Hybrid Data-Driven Parameters Estimation for Communication-Less WPT System With Reduced Primary Sampling Data
Bing Cheng, Liangzong He, Houxuan Liu
Abstract
Wireless power transfer (WPT) systems advantaging environment-friendly and highly efficient, and accurate parameter estimation is a premise for precise control and a guarantee for system safety. However, the conventional parameter estimation method, which is based on the primary impedance angle is easily affected by the components delay, current distortion, and other issues during the detection process. Meanwhile, conventional data-driven parameter estimation requires a large number of data and dual-side commutation, resulting in higher execution costs. To address those issues, an online/offline hybrid data-driven parameters estimation method utilizing the backpropagation neural network (BPNN) is proposed for the communication-less WPT system. The primary dc input voltage and current are taken as the input variables of the BPNN model to reduce the noise of training data. Further, to overcome the difficulty of a large account of training data collection, an approximate BPNN model is built first by the offline training under a small sampling set. Then, online data-driven parameter estimation by monitoring the primary resonant current is implemented to optimize the built BPNN model constantly, resulting in enhancing estimation accuracy and avoiding complex wireless communication. Finally, an experimental setup is constructed to verify the feasibility of the proposed data-driven estimation method.