Re: [openmod-initiative] HELM for real
Robbie Morrison
Hello Santiago
That all sounds really interesting. I think it would be useful to start a openmod forum topic on this theme. I can help if you like.
Also to note there is a good article on wikipedia on HELM:
The wikipedia article also quotes US patents 7519506 and 7979239. I presume these patents do not cover aspects of the core method. Do you know anything about their scope and consequences.
Two of the early papers on HELM listed below.
I recall PyPSA was also looking at the method. Any news there from KIT?
with best wishes, Robbie
References
Trias, Antonio (July 2012). The holomorphic embedding load flow method. 2012 IEEE Power and Energy Society General Meeting. New Jersey, USA: IEEE. doi:10.1109/PESGM.2012.6344759.
Trias, Antonio
(2015). “Fundamentals
of the holomorphic embedding load-flow method”. arXiv.
Abstract.
On 09/03/2020 19.58, Santiago Peñate wrote:
Dear all,--
It has been a while since I started reading and attempting to crack a working version of the Holomorphic Embeding Load Flow (HELM), being fascinated by it's almost magical properties.
In this search I found plenty of half-way papers, or papers with straight wrong formulations claiming otherwise. Last month, out of the blue a guy from Girona (Josep Fanals-Batllori) wrote me with his formulation and he wanted me to check it out. It worked! This was the first PQ+PV nodes formulation that I had seen working, providing almost perfectly convergent series. But it was not competitive in terms of speed or grid cases coverage (transformer taps, sparse matrix formulation, etc.)
After quite some work by Josep and myself, Today GridCal features a fully fledged holomorphic embedding power flow algorithm, with the "magical" convergence properties, that beat Newton-Raphson, and also beat the fast-decoupled algorithm in large grids. And that is a lot to say. However now, and unlike others before us, we provide hard evidence:
Enjoy.
You received this message because you are subscribed to the Google Groups "openmod initiative" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openmod-initiat...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/openmod-initiative/4c52c178-82be-41a8-be0f-fbfe6f15dc6f%40googlegroups.com.
-- Robbie Morrison Address: Schillerstrasse 85, 10627 Berlin, Germany Phone: +49.30.612-87617
Robbie Morrison
RESEND: for some reason the original send was not delivered after 00:50:00. It could be because of some weird signature fields in the original posting?
Hello Santiago
That all sounds really interesting. I think it would be useful to start a openmod forum topic on this theme. I can help if you like.
Also to note there is a good article on wikipedia on HELM:
The wikipedia article also quotes US patents 7519506 and 7979239. I presume these patents do not cover aspects of the core method. Do you know anything about their scope and consequences.
Two of the early papers on HELM listed below.
I recall PyPSA was also looking at the method. Any news there from KIT?
with best wishes, Robbie
References
Trias, Antonio (July 2012). The holomorphic embedding load flow method. 2012 IEEE Power and Energy Society General Meeting. New Jersey, USA: IEEE. doi:10.1109/PESGM.2012.6344759.
Trias, Antonio (2015). “Fundamentals of the holomorphic embedding load-flow method”. arXiv. Abstract.
-- Robbie Morrison Address: Schillerstrasse 85, 10627 Berlin, Germany Phone: +49.30.612-87617
Tom Brown
Yes, we wanted to implement it in PyPSA but were put off by exactly the
issue Josep and Santiago have solved, i.e. getting it to work with a
large number of PV buses, which break holomorphicity. (Finally, a use
for my math degree...)
Lay explanation: The method is for solving the power flow problem. The
standard Newton-Raphson method can run into numerical problems where it
fails to converge to a solution, despite a physical solution existing.
HELM is guaranteed not to have these types of convergence problems
thanks to the beautiful theory of complex functions. This is important
in a real-time control setting, where you don't want convergence problems.
Patent: I'm not a lawyer. My understanding was that the patents skirt
the issue that math should not be patentable in the US by patenting the
combination of algorithm and SCADA system. Either way, they're not valid
in Europe as I understand it. Please someone correct me if I'm wrong.
Performance: @Santiago: How well does the algorithm perform on networks
with 100 or 1000 buses?
Best wishes,
Tom
On 10/03/2020 07:37, Robbie Morrison wrote:
> Hello Santiago
>
> That all sounds really interesting. I think it would be useful to start
> a openmod forum topic on this theme. I can help if you like.
>
> Also to note there is a good article on wikipedia on HELM:
>
> The wikipedia article also quotes US patents 7519506 and 7979239. I
> presume these patents do not cover aspects of the core method. Do you
> know anything about their scope and consequences.
>
> Two of the early papers on HELM listed below.
>
> I recall PyPSA was also looking at the method. Any news there from KIT?
>
> with best wishes, Robbie
>
>
> Trias, Antonio (July 2012). /The holomorphic embedding load flow
> method/. 2012 IEEE Power and Energy Society General Meeting. New Jersey,
> USA: IEEE. doi:10.1109/PESGM.2012.6344759
> <https://dx.doi.org/10.1109/PESGM.2012.6344759>.
