In recent years, stochastic geometry tools have been widely used for the deployment of base stations (BSs) in ultra-dense multi-tier heterogeneous networks. One of these tools, the independent Poisson point process (PPP), is favored due to its tractability, but in reality, the BSs are not completely independent of each other within tier and across tiers. In this study, a two-tier cellular network model including macro and pico base stations (MBSs and PBSs), is proposed. In this model, MBSs are distributed based on the PPP, while the PBSs are deployed according to the Poisson hole process. Then, the effect of the deployment and the density of BSs on energy efficiency are investigated. The minimum achievable data rate for each layer is formulated and the minimum achievable throughput for the entire cellular network is obtained. Through the optimized BS distribution, it is shown with simulations that energy efficiency can be maximized.