With the development of power grid automation technology, massive multi-source and multi-spatial-temporal data can be obtained by a data center. On this basis, multi-source measurement data can be checked reciprocally. This is conducive to later high-level big data applications. Given a situation where PMU and SCADA measurements of the same node exist at the same time, a front-end data identification framework is proposed. To overcome a lack of negative samples of data, an improved One-Class Support Vector Machine (OCSVM) identification method based on particle swarm optimization is used to identify outliers according to the difference between two-source measurements. Any data which may be misjudged because of being close to the OCSVM boundary can be corrected by a gap statistical algorithm and the bad data can be found. Then the PMU data at the time point of the bad data is checked to determine the location. With the actual power grid data of a province, the bad data identification framework based on mutual checking between PMU and SCADA is verified and analyzed. The results show that the proposed method can effectively identify front-end bad data of two sources, with less computation and consumption of time, and is more reliable than the results which use single-source data only. This work is supported by National Key Research and Development Program of China (No. 2017YFB0902900).