An Iterative Method for Detecting and Localizing Islands Within Sparse Matrixes Using DSSim-RT

Nowadays, the simulation tools used to perform power system analysis are evolving into the many-core computation era; some of these techniques propose to tear the power system network into several subnetworks (islands) for its parallel processing. The island's detection is an issue considered by the power flow analysis due to management activities such as feeder reconfiguration, fault detection, and isolation, among others, that generate topological changes. These methods include graph theory, checking the circuit breaker status and on-site measurements, decision trees, frequency deviation, and pattern recognition, among other techniques, which are focused to detect topology changes that may generate islands and affect the state estimation of the system. In contrast, other techniques such as the approximate minimum degree perform a reorganization of the YBUS matrix that describes the network components and their connectivity, looking to factorizing the YBus matrix to solve the power flow problem. These techniques are very functional and efficient; however, some of them require detailed information of the network, and are addressed to cover meshed and radial network configurations separately. This paper presents an iterative algorithm that uses the compressed coordinate branch-to-node matrix for detecting, classifying, and grouping islands within sparse matrixes, describing mesh or radial networks. This algorithm is a valuable tool to simplify islands location and can be implemented using any programing language due to its simplicity. This method is used in Distribution System Simulator - Real Time Version (DSSim-RT), which is a simulator based in OpenDSS, for tearing the power system network to allow the multithread power flow analysis of distribution systems in real time.