Wi-Fi devices are ubiquitous, thus they have been extensively studied to understand, for example, the impact of different channel conditions and network properties over network performance. However, improving network performance without considering energy consumption can lead to critical issues: battery depletion, higher costs, and increased latency. Existing works provide algorithms and techniques for more efficient use of energy for Wi-Fi communication, especially in the case of IoT networks, limited by battery capacity. But the evergrowing number of Wi-Fi devices along with the increase in traffic and heterogeneity of current networks make measuring the energy footprint of Wi-Fi communication particularly complex, especially at a large scale. Existing simulation models to study the energy consumption of Wi-Fi devices either suffer from scalability issues due to their fine granularity, or lack realism hindering their usage in practice. In this paper, we propose a power model tackling these scalability and accuracy issues through the use of a flow-based simulation model. By comparing the accuracy and performance of our model to state-of-the-art solution, we show that our approach achieves accurate energy predictions on largescale and heterogeneous network infrastructures. Our flow-level model allows us to simulate the energy consumption of 800 nodes in a few seconds compared to more fine-grained simulators such as ns-3 that require more than 8 hours under the same scenario, with similar accuracy.