Fault Detection in Active Hybrid Distribution Networks: Overcoming Uncertainty

Abstract

Fault detection in active hybrid distribution networks that contain distributed energy resources and employ both alternating current and direct current is a highly complex and challenging task. Such networks are inherently stochastic, partially observable, and suffer from noisy or corrupt data. This paper proposes a fault detection method based on Bayesian inference paradigm and employs its corresponding graphical representation, that is Bayesian Belief Network (BN), to detect faults. The BN takes advantage of causal data produced by a distributed state estimation algorithm and correlational redundant data gathered from various devices to overcome uncertainty in making plausible decisions about the status of the system. Simulation results prove the value of the proposed technique in improving the reliability of conventional protection and relaying schemes.