Extreme weather poses a problem to the operation and scheduling of a large proportion of modern energy generation systems due to its low chance of occurrence and high destructiveness. To that aim, this paper provides a day-ahead optimal dispatch model for large-scale wind power (LSWP) integration in power systems that takes cold wave weather into account, directly linking system outages and unit up/down with cross-section current fluctuations. Firstly, Wasserstein- Generative Adversarial Network WGAN is constructed to generate the output of wind farms during cold wave weather to solve the sample shortage problem; secondly, a wind-thermal power joint optimal dispatch model based on the low output of wind power is constructed to minimize the expected total system cost to cope with all the uncertainties caused by cold weather and transform the model into a mixed integer programming model; finally, the model is solved using the nonlinear transformation technique of segmental linearization. The suggested method's effectiveness is validated using an enhanced 24 node standard test system, and the study demonstrates that it successfully handles the problem of LSWP grid-connected scheduling with substantial variations during cold-wave weather.