Résumé : Paper is a material exhibiting a complex microstructure that is composed of a network of fibres at the micro-level. When subjected to external loading or variations in moisture conditions over different time scales, changes in strain that are non-linear with respect to time are observed at the sheet level (macro-scale). In order to investigate this time-dependent behavior of paper, a creep power law model is implemented within a finite element approach at the level of single fibres. This rate-dependent model is found to capture experimental results available in literature for single fibres with a good agreement (both quantitatively and qualitatively). Based on the identified model at the level of single fibres, the time-dependent hygro-mechanical response is upscaled towards the network scale. To this end, random model networks of ribbon shaped fibres are generated and their response is simulated. The network-scale response, emerging from the rate-dependent fibre model, demonstrates the ability to predict the response of networks subjected to relaxation at a constant moisture level. The developed numerical model predicts lower values of overall stress response in single fibres as compared to networks. Also, stress relaxation predicted by the rate-dependent model in the cross-direction of the networks is in agreement with the experimental observations. Therefore, one of the remarkable findings of the present work is that the developed rate-dependent model is robust enough to capture the sheet scale response also qualitatively. Based on the study of these computational results, a better understanding is achieved regarding the influence of mechanical and rate-dependent properties of single fibres on the hygro-expansion of complete fibre networks, and in particular of paper sheets.