A large number of coils are able to provide enhanced signal-to-noise ratio and improve imaging performance in parallel imaging. As the increasingly grow of coil number, however, it simultaneously aggravates the drawbacks of data storage and reconstruction speed, especially in some iterative reconstructions. Coil compression addresses these issues by generating fewer virtual coils. In this work, a novel variable augmented network for invertible coil compression (VAN-ICC) is presented, which utilizes inherent reversibility of normalizing-flow-based models, for better compression and invertible recovery. VAN-ICC trains invertible network by finding an invertible and bijective function, which can map the original image to the compression image. In the experiments, both fully-sampled images and under-sampled images were used to verify the effectiveness of the model. Extensive quantitative and qualitative evaluations demonstrated that, in comparison with SCC and GCC, VAN-ICC can carry through better compression effect with equal number of virtual coils. Additionally, its performance is not susceptible to different num-ber of virtual coils.