Equivalent Parameter Extraction Method of QSCR-WPT System Based on LSTM and NLS

Quasi-static cavity resonance wireless power transfer (QSCR-WPT) technology can achieve pervasive power delivery in room-scale range. The coupling model of the system has been established by several approaches. However, the accuracy of mathematical model is seriously limited due to the lack of direct measurement of the equivalent parameters of the cavity, which hinders the analysis and design optimization of the system. Additionally, the coupling model is difficult to analyze key characteristics, such as the voltage and current. To address this issue, a hybrid parameter extraction methodology is presented firstly and it combines long short-term memory (LSTM) networks with nonlinear least squares (NLS) optimization. Specifically, approximate equivalent circuit parameters are obtained using an LSTM network from the input impedance spectrum data. These parameters serve as initial values for the NLS algorithm, which then performs precise regression to yield precise circuit parameters. Thus, system performance can be represented and analyzed clearly. Experimental results demonstrate that the implemented methodology achieves 95.78% voltage prediction accuracy and 98.13% transmission efficiency estimation precision across the operational spectrum. This paper provides a novel perspective on the QSCR-WPT modeling and paves the way for further exploration of the parameter optimization and control methods of the QSCR-WPT.