Safe mining of coal has an important impact on energy security, while effective control of methane hazard is the key to ensuring safe coal mining. Methane concentration is the main factor determining the hazard of methane in coal mines, and in order to limit the impact of methane on coal mine safety, this study proposes a methane concentration interval prediction method based on a hybrid “model-data” driven idea. Firstly, by analyzing the data and constructing a methane concentration prediction method based on model-driven, which reduces the influence of multicollinearity in the methane concentration series on the prediction effect, and then, in combination with the deep learning technique, a method based on the Wasserstein distance to improve the Informer model is proposed, and finally a hybrid-driven methane concentration interval prediction model is established by introducing the IOWGA operator and the statistical method. After an example analysis of a coal mine in Guizhou Province, China, the hybrid-driven model proposed in this study has better applicability and prediction accuracy in the methane concentration prediction task, which can effectively prevent the occurrence of coal mine accidents and is more in line with the needs of coal mine safety production.