The time-dependent origin-destination (TDOD) demand estimation problem aims at estimating dynamic demand that represents the observed traffic flow patterns in a transportation network. Errors in TDOD demand are often propagated into the network outputs causing unreliable planning and operational policies. In this study, a bi-level optimization problem is proposed where the upper level is an Ordinary Least-Squared (OLS) error minimization problem that minimizes the deviation between the estimated and observed traffic volumes from SCATS, while the lower level generates assignment proportions matrix using a mesoscopic simulation-based dynamic user equilibrium model. The interior point conjugate gradient method, as an exact gradient method, is applied to solve the TDOD demand estimation problem. The obtained results highlight the capability of the proposed approach in improving the performance of a dynamic large-scale network model of Melbourne CBD.