Active disturbance rejection learning control for inverter systems with non-repetitive features

Error is the main source to generate the control input and directly decides the effect of the current time control. In this paper, the error effective value is utilized to describe the variation of error energy. According to the preset threshold of error variation intensity, the non-repetitive information is selected/eliminated to improve the quality of information supplied for the learning controller. Learning controller is implemented by means of a unit positive feedback component with only one 2N-orders low-pass filter, where convergence condition along the period domain is analyzed, and the filter cut-off frequency can be obtained based on the frequency of inverter fundamental wave and each harmonics. The state influence caused by non-repetitive disturbance information is estimated by an extended state observer and rejected by the equivalent control input in advance, and negative effects from non-repetitiveness is avoided for the learning process. Matlab simulation shows that a better tracking performance can be obtained when suppressing short-time disturbance by using the proposed algorithm, and inverter stability is enhanced. © 2018, Editorial Department of Control Theory & Applications South China University of Technology. All right reserved.