Conundrum of fault detection in active hybrid AC–DC distribution networks

Abstract

Fault detection in hybrid AC–DC distribution networks is a challenging problem due to various sources of uncertainty and high degrees of complexity. A few well-known sources that instil uncertainty in the system are stochasticity of energy injected by distributed energy resources, noisy or corrupt data, heterogeneity of agents, problems with the automated mapping of equipment connectivity, and partial knowledge of the system. This study presents a distinctive approach that draws upon the use of Bayesian belief network to overcome uncertainties. The key advantage of Bayesian inference methodology is its capability to leverage both causal and correlational data in formulating a plausible conclusion. The proposed method uses state variables produced by distributed state estimation along with data collected from self-aware agents as the main sources of causal information. The rationale for using state estimation is its capability to overarch heterogeneity of AC and DC agents. It is shown that probabilistic graphical models can be employed successfully to detect faults in active hybrid distribution networks. An augmented version of IEEE 13-bus network is utilised to simulate and verify the suitability and effectiveness of the proposed technique.