Power system state estimation: ANN application to bad data detection and identification

Abstract

The state estimation problem in power systems consists of four distinct basic operations: hypothesis structure; estimation; detection; identification. This paper solves this problem based on a proposed artificial neural network (ANN) scheme. The state estimation/bad data detection and identification (SE/BDDI) process is conducted via a two-stages (cascaded) neural network. The first stage is devoted to the estimation of the system states, using the raw measurements and network information. The second stage projects the estimated state vector, resulting from the first stage, onto the set of measurements that originates the estimated state vector. The neural computing is followed by a bad data detection block that detects and identifies the presence of bad data, if any. Bad data replacement is also suggested to enhance the state estimator reliability. Theoretical results are illustrated by means of a simple power network example.