Optimal Design of Experiments for Identification of PMSM Thermal Models

Abstract

Precise thermal models of permanent magnet synchronous motors (PMSMs) are required for several purposes, like mechanical design, thermal management, drive control, and integration into electric vehicles. Identifying a precise thermal model based on measurements is typically a time-consuming and complex task, and the result depends considerably on the measurements available. To simplify this task, the current article develops a method for optimally designing an experimental cycle in favor of parameter identification. The aims are to maximize the information contained in the measurements, to make the parameter estimation accurate and robust, and to reduce the total cycle time. The method uses a tailored optimization problem and an existing gray-box model of the PMSM to find optimum input trajectories for the experiment. The objective function considers the A-optimality criterion based on the Fisher information matrix. The method is validated based on both computer simulations and test bench measurements. The obtained results demonstrate that the cycle designed using the proposed method leads to reduced parameter variance in simulation and improved model accuracy when evaluated on test data from experiments.