Abstract
LCL–LC filters attenuate switching-frequency harmonics, but distorted grid voltages still induce low-order harmonic currents. Combining repetitive compensation with a higher-order filter also requires an implementable inverse and a stability analysis that retains the controller’s internal dynamics. This paper develops an odd-harmonic repetitive control (OHRC) and extended state observer (ESO)-based active disturbance estimation–compensation controller (OHRC-ESO-ADECC) for this problem. Its contribution is a discrete realization in which the predictor, correction, and voltage-command sequence defines the internal compensation channel. Half-period memory permits causal finite approximation of its nonminimum-phase inverse, while complete-mode analysis relates inverse length to the limits of closed-loop decay. At 5 kHz, averaged-model comparisons use the same harmonic-admittance objective, stability test, and noise budget. At the original benchmark budget, the selected five-term inverse reduces the objective from 19.071 to 12.550 mA/V relative to the selected phase-lead compensator, a 34.2% reduction, with higher calculated noise gain. Tighter budgets can instead favor phase lead. Hardware records compare four controller architectures during grid-voltage and current-reference steps; recovery rankings depend on the evaluation tolerance. The results provide a constraint-based selection procedure that balances harmonic-current suppression, recovery, noise amplification, and implementation cost rather than an unconditional preference for a longer inverse.