High-Frequency Impedance Modeling of Induction Motors Using Adaptive Multistage RLC Circuit and Neural Network-Based Stagewise Parameter Identification

Abstract

Induction motor high-frequency impedance modeling facilitates predictive analysis for critical high-frequency or transient behaviors in motor drive systems. Current modeling methods typically demonstrate high accuracy only below 30 MHz and some even involve complicated parameter identification or lack versatility. This article proposes a novel modeling method, simultaneously achieving high accuracy, simplified parameter identification, and good versatility across a wider frequency range beyond 30 MHz. The method introduces an adaptive multistage RLC circuit model that dynamically adjusts its number of stages based on the desired accuracy, enabling precise prediction of motor impedances across 100 Hz–120 MHz using the fewest stages. Additionally, the model’s symmetric circuit structure accommodates both star- and delta-connected motors without the need to recalculate model parameters’ values when changing motor connections. Furthermore, the method presents a neural network-based stagewise parameter identification algorithm, which can determine all model parameters’ values within 7 min using a standard 12-core computer. Experimental validation using three induction motors (1.1, 5.5, and 22 kW) affirms the method’s high accuracy (average magnitude error ≤ 2.96 dB and average phase error ≤ 8.91°) in estimating magnitude and phase information of four types of motor impedances (common-mode, differential-mode, single-phase, and phase-to-ground) from 100 Hz to 120 MHz and demonstrates its capabilities in terms of parameter identification and versatility.