies, concepts, and issues under discussion become clearer and perhaps converge. Regarding my "might have" remarks, I have no recollection of how I might have formulated individual edits over the last four years — that's not obfuscation, it's simply how human memory works. I think WP:LLM should be our point of reference — and bear in mind that there is no Wikipedia policy on absolute prohibition, as much as some might wish. Best, RobbieIanMorrison (talk) 21:23, 11 May 2026 (UTC)

For the record, DeepL Write has been LLM-based since mid-2024, according to DeepL. Gnomingstuff (talk) 17:51, 23 June 2026 (UTC)

Just some suggestions if your goal is to remove the tag. The current state of the art in AI cleanup is manual edits with no help from even the tools use to detect AI. @MarkBernstein, based on your assertion (which makes sense) about the first possible use of DeepL.Write, the changes to the article since that time are in this diff. The vast majority of the changes are minor ones unrelated to AI. @RobbieIanMorrison perhaps you can check each text addition in the diff to help recall which ones need to be edited to evade the AI detector. (Of course the AI detection report could have noted the suspect text but that is not yet a standard requirement). If each added source is also verified, that would cover the issues flagged by the tag.

(I encourage you to focus on that task in this Talk page. This isn't the place or time to fight a meta-battle against the editor/s who set the tag.) Johnjbarton (talk) 18:33, 28 May 2026 (UTC)

Keep in mind that the editor that has added AI-generated text has been less than forthcoming, and it is entirely possible that they have used tools other than DeepL.Write as well. We normally use the public release of ChatGPT in late 2022 as a cutoff for AI edits, so the set of diffs that you have picked makes sense. A bunch of those edits are touching references, which are very time consuming to go through and validate. This needs to be checked for both source integrity issues (by reading all of the sources and checking that they support what is being sourced to them) and also plainly check if they are real and make sense. --gurkubondinn 18:37, 28 May 2026 (UTC)

Keep in mind that the effort against Artificial intelligence in Wikimedia projects did not even start until August 2025 and was not official until March 2026. No edits with AI before these times are against AI guidelines. Based on my reading of the Talk page and the content of this article, these edits were made in WP:Good faith and in a sincere effort to improve the article. (I encourage you to focus on improving this article in this Talk page. This isn't the place to impune editor actions). Johnjbarton (talk) 19:30, 28 May 2026 (UTC)

Yes they are, there are no "retroactive exemptions" to guidelines. Editors won't be sanctioned for it, but the edits should be removed/articles deleted/etc as appropriate. LLM-generated edits are not "grandfathered in" just because they may have been saved before we passed the guidelines. Since I mentioned references, it should go without saying that there are no exceptions to WP:V just because something was generated by an LLM before some arbitrary date. --gurkubondinn 20:01, 28 May 2026 (UTC)

@Gurkubondinn: Why do you keep saying that I am not forthcoming? I have been entirely candid since my first reply in this thread. And ever since and elsewhere too. You say it is entirely possible that I have used tools other than DeepL.Write. I have stated multiple times that I did not. I reiterate here that I have not used other AI tools. Nor do I have, or have ever had, accounts for any other such tools. Are you accusing me of lying about my usage? Do you have any evidence to support your claim? Real evidence, not speculation. Surely I would deserve to know. A ping to @Johnjbarton for their efforts. RobbieIanMorrison (talk) 13:07, 29 May 2026 (UTC)

I would like to make some observations regarding the timing of my alleged use of LLM tooling, whether supported by Wikipedia exceptions at some point or not. The RAMP section I added here spanned edits from 20–24 June 2024. My DeepL.Write support subscription began on 9 August 2024 (the Write service itself launched on 25 April 2024 while my DeepL Pro account dates back to 21 July 2023). I did not use unsubscribed services for copyediting support due to the retention of text (despite the public beta testing for the Write service having been available from 17 January 2023). There is no overlap between my late‑2022 edits and my potential use of DeepL.Write. I therefore conclude that none of my edits to this article could have involved the use of LLM tooling under any possible scenario. This result is fully consistent with my earlier statements on this Talk page and elsewhere on Wikipedia. If you respond, please note that suspicion is not evidence. RobbieIanMorrison (talk) 13:04, 30 June 2026 (UTC)

