Network frequency variations cause a dramatic performance decay in repetitive controller-based shunt active power filters. This problem may be solved by adapting the sampling period in order to keep the ratio between the network period and the sampling period at a constant value. However, these changes may yield closed-loop instability. The introduction of a precompensator that forces the plant to remain invariant despite sampling rate changes allows the use of standard LTI methods in control design and stability analysis as well. Moreover, in order to improve robustness in the face of network frequency estimation uncertainty and sampling time quantization, the regular repetitive controller is replaced by a high order one. Experimental results show the validity of the proposal.