Abstract-- With the rapid advancement of phasor measurement units (PMUs) technology, system operators in different level of power systems have access to new and abundant measurements. Taking into account these measurements in active distribution systems (ADNs), a new algorithm for short-circuit fault detection and identification based on state estimation (SE) is introduced in this paper. In this regard, as the first step, traditional SE process is revised to be compatible with fault conditions. Then, a fault location algorithm (FLA) based on the revised SE (RDSSE) is presented which attends to detect the location of fault after diagnosing faulted zone. For this purpose, current and voltage synchrophasors captured by PMUs as well as pre-fault SE results are used and according to calculated measurement residual indexes, the correct location of fault is diagnosed. The performance of RDSSE and SE based fault location method are tested by applying on an ADN, considering different fault scenarios in the network. The results proved that the proposed method is more accurate and reliable than traditional SE based methods in fault conditions and can precisely determine the real location of fault at lower SE execution times. Index Terms--Active distribution networks, short-circuit fault location, phasor measurement units, pseudo-measurements, state estimation. I. INTRODUCTION ORE than 80% of outages in power systems are due to failures occur in distribution systems. Therefore, distribution companies (DisCos) around the world have tried to utilize some efficient technologies, maintenance actions and corrective procedures for decreasing the rate and destructive effects of possible faults. Although, most of the transient faults in distribution networks can be cleared by reclosers, sustained faults may require intervention from the technical crew which in result increase the outage time significantly. Since it is crucial for the utilities to minimize outage times and provide high quality service, determining the location of fault in a timely manner becomes a priority for system planners and decision makers [1], [2]. On the other hand, distribution grids are facing with some upcoming events, i. e. advent of new distributed energy resources (DG) and gradually reshaping of distribution networks from a mainly M. Gholami, A. Abbaspour, and M. Fotuhi-Firuzabad are with the Center of Excellence in Power System Control and Management, Electrical Engineering Department, Sharif University of Technology, Tehran 11365- 11155, Iran (e-mail: gholami_mohammad@ee.sharif.edu; abbaspour@sharif.edu; fotuhi@sharif.edu). radial to a meshed topology. This can be translated to more complex structure of distribution networks [3]. These changes are all calling for more advanced and effective fault location algorithm (FLA) for fast restoration of networks. Considerable efforts have been devoted in the literature to propose some efficient algorithms for detecting the location of different faults in distribution networks. These techniques can be categorized into two main classes: Signal Analysis Methods (impedance, traveling wave and wavelet based) and Knowledge-based Methods (artificial neural network, fuzzy and expert system) [4]. Each of these algorithms has some drawbacks which can affect their capability [1], [4]. For the sake of comparison, these drawbacks are itemized in Table I. TABLE I MAIN CLASSES OF FLA IN LITERATURE Fault Location Techniques Drawbacks [1], [14] Impedance-Based [5], [6] -Inability in presence of multi terminal lines. -Provide multiple fault locations when a single fault occurs in the network. -Lower accuracy in facing with high unbalanced load flow and DG penetration. Travelling Wave [7], [8] and wavelet [9], [10] Based -High computational burden. -Difficult to guarantee the reliability of these methods because of variety of load characteristics and fault cause. -Requires costly installations of high frequency measurement devices and possibly equipment for pulse generation. Knowledge-Based [11]-[13] -Requires continues training and extensive data for training (voltage and current data at secondary substation level). -Slow response time. As can be traced in Table I, the main challenges in diagnosing fault events in ADNs are the limited number of measurements (lack of data) and the weakness(es) of applied FLA in computational burden, implementing cost, reliability and situation with high DG penetration, unbalanced load flow and network topology change. To deal with these issues, state estimation (SE) can be a key tool. SE is the main function of distribution management system (DMS) executed in real time form and it maps inaccurate and available measurements of the network to best estimation of system condition [3]. The SE process can be implemented based on three-phase formulation M. Moeini-Aghtaie is with Department of Energy Engineering, Sharif University of Technology, Tehran, Iran (e-mail: moeini@sharif.edu). M. Lehtonen is with the Department of Electrical Engineering, Aalto University, Espoo 11000, Finland (e-mail: matti.lehtonen@aalto.fi). Detecting the Location of Short-Circuit Faults in Active Distribution Network Using PMU based State Estimation Mohammad Gholami, Ali Abbaspour, Moein Moeini-Aghtaie, Member, IEEE, Mahmud Fotuhi- Firuzabad, Fellow, IEEE, Matti Lehtonen M