<p>PWM rectifiers have gained widespread adoption in new energy grid integration and industrial drive systems owing to their high efficiency and low harmonic distortion. However, the inherent common-mode voltage (CMV) can induce electromagnetic interference and bearing current problems, while neutral-point (NP) potential imbalance may subject devices to overvoltage risks, posing significant challenges to system reliability and safety. Although existing model predictive direct power control (MPDPC) methods offer fast dynamic responses, they are limited by two key issues. Traditional MPDPC necessitates evaluating all 27 voltage vectors, leading to high computational complexity. In addition, current strategies fail to effectively suppress CMV while simultaneously achieving precise NP potential balance. To tackle these challenges, this paper proposes a low common-mode voltage model predictive direct power control (LCMV-MPDPC) strategy. Specifically, the space voltage vector diagram (SVVD) is segmented into a 6-sector, 12-subsector structure. By employing hierarchical screening (from sector to subsector) and excluding zero vectors as well as small vectors with high CMV, the CMV amplitude is constrained to less than 1/6 of the DC-link voltage. This approach reduces the number of candidate vectors from 27 to 19, significantly lowering computational demands. Furthermore, a pair of virtual vectors with identical directions but opposing effects on the NP potential is constructed. Dynamic compensation is then applied to achieve autonomous balancing of the NP potential. The effectiveness of this method is substantiated through comparative simulation analysis and experimental validation.</p>