Most large-eddy simulation studies related to wind energy have been carried out either by using a fixed pressure gradient to ensure that mean wind direction is perpendicular to the wind turbine rotor disk or by forcing the flow with a geostrophic wind and timely readjusting the turbines’ orientation. This has not allowed for the study of wind farm characteristics with a time-varying wind vector. In this paper, a new time-adaptive wind turbine model for the large-eddy simulation framework is introduced. The new algorithm enables the wind turbines to dynamically realign with the incoming wind vector and self-adjust the yaw orientation with the incoming wind vector similar to real wind turbines. The performance of the new model is tested first with a neutrally stratified atmospheric flow forced with a time-varying geostrophic wind vector. A posteriori, the new model is used to further explore the interaction between a synthetic time-changing thermal atmospheric boundary layer and an embedded wind farm. Results show that there is significant potential power to be harvested during the unstable time periods at the cost of designing wind turbines capable of adapting to the enhanced variance of these periods. Stable periods provide less power but are more constant over time with an enhanced lateral shear induced by an increased change in wind direction with height.