With the rapid development of power system, it is necessary for the grid to analyze and identify the massive, heterogeneous, polymorphic data. The efficiency of the traditional identification method is low, and it is not able to use all known data information efficiently. In order to solve this problem, a new bad data identification method based on load forecasting and the correction of association rule is proposed. At first, a three dimensional matrix data storage method is put forward according to the intrinsic relationship between the data. The forecast model of the regression analysis and the correlation analysis model based on grey correlation are established to analyze the relationship between the node power injection and the temperature. The association rules and special profile correction method are used to modify the predicted value, and then the identification of the injection power is completed. On this basis, the identification of the branch power flower is completed according to the Kirchhoff's law and the residual error identification method. At last, the simulation results of the practical system prove that the method proposed can overcome the “residual pollute” and “residual submerge” phenomenon to identify all the bad data accurately. This work is supported by National Natural Science Foundation of China (No. 61374098) and Research Fund for the Doctoral Program of Higher Education of China (No. 20131333110017).