[Objective] Short-term intense fluctuations in wind gusts at hub height(90 m) present critical challenges to the efficient grid integration and safety operation of wind farms. However

a significant gap exists in the hubheight gust forecasting because the conventional gust models based on the surface(approximately 10 m) wind observations exhibit considerable uncertainty in predicting hub-height gusts due to the complex vertical structure which modifies the gust major characteristics across the lower atmospheric boundary layer. To address this limitation

this study systematically evaluates and optimizes multiple gust models

aiming to establish a robust methodological reference for accurate hub-height gust prediction. [Methods] To overcome the lack of high-precision gust observations

the weather research and forecasting model with large eddy simulation(WRF-LES) is utilized to conduct an ultra-high-resolution meteorological numerical simulation. Operating at a second-level temporal resolution and a hectometer-level spatial scale

this simulation effectively reproduces the rapid variation of hubheight gusts at a target wind farm and also generates physically consistent three-dimensional atmospheric variables(e.g.

turbulence kinetic energy

vertical wind profile and stability parameters) for model validation. Multiple gust models are evaluated

including turbulence kinetic energy(TKE)-based models which directly quantify the turbulent intensity

and wind-profile-based models which rely on the vertical wind shear relationships. Furthermore

an improved hybrid model integrating both algorithms is subsequently proposed. The applicability

accuracy and limitations of all models are thoroughly assessed based on long-term observations of hub-height gusts across diverse atmospheric stability conditions. [Results] The results demonstrate that the conventional wind-profile-based models exhibit relatively poor performance with R2 of 0.64 and mean bias of 1.43 m/s

especially under unstable atmosphere conditions. In contrast

the TKE-based model demonstrates stable and superior performance across different atmospheric stability conditions

achieving an R2 of 0.93 and a mean bias of 0.64 m/s since the TKE captures the major features of the turbulent intensity. Compared to both individual gust models

the optimized hybrid model yields the best forecasting performance with R2 of 0.94 and mean bias of 0.12 m/s. [Conclusion] Therefore

the optimized hybrid model exhibits significant potential for operational application in hub-height gust forecasting and wind farm management. However

the hybrid model still tends to underestimate the rapid gust fluctuations during strong convection and extreme weather events

suggesting a key area for future refinement.