I have yet to see convincing evidence that the edits in question were generated by an LLM. I am concerned that we’re approaching WP:ASPERSIONS and WP:NPA territory. No, the prose about (checks notes) competing power system products is not always lively; that is only to be expected, given the subject matter, and the natural remedy would be to polish the text. MarkBernstein (talk) 13:22, 30 June 2026 (UTC)

Proposed addition to Further reading in PyPSA section

I am one of the authors of a recently published Elsevier book on PyPSA and would like to propose a neutral addition under literature/further reading. Suggested wording below:

A comprehensive practical reference on modeling power systems and renewable energy systems with PyPSA is provided in The PyPSA Handbook (Elsevier, 2025).

Ref: https://www.sciencedirect.com/book/monograph/9780443266317/the-pypsa-handbook ~2025-43669-23 (talk) 06:38, 31 December 2025 (UTC)

Sources added since Dec. 2022

Based on this diff, the following sources have been added and should be verified. Please feel free to edit my post to add indented comments on these sources.

• Fairley, Peter (27 June 2023). "EU: Climate-proof grids require more transparency > 'Black-box' US energy planning hinders renewable energy development". IEEE Spectrum. New York, USA. ISSN 0018-9235. Retrieved 2023-06-27. "Europe’s network planners have already begun using open tools." "In North America, experts cite a few moves toward open tools,..."  Done
• Schimeczek, Christoph; Nienhaus, Kristina; Frey, Ulrich; Sperber, Evelyn; Sarfarazi, Seyedfarzad; Nitsch, Felix; Kochems, Johannes; Ghazi, A. Achraf El (2023-04-17). "AMIRIS: Agent-based Market model for the Investigation of Renewable and Integrated energy Systems". Journal of Open Source Software. 8 (84): 5041. Bibcode:2023JOSS....8.5041S. doi:10.21105/joss.05041. ISSN 2475-9066. Verifies after moving to the first sentence of the paragraph.  Done
• Schimeczek, Christoph; Nitsch, Felix; Kochems, Johannes; Nienhaus, Kristina (2026-02-28). "Avoiding avalanches: Effective dispatch planning for competing storage units in day-ahead electricity market simulations". Journal of Energy Storage. 148 120054. doi:10.1016/j.est.2025.120054. ISSN 2352-152X.  Done
• Nitsch, Felix; Schimeczek, Christoph; Frey, Ulrich; Fuchs, Benjamin (2023-04-17). "FAME-Io: Configuration tools for complex agent-based simulations". Journal of Open Source Software. 8 (84): 4958. Bibcode:2023JOSS....8.4958N. doi:10.21105/joss.04958. ISSN 2475-9066.  Done
• Schimeczek, Christoph; Deissenroth-Uhrig, Marc; Frey, Ulrich; Fuchs, Benjamin; Ghazi, A. Achraf El; Wetzel, Manuel; Nienhaus, Kristina (17 April 2023). "FAME-Core: An open Framework for distributed Agent-based Modelling of Energy systems". Journal of Open Source Software. 8 (84): 5087. Bibcode:2023JOSS....8.5087S. doi:10.21105/joss.05087. S2CID 258195264.  Done
• https://www.science.org/doi/10.1126/science.1215539 A trivial correction, I deleted this it is not encyclopedic.  Done
• Löffler, Konstantin; Hainsch, Karlo; Burandt, Thorsten; Kemfert, Claudia; von Hirschhausen, Christian (2017). "Designing a model for the global energy system — GENeSYS-MOD: an application of the Open-Source Energy Modeling System (OSeMOSYS)". Energies. 10 (1468): 1468. doi:10.3390/en10101468. hdl:10419/200750. ISSN 1996-1073. S2CID 6944356.  Done
