Model predictive control (MPC) has been recently suggested as an interesting alternative for the regulation of multiphase electric drives because it easily exploits the inherent advantages of multiphase machines. However, standard MPC applies a single switching state during the whole sampling period, inevitably leading to an undesired x-y voltage production. Consequently, its performance can be highly degraded when the stator leakage inductance is low. This shortcoming has been however mitigated in recent works with the implementation of virtual/synthetic voltage vectors (VVs) in MPC strategies. Their implementation reduces the phase current harmonic distortion since the average x-y voltage production becomes null. Nevertheless, VVs have a static nature because they are generally estimated offline, and this implies that the flux/torque regulation is suboptimal. Moreover, these static VVs also present some limitations from the point of view of the DC-link voltage exploitation. Based on these previous limitations, this work proposes the implementation of dynamic virtual voltage vectors (DVVs), where VVs are created online within the MPC strategy. This new concept provides an online optimization of the output voltage production depending on the operating point, resulting in an enhanced flux/torque regulation and a better use of the DC-link voltage. Experimental results have been employed to assess the goodness of the proposed MPC based on DVVs.

Multiphase drives have been important in particular industry applications where reliability is a desired goal. The main reason for this is their inherent fault tolerance. Different nonlinear controllers that do not include modulation stages, like direct torque control (DTC) or model-based predictive control (MPC), have been used in recent times to govern these complex systems, including mandatory control reconfiguration to guarantee the fault tolerance characteristic. A new reconfiguration-less approach based on virtual voltage vectors (VVs) was recently proposed for MPC, providing a natural healthy and faulty closed-loop regulation of a particular asymmetrical six-phase drive. This work validates the interest in the reconfiguration-less approach for direct controllers and multiphase drives.

Multiphase machines are interesting options for high‐reliability applications due to their inherent fault tolerance against open‐circuit faults (OCFs). Moreover, if the regulation of the x – y currents is realised in open‐loop mode using virtual voltage vectors (VVs), the mandatory post‐fault control reconfiguration is avoided. The new reconfigurationless approach was recently defined as a natural/passive fault‐tolerant strategy, offering good prospects for industry applications. This work extends the idea to the fault detection procedure and suggests new settings for the current imbalance localisation method. The proposal is based on the vector space decomposition approach that allows the joint detection of OCFs and stator resistances dissymmetry. Experimental results in a five‐phase induction motor drive using VVs confirm the viability of the technique.

Finite Control-Set Model Predictive Control (FCS-MPC) appears as an interesting alternative to regulate multiphase electric drives, thanks to inherent advantages such as its capability to include new restrictions and fast-transient response. Nevertheless, in industrial applications, FCS-MPC is typically discarded to control multiphase motors because the absence of a modulation stage produces a high harmonic content. In this regard, multi-vectorial approaches are an innovative solution to improve the electric drive performance taking advantage of the implicit modulator flexibility of Model Predictive Control (MPC) strategies. This work proposes the definition of a new multi-vectorial set of control actions formed by a couple of adjacent large voltage vectors and a null voltage vector with an adaptative application ratio. The combination of two large voltage vectors provides minimum x-y current injection whereas the application of a null voltage vector reduces the active voltage production. Moreover, the optimum selection of the null voltage vector for each couple of large voltage vectors permits reducing the switching frequency. On the other hand, the active application time for this couple is estimated through an analytic function based on the operating point. This procedure avoids the use of an iterative process to define the duty cycles, hence significatively decreasing the computational burden.