Skip to content

minjiechen/magnetchallenge-2

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

462 Commits
.github/workflows
.github/workflows
 
 
Pretest
Pretest
 
 
Test
Test
 
 
docs
docs
 
 
img
img
 
 
tutorials
tutorials
 
 
.DS_Store
.DS_Store
 
 
README.md
README.md
 
 

Repository files navigation

MagNet Challenge 2 GitHub

Presentation Sign-Up

  • Feb 9: Princeton, Eindhoven, Hefei, Fuzhou, Southeast1, Tianjin
  • Feb 10: IIT, Hangzhou, UESTC, ASU, Southeast2, Cambridge, AAU1, NITC
  • Feb 11: Gatech, XJTU2, AAU2, SiegenPaderborn, Tsinghua, Zhejiang, NTU2, Bristol
  • Feb 12: Leuven, SAL, Sydney, Mizzou, NTU1, HUST, Tokyo
  • Unconfirmed: FJUT, PolyU, XJTU1

[News #14] Code evaluation starts NOW.

  • Please email us your inference code and make them as compact and as easy to use as possible.
  • 4-Page Report due Jan 30.
  • Please Sign Up for the upcoming four webinars on Feb 9~12.
  • Email us your preferred slot and prepare a 5-min presentation.

[News #13] Preliminary ranking of 32 teams. All submitted results available Here for cross-checking. Let us know if you find any error...

Team Avg. Seq. Error Ranking Avg. Energy Error Ranking Avg. Size Ranking Avg. 95% Seq. Error [%] Avg. 95% Energy Error [%] Avg. Parameters
Aalborg1 25 27 19 87.10 39.67 111901
Aalborg2 21 14 24 59.49 7.36 415753
ASU 20 21 14 54.08 20.92 18337
Bristol 7 7 26 25.22 8.40 860929
Cambridge 14 16 4 37.01 9.24 1182
Eindhoven 4 3 17 17.71 2.38 81893
FJUT 29 24 31 279.37 43.09 4447233
Fuzhou 2 2 16 12.42 1.91 63022
Gatech 17 15 1 38.76 7.97 83
Hangzhou 27 28 32 166.77 45.08 13185504
Hefei 1 1 10 9.17 1.57 6564
HUST 31 29 30 239.56 38.95 3063741
IIT 11 12 18 36.52 26.59 93876
Leuven 12 10 21 34.54 5.37 111985
Mizzou 18 20 22 50.66 21.57 153462
NITC 10 9 25 24.85 4.73 347237
NTU1 30 30 20 293.78 61.30 108848
NTU2 23 24 6 75.11 27.43 2399
PolyU 32 32 28 404.84 525.89 1478785
Princeton 23 19 5 66.85 15.89 1443
SAL 15 18 7 39.48 30.05 3713
SiegenPaderborn 6 6 3 20.43 3.41 325
Southeast1 3 5 15 14.51 2.69 22197
Southeast2 28 30 11 248.88 66.05 6913
Sydney 9 11 8 24.63 5.44 4777
Tianjin 13 13 13 32.56 6.99 16391
Tokyo 26 23 23 800.00 800.00 185729
Tsinghua 5 4 29 19.93 2.52 2177283
UESTC 18 22 9 51.95 21.33 5202
XJTU1 8 8 27 22.48 3.64 1350913
XJTU2 22 26 2 836.88 32.57 99
Zhejiang 15 17 12 44.60 9.39 16161

[News #12] We have received final submissions from 32 teams ...

