Optimization of an Ann-based speed and position estimator for an FOC-controlled PMSM using genetic algorithm

• Main Author: Quismundo, Juan Paolo B.
• Format: text
• Language: English
• Published: Animo Repository 2022
• Subjects: genetic algorithm, speed and position estimator, field-oriented control, permanent magnet synchronous motor, optimization; Computational Engineering; Controls and Control Theory; Electrical and Electronics; Signal Processing
• Online Access: https://animorepository.dlsu.edu.ph/etdm_ece/17; https://animorepository.dlsu.edu.ph/context/etdm_ece/article/1017/viewcontent/2022_Quismundo_CompleteVersionETD.pdf
• Institution: De La Salle University

This study develops a neural network-based estimator for the speed and position of a field-oriented-controlled permanent magnet synchronous motor optimized using a genetic algorithm. An estimator based on a neural network provides an alternative to conventional methods that require accurate information on the motor parameters. Genetic Algorithm provides an avenue to optimize the hyperparameters for optimal performance. A training dataset is obtained from the motor operating points consisting of the alpha- beta voltages and currents with the sin and cosine of the rotor position as the targets. A genetic algorithm was used to determine the optimal hyperparameters for the network’s batch size, the training algorithm parameters, and the number of hidden layers and its respective number of neurons. In this study, the genetic algorithm developed was able to optimize the hyperparameters for the neural network to achieve a high accuracy over the operating range. The neural network-based estimator can estimate the speed and position of the PMSM required in executing the field-oriented control scheme. The optimized neural network proved to have more accurate estimations than conventional methods such as the SMO and MRAS as well as other neural network estimators during steady-state and dynamic conditions, including when qualified using a UAV Flight Plan. The efficiency of the proposed estimator proved to be relatively higher than the conventional estimators but still fall short of the efficiency when using sensors.