Detection of Power Quality Disturbances Based on Residual Analysis Using Kalman Filter Based on Maximum Likelihood

Abstract

This paper presents the estimated residuals for the detection of power quality (PQ) disturbances using Kalman filter (KF) based on the maximum likelihood (KF-ML), which uses the ML method to adaptively optimize the error covariance matrices and the initial conditions as the parameters. Aiming at the sensitiveness to noise, residuals between the observed values and the estimated values by the KF-ML are proposed to detect the disturbances, and the estimated residuals exhibit mutation at the starting point and the ending point of disturbances. Thus, the singularities of residuals can be used for exactly detecting disturbances. Simulations on a variety of disturbances, such as voltage sag, impulse interruption, and harmonics with sag, are performed in the presence of noise. Simulation results verify that the detection method based on residuals can exactly determine the starting and ending time of the disturbances. Also, this method is highly sensitive to a variety of disturbances and less sensitive to random noise. Therefore, the method is a better choice for disturbance detection and is especially appropriate for estimation of unknown measurement noise in real applications.