In order to solve the time-consuming problem of multi-source heterogeneous power grid data integration, a data integration model based on decision rough set is proposed. Through vertical fusion and horizontal fusion, the fusion processing of multi-source heterogeneous power grid data is completed, and then the parallel forward maximum matching redundancy removal algorithm is applied to delete the redundant information in the fusion data; Based on decision rough set, the attribute reduction of power grid data is carried out to remove the noise condition attributes with interference, and then the fuzzy classification rules are extracted from the reduction decision table to realize the classification and integration management of power grid data. The integration data repair scheme based on data correlation is created to complete the design of data integration model. The experimental results show that the proposed model is applied to the integration of 43 700 multi-source heterogeneous power grid data, and the required data integration time is 5.9 s, which meets the real-time requirements.