Night flush cooling is an effective passive strategy to condition spaces and provide larger capacity to store heat. A common technique to quantify the benefit from this cooling strategy is using building energy simulation (BES) modelling, which, however, is weak in capturing the airflow patterns and convective heat transfer under the complex urban contexts. Therefore, dynamic coupling techniques of BES to computational fluid dynamics (CFD) are often used to improve the simulation performance of the night flush cooling strategies.Nonetheless, dynamic coupling techniques always encounter a high computational cost to achieve accurate CFD calculations for naturally ventilated scenarios, as high-resolution grids are required to be prepared. Therefore, a framework of integration of a fine-resolution CFD grid (CFDf), a coarse-resolution CFD grid (CFDc) and BES is proposed in this work to study night-flush cooling. In this framework, the CFDf domain performs as the off-line module that generates the boundary conditions for the CFDc domain at its openings before the start of the iterative calculations. CFDc and BES domains are then performing a fully dynamic external coupling to achieve a convergence for the energy simulation using the updated convective heat transfer coefficient (CHTC) profiles at the exterior surfaces. A case study is investigated for a simple urban morphology in Los Angeles on a typical hot day. The results highlight an improvement in representation of the neighbourhood effect in energy calculations obtained in a feasible calculation run-time.