Generalized persistent fault detection in distribution systems using network flow theory M´ arton Greber ∗ Attila Fodor ∗ Attila Magyar ∗ ∗ Department of Electrical Engineering and Information Systems, Faculty of Information Technology, University of Pannonia, Egyetem u. 10., Veszpr´ em, H-8200, Hungary, ({greber.marton;fodor.attila;magyar.attila}@virt.uni-pannon.hu) Abstract: Persistent faults are steady state anomalies with a magnitude which does not necessary trigger general protective gear. It is present in various types of distribution networks, as leak in pipe networks or as high-impedance fault in electric systems. As smart meters come into general use, distribution systems are upgraded to have advanced metering infrastructure which can be used for diagnostic purposes. Different kind of detection methods are presented in different physical domains therefore comparison is cumbersome. The main achievement of the presented work is the formulation of an abstract system description, in order to tackle problems from various domains on a common ground. The notions of the well established general network theory are being used as a solid foundation. In this framework a general extension of flow networks is presented for distribution systems with measurement data available. Detecting faults at metered points is tackled in the literature, this problem is translated into the proposed representation. On the other hand a novel problem is described, the fault at unknown location between two metered points. The applicability of the abstract description to specialized distribution systems is presented through a simple case study. Keywords: Distribution networks; Smart power applications; Steady-state errors ; Losses; Networks 1. INTRODUCTION Distribution networks have the soul purpose to deliver re- sources from some supply entity to end consumers. Accord- ing to the type of the delivered resource, the networks take different appearances in the real world. Water and gas are transported through pipes, electricity is delivered through conductors. Effects from various sources like weather, ge- ographical location or human activity cause distribution lines to be vulnerable to faults. Persistent faults are steady state type faults. They can last relatively long because the magnitude of the faults is not big enough to trigger general protective gear. Persistent faults can take many shapes according to the physical system. In electrical power distribution systems high impedance faults (HIF) can happen when at some point the distribution conductor is grounded through a high impedance, (Ghaderi et al. (2017)). Another example is the theft of electricity, also called non-technical losses (NTL). In this case live wires are manipulated to tap electricity or the measurement equipment is tampered, in order to hide consumption (Viegas et al. (2017)). Since more current needs to be supplied in order to keep existing demand satisfied, both cases lead to excessive load of the line and heat build up, in extreme case fire and blackout. In case of pipe network there are two main definitions called unaccounted for gas (UFG) and non- revenue water (NRW), both of these cover the same topics, just in different technological context. In case of both, gas and water, resources are lost before reaching end customers. This is influenced by many factors like leakages, measurement variations and meter tampering, in case of gases there are also emissions . Real world case studies show, that leakage plays a considerable role both in UFG (Shafiq et al. (2018)) and NRW (Kanakoudis and Muhammetoglu (2013)). Distribution systems nowadays are getting more and more support in terms of measurement equipment. Advanced metering infrastructure (AMI) relies on two way com- munication between the utility and the consumers (Jha et al. (2014)). This is enabled through smart meters, which are measurement devices capable of communication, to provide consumption data periodically. According to the type of distribution network, these are prominent in both electrical (Bahmanyar et al. (2016)), and water distribu- tion networks (Singapore (2016)). This data can also be used to perform various diagnostics on the network. In the case of NTL the general power balance equations need to be modified in order to account for wasted load. Using the extended expressions it was noted that observing and comparing the voltage drop on line segments is an indicator, and can be used for NTL detection (Bula et al. (2016)). Since distribution networks can have large sizes, the question of structural decomposition and splitting a Preprints of the 21st IFAC World Congress (Virtual) Berlin, Germany, July 12-17, 2020 Copyright lies with the authors 13754