Piecewise Linear Fuzzy Adaptive Speed Observer with Torque Error Compensation for Electric Vehicle Traction Drives

This paper presents a piecewise-linear single-input fuzzy-controller-based model reference adaptive system (SFC-MRAS) for speed estimation in speed-sensorless vector-controlled induction motor drives for electric vehicle (EV) traction applications. Conventional PI-MRAS observers employ fixed-gain adaptation, which may limit estimation performance under rapid speed variations, load disturbances, and motor parameter uncertainties. To enhance nonlinear adaptation capability while reducing computational requirements, a conventional dual-input fuzzy adaptation mechanism is transformed into a signed-distance-based single-input structure and subsequently reformulated as a piecewise-linear control law. An auxiliary electromagnetic torque-error compensation loop is also incorporated to provide a filtered speed-equivalent correction during dynamic operating conditions. The proposed SFC-MRAS is comprehensively evaluated against conventional PI-MRAS and dual-input fuzzy-controller-based MRAS (DFC-MRAS) under acceleration, load disturbances, speed transitions, forward-reverse operation, and motor parameter variations. In addition, an EV-oriented validation case derives the motor reference speed and load torque from vehicle longitudinal dynamics, incorporating acceleration, cruising, deceleration/braking, re-acceleration, road grade, and drivetrain characteristics. The comparative results demonstrate that the proposed SFC-MRAS achieves improved speed-estimation and tracking performance, faster convergence, and enhanced robustness compared with PI-MRAS, while providing competitive estimation performance with substantially lower computational burden than DFC-MRAS. The results further demonstrate that the proposed single-input piecewise-linear formulation preserves the essential nonlinear adaptation behavior while simplifying the online implementation. Overall, the proposed SFC-MRAS provides an effective balance among estimation accuracy, dynamic robustness, computational efficiency, and EV-oriented operating capability for induction-motor-based sensorless traction drives.