• Burandt, Thorsten; Löffler, Konstantin; Hainsch, Karlo (2018). GENeSYS-MOD v2.0 — Enhancing the Global Energy System Model: model improvements, framework changes, and European data set (Report). DIW Berlin Data Documentation. Berlin, Germany: DIW Berlin (Deutsches Institut für Wirtschaftsforschung). ISSN 1861-1532. Retrieved 2024-10-24.  Done
• Löffler, Konstantin; Burandt, Thorsten; Hainsch, Karlo; Oei, Pao-Yu (16 November 2019). "Modeling the low-carbon transition of the European energy system — a quantitative assessment of the stranded assets problem". Energy Strategy Reviews. 26 100422. Bibcode:2019EneSR..2600422L. doi:10.1016/j.esr.2019.100422. hdl:10419/214633. ISSN 2211-467X. Retrieved 2024-10-25.  Done
• Burandt, Thorsten; Xiong, Bobby; Löffler, Konstantin; Oei, Pao-Yu (1 December 2019). "Decarbonizing China's energy system — Modeling the transformation of the electricity, transportation, heat, and industrial sectors". Applied Energy. 255 113820. Bibcode:2019ApEn..25513820B. doi:10.1016/j.apenergy.2019.113820. hdl:10419/214646. Retrieved 2024-10-22.  Done
• Hainsch, Karlo; Löffler, Konstantin; Burandt, Thorsten; Auer, Hans; Crespo del Granado, Pedro; Pisciella, Paolo; Zwickl-Bernhard, Sebastian (15 January 2022). "Energy transition scenarios: what policies, societal attitudes, and technology developments will realize the EU Green Deal?". Energy. 239 122067. Bibcode:2022Ene...23922067H. doi:10.1016/j.energy.2021.122067. Retrieved 2024-10-22.  Done
• Schwartz, Aaron; Dyson, Mark; Shwisberg, Lauren (December 2022). Power planning to the people: how stakeholder-driven modeling can help build a better grid. Basalt, Colorado, United States: RMI. Retrieved 2023-06-30.  Done
• Schwartz, Aaron (12 December 2022). "Free software, more voices, better plans". RMI. Basalt, Colorado, USA. Retrieved 2023-06-29.  Done
• Kubiczek, Patryk (19 January 2023). "Modelling the Polish energy system — new findings of PyPSA-PL". Instrat. Warsaw, Poland. Retrieved 2023-01-23.  Done
• Lyden, Andrew; Sun, Wei; Struthers, Iain; Franken, Lukas; Hudson, Seb; Wang, Yifan; Friedrich, Daniel (1 May 2024). "PyPSA-GB: an open-source model of Great Britain's power system for simulating future energy scenarios". Energy Strategy Reviews. 53 101375. Bibcode:2024EneSR..5301375L. doi:10.1016/j.esr.2024.101375. ISSN 2211-467X. Retrieved 2024-08-23.  Done
• Parzen, Maximilian (27 June 2023). "Launch of "Open Energy Transition" and article "'Black-box' US energy planning hinders renewable energy development"". Google Groups. Retrieved 2023-07-01. This verifies but it is a WP:SELFPUB.  Done
• Abdel-Khalek, Hazem; Schumm, Leon; Jalbout, Eddy; Parzen, Maximilian; Schauß, Caspar; Fioriti, Davide (2025). "PyPSA-Earth sector-coupled: A global open-source multi-energy system model showcased for hydrogen applications in countries of the Global South". Applied Energy. 383 125316. doi:10.1016/j.apenergy.2025.125316. ISSN 0306-2619.  Done
• Wetzel, Manuel; Gils, Hans Christian; Bertsch, Valentin (2023). "Green energy carriers and energy sovereignty in a climate neutral European energy system". Renewable Energy. 210: 591–603. Bibcode:2023REne..210..591W. doi:10.1016/j.renene.2023.04.015. Retrieved 2024-09-13. Replaced with a better ref  Done
