Deep ANN-based RF-MRAS for Speed-Sensorless Indirect Field Oriented Control of Induction Motor Drive for Electric Vehicles

This paper applies a deep neural network to rotor flux-based field-oriented control of an induction motor drive for electric vehicles. The proposed technique uses a deep neural network in a model reference adaptive system (MRAS) to estimate rotor speed. The proposed MRAS consists of a reference and adaptive model with a learning mechanism. The proposed MRAS with a deep neural network allows an accurate approximation of non-linearity and increases the system's response time. The performance of the proposed approach is validated on different scenarios, namely, vehicle acceleration conditions, constant speed, uneven road surface conditions, and deceleration modes. The dynamic performance is simulated in MATLAB/Simulink. The experimental results demonstrate the proposed method's advantages over conventional PI-based approaches for controlling sensorless induction drives for electric vehicles.