Core Loss Modeling of Magnetic Components Using a Data-Driven Method

Area of Science:

• Power Electronics and Electrical Engineering
• Materials Science and Engineering
• Computational Intelligence and Machine Learning

Background:

• Magnetic component core loss significantly impacts power conversion efficiency.
• Existing core loss models often show discrepancies with practical applications.
• Accurate modeling is crucial for optimizing power electronic systems.

Purpose of the Study:

• To develop a high-precision core loss evaluation model for magnetic components.
• To investigate the influence of temperature, material, and excitation waveform on core loss.
• To optimize magnetic energy transmission and efficiency through multi-objective optimization.

Main Methods:

• Utilized eXtreme Gradient Boosting (XGBoost) for core loss modeling.
• Employed confusion matrices, boxplots, and regression metrics (R², MSE, MAE) for model evaluation.
• Applied simulated annealing algorithms for multi-objective optimization.

Main Results:

• The developed XGBoost model demonstrated high accuracy and practical relevance in predicting core loss.
• Single-factor and synergistic analyses revealed key drivers of core loss.
• Optimization identified conditions for minimizing core loss and maximizing transmission efficiency.

Conclusions:

• The eXtreme Gradient Boosting model offers a superior approach to core loss evaluation in power conversion.
• The research provides a framework for optimizing magnetic component performance.
• This work extends to multi-objective optimization for enhanced power system design.