> Trias, Antonio (2015). “Fundamentals of the holomorphic embedding
> <https://arxiv.org/abs/1509.02421>.
>> https://groups.google.com/d/msgid/openmod-initiative/4c52c178-82be-41a8-be0f-fbfe6f15dc6f%40googlegroups.com
>
>
> --
> Robbie Morrison
> Address: Schillerstrasse 85, 10627 Berlin, Germany
> Phone: +49.30.612-87617
>
> You received this message because you are subscribed to the Google
> Groups "openmod initiative" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to openmod-initiat...@googlegroups.com
> <https://groups.google.com/d/msgid/openmod-initiative/e936d5de-8a31-a8bf-b59a-1316e1925c1f%40posteo.de?utm_medium=email&utm_source=footer>.
--
Karlsruhe Institute of Technology (KIT)
Institute for Automation and Applied Informatics (IAI)
Tom Brown (he/him)
Research Group Leader, Energy System Modelling
Phone: +49 721 608 25737
Fax: +49 721 608 22602
Group website: https://www.iai.kit.edu/english/ESM.php
Personal website: https://nworbmot.org/
Visitor Address:
Office 309
Campus North Building 445
Hermann-von-Helmholtz-Platz 1
76344 Eggenstein-Leopoldshafen
Santiago Peñate
| IEEE 118 Bus | ||||
| Method | Converged? | Error | Elapsed(s) | Iterations |
| SolverType.HELM | True | 1.51E-11 | 0.0469 | 19 |
| SolverType.HELM | True | 1.51E-11 | 0.0625193 | 19 |
| SolverType.HELM | True | 1.51E-11 | 0.0533915 | 19 |
| SolverType.HELM | True | 1.51E-11 | 0.0505242 | 19 |
| SolverType.HELM | True | 1.51E-11 | 0.0523407 | 19 |
| SolverType.NR | True | 3.93E-12 | 0.0156519 | 3 |
| SolverType.NR | True | 3.93E-12 | 0.031256 | 3 |
| SolverType.NR | True | 3.93E-12 | 0.0221415 | 3 |
| SolverType.NR | True | 3.93E-12 | 0.0271504 | 3 |
| SolverType.NR | True | 3.93E-12 | 0.0229998 | 3 |
| Pegase 2869 Bus | ||||
| Method | Converged? | Error | Elapsed(s) | Iterations |
| SolverType.HELM | True | 7.37E-11 | 2.48401 | 49 |
| SolverType.HELM | True | 7.37E-11 | 2.44854 | 49 |
| SolverType.HELM | True | 7.37E-11 | 2.63507 | 49 |
| SolverType.HELM | True | 7.37E-11 | 2.49048 | 49 |
| SolverType.HELM | True | 7.37E-11 | 2.4523 | 49 |
| SolverType.NR | True | 3.52E-08 | 6.91317 | 5 |
| SolverType.NR | True | 3.52E-08 | 6.88511 | 5 |
| SolverType.NR | True | 3.52E-08 | 6.74982 | 5 |
| SolverType.NR | True | 3.52E-08 | 6.8204 | 5 |
| SolverType.NR | True | 3.52E-08 | 6.84771 | 5 |
El martes, 10 de marzo de 2020, 7:37:22 (UTC+1), Robbie Morrison escribió:
Hello Santiago
That all sounds really interesting. I think it would be useful to start a openmod forum topic on this theme. I can help if you like.
Also to note there is a good article on wikipedia on HELM:
The wikipedia article also quotes US patents 7519506 and 7979239. I presume these patents do not cover aspects of the core method. Do you know anything about their scope and consequences.
Two of the early papers on HELM listed below.
I recall PyPSA was also looking at the method. Any news there from KIT?
with best wishes, Robbie
References
Trias, Antonio (July 2012). The holomorphic embedding load flow method. 2012 IEEE Power and Energy Society General Meeting. New Jersey, USA: IEEE. doi:10.1109/PESGM.2012.6344759.
Trias, Antonio (2015). “Fundamentals of the holomorphic embedding load-flow method”. arXiv. Abstract.
On 09/03/2020 19.58, Santiago Peñate wrote:
Dear all,--
It has been a while since I started reading and attempting to crack a working version of the Holomorphic Embeding Load Flow (HELM), being fascinated by it's almost magical properties.
In this search I found plenty of half-way papers, or papers with straight wrong formulations claiming otherwise. Last month, out of the blue a guy from Girona (Josep Fanals-Batllori) wrote me with his formulation and he wanted me to check it out. It worked! This was the first PQ+PV nodes formulation that I had seen working, providing almost perfectly convergent series. But it was not competitive in terms of speed or grid cases coverage (transformer taps, sparse matrix formulation, etc.)
After quite some work by Josep and myself, Today GridCal features a fully fledged holomorphic embedding power flow algorithm, with the "magical" convergence properties, that beat Newton-Raphson, and also beat the fast-decoupled algorithm in large grids. And that is a lot to say. However now, and unlike others before us, we provide hard evidence:
Enjoy.
