Controle MTPA de máquinas de relutância chaveada com modelagem no referencial dq0 e proposta do referencial dqe baseado em autovetores

Abstract

Switched reluctance machines present attractive features for electric traction applications when compared to permanent magnet-based machines, particularly due to their simple construction, reduced dependence on rare-earth materials, and high fault tolerance. However, the strong angular dependence of phase inductances leads to significant torque ripple, posing challenges to the application of vector control strategies. Within the maximum torque per ampere (MTPA) control framework, the use of the dq0 reference frame results in approximate solutions, since the explicit angular dependence of inductances preserves magnetic coupling in this coordinate system, making the exact formulation computationally demanding for real-time implementation. Consequently, state-of-the-art techniques based on this reference frame rely on structural simplifications of the electromagnetic model. This thesis proposes two MTPA control methods based on the dq0 reference frame and a novel coordinate system, denoted as the dqe reference frame, obtained from the eigenvalues of the inductance derivative matrix. In this new reference frame, the structure of the optimization problem is reformulated in order to simplify the determination of optimal current references, enabling an analytical solution and reducing the computational burden associated with its implementation. A compensation strategy associated with the non-instantaneous current decay of phases during turn-off is also proposed, aiming to mitigate the production of negative torque by these phases. Experimental validation was carried out on a laboratory test bench under different operating conditions, comparing the proposed methods with classical drive strategies, including firing angle optimization, torque sharing functions, and vector control based on the non-sinusoidal dq transformation (dqx). The results show a reduction of up to 35% in the torque ripple factor and an increase of up to 46.3% in torque per ampere compared to reference methods, indicating viability for embedded implementation.