ANN-based Control of Three-Phase Open-End Winding Induction Motor Drive for EV Applications

Abstract

Rare-earth permanent magnet synchronous motors (PMSMs) are preferred for electric vehicles (EVs) due to their high power density and efficiency. However, the limited availability of rare-earth magnets, increasing demand for EVs, and high power rating EVs make these motors uneconomical. Therefore, the EV industry is looking for alternative motor drives that do not use rear earth magnets. Simple construction, low cost, and well-established control algorithms make the induction motor an economical solution to this problem. However, it is necessary to improve torque density, efficiency, and torque ripple to make them attractive for EVs. This thesis addresses these issues by developing an all-wheel-drive (AWD) for EVs based on a dual open-end winding induction motor (OEWIM) powertrain. An artificial neural network (ANN)-based direct torque control (DTC) algorithm is developed for the above application to achieve improved performance.

Induction motor torque density can be improved by using higher motor voltages, which reduces current and allows for smaller diameter windings. However, higher voltages are not desirable for safety reasons in EVs. Therefore, this thesis proposes an open-end winding(OEW)configurationwheretheDCbusvoltagerequirementisreducedby50%. The OEWIM has 19 distinct switching vectors equivalent to three-level inverters. An ANN-based DTC algorithm is proposed with 40 levels in torque error and 20 levels in f lux error for every degree position of the flux vector. This improves torque ripple, flux ripple and narrows variations in switching frequency. Performance is further improved by developing an ANN-based duty-DTC algorithm, where the voltage vectors and their duty ratios are selected simultaneously. The ANN structure is simplified, and training effort is reduced by exploiting the symmetrical features of switching vectors, where the ANN is trained for only one sector of 𝜋/3radians. A novel duty ratio calculation method is proposed, compensating for the effect of motor speed on torque.

The above algorithms are applied to an AWD dual OEWIM powertrain, resulting in reduced torque ripple, flux ripple, and steady-state errors. This powertrain offers several advantages, such as fault-tolerant operation, battery mass distribution, and improved power management. A laboratory prototype of the proposed powertrain is designed and fabricated. The above algorithms are verified through computer simulations and experimental validation using federal test procedure (FTP-75) drive test cycles. The simulation and experimental results are presented.

Date of Award	8 Dec 2024
Original language	American English
Supervisor	BALANTHI BEIG (Supervisor)