Spectral entropy prior-guided deep feature fusion architecture for magnetic core loss

2025-12-12

AI Summary

To address the longstanding bottlenecks in core loss modeling—namely, low prediction accuracy, poor interpretability, and weak out-of-distribution generalization—this paper proposes a physics-informed hybrid modeling framework. Methodologically, it introduces (1) a novel spectral entropy-guided prior mechanism that dynamically selects the optimal empirical model, and (2) an adaptive multimodal feature fusion module that jointly integrates physics-based constraints with deep learning representations—including CNN-extracted spatial features, bidirectional LSTM-captured temporal dynamics, and multi-head attention-refined global dependencies. Evaluated on the MagNet benchmark, the framework consistently outperforms all 21 models from the 2023 IEEE ICPE Core Loss Challenge and three state-of-the-art methods published in 2024–2025. It achieves significant improvements in prediction accuracy, robustness to distributional shifts, and model interpretability. This work establishes a reliable, physically grounded modeling paradigm for high-efficiency power electronic system design.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Low Level & Physics-based Vision

Application Category

• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Accurate core loss modeling is critical for the design of high-efficiency power electronic systems. Traditional core loss modeling methods have limitations in prediction accuracy. To advance this field, the IEEE Power Electronics Society launched the MagNet Challenge in 2023, the first international competition focused on data-driven power electronics design methods, aiming to uncover complex loss patterns in magnetic components through a data-driven paradigm. Although purely data-driven models demonstrate strong fitting performance, their interpretability and cross-distribution generalization capabilities remain limited. To address these issues, this paper proposes a hybrid model, SEPI-TFPNet, which integrates empirical models with deep learning. The physical-prior submodule employs a spectral entropy discrimination mechanism to select the most suitable empirical model under different excitation waveforms. The data-driven submodule incorporates convolutional neural networks, multi-head attention mechanisms, and bidirectional long short-term memory networks to extract flux-density time-series features. An adaptive feature fusion module is introduced to improve multimodal feature interaction and integration. Using the MagNet dataset containing various magnetic materials, this paper evaluates the proposed method and compares it with 21 representative models from the 2023 challenge and three advanced methods from 2024-2025. The results show that the proposed method achieves improved modeling accuracy and robustness.

Problem

• Improves magnetic core loss modeling accuracy
• Enhances interpretability and generalization of data-driven models
• Integrates empirical models with deep learning for robustness

Innovation

• Hybrid model integrates empirical models with deep learning
• Spectral entropy prior selects suitable empirical models adaptively
• Adaptive fusion module enhances multimodal feature interaction