NEURAL SWITCHING TABLES FOR A FOUR-LEVEL MULTICELLULAR INVERTER TO ENHANCE DIRECT TORQUE CONTROL OF A PMSM
Abstract
This paper proposes neural switching tables (NCST, NRST) to improve direct torque control (DTC) of a permanent magnet synchronous motor (PMSM) driven by a four-level multicellular inverter (4LMI). Classical DTC suffers from high torque and flux ripples, which can cause mechanical stress and reduce reliability. The neural approach replaces conventional and reduced lookup tables with compact neural network models, allowing seamless integration in Simulink without large precomputed datasets. Simulation results confirm identical control behavior to that in classical tables, while reducing torque ripple by approximately 50% and maintaining flux within hysteresis limits. This solution enhances modularity, simplicity, and scalability, making it a promising step toward AI-based DTC strategies.
Fields of Study & Subjects
Sensorless Control of Electric Motors
Multilevel Inverters and Converters
Electric Motor Design and Analysis
Publication details
- Journal / source
- REVUE ROUMAINE DES SCIENCES TECHNIQUES — SÉRIE ÉLECTROTECHNIQUE ET ÉNERGÉTIQUE
- Volume
- 71
- Issue
- 1
- Pages
- 27–32
- Publication date
- 8 March 2026
- Language
- EN