A novel online algorithm for missing data imputation for networks with signals defined on the edges is presented in this paper. Leveraging the prior knowledge intrinsic to most real-world networks, we propose a bi-level optimization scheme that includes (i) a sparse line graph identification strategy by solving a group-Lasso-based optimization framework via composite objective mirror descent to exploit the causal dependencies among the signals and (ii) a Kalman filtering-based signal reconstruction strategy developed using simplicial complex (SC) formulation to exploit the flow conservation. To the best of our knowledge, this is the first SC-based attempt for time-varying signal imputation, whose advantages have been demonstrated through numerical experiments conducted using EPANET models of both synthetic and real water distribution networks.