Power Unavailability Reduction in Distribution Grid Fault Management with Entropy Minimization Michele Garau Department of Energy Systems SINTEF Energy Research Trondheim, Norway michele.garau@sintef.no Bjarne E. Helvik Dept. of Information Security and Communication Technology - NTNU Trondheim, Norway bjarne@ntnu.no Abstract—Smart automation is acquiring a high importance in current distribution systems. The high number of buses, the ra- dial topology, the small number of sensors and automated devices, require new approaches in managing fault conditions. These approaches must be able to deal with a high level of uncertainty of the state of the system and the measurement data. In this paper a novel method for fault location and isolation is proposed, which is based on the principle of entropy minimization. The algorithm builds a switch operation strategy which is able to locate the fault in a minimum number of manoeuvres, and therefore to reduce the impact of blackouts in terms of power unavailability. The application of the method on different distribution network topologies, with different levels of automation in terms of fault indicators and remotely controlled switches, demonstrates the potential of the method for distribution system analysis and supporting system automation planning. Index Terms—Power grid availability, Smart Grid, Distribution System Protection, Reliability Modelling I. I NTRODUCTION Nowadays, distribution systems are acquiring a high atten- tion in current power systems. The spread of low size gener- ation power plants among domestic appliances, the increasing interest towards electricity as energy vector for different usages (buildings heating and cooling with heat pumps, electric vehicles transportation, etc.), the trend towards an horizontal participation to the energy market, increase the necessity of new solutions that enhance the quality of service of the power system at the distribution level. In this scenario, the reliability of distribution networks assumes a paramount importance. To minimize the impact of outages among distribution network customers, it becomes fundamental to introduce smart automation that allows a fast detection, location and isolation of network faults, and to elaborate strategies for a quick restoration of the network. In fact, currently in Europe about 80% of the blackouts are due to faults that occur on the distribution grid, and the yearly average blackout duration ranges from 15 to 400 minutes per customer, mostly due to the weak automation of distribution systems which makes fault location mostly a process based on grouping This paper has been funded by CINELDI - Centre for intelligent electricity distribution, an 8 year Research Centre under the Research Council of Norway’s FME-scheme (Centre for Environment-friendly Energy Research, 257626/E20). The authors gratefully acknowledge the financial support from the Research Council of Norway and the CINELDI partners. of customer outage calls and experience-based power lines patrolling. [1]. In this context, this paper proposes a novel, efficient method to locate and isolate faulted line segments in distribution networks, and minimize the unavailability of the power supply to the unaffected part of the network, i.e., restoring it in a minimum time. Both these two objective formulations will be used synonymously in the paper. Traditionally, fault location methods are developed for pro- cessing the raw data generated from specific measurement devices and returning possible fault locations, with a specific level of uncertainty. Essentially, these methods are meant to provide a useful information to Distribution System Operators (DSOs) that, upon this information, will build a plan for inspecting the network and reduce the overall sectioning and repairing time, isolating the fault and restoring the power supply on the healthy portion of the network. Fault isolation and service restoration can be considered as processes that lay on top of the data-information stack, which elaborate a strategy for network inspection in failure conditions; this stage of fault management typically relies on a solid em- pirical knowledge of the power system. The automation of this process allows to extend the knowledge of the system, by elaborating an optimal isolation and restoration strategy that takes into account different factors, e.g. the accuracy of measurement data and the consequent uncertainty in the information elaborated by the fault location algorithms, and provides a fundamental support to DSOs in decision making during fault conditions (Fig. 1). In transmission systems different techniques have been developed and successfully applied for allowing an efficient and fast location of faults along the network. However, these techniques may not directly be applied in distribution systems. The high number of buses, often weakly monitored and automated, and the high number of lateral branches, make the application of transmission system fault location techniques to distribution networks unsuited. Achieving in distribution systems a level of monitoring comparable with transmission systems would require a massive investment. For this reason, in the last decades a large number of works have focused on formulating new methodologies for fault location that are specific for distribution networks [2], [3]. Despite being purposely designed for distribution systems, these methods