Towards a Safer Smart Grid: Bayesian Prediction of Non Detectable Electrical Faults at the Edge

Abstract:
Distributed energy resources (DERs), while enhancing the distribution grid efficiency and stability, also challenge the current grid protection systems. In fact, a reverse power flow produced by the DERs can mask an electrical fault by feeding it. This issue can be solved by pervasively monitoring the grid and assessing its state at any given time. In this paper, we address the problem of optimizing the share of network resources required by such a monitoring system. To do that, we predict the probability that a fault would not be detected by the legacy protection system based on the near future grid state. Experimental results show that this system allows to successfully identify future risk conditions, and set the monitoring accuracy accordingly.