Gradient Descent-Based Online Mutual Inductances Identification and Transfer Route Optimization for Dynamically Reconfigurable WPT Network

Wireless Power Transfer Network (WPTN), with their multidimensional dynamic magnetic coupling capabilities, address the limitations of traditional wireless power transfer (WPT) technologies in terms of degrees of freedom, thus advancing the development and application of WPT technologies. Accurate mutual inductance identification is essential for establishing and optimizing paths in WPTN systems. However, the dynamic nature of the network causes frequent changes in mutual inductance parameters as nodes join or leave. This makes traditional calculation methods inadequate for complex multi-node network. To address this challenge, this paper proposes a dynamic mutual inductance identification method based on the gradient descent algorithm. The method effectively handles simultaneous changes in multiple mutual inductance parameters caused by network topology variations. By constructing a mathematical model of the WPTN system and designing a gradient descent algorithm with adaptive learning rates, the method achieves synchronous identification of multiple mutual inductance parameters. It does not require pre-training or extensive computational resources, and supports both online and offline identification. Experimental results on a four-node system platform show that the proposed method can identify three new mutual inductances in real time during node switching, while dynamically selecting the most efficient transmission path.