Deep Learning Observer for Position Estimation in Field-Oriented Control of Permanent Magnet Synchronous Motors (PMSM)

This paper presents a comprehensive comparison of various neural network architectures for creating a sensorless observer for rotor position estimation in permanent magnet synchronous motors (PMSMs) field-oriented control. It provides a workflow, encompassing data collection using simulation environment, implementation and training of neural networks, validating the network estimated position using simulation, and hardware deployment and testing. The architectures used include a feedforward network (FFNN), a nonlinear auto regressive network (NARX), and a long short-term memory (LSTM) network to estimate the PMSM rotor position. Simulation and experimental results demonstrate that the proposed approach provides accurate estimation of rotor speed and position.