High-Frequency Characterization and Modeling of Pulse Magnet via High-Precision Impedance Measurement and Neural Network-Based Parameter Identification

Pulsed high magnetic fields (PHMFs) have emerged as essential tools in scientific research. Consequently, the design and analysis of pulsed magnets, as the key to generating PHMFs, have garnered significant research attention. Among them, the circuit model of pulsed magnets is critical for simulating their electromagnetic behavior and predicting the discharge characteristics. In this paper, a high-frequency characterization and modeling method via high-precision impedance measurement and neural network-based parameter identification is proposed. This study eliminates the influence of parasitic parameters and performs de-embedded impedance measurements of the pulsed magnet using a vector network analyzer (VNA) in conjunction with a custom-designed fixture adapter. Based on the measured data, a circuit model of the pulsed magnet was constructed, and the values of individual components in the model were subsequently identified using a neural network. For validation, the proposed circuit model was implemented in a simulation to predict the terminal-voltage variation and discharge behavior of the magnet. The results demonstrate a close agreement between the simulated and experimental voltage data, confirming the high accuracy and strong predictive capability of the model derived using the proposed method. In specific, under broadband square-wave with excitation frequencies of 1 MHz and 2 MHz, the model achieves a time-domain mean absolute errors (MAE) of 0.058/0.038 and a frequency-domain mean absolute relative error (MARE) of 0.066/0.064.