Neural network-enhanced space vector PWM Control of permanent magnet BLDC motors in compact electric vehicles

Abstract

This study investigates the integration of Neural Network-Enhanced Space Vector Pulse Width Modulation (SVPWM) control for permanent magnet Brushless DC (BLDC) motors in compact electric vehicles (EVs) to address efficiency losses, torque ripple, and harmonic distortion issues observed with conventional motor control methods. The primary objective was to design, implement, and validate a neural network-based SVPWM control system capable of optimizing motor performance across varying load and speed conditions. The experimental setup included a 48V BLDC motor, a TMS320F28379D DSP microcontroller, and an IGBT-based inverter circuit. Neural networks were trained using MATLAB/Simulink, employing real-time sensor feedback for dynamic adjustments in SVPWM parameters. Results revealed significant improvements in motor performance, with efficiency values reaching 85-95%, torque ripple reduced to 0.5-1.0%, and Total Harmonic Distortion (THD) minimized to 2-5%, outperforming both Conventional SVPWM and PID Control methods. Statistical analysis confirmed these improvements as significant (p < 0.05). The findings underscore the potential of neural networks in enhancing real-time motor control precision while overcoming computational constraints. Practical recommendations include optimized DSP integration, hybrid control approaches, and standardized sensor protocols. This research bridges the gap between theoretical advancements and real-world applications, contributing to sustainable and efficient compact EV technologies.