Search

Predictive errors restrict high performance of finite control set model predictive control (FCS-MPC) in the cascaded H-bridge multilevel inverter.In order to solve this problem, the influence mechanisms of parameter mismatch, model discretization, and sampling error on voltage and current prediction errors are analyzed in this article. A hierarchical multifactorial error correction method is proposed. The known voltage prediction errors of the previous adjacent prediction periods are first fed back to calculate the voltage predictive error factors and self-correct the next voltage prediction errors. And then the current prediction errors are calibrated by the corrected voltage and the calculated current predictive error factor. This proposed method involves a simple structure and a good inhibitory ability on multifactorial prediction errors. Finally, experimental results in a three-phase five-level cascaded H-bridge multilevel inverter support the theoretical analysis and the proposed method. [ABSTRACT FROM AUTHOR]

The extensive application of vector phase shift modulation method in permanent magnet synchronous motor drive systems with single dc-link current sensor results in obvious current ripple in the low-speed region. The patterns of these ripples vary greatly in different sectors, leading to nonnegligible phase current reconstruction errors. In this article, an estimation of the local average current change rate (ELAC2R) based reconstruction error suppression method within the switching cycle is proposed to resolve this problem. By comparing the actual and the ideal sampling points, the increase of current ripple and the reconstruction error caused by the combination of the vector phase shift and the single current sensor sampling in the low-speed region are modelled and analyzed. Then according to equivalent modelling of the current ripple between the ideal and the actual sampling points, the more accurate phase current information can be quickly obtained based on the sampling value. The compensation process is significantly simplified and the computation burden of the processor is reduced. Experimental results show that the performance of current reconstruction can be improved effectively with less harmonic content with the proposed method. [ABSTRACT FROM AUTHOR]