A New Paradigm in AC Drive Control: Data-Driven Control by Learning Through the High-Efficiency Data Set—Generalizations and Applications to a PMSM Drive Control System

Abstract

This paper presents a new means to control the processes involving energy conversion. Electric machines fed by electronic converters provide a useful power defined by the inner product of two generalized energetic variables: effort and flow. The novelty in this paper is controlling the desired energetic variables by a Data-Driven Control (DDC) law, which comprises the effort and flow and the corresponding process control. The same desired useful power might be obtained with different controls at different efficiencies. Solving the regularization problem is based on building a knowledge database that contains the maximum efficiency points. Knowing a reasonable number of optimal efficiency operation points, an interpolation Radial Base Function (RBF) control was built. The RBF algorithm can be found by training and testing the optimal controls for any admissible operation points of the process. The control scheme developed for Permanent Magnet Synchronous Motor (PMSM) has an inner DDC loop that performs converter control based on measured speed and demanded torque by the outer loop, which handles the speed. A comparison of the DDC with the Model Predictive Control (MPC) of the PMSM highlights the advantages of the new control method: the method is free from the process nature and guarantees higher efficiency.

Keywords

• data-driven control (DDC)
• model predictive control (MPC)
• permanent magnet synchronous machine (PMSM)
• searching algorithm
• radial basis function (RBF) neural networks (NN)
• interpolation technique