Low-Complexity Model Predictive Control Techniques For Five-Phase PMSM Drives

Model predictive control(MPC)is considered as a promising control strategy for power electronic converters and drive systems thanks to its merits of simplicity,fast-dynamic response,and multi-variable control flexibility.However,the computational burden is the main challenging for applying the conventional finite control set(FCS)-MPC on five-phase PMSM drives due to the large number of the voltage vectors and the required control objectives.This thesis focuses on modifying the conventional FCS-MPC scheme to be applied on five-phase permanent magnet synchronous motor(PMSM)drives.Therefore,new low-complexity FCS-MPC schemes are proposed to achieve the required control objectives with the minimum computational burden using the deadbeat(DB)principles.In addition,the proposed scheme with low computational burden can provide high time-space for adding an extended observer to improve the controller robustness against parameter mismatches.Firstly,for five-phase two-level(2L)PMSM drives,low-complexity FCS-MPC schemes are developed for direct selection of optimal voltage vector(s)using the virtual voltage vector(V~3s)concept.The designed V~3s are utilized to minimize the x-y harmonic components.Therefore,the difficulty of applying explicit voltage vector selection in two different planes is removed with the aid of V~3s.The proposed scheme calculates the reference voltage vector using the DB principles,then it selects the best approximation for the fundamental reference voltage from ten V~3 to be applied in the next sampling interval,since the average x-y components are set to zero.Furthermore,a duty ratio optimization(DRO)is proposed to reduce the tracking errors and to improve the steady state performance.Secondly,compared with conventional 2L-VSIs,great potentials achieved by multi-level(ML)-VSIs.The ML-VSIs present higher efficiency,smaller filter size,and lower stresses on the power electronic switches compared with the conventional 2L-VSI structure.Therefore,simplified FCS-MPC techniques are proposed for five-phase three-level(3L)neutral point clamped(NPC)drives,utilizing the V~3s concept.The V~3s with zero-averaged(x-y)components are designed using all eligible voltage vectors for improving the steady state performance.Furthermore,the DB principles are used to select the most appropriate candidate V~3s from the control-set during every control interval.Moreover,the neutral point(NP)voltage fluctuation is regulated using two different techniques namely,a hysteresis regulating function or proper classification and selection of candidate vector according to their effects on the capacitor voltages.Thirdly,the V~3s are always designed offline,which is not considered as an optimized solution.Therefore,an improved double-vector-based FCS-MPC(FCS-MPC-2V)scheme is proposed for five-phase 2L PMSM drives.The proposed FCS-MPC-2V scheme remedies the limitations of the conventional V~3s using online selection of the two vectors combination and duty ratio calculation.To keep low computational burden,ten possible double-vector combinations are examined during every sampling interval for each fundamentalα-βsector,which is defined using the DB calculations.However,the computational burden of proposed FCS-MPC-2V scheme implementation is relatively high due to the online optimization of the duty ratio and the third harmonic current predictions.To deal with the conflict between the high performance and the computational burden,a generalized low-complexity DB solution is proposed for MPC schemes of five-phase PMSM drive,utilizing the direct selection method of basic voltage vectors.In the proposed solution,five candidate vectors are defined with their duties during every sample interval.The best candidate vector(s)are selected to be modulated in the next sampling interval according to the highest duty cycle(s).Hence,six different MPC modulations can be obtained from this solution.The proposed DB solution improves the flexibility of the existing MPC schemes.Compared with the conventional MPC schemes,the salient features of the proposed DB solution are summarized as:low computational burden,improved steady state performance,high DC-bus voltage utilization,simple implementation,and generalized concept.Finally,MPC schemes are proposed to control other multi-objective drive systems using direct-vector-selection strategy.Two different case studies are investigated namely,the three-phase and five-phase open-windings(OW)-PMSM drive systems,when they were controlled using low-complexity MPC.These techniques achieve the system control objectives with the minimum possible complexity.