Plug-and-Play Architecture for Non-Identical Dual-Parallel PMSMs: A Hybrid Learning Control Approach

Abstract

The mono-inverter dual-parallel permanent magnet synchronous motor (MIDP) drive system is a cost-effective solution for compact multi-motor applications, but existing control schemes suffer from strict constraints of identical motor parameters and precise prior parameter knowledge, limiting their practical implementation. To address this challenge, a novel plug-and-play control architecture integrated with hybrid learning and adaptive control is proposed for non-identical MIDP systems in this paper. First, a strict theoretical proof based on the Unknown Input Observer (UIO) error dynamic model is presented, verifying that the reconstructed signals of the supplementary motor are independent of its inductance and flux linkage under steady-state conditions, which decouples signal reconstruction and parameter identification. On this basis, a two-stage hybrid parameter learning mechanism is designed to circumvent the rank–deficiency problem such that the stator resistance is identified at standstill, and the inductance and rotor flux linkage are estimated online by the Recursive Least Squares (RLS) with a forgetting factor. Furthermore, the non-master-slave control strategy is extended with a newly derived stability criterion to accommodate non-identical motors. Comprehensive experiments covering start-up, sudden speed fluctuations, and extreme unbalanced load torques are conducted on a hardware platform with two non-identical PMSMs. Experimental results show that the proposed architecture enables plug-and-play operation of the supplementary motor without prior parameter knowledge, significantly improving the flexibility and practicality of the MIDP drive system.