Roughly 85% of the matter in the Universe is in the form of invisible matter, i.e. dark matter. Under the gravitational interaction, dark matter clusters and forms hierarchical structure. Studying the evolution of dark matter substructure, i.e. subhalos, and predicting their abundance and spatial distribution in the host halo are extremely important for understanding the nature of dark matter. A recent study by Daniel et al. 2019 shows that with 8 quadruple-image strong gravitational lenses, it is possible to put constraints on the abundance of subhalos in the mass range of 10^6~10^9 solar mass. However, simulating the evolution of subhalos has been computationally challenging and may suffer from artificial disruption if the resolution is not sufficiently high (van den Bosch et al. 2018). In this talk, I will introduce another approach to tackle this challenge, the semi-analytic model. With analytic and empirical models for the physical processes such as dynamical friction, tidal stripping and tidal heating, it has been shown previously that semi-analytic models can give statistically consistent results with cosmological simulations. By performing idealized simulations, we look at more details of the evolution of individual subhalo. I will show a few improvements we have made in modeling the orbit evolution and mass loss of subhalos. I will also show how our models can be applied to the future study on substructure gravitational lensing.