  • Aalborg University 1 (Zhao), Denmark 🇩🇰 🌸 📝 💾
  • Aalborg University 2 (Davari), Denmark 🇩🇰 🌸 📝 💾
  • Arizona State University, USA 🇺🇸 🌸 📝 💾
  • University of Bristol, UK 🇬🇧 🌸 📝 💾
  • University of Cambridge, UK 🇬🇧 🌸 📝 💾
  • Eindhoven University of Technology, Netherland 🇳🇱 🌸 📝
  • Fujian University of Technology, China 🇨🇳 🌸 📝 💾
  • Fuzhou University, China 🇨🇳 🌸 📝 💾
  • Georgia Institute of Technology, USA 🇺🇸 🌸 📝 💾
  • Hangzhou Dianzi University, China 🇨🇳 🌸 📝 💾
  • Hefei University of Technology, China 🇨🇳 🌸 📝 💾
  • Huazhong University of Science and Technology, China 🇨🇳 🌸 📝
  • Indian Institute of Technology Dharwad, India 🇮🇳 🌸 📝 💾
  • KU Leuven, Belgium, 🇧🇪 🌸 📝 💾
  • University of Missouri Columbia, USA 🇺🇸 🌸 📝 💾
  • National Institute of Technology Calicut (Muhammed), India 🇮🇳 🌸 📝 💾
  • Nanyang Technological University 1 (Yang), Singapore 🇸🇬 🌸 📝 💾
  • Nanyang Technological University 2 (Tang), Singapore 🇸🇬 🌸 📝 💾
  • Princeton University, USA 🇺🇸 🌸 (not competing)
  • The Hong Kong Polytechnic University, Hong Kong SAR 🇭🇰 🌸 📝
  • Silicon Austria Labs, Austria 🇦🇹 🌸 📝 💾
  • Seigen & Paderborn, Germany 🇩🇪 🌸 📝 💾
  • Southeast University 1 (Cheng), China 🇨🇳 🌸 📝 💾
  • Southeast University 2 (Xu), China 🇨🇳 🌸 📝 💾
  • Tianjin University, China 🇨🇳 🌸 📝 💾
  • Tsinghua University, China 🇨🇳 🌸 📝 💾
  • Tokyo Metropolitan University, Japan 🇯🇵 🌸 📝 💾
  • University of Sydney, Australia 🇦🇺 🌸 📝 💾
  • University of Electronic Science and Technology of China, China 🇨🇳 🌸 📝 💾
  • Xi'an Jiaotong University 1 (Wei), China 🇨🇳 🌸 📝 💾
  • Xi'an Jiaotong University 2 (Zhu), China 🇨🇳 🌸 📝 💾
  • Zhejiang University, China 🇨🇳 🌸 📝 💾

[News #11] Award Ceremony in APEC 2026

We are planning an Award Ceremony of MagNet Challenge 2 in APEC 2026. Please send an email to pelsmagnet@gmail.com with the following information:

  • Name of your team? XXX
  • Are you on track to submit your final results to us before Jan 15? Yes or No
  • Will your team send one or more representatives to attend the Award Ceremony during APEC? Yes or No
  • How many people (team members, guests, colleagues) do you plan to bring to the award ceremony? Expected Total Number.

We look forward to seeing many of you in San Antonio in March!

[News #10] MagNet Challenge 2 - Final Data Released

Training and testing data released for 5 new materials [A, B, C, D, E]:

  • MagNet Challenge 2 Data under "Final Evaluation"
  • ** We updated the testing data for Material C on Nov. 21. Please use the updated version in Dropbox.**
  • Please note there is a "Template.zip" folder which serves as a template for final submission
  • Final Submission Due January 15th (CSV Results and Num. of Parameters)
  • 4-Page Report Due January 30th (Methods and Code Verification for Winning Teams). Please list the names of your team members in the right order for future acknowledgement
  • Pretest Results: Note the submitted pretest results are self-reported, are not verified and have no influence on the final evaluation, the model parameters are also unknown
  • Please Sign-Up to PELS TC10 as a Member or Student Member (for Free) to support MagNet Challenge

The MagNet community mourns the passing of Haoran Li, a wonderful person and a key contributor to the MagNet project. Haoran’s passion, creativity, and kindness will continue to influence our community. With the consent of his family and IEEE PELS, one of the MagNet Challenge 2 awards will be named in his honor.

[News #9] MagNet Tutorial 6

  • Tutorial 6 (October 30 10am EST): Video, Slides
  • Present final testing data and rules
  • Final testing data releasing date: November 15th, 2025
  • Optional pre-test evaluation due: November 15th, 2025
  • Final testing result due: January 15th, 2026
  • Final 4-page report due: January 30th, 2026

IEEE PELS International Challenge on Design Methodologies [Prize Pool $60,000]

  • This site provides the latest information about the MagNet Challenge 2.
  • Please contact pelsmagnet@gmail.com for all purposes.
  • Sign-Up to MagNet Challenge 2 before May 1st PDF, extended to May 8th.
  • 2-Page concept proposal due June 1st PDF, DOC, Latex, extended to June 15th.
  • The purpose of the concept proposal is to ensure all teams understand the rules and complete team forming.