You received this message because you are subscribed to the Google Groups "openmod initiative" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openmod-initiative+unsub...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/openmod-initiative/4c52c178-82be-41a8-be0f-fbfe6f15dc6f%40googlegroups.com.
Thurner, Leon, Dr.-Ing.
Hi Santiago,
congratulations on getting HELM up an running, great work!
The performance is of course very difficult to measure, because you are always measuring a specific implementation of an algorithm. For reference, the pegase2869 network that you measured at ~2.5s with HELM and ~7s with NR takes just 75ms to simulate with pandapower:
Leon
Gesendet: Dienstag, 10. März 2020 13:57
An: openmod initiative
Betreff: Re: [openmod-initiative] HELM for real
To view this discussion on the web, visit https://groups.google.com/d/msgid/openmod-initiative/c98e9f1a-a496-4e8d-ac06-6101575e1aaf%40googlegroups.com.
Santiago Peñate
El martes, 10 de marzo de 2020, 7:37:22 (UTC+1), Robbie Morrison escribió:
Hello Santiago
That all sounds really interesting. I think it would be useful to start a openmod forum topic on this theme. I can help if you like.
Also to note there is a good article on wikipedia on HELM:
The wikipedia article also quotes US patents 7519506 and 7979239. I presume these patents do not cover aspects of the core method. Do you know anything about their scope and consequences.
Two of the early papers on HELM listed below.
I recall PyPSA was also looking at the method. Any news there from KIT?
with best wishes, Robbie
References
Trias, Antonio (July 2012). The holomorphic embedding load flow method. 2012 IEEE Power and Energy Society General Meeting. New Jersey, USA: IEEE. doi:10.1109/PESGM.2012.6344759.
Trias, Antonio (2015). “Fundamentals of the holomorphic embedding load-flow method”. arXiv. Abstract.
On 09/03/2020 19.58, Santiago Peñate wrote:
Dear all,--
It has been a while since I started reading and attempting to crack a working version of the Holomorphic Embeding Load Flow (HELM), being fascinated by it's almost magical properties.
In this search I found plenty of half-way papers, or papers with straight wrong formulations claiming otherwise. Last month, out of the blue a guy from Girona (Josep Fanals-Batllori) wrote me with his formulation and he wanted me to check it out. It worked! This was the first PQ+PV nodes formulation that I had seen working, providing almost perfectly convergent series. But it was not competitive in terms of speed or grid cases coverage (transformer taps, sparse matrix formulation, etc.)
After quite some work by Josep and myself, Today GridCal features a fully fledged holomorphic embedding power flow algorithm, with the "magical" convergence properties, that beat Newton-Raphson, and also beat the fast-decoupled algorithm in large grids. And that is a lot to say. However now, and unlike others before us, we provide hard evidence:
Enjoy.
You received this message because you are subscribed to the Google Groups "openmod initiative" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openmod-initiative+unsub...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/openmod-initiative/4c52c178-82be-41a8-be0f-fbfe6f15dc6f%40googlegroups.com.
Robbie Morrison
Hi Tom
If someone needs to dig into the applicability of United States software patents in Europe, I suggest Haapanen (2017). Haapanen reviews software patents in Europe (p72–79) and software patents in the US (p79–86). FOSS software licensing is covered too relative to legal theory.
Robbie
References
Haapanen, Anna (1 April 2017). Free and open source software licensing: and the mystery of licensor’s patents — PhD thesis. Helsinki, Finland: Faculty of Law, University of Helsinki.
On 10/03/2020 10.42, Tom Brown wrote:
Patent: I'm not a lawyer. My understanding was that the patents skirt the issue that math should not be patentable in the US by patenting the combination of algorithm and SCADA system. Either way, they're not valid in Europe as I understand it. Please someone correct me if I'm wrong.
-- Robbie Morrison Address: Schillerstrasse 85, 10627 Berlin, Germany Phone: +49.30.612-87617
Thurner, Leon, Dr.-Ing.
Hi Santiago, all,
I was really curious about the HELM implementation, so I gave it a try and integrated into pandapower. Here is the short version:
- It works and gives almost the same results as the pandapower standard NR solver (<1e-4pu voltage difference on the first try)! Thanks and kudos for this implementation!
- For the ill-conditioned case11_iwamoto system thats included in pandapower, neither NR nor HELM find a valid solution. I would be interested in finding out why that is...
For more discussion, I suggest we move to the openmod forum to not spam everyone: https://forum.openmod-initiative.org/t/the-new-helm-powerflow-in-gridcal/1952
Cheers
Leon
An: openmod initiative
Betreff: Re: [openmod-initiative] HELM for real
To view this discussion on the web, visit https://groups.google.com/d/msgid/openmod-initiative/952b4755-b39b-45a2-aac4-c91028a37fb7%40googlegroups.com.