The power grid mandates that wind farms must have a certain level of reactive support capability. To effectively utilize this capability

it is necessary to quantitatively assess it to prevent voltage limit violations due to insufficient capability or resource wastage due to excess capability. However

accurately assessing this capability is challenging due to the variability of wind resources and complex turbine operations. To address this issue

the paper first provides a reasonable quantification of the reactive support capability at two assessment levels: theoretical support capability and actual support demand. It then proposes an online assessment method based on the Koopman operator. During offline training

this method linearizes complex nonlinear structures into Koopman operators in high- dimensional space. In the online phase

real-time data optimizes these assessments. Finally

a case study analysis of grid-connected doubly-fed wind farm is conducted to evaluate the reactive support capability under different control schemes and wind speed conditions. The simulation shows that the proposed method can accurately and quickly quantify assessments under various complex conditions

and the evaluation results can further enhance voltage support at both the wind unit level and the station level.