A Low Computational Burden Model Predictive Control for Dynamic Wireless Charging

Dynamic wireless charging (DWC) technology can help alleviate the problem of short driving range for battery-powered vehicles. In this article, a model predictive control (MPC) is applied to the buck converter on the secondary side of a DWC system to address fast output fluctuations. This approach features a fast-dynamic response, and no communication link is required. To solve the key issue of MPC, which is the computational burden, a polynomial fitting method based on the parsing solution of the sampled-data model is proposed. The complex matrix exponential calculation is replaced by simple polynomial operations, and the optimal duty cycle can be calculated directly by solving a quadratic function. This significantly reduces the computational burden. A DWC experimental setup is constructed, and results show that the proposed MPC has a better dynamic performance compared to proportional-integral control. The adjustment time is only 140 μs (around seven switching cycles) when the reference voltage is stepping. Moreover, the computational burden for matrix calculation in two-step prediction can be reduced by 50.6% and 79.7% compared to the lookup table and Taylor series approximation, respectively. Meanwhile, MPC with current limitation is analyzed and demonstrates a neat spectrum, small ripple but large response time.