A real-time filtration model is developed in this paper, and numerical simulations are presented for filtration performance of a randomly packed granular filter. Fluid flow characteristics and filtration performance at the initial state are first investigated. Correlations of initial filtration efficiency and initial pressure drop are obtained with good prediction accuracy. Particle filtration characteristics on the granule surface show that the real-time filtration model can well predict particle deposition and accumulation characteristics on the granule surface. The particle deposition fraction in the filter decreases with filter depth, and for longer filtration time, inhomogeneity appears in the axial deposition fraction. Correlations of filtration efficiency and pressure drop versus time are presented and both are of sufficient accuracy for engineering purposes. The expression of the cleaning time is obtained, which can be used to determine the optimized cleaning time for economic operation of granular fillers.