Recent applications of Wireless Sensor Network require small yet sustainable battery-powered devices. As a consequence, it becomes crucial to accurately and efficiently compute a node???s power consumption in order to estimate its lifetime. Existing wireless network simulators either implement simplistic energy consumption and battery models, or very complex and general ones that hinders scalability. In this paper, we (i) present WiSeBat, a module to estimate devices lifetime using realistic energy consumption and battery models, that has specially been optimized for wireless sensor network simulations. We then (ii) validate it through real measurements. Finally, we used it (iii) to compare wireless sensor lifetime in several realistic scenarios. Firstly, we review existing techniques to simulate a battery and discuss what behaviors are important to get realistic and fast simulations. We then propose simulator-independent models for the battery and for the energy consumption of sensors, and implement this model in the WSNet simulator. Secondly, we compare measured and simulated lifetimes of sensors. On the one hand, our experiments show that our models provide an 86 -- 96% accurate lifetime estimation. On the other hand, the previous default WSNet models overestimate lifetime by more than 2600%. Once validated, we used our approach to benchmark the energy consumption of different protocol stacks of wireless sensor networks, under different scenarios. These simulations match well-known results in simple scenarios, as we demonstrate better performance of ContikiMAC over X-MAC. They also provide an accurate comparison of sensor lifetime in more complex scenarios.