28th Signal Processing and Communications Applications Conference (SIU), ELECTR NETWORK, 5 - 07 October 2020
Vast and unauthorized penetration of the photovoltaic systems in distribution systems imposed inadequate monitoring challenge to system operators. Although the installation of required monitoring infrastructure does seem as a solution, due to the high installation cost, it is not considered a feasible one, at least for the near future. As an alternative approach, a two-stage method composed of an ANN-based mapping function, and a regression model is presented to detect unauthorized PV connections to the system. A gradient boosting machine is implemented to build the regression model and the corresponding confidence interval. The Monte Carlo simulations are utilized to generate synthetic training data-sets for both models. Finally, the performance of the approach is assessed with real data.