Recursive Identification With Multiple Forgetting Factors for Time-Varying Wireless Power Transfer Systems
Jie Hou, Junwei Jiang, Shiwei Wang, Fengwei Chen, Ke Shi, Sheng Xiang, Tianxu Feng
Abstract
To better track time-varying characteristics in the dynamic modeling of Wireless Power Transfer (WPT) systems, a Recursive Least Squares (RLS) algorithm with multiple adaptive forgetting factors is proposed. Firstly, the dominant mode analysis reveals that the time-varying parameters of the data-driven model always have different rates of change when the system load and mutual inductance undergo changes. On this basis, we propose an RLS method with multiple forgetting factors that assigns different forgetting factors to data-driven model parameters varying at different rates, resulting in superior tracking capability and dynamic performance compared to standard RLS which employs a single forgetting factor for all parameters with different rates of change. Furthermore, to improve the accuracy of parameter identification, an adaptive strategy is proposed that leverages parameter variation rates to dynamically adjust multiple forgetting factors and uses auxiliary models to estimate data corrupted by noise. Experiments conducted in real-world time-varying scenarios of the WPT system validated the effectiveness and merits of the proposed method.