[News #8] MagNet Tutorial 5

  • Tutorial 5 (September 12 10am EST): Video, Slides
  • Release new benchmark models, examples, and testing results
  • MagNet Challenge 2025 November Preliminary Test Data Released Pretest

[News #7] Useful Resources

[News #6] MagNet Tutorial Series

[News #5] 40 Teams Confirmed, ✅:Concept Paper, 👍: Pre-test, 🌸: Final

  • Aalborg University (Zhao), Denmark 🇩🇰 ✅ 👍
  • Aalborg University (Davari), Denmark 🇩🇰 ✅ 👍
  • University of Auckland, New Zealand 🇳🇿 ✅
  • Eindhoven University of Technology, Netherland 🇳🇱 ✅
  • Tianjin University, China 🇨🇳 ✅ 👍
  • Xi'an Jiaotong University (Zhu1), China 🇨🇳 ✅
  • Xi'an Jiaotong University (Zhu2), China 🇨🇳 ✅
  • Xi'an Jiaotong University (Wei), China 🇨🇳 ✅ 👍
  • Xi'an Jiaotong University (Chen), China 🇨🇳 ✅
  • Fuzhou University, China 🇨🇳 ✅ 👍
  • Hangzhou Dianzi University, China 🇨🇳 ✅
  • Southeast University (Cheng), China 🇨🇳 ✅ 👍
  • Southeast University (Xu), China 🇨🇳 ✅
  • Tsinghua University, China 🇨🇳 ✅
  • Fujian University of Technology, China 🇨🇳 ✅
  • Hefei University of Technology, China 🇨🇳 ✅
  • Huazhong University of Science and Technology, China 🇨🇳 ✅
  • University of Electronic Science and Technology of China, China 🇨🇳 ✅
  • Zhejiang University, China 🇨🇳 ✅
  • The Hong Kong Polytechnic University, Hong Kong SAR 🇭🇰 ✅
  • Seigen & Paderborn, Germany 🇩🇪 ✅ 👍
  • University of Kassel, Germany 🇩🇪 ✅
  • TU Munich & FU Santa Catarina, Germany & Brazil 🇩🇪 🇧🇷 ✅
  • Leeds & Wuerth Elektronik, UK & Germany 🇬🇧 🇩🇪 ✅
  • KU Leuven, Belgium, 🇧🇪 ✅ 👍
  • University of Bristol, UK 🇬🇧 ✅ 👍
  • University of Cambridge, UK 🇬🇧 ✅
  • Politecnico di Torino, Italy 🇮🇹 ✅
  • Tokyo Metropolitan University, Japan 🇯🇵 ✅
  • Nagoya Institute of Technology, Japan 🇯🇵 ✅
  • Nanyang Technological University (Yang), Singapore 🇸🇬 ✅
  • Nanyang Technological University (Tang), Singapore 🇸🇬 ✅
  • University of Sydney, Australia 🇦🇺 ✅ 👍
  • National Institute of Technology Calicut (Mohan), India 🇮🇳 ✅
  • National Institute of Technology Calicut (Muhammed), India 🇮🇳 ✅
  • Indian Institute of Technology Dharwad, India 🇮🇳 ✅
  • Silicon Austria Labs, Austria 🇦🇹 ✅ 👍
  • Arizona State University, USA 🇺🇸 ✅ 👍
  • University of Missouri Columbia, USA 🇺🇸 ✅
  • Georgia Institute of Technology, USA 🇺🇸 ✅ 👍
  • Princeton University, USA 🇺🇸 (not competing)
  • Dartmouth College, USA 🇺🇸 (not competing)

[News #4] Sponsors Confirmed

  • Nvidia (Top GPUs)
  • Texas Instruments ($5000)
  • Wurth Electronik ($5000)
  • ITG Electronics ($5000)
  • pSemi ($5000)
  • IEEE Power Electronics Society ($35000)
  • Princeton University (as much as needed!)

