On Rejection of Periodical Disturbances Based on Disturbance Learning and Extended State Observer

In this paper, a novel addition is proposed to the extended state observer (ESO) for a class of systems that subjects to periodic disturbance. It is shown that given the repeating nature of the external disturbance, it can be reconstructed via disturbance learning (DL) during the initial periods of operation and cancelled timely in the subsequent periods. The ESO assisted with DL, will only need to estimate any residue of the disturbance not fully cancelled with DL. Simulation results show that after a few iterations, the external disturbance is almost completely cancelled by DL and the effectiveness of disturbance cancellation is greatly improved compared to the standard active disturbance rejection control (ADRC). Such combination of ADRC and DL enables the controller to "remember" the past disturbance information and to reuse it, which is an ability not present in current ADRC framework.