Abstract: Cellular networks in 5G are tending to be heterogeneous and ultra-dense. Simulation methods used in a traditional hexagonal grid model are too idealized and inaccurate, and they are no more adaptive to heterogeneous cellular networks today. To settle optimal deployment of base stations, a popular approach for analyzing heterogeneous cellular networks is Poisson point process (PPP) based on stochastic geometry, which assumes that the locations of base stations are completely space random. According to the model of PPP, a lower bound of the probability of coverage is gotten. However, the base stations or users may be clustered at specific areas (e.g. cell edge and hot spot areas) so that the space distribution based on PPP will be inaccurate. In order to settle this problem, Poisson clustered process (PCP) is used to model deployment and planning of the base stations in three-layer heterogeneous cellular networks including macrocells, picocells and femtocells. First of all, a model of the locations of base stations based on PCP was proposed and the formation of clusters of the base stations was discussed. Second, based on the analysis of aggregated interference received by users, the cell-association mechanism of instantaneous signal to interference plus noise ratio (SINR) was adopted, the model of outage probability using SINR was derived and three special cases of the model were given. Finally, according to the model, the difference of outage probability between using PCP and PPP was compared by simulation, and the curve of outage probability is changing as the SINR threshold changes. It shows that using PCP can get a lower outage probability than using PPP.