Real-Time Data-Driven System Identification of Motor Drive Systems Using Online DMDc

Real-time parameter and state estimation is an interesting topic in motor drive applications, especially with more autonomous systems employing motor drives. With no parameter dependency, data-driven methods have become more popular. One of these methods is dynamic mode decomposition (DMD). Its controllable variant, DMDc, and online variant, online-DMDc, are of interest in this paper. DMD and DMDc require large datasets for training, but ODMDc requires smaller datasets to be initialized. The comparisons for these algorithms are made by looking into MATLAB/Simulink simulations and experiments. It is shown that DMD and DMDc are powerful algorithms but online implementation is comparable to the offline versions in terms of estimation performance.