Bad Data Processing for Hybrid Power System State Estimation based on Block Orthogonal Methods

Recently proposed hybrid power system state estimators allow the imbedding of synchronized phasor measurements into the estimation process by using a two-stage architecture. Such a scheme preserves existing SCADA-based estimators and enhances computational efficiency by processing phasor data through a linear second-stage estimator that exhibits superior accuracy properties. The latter is achieved through the use of a block version of orthogonal Givens rotations able to properly take data correlation into account. Despite those advances, there is a clear need to deepen efforts towards equipping those new estimators with bad data processing tools. This paper is aimed at developing advanced methods for bad data detection and identification associated to that new class of hybrid estimators. It is shown that the proper partition of large networks into subnetworks provide a powerful accessory means to improve the performance of hypothesis testing procedures for gross measurement detection. The performance of the proposed procedures is evaluated through several tests conducted on three benchmark networks, namely, the IEEE 57-bus, 118-bus, and 300-bus test systems, for different gross measurements types. The results reveal significant improvements in terms of higher detection rates with respect to conventional bad data detection methods. Moreover, such gains tend to increase with system size. Those conclusions are supported by the superior success rates obtained with the proposed approach when massive statistical tests are conducted.