With the high evolution of society and the continuous change in the appearance of of people's living standards, the Person Re-Identification (Re-ID) model has begun to be commonly employed in safety deployment. However, in the process of re-identification, different weather conditions will affect the robustness of the model, for example, foggy weather will cause many person details to be lost, which brings great challenges to person matching. To learn the invariant representation of foggy query in the re-identification task, we proposed a plug-and-play Adaptive Recognition of Fog Attributes (ARFA), which inserts this module from the adaptive recognition fog attributes in the person re-identification model, so that even for the query in adverse weather, the matching target can still be retrieved in the gallery. Experimental results on multiple widely used benchmark datasets, such as market-1501, dukemtmc-reid, and person x, reveal that our ARFA module carries out advanced performance and also shows superiority in unsupervised domain adaptation.