To reduce the thrust force ripple of a complementary and modular linear flux-switching permanent magnet machine, a model predictive thrust force control (MPTFC) is developed and implemented in this paper based on the active voltage vectors selection (AVVS) and two-voltage-vector synthesis (TVVS). First, the active voltage vectors are screened and assigned to the specific sector, instead of traversing all the possible vectors, and consequently, significantly decrease prediction workload. Second, for the elimination of the weighting factor in the conventional MPTFC, the control object is transformed from the thrust force and stator flux to the stator flux only by using the load angle. Third, to achieve the steady-state performance improvement, the TVVS approach is applied in each control period, where the optimal TVVS and its dwell time are simultaneously obtained by only one cost function. Both simulated and experimental results verify that the steady-state performance improvement can be realized by the developed MPTFC through comparing with normal hysteresis current control and conventional MPTFC strategies.