• Wetzel, Manuel; Salvador Arellano Ruiz, Eugenio; Witte, Francesco; Schmugge, Jens; Sasanpour, Shima; Yeligeti, Madhura; Miorelli, Fabia; Buschmann, Jan; Cao and, Karl-Kiên; Wulff, Niklas; Gardian, Hedda; Rubbert, Alexander; Fuchs, Benjamin; Scholz, Yvonne; Gils, Hans Christian (2024). "REMix: A GAMS-based framework for optimizing energy system models". Journal of Open Source Software. 9 (99): 6330. Bibcode:2024JOSS....9.6330W. doi:10.21105/joss.06330. I marked this one with {{failed verification}} in the article because the claims are beyond what the article supports.  Done
• Gils, Hans Christian; Scholz, Yvonne; Pregger, Thomas; de Tena, Diego Luca; Heide, Dominik (2017). "Integrated modelling of variable renewable energy-based power supply in Europe". Energy. 123: 173–188. Bibcode:2017Ene...123..173G. doi:10.1016/j.energy.2017.01.115. Retrieved 2024-09-13.  Done
• Gils, Hans Christian; Gardian, Hedda; Schmugge, Jens (2021). "Interaction of hydrogen infrastructures with other sector coupling options towards a zero-emission energy system in German". Renewable Energy. 180: 140–156. Bibcode:2021REne..180..140G. doi:10.1016/j.renene.2021.08.016. Retrieved 2024-09-13.  Done
• Sasanpour, Shima; Gils, Hans Christian; Cao and, Karl-Kiên; Jochem, Patrick (2021). "Strategic policy targets and the contribution of hydrogen in a 100% renewable European power system". Energy Reports. 7: 4595–4608. Bibcode:2021EnRep...7.4595S. doi:10.1016/j.egyr.2021.07.005. Retrieved 2024-09-13.  Done
• Gils, Hans Christian; Gardian, Hedda; Kittel, Martin; Schill, Wolf-Peter; Zerrahn, Alexander; Murmann, Alexander; Launer, Jann; Fehler, Alexander; Gaumnitz, Felix; van Ouwerkerk, Jonas; Bußar, Christian; Mikurda, Jennifer; Torralba-Díaz, Laura; Janßen, Tomke; Krüger, Christine (2022). "Modeling flexibility in energy systems — comparison of power sector models based on simplified test cases". Renewable and Sustainable Energy Reviews. 158 111995. Bibcode:2022RSERv.15811995G. doi:10.1016/j.rser.2021.111995. Retrieved 2024-09-13. Marked failed verification in article  Done
• Lombardi, Francesco; Balderrama, Sergio; Quoilin, Sylvain; Colombo, Emanuela (15 June 2019). "Generating high-resolution multi-energy load profiles for remote areas with an open-source stochastic model". Energy. 177: 433–444. Bibcode:2019Ene...177..433L. doi:10.1016/j.energy.2019.04.097. hdl:11311/1087101. Retrieved 2024-06-19.  Done
• Mangipinto, Andrea; Lombardi, Francesco; Sanvito, Francesco Davide; Pavičević, Matija; Quoilin, Sylvain; Colombo, Emanuela (15 April 2022). "Impact of mass-scale deployment of electric vehicles and benefits of smart charging across all European countries". Applied Energy. 312 118676. Bibcode:2022ApEn..31218676M. doi:10.1016/j.apenergy.2022.118676. hdl:11311/1218999. Retrieved 2024-06-19.  Done
• Lombardi, Francesco; Duc, Pierre-François; Tahavori, Mohammad Amin; Sanchez-Solis, Claudia; Eckhoff, Sarah; Hart, Maria CG; Sanvito, Francesco; Ireland, Gregory; Balderrama, Sergio; Kraft, Johann; Dhungel, Gokarna; Quoilin, Sylvain (12 June 2024). "RAMP: stochastic simulation of user-driven energy demand time series". Journal of Open Source Software. 9 (98): 6418. Bibcode:2024JOSS....9.6418L. doi:10.21105/joss.06418. ISSN 2475-9066.  Done
• "RAMP: NESSI web interface". NESSI:RAMP. Retrieved 2024-06-19. Marked failed verification, its a login page of some kind.  Done

Johnjbarton (talk) 20:21, 28 May 2026 (UTC)

I check every source added in the diff above. Despite claims, there is no evidence of AI involved in the sources. A couple of sources were poorly summarized and I marked these in the article as {{failed verification}} for someone else to follow up. Johnjbarton (talk) 22:01, 28 May 2026 (UTC)