Model Free Artificial Neural Network for an Induction Machine

This paper presents the design, training, and implementation of a Feed-Forward Neural Network based on Model-Free Predictive Torque Control (MF-PTC) to control an Induction Machine (IM) powered by a two-level Voltage Source Inverter (2L-VSI). This control uses an algorithm capable of estimating the dynamic behavior of an IM without having a detailed knowledge of the system using the past measurements of the control variables (stator currents and flux) and input variables (stator voltages). With this approach, it is possible to design a neural network that mimics the behavior of this control just by measuring the current and estimating the flux. The control topology, neural network architecture, network training, and implementation are detailed in this work. Simulation results are presented for different operating points of the machine to validate the performance of the proposed strategy.