To study different wind demand exceeding probability assessment of the membrane roof, Chongqing was taken as an example. Taking PEER performance-based seismic design method as reference and considering the randomness of wind field, the probability distribution characteristics of monthly maximum wind speed samples in recent ten years were calculated, and the designed maximum wind speed risk curve was obtained by polynomial fitting. The horizontal wind speed spectrum was represented by Davenport spectrum. The nonlinear characteristics of membrane roof structure were considered, and the relationship between resistance requirements (vertical displacement limit) and the average wind speed spectrum values was approximately represented by a power function. The expression of the mean annual frequency of exceedance for displacement demand was derived. A flat membrane roof was taken as an example, 30 pulsating wind speed time intervals were generated with the AR method, and structural response analysis was carried out. Through calculation, the transcendental probability and recurrence interval corresponding to different performance levels of the roof were obtained. The results show that polynomial fitting has a better effect on the derivation of the expression of annual average exceeding probability. Performance-based wind resistance probabilistic evaluation method for membrane roofs can obtain the transcendental probabilistic evaluation values of different displacement requirements of structures through a small number of structural analyses.