[News #3] MagNet Challenge 2 - Training Data Release

Training data for 10 materials (the same 10 materials as MagNet Challenge 1):

You may want to reuse the steady-state data from MagNet Challenge 1:

Note:

  • Testing data on 5+ new materials will be released on Nov. 1st;
  • Data sequences are sampled at a fixed rate (16 MHz) in different lengths;
  • Data sequences in different rows are independent from each other;
  • Let us know if you find any problems with the data.

[News #2] MagNet Challenge 2 - APEC Launch

  • Time: 2025-3-19 Wed 2pm EST
  • Atlanta Omni Hotel, Room Grand A

[News #1] MagNet Challenge 2 - Public Launch

Overview

Build upon the success of MagNet Challenge 1, the goal of the MagNet Challenge 2 is to develop intelligent software tools that can learn and predict magnetic characteristics in transient. For each magnetic material of interest, we are looking for a MATLAB or Python function that takes the following three inputs:

  • A pair of B(t) and H(t) waveforms documenting the excitation history from t_0 to t_1;
  • A future flux density excitation wave B’(t) from t_1 to t_2;
  • Temperature: T.

And produce the following one output:

  • The corresponding field strength wave H’(t) from t_1 to t_2 paired with B’(t).

This function should be packaged as: H'(t)=function (B(t),H(t),B'(t),T).

In order to capture the physical behaviors of the magnetic material in transient, the models should be frequency agnostic (no frequency information), time-step agnostic (short or long time-steps), and initial-state agnostic (always converging after a long time). We encourage using the latest stable version of commonly used MATLAB and Python packages. Analytical methods and machine learning methods are both encouraged.

There are intrinsic correlations between the materials behavior in steady-state and in transient. In fact, a model operates well for transient conditions must operate well in steady states. As a result, student teams are encouraged to reuse the data and models made available for the MagNet Challenge 1 in 2023 and leverage the physical and analytical understandings of the models developed for the MagNet Challenge 1 in 2023 for the MagNet Challenge 2 in 2025.

Please refer to the Handbook and Slides for more details.

Timeline

  • 02-14-2025 Initial Call for Participation Announcement Handbook Slides SignUp
  • 03-19-2025 APEC Official Annoucement
  • 04-01-2025 Training Data for 10 Materials Available
  • 05-01-2025 1-Page Letter of Intent Due with Signature
  • 06-01-2025 2-Page Concept Proposal Due PDF DOC Latex
  • 07-01-2025 Notification of Acceptance
  • 08-01-2025 Expert Feedback on the Concept Proposal
  • 11-15-2025 Preliminary Submission Due (postponed to 11-15)
  • 11-15-2025 Testing Data for 5 New Materials Available (postponed to 11-15)
  • 01-15-2026 Final Submission Due (postponed to 01-15)
  • 03-01-2026 Winners Selected

Evaluation

  • 06-15-2025 Evaluate the concept proposals and ensure all teams understand the competition rules.
  • 12-01-2026 Evaluate the 10 models the teams developed for the 10 materials and provide feedback for improvements.
  • 01-15-2026 Evaluate the 5 new models the teams developed for the 5 new materials and announce the winners.

Judge Committee

  • Prof. David Perreault, MIT, USA
  • Prof. Jürgen Biela, ETH Zurich, Switzerland
  • Prof. Dragan Maksimovic, CU Boulder, USA
  • Prof. SY Ron Hui, CityU, Hong Kong
  • Prof. Charles Sullivan, Dartmouth, USA

The judging committee will evaluate the results of each team with the following criterias.

  • Model accuracy: core loss and B-H trajectory prediction accuracy (lower error better)
  • Model size: number of parameters the model needs to store for each material (smaller model better)
  • Model explanability: explanability of the model based on existing physical insights (more explainable better)
  • Model novelty: new concepts or insights presented by the model (newer insights better)
  • Software quality: quality of the software engineering (more concise better)

Discussions

  • Discussion Forum Link

Awards (pending)

  • Model Performance Award, First Place $10,000
  • Model Performance Award, Second Place $5,000
  • Model Novelty Award, First Place $10,000
  • Model Novelty Award, Second Place $5,000
  • Outstanding Software Engineering Award $10,000
  • Honorable Mentions Award multiple x $1,000

Useful Links

Other Resources

Reference Papers

  • S. Wang et al., "Unified Time Domain Foundation Models for Hysteretic Passive Components," in IEEE 26th Workshop on Control and Modeling for Power Electronics (COMPEL), Knoxville, TN, USA, 2025, pp. 1-8, Paper
  • M. Chen, H. Li, S. Wang, T. Guillod, D. Serrano, N. Forster, W. Kirchgassner, T. Piepenbrock, O. Schweins, O. Wallscheid, Q. Huang, Y. Li, Y. Dou, B. Li, S. Li, E. Havugimana, V. T. Chacko, S. Radhakrishnan, M. Ranjram, B. Sauter, S. Reese, S. Sinha, L. Zhang, T. McKeague, B. Cui, N. Rasekh, J. Wang, S. Liu, A. Martinez, X. Liu, C. Mei, R. Zhao, G. Wu, H. Wu, R. Zhang, H. Song, L. Zhang, Y. Lu, L. Hang, N. Rajput, H. B. Sandhibigraha, N. Agrawal, V. M. Iyer, X. Shen, F. Tian, Q. Sui, J. Kong, W. Martinez, A. Arruti, B. Alberdi, A. Agote, I. Aizpuru, M. Zhang, X. Chen, Y. Dong, D. Wang, T. Shen, Y. Zhou, Y. Li, S. Wang, Y. Wu, Y. Jiang, Z. Xiao, Y. Tang, Y.-S. Hsieh, J.-D. Li, L.-C. Yu, T.-C. Hsu, Y.-C. Liu, C.-H. Hsia, C. Chen, A. Giuffrida, N. Lombardo, F. Marmello, S. Morra, M. Pasquale, L. Solimene, C. S. Ragusa, J. Reynvaan, M. Stoiber, C. Li, W. Qin, X. Ma, B. Zhang, Z. Wang, M. Cheng, W. Xu, J. Wang, Y. Hu, J. Xu, Z. Shi, D. B. Sapkota, P. Neupane, M. Joshi, S. Khan, B. Su, Y. Xiao, M. Yang, K. Sun, Z. Li, R. Mirzadarani, R. Liu, L. Wang, T. Luo, D. Lyu, M. G. Niasar, Z. Qin, S. I. A. Meerza, K. Froehle, H. H. Cui, D. Costinett, J. Liu, Z. Liu, C. Zhan, Y. Dang, Y. Zhang, N. Wang, Y. Chen, Y. Zhang, C. Li, Y. Yao, T. Hu, L. Xu, Y. Wang, S. Wang, S. Jiang, D. Shumacher, D. Maksimovic, R. S. Y. Hui, J. W. Kolar, D. J. Perreault, and C. R. Sullivan, "Magnet Challenge for Data-Driven Power Magnetics Modeling," in IEEE Open Journal of Power Electronics, doi: 10.1109/OJPEL.2024.3469916. Paper
  • S. Wang, H. Kwon, H. Li, et al. "MagNetX: Foundation Neural Network Models for Simulating Power Magnetics in Transient," IEEE Applied Power Electronics Conference and Exposition (APEC), Atlanta, GA, USA, 2025, pp. 2438-2445. Paper
  • H. Kwon, S. Wang, H. Li, et al. "MagNetX: Extending the MagNet Database for Modeling Power Magnetics in Transient," IEEE Applied Power Electronics Conference and Exposition (APEC), Atlanta, GA, USA, 2025, pp. 566-572. Paper
  • D. Serrano et al., "Why MagNet: Quantifying the Complexity of Modeling Power Magnetic Material Characteristics," in IEEE Transactions on Power Electronics, doi: 10.1109/TPEL.2023.3291084. Paper
  • H. Li et al., "How MagNet: Machine Learning Framework for Modeling Power Magnetic Material Characteristics," in IEEE Transactions on Power Electronics, doi: 10.1109/TPEL.2023.3309232. Paper
  • H. Li, D. Serrano, S. Wang and M. Chen, "MagNet-AI: Neural Network as Datasheet for Magnetics Modeling and Material Recommendation," in IEEE Transactions on Power Electronics, doi: 10.1109/TPEL.2023.3309233. Paper

Organizers

About

MagNet Challenge-2 Official Site

Resources

Readme
Activity

Stars

37 stars

Watchers

5 watching

Forks

6 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages