Restructuring of distribution systems through clustering of voltage sensitivity to PV presence

Distributed photovoltaic micro and mini-generators play a prominent role as they become increasingly accessible to end consumers. They promise to reduce expenses related to electrical energy and potential financial returns. One way to improve the application and planning of the distribution network and its hosting capacity is partitioning these systems, using the impact of distributed generation as a criterion. Therefore, this work aims to construct a generic network partitioning method using the Spectral Clustering algorithm adapted with a balancing algorithm to guarantee a certain similarity in the output cluster size. The input data for this method are obtained from modeling actual standardized electrical systems and the simulation of photovoltaic generation individual impact on each low-voltage feeder connected to the primary grid. The voltage sensitivity coefficient is a metric that quantifies the voltage change at one node with the variation of active power at another one. Considering all simulation states, the ratios of these coefficients per node are the Sensitivity matrix elements. By applying Balanced Spectral Clustering to this Sensitivity matrix, the method can provide continuous electrical regions that highlight the optimal partitioning of the tested distribution system. This proposed method may serve as an initial framework to define, for example, the hosting capacity of smaller regions, considering the characteristics of the electrical system as a whole but managing its restrictive factors locally.