Research On A State Estimation Based On The Maximum Observable Collection With Identifing Bad Data

The distributed state estimation has been widely used with the expansion scale of power system and the grid interconnection. Because it can be able to solve the state estimation problem for the high-dimensional power system by transforming it into a series of smaller dimension state estimation problem, and it can realize parallel computing. The research object of traditional distributed state estimation is the interconnected system, and each administrative area will be regard as a partition. In this case, when this partition is not observable, this partition can not do state estimation calculation even though the known measurement is enough to calculate partial node status. What is more, the traditional distributed state estimation focus on the question of how to coordinate the regional, so the effect for bad data identification is relatively weak. The existing methods for identification are difficult to overcome “residual pollute” and “residual submerge” phenomenon.This article mainly aims at the above two problems of calculate the maximum observable collection and data method identification. The concrete contents are as follows:Firstly, this paper put forward a coefficient matrix method that can solve all observable branches and observable nodes, that is the algorithm to calculate the maximum observable collection. This method is able to accurately calculate the status or power flow of all observable nodes when this partition is not fully observable. On the basis of this collection, the author puts forward the concept of node correlative degree. Combining with the Dempster-shafer theroy, this paper offers a division principle which can overcome “residual pollute” and “residual submerge” phenomenon. In this case, the object of study that is the whole grid system at the first place transform into a partition of the traditional state estimation.Secondly, this paper establishes a calculation model of division state estimation and a multi-agent system framework based on blackboard model. On the basis of the models and aiming at the relatively weak effect of the identification, a bad data identification method is presented based on the linear relationship between measurements, which can overcome “residual pollute” and “residual submerge” phenomenon.Finally, a bad data identification method based on association rules. At first, a big data storage method is given. Then forecast the node power injection according to the regression analysis method, and establish the correlation analysis model according to the grey correlation method to find the node power injection which is sensitive to temperature changes. Complete the identification of power injection combining the association rules and the special profile correction. In this case, complete the identification of branch power flow according to the Kirchhoff’s law and the residual error identification method.