The invention relates to the technical field of train maintenance, and discloses a train bearing temperature processing method and device and a storage medium. By the train bearing temperature processing method and device and the storage medium, the problem that early warning cannot be performed on train hot shaft in the prior art is solved. The method comprises the steps of continuously acquiring detection records of detection stations in preset numbers from current THDS detection data; performing data pre-processing on the detection records of the detection stations in preset numbers; performing sliding window processing to obtain a sliding window accumulation value of a preset characteristic parameter according to bearing temperature rising data in pre-processed detection records and the preset characteristic parameter of the temperature rising data; obtaining a bearing state forecast value corresponding to current detection data according to the sliding window accumulation value of the preset characteristic parameter and a corresponding preset weighted value; and judging whether the bearing state forecast value is larger than or equal to an early-warning threshold or not, if yes, prompting that a potential fault exists in the bearing. The method disclosed by the embodiment of the invention is applicable to the fault detection process of the train bearing.