Online topology determination and bad data suppression in power system operation using artificial neural networks

Abstract

The correct assessment of network topology and system operating state in the presence of corrupted data is one of the most challenging problems during real-time power system monitoring, particularly when both topological (branch or bus misconfigurations) and analogical errors are considered. This paper proposes a new method that is capable of distinguishing between topological and analogical errors, and also of identifying which are the misconfigured elements or the bad measurements. The method explores the discrimination capability of the normalized innovations, which are used as input variables to an artificial neural network whose output is the identified anomaly. Data projection techniques are also employed to visualize and confirm the discrimination capability of the normalized innovations. The method is tested using the IEEE 118-bus test system and a configuration of a Brazilian utility.