To deal with problems of low reuse and sharing difficulties of heterogeneous models in the process of hydrological forecasting, this paper proposes a service-based encapsulation and combination method for hydrological models. Through the design of standardized service description, operation, deployment and other encapsulation interfaces, the hydrological model's data-algorithm-module-model multi-granularity splitting and service-oriented encapsulation are realized. Taking the grid Xin’anjiang model (GXM) as an example to verify the proposed method, the experimental result on the Tunxi watershed shows that the method in this paper can effectively improve the reuse ability of model resources and realizes the dynamic configuration technology of the model driven by business requirements, and thus provides basic support for the service encapsulation and sharing application of the hydrological model driven by business needs.