The innovation graph approach is effective on identification of bad data and topology change. For the radial regional power grid

this paper presents a new hierarchical identification method using innovation graph approach

and establishes innovation graph models for substation and tieline layers. It uses pseudomeasurement data to revise bad data in link line

which overcomes the lack of rebuilding trees after replacing link line in traditional method. The accuracy of forecasting vector and innovation vector is interrelated. This paper defines the variable ΔP

which can be gained by subtracting measurement data from power flow calculated in power flow checking model of automatic voltage control (AVC) system at the time k

to compensated the forecasting vector at time k+1 by the rate α. The result is valid that corrected power flow curve after compensation is closer to the true value curve and the accuracy of topology change identification is improved. The proposed method is applied online in several regional AVC systems of Sichuan Grid. It proves that the method is effective to identify the bad data and topology change.