sir/madam
I want the ppt ,abstract, and report for the topic-automated fault location system for primary distribution.
Sriya

ppt on advanced automated fault location system for primary dristrubution networ

System reliability and resilience are major goals of the modernization and creation of a smart grid. In the United States, power outages are estimated to cause $79 billion in annual economic losses, and the cost of each outage increases with duration[1]. Automated fault location systems have the potential to reduce outage time and costs to customers by allowing fast and efficient deployment of repair crews. Furthermore, 90 percent of outages originate in distribution systems[2], highlighting the need for fault location algorithms tailored to generally opaque distribution feeders. The problem illustrated by these statistics is this: the costs of outages demand investments to reduce outage time, but the problems occur on the distribution system, where the cost of undergrounding or heavily instrumenting every feeder also becomes very high[2]. An effective fault location system should be accurate enough to direct maintenance crews directly to the point of the fault, and be cost effective so that significant investments are not required to enable the system.
Power system faults can be classified as either series or shunt faults. A series fault (also known as an open circuit fault) is a fault for which the impedances of each of the three phases are not equal, usually caused by the interruption of one or two phases[3]. A shunt fault (also known as a short circuit fault) is defined as a fault that is characterized by the flow of current between two or more phases or between phase(s) and earth at the frequency of the associated power system[3]. Shunt faults are generally more severe than series faults and can cause fires or damage to equipment from high short circuit currents[4]. From 1992-2011, weather alone caused 78 percent of customer outages in the United States[6]. Not only are overhead distribution circuits highly susceptible to shunt faults, but the circuits in rural areas that are exposed to the elements are more likely to be long, sparsely loaded and in difficult to access areas. This makes an important class of faults difficult to locate by visual inspection.
Existing fault location methodologies
Existing methods of fault location vary in complexity and in feasibility. The traditional brute force method consists of obtaining an initial estimate of the location of a fault from mapping customer outages, performing switching operations and then sending repair crews to visually inspect long sections of line[7]. While reliable and easily performed, this method is antiquated and prolongs outage time.
Another class of fault location systems, often referred to as traveling wave methods, relies on a high sampling rate and synchronized measurements to calculate the location of the fault[8]. This type of approach involves using signal processing techniques (particularly wavelet transforms) to identify the time at which the disturbance from a fault is observed. Because the time for the signal to propagate from the disturbance to the measurement equipment is a function of the distance from the fault, the difference in observed disturbance times at different synchronized measurement devices can be used to locate the fault[9]. Another wave based approach is presented in Ref. 10 that is specifically tailored for tree structured distribution systems with distributed generation and makes use of the difference in arrival times of reflected waves, rather than synchronized measurements. These techniques have been shown to be highly accurate and robust to variations in system conditions (including loading and distributed generation) since they are not dependent on any assumptions about the system besides the topology and line parameters.
A third class of fault location systems for distribution networks, which are broadly known as impedance based methods, involve calculations using line impedances and observed changes in voltage and current at the substation[5]. These methods vary in subtle ways, but the general approach is to use pre-fault and during-fault voltage and current measurements to estimate the fault loop impedance, which depends on the location of the fault. A comprehensive review of some of the more cited impedance based approaches is given in Ref. 11 and a comparison of the performance of the leading algorithms is presented in Ref. 12.
