Identification of bad measurements in power system from historical measurement data

Abstract

Bad measurements identification is an important problem for system monitoring and controlling in electric power system. In recent years, some studies have shown that the bad measurements can be corrected by some state estimation methods. However, most of the studies mainly focus on a single measurement snapshot but ignoring the possibility of the continuous measurement errors existence in historical measurement data. In this paper, we propose an algorithm for identification of bad measurements from historical measurements. And considering the different occurrence pattern of bad measurements, we propose a more common situation of bad measurements which is a continuous error in multi-snapshots of measurements rather than a single one. In our method, we use a least-square optimization model as the basic estimator to get the initial results of measurement residuals. After that we set a searching strategy based on hypothesis test and normalized measurement residuals to find a continuous errors in historical measurement data. Then we update the computation of the basic estimator to correct the biases in state variables caused by bad measurements, which lead to more precise results of system states. The performance of the developed identification algorithm is demonstrated and compared with conventional estimation methods based on experiments using IEEE 30-bus systems.