MagLearn 2: High-Fidelity, Saturation-Aware, and History-Efficient Sequence-to-sequence Modeling of Transient B-H Behaviour Under PWM Excitation

📄 arXiv: 2610.03504v1 📥 PDF

作者: Jingrong Yang, Binyu Cui, Yuming Huo, Lizhong Zhang, Alfonso Martinez, Song Liu, Jun Wang

分类: eess.SY

发布日期: 2026-10-02

备注: Submitted to IEEE Transactions on Power Electronics (TPEL) for peer review. An earlier conference version of this work was published in IEEE Xplore: https://ieeexplore.ieee.org/document/11597291/


💡 一句话要点

提出MagLearn 2以解决PWM激励下瞬态B-H建模问题

🎯 匹配领域: 支柱八:物理动画 (Physics-based Animation)

关键词: 瞬态建模 脉宽调制 LSTM 饱和效应 机器学习 电磁材料 序列预测

📋 核心要点

  1. 现有的瞬态B-H建模方法未能有效处理磁历史的需求、饱和区域的处理及模型对新材料的适应性。
  2. 本文提出MagLearn 2框架,利用短窗口LSTM模型重建H(t)响应,并引入饱和感知模块以优化预测结果。
  3. 在实验中,MagLearn 2显著降低了预测误差,最佳模型在MagNet Challenge II中实现了2.2%的平均RMSE,表现优异。

📝 摘要(中文)

在脉宽调制(PWM)激励下,准确的瞬态B-H建模面临挑战,因为磁场响应依赖于瞬时磁通密度和先前的磁化轨迹。现有研究未能共同解决所需的磁历史量、饱和区域处理及模型对未见材料的适应性。为此,本文提出了MagLearn 2,一个基于短窗口LSTM的序列到序列框架,能够从历史B(t)和H(t)轨迹重建未来的H(t)响应。一个饱和感知辅助模块选择性地替换完整波形预测中的饱和主导预测。在代表性的饱和3C90案例中,该辅助模块将序列级相对RMSE从67.84%降低至9.19%。通过基于遮蔽的敏感性分析,量化了磁历史对预测的贡献,发现与预测最相关的信息集中在预测边界附近。在MagNet Challenge II的独立评估中,所提出框架的最佳模型在序列预测任务中实现了迄今为止最低的平均RMSE,达到2.2%。

🔬 方法详解

问题定义:本文旨在解决在PWM激励下瞬态B-H建模的复杂性,现有方法未能有效处理磁历史、饱和效应及对新材料的适应性问题。

核心思路:提出的MagLearn 2框架通过短窗口LSTM模型重建未来的H(t)响应,并引入饱和感知模块,选择性地优化饱和区域的预测,提升模型的准确性和适应性。

技术框架:整体架构包括历史B(t)和H(t)轨迹的输入,LSTM网络进行序列建模,以及饱和感知辅助模块用于优化输出。模型通过训练学习历史信息与当前状态的关系。

关键创新:最重要的创新在于引入饱和感知模块,该模块能够有效识别并替换饱和主导的预测,显著提高了模型在复杂波形下的预测能力。

关键设计:模型采用短窗口LSTM结构,损失函数设计为考虑饱和效应的加权RMSE,确保在训练过程中对重要特征的关注。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在实验中,MagLearn 2的饱和感知模块将序列级相对RMSE从67.84%降低至9.19%,并在MagNet Challenge II中实现了2.2%的平均RMSE,创下了该任务的最低记录,显示出显著的性能提升。

🎯 应用场景

该研究在电磁材料的建模与预测方面具有广泛的应用潜力,尤其是在电动机、变压器及其他电磁设备的设计与优化中。通过提高瞬态响应的预测精度,能够有效提升设备性能和能效,推动智能电网及可再生能源技术的发展。

📄 摘要(原文)

Accurate transient B-H modeling under pulsewidth-modulated (PWM) excitation is challenging since the magnetic-field response depends on both the instantaneous flux density and the preceding magnetisation trajectory. Although sequence-to-sequence models enable waveform prediction, existing studies have not jointly addressed how much magnetic history is required, how saturation region should be handled, and how trained models can be adapted to unseen materials. To address these gaps, this paper presents MagLearn 2, a condition-aware, short-window LSTM-based sequence-to-sequence framework for reconstructing the future H(t) response from the historical B(t) and H(t) trajectories. A saturation-aware auxiliary module selectively replaces saturation-dominated predictions within the complete waveform forecast. In a representative saturated 3C90 case, the auxiliary module reduces the sequence-level relative RMSE from 67.84% to 9.19%. To understand how much the history samples contributes to the predictions, an occlusion-based sensitivity analysis is conducted to quantifies the contribution of magnetic history and reveals that that the information most relevant to prediction is concentrated near the forecast boundary. In the independent evaluation for MagNet Challenge II, the proposed framework's best-performing model achieved the lowest reported average RMSE to date on the sequence-prediction task at 2.2%.