The invention discloses a bad data identifying method based on load prediction, and belongs to the technical field of bad data detection and identifying based on state estimation in a power system. The method herein includes the following steps: by counting actual measurement of each physical busbar, making an algebraic sum of respective measurement associated with the physical busbar, in accordance with the KCL theory, determining whether the actual remote metering associated with respective busbar is correct, if the KCL theory is not satisfied, the measurement is incorporated into a suspected remote metering set; by resolving correlation coefficient between each suspected remote metering point in the suspected remote metering set and measurement values of other remote metering points, obtaining real bad data; for the detected bad measurement, using the KCL theory to correct the bad data. According to the invention, the method herein can detect the bad data in the measurement data of large power grids, increases the practicality and accuracy of the measurement data and meets basic requirement of state estimation.