Bad Data Detection and Identification Using Artificial Neural Networks

Abstract

This paper explores the capability of the artificial neural networks to detect and identify the single and multiple bad data which can be found within the measurements provided to energy management system centers. A reduced model for state estimation was developed in the MATLAB environment using NN-TOOL. A comparison of the single bad data detection and identification between the proposed state estimator and the Weighted Least Squares state estimator on IEEE 14-bus power systems is provided. The results show that the proposed model is more accurate and faster than the WLS state estimator. Furthermore, the proposed methodology is a great alternative to the conventional techniques and is therefore well suited for smart grid applications.

This research presents a different approach for handling bad data compared to the iterative and lengthy statistical methods used previously. It provides a streamlined model for the state estimator that allows for effective and reliable monitoring and operation of the power system with lower computational requirements and the implementation of this procedure in real time, which positively impacts the overall performance of the energy management system. Its main objective is to enhance the functioning of the electric power management system by utilizing artificial neural networks.

Typically, a state estimation is performed prior to the process of detecting, identifying, and excluding bad data, where the traditional state estimator relies on a set of redundant measurements to describe the system through a set of over-specified nonlinear equations, followed by a series of iterative numerical mathematical operations aimed at minimizing measurement errors until the optimal value of the system state variables is achieved. The theory of weighted least squares is one of the most commonly used methods for this purpose.