Control Directo de Par de un Motor Síncrono de Imanes Permanentes Aplicado en Electromovilidad Utilizando Redes Neuronales Artificiales

Resumen

This article presents a comparison between two direct torque control (DTC) strategies applied to permanent magnet synchronous motors (PMSMs), both employing space vector modulation (SVM): a conventional DTC scheme based on PI controllers (SVM-DTC), and a proposed method integrating artificial neural networks (ANN-DTC). In the conventional approach, torque and flux control are handled by PI regulators, while in the proposed architecture, the ANN replaces these controllers by directly estimate the stationary reference frame voltage components (α and β) required by the SVM block. Additionally, the network infers the rotor position implicitly, eliminating the need for a separate angle estimator. Several related works on neural-network-based DTC are reviewed to support the development of an efficient control topology. Simulations under various operating conditions—including startup and load variation—demonstrate that the proposed ANN-DTC scheme achieves a significant reduction in torque and flux ripple, along with more accurate reference tracking.