Using observational data from three floating wind lidar systems deployed in the northern South China Sea (including the Beibu Gulf) during the 2025 typhoon season, this study evaluates the fifth-generation global climate and weather reanalysis dataset (ERA5) reanalysis wind fields at four height levels—80 m, 120 m, 160 m, and 200 m—which are closely related to wind energy resource assessment and turbine load calculation, for three typhoon events: “Kajiki,” “Ragasa,” and “Matmo”. The actual altitude of pressure levels is determined from the ERA5 geopotential height field, and spatiotemporal matching and height adjustment are performed using the power law wind profile model prior to quantitative comparison. The results indicate that ERA5 effectively reproduces the evolution trends of typhoon wind fields, with correlation coefficients generally exceeding 0.84 and root mean square errors (RMSE) ranging from 1.4 to 4.3 m·s?1.However, a systematic underestimation of peak wind speeds exists, especially under extreme conditions where wind speeds exceed 20 m·s?1, with underestimations reaching 1–4 m·s?1.The accuracy for wind direction is generally lower than that for wind speed, with significant errors during periods of rapid directional shifts. Regarding vertical distribution, the RMSE of wind speed decreases with increasing height, with the 200 m level exhibiting smaller errors than the 80 m level. This study reveals that while ERA5 can capture the evolution of typhoon wind fields, its systematic biases introduce uncertainties in offshore wind turbine load calculation, power forecasting, and precise typhoon risk assessment. Therefore, bias correction based on observational data is recommended prior to its application in such contexts.