PERFORMANCE ENHANCEMENT OF PERMANENT MAGNET SYNCHRONOUS MOTOR (PMSM) FOR ELECTRIC VEHICLE APPLICATIONS

Abstract

The electrification of transportation is accelerating globally as a response to fossil fuel depletion and environmental degradation. At the core of modern electric vehicles (EVs), Permanent Magnet Synchronous Motors (PMSMs) offer high torque density, superior efficiency, and compactness. Despite these advantages, their nonlinear dynamics, sensitivity to parameter variations, and susceptibility to torque and current ripple pose serious challenges under dynamic EV driving conditions. This thesis presents a detailed investigation into intelligent current control strategies aimed at enhancing the performance and robustness of PMSM drives for EV propulsion. Three advanced control strategies are developed and validated: Adaptive Neuro-Fuzzy Inference System-based Hysteresis Current Control (ANFIS-HCC), Adaptive Neuro-Fuzzy Model Predictive Current Control (ANFIS-MPCC), and Adaptive Neuro-Fuzzy Sliding Mode Current Control (ANFIS-SMCC). These hybrid frameworks integrate the adaptive learning capabilities of ANFIS with classical control structures to achieve real-time performance improvements in ripple suppression, current tracking, and energy efficiency. The ANFIS-HCC controller adaptively adjusts the hysteresis band based on motor speed and current error using a fuzzy rule base, leading to improved switching decisions and minimized torque ripple. Experimental results on a 3 kW PMSM prototype show that ANFIS-HCC reduces torque ripple from 1.8 Nm (conventional HCC) to 1.0 Nm and current ripple from 1.1 A to 0.7 A, with a 12 % improvement in energy efficiency under urban and highway drive cycles. ANFIS-MPCC combines the model-based prediction of current trajectories with ANFIS-tuned weighting factors for the cost function. It enhances steady-state performance and switching efficiency, especially during high-speed operations. Simulation and hardware-in-loop results reveal that torque ripple is reduced to 0.68 Nm and current ripple to 0.5 A, achieving 25–30% better dynamic v response than classical MPC while reducing computation time through precomputed ANFIS lookup tables. The ANFIS-SMCC strategy merges the robustness of sliding mode control with adaptive neuro-fuzzy tuning of switching gains. It incorporates a saturation-based switching law to mitigate chattering while ensuring finite-time convergence of current tracking errors. Experimental validation on the same PMSM platform under load steps and temperature variations demonstrates torque ripple suppression to 0.5 Nm and current ripple below 0.5 A, outperforming both ANFIS-HCC and ANFIS-MPCC. Under sensor noise (σ = 0.05), temperature drift (+30% Rs), and inverter nonlinearities, ANFIS-SMCC showed only 10–15% performance degradation, compared to 30–70% for other methods. All proposed controllers were implemented on a dSPACE DS1104 real-time hardware platform, interfaced with a laboratory-grade PMSM, Voltage Source Inverter, and high-precision sensors. Real-time execution met the 5 μs sampling constraint, with ANFIS-SMCC operating within 120 μs per cycle. Comprehensive testing using IM240 and Modified Indian Drive Cycles(MIDC) revealed that ANFIS-SMCC provided the best trade-off between ripple suppression, computational efficiency, and robustness, while ANFIS-MPCC delivered superior predictive accuracy, and ANFIS-HCC offered simplicity and rapid convergence. In conclusion, this thesis presents a unified framework of intelligent PMSM current controllers tailored for electric vehicle propulsion, with thorough validation through simulation and real-time experimentation. The proposed ANFIS-based hybrid strategies yield substantial improvements in torque smoothness, thermal stability, control precision, and overall motor efficiency, contributing to the performance, reliability, and energy economy of EVs.