Abstract: Multi-level multiphase motor drive systems usually have problems of numerous output voltage vectors and multi-objective constraints. A neutral point clamped (NPC) three-level six-phase inverter-fed dual three-phase permanent magnet synchronous motor (PMSM) has been taken as research objective, and an improved model predictive voltage control strategy based on voltage vectors' triangular region classification has been proposed to suppress the harmonic current components and balance the neutral-point potential. First, 729 output voltage vectors are preliminary screened and those vectors with larger amplitudes in α-β fundamental subspace have been chosen to suppress harmonic current components. Then, these screened voltage vectors are classified based on the vectors' triangular region classification and the expected voltage vector is located based on this classification method to choose the candidate vectors, on account of which an objective function is designed to satisfy the harmonic suppression and the neutral-point potential balancing. To address the difficulty in the weight factor designing for the multi-objective control, this paper provides a theoretical design basis based on the mathematical models of the motor and the characteristics of the inverter. The confirmatory and comparative experiments under different operating conditions show that this control strategy can not only improve the multi-objective control performance of the multilevel multiphase motor drive system, but also effectively reduce the computational burden of the controller.