Multi-channel active noise control (MCANC) systems have been widely investigated for acoustic noise attenuation over a spatial region. In MCANC systems, incorporating more reference microphones (RMs) can boost noise reduction performance but increases computational complexity, which may affect real-time processing capability. To improve computational efficiency, we propose a joint approach for RM selection and filter order (FO) determination in MCANC systems. Under the specified performance constraint, the optimal RM subset is selected by promoting the group sparsity in filter coefficients, and the optimal FO is determined by encouraging the consecutive highest-order filter coefficients to approach zero. By combining the above two criteria, we can simultaneously select the RM subset and determine the FO. In addition, we present a re-weighting strategy to further reduce both the number of selected RMs and the FO value. Numerical simulations confirm the validity of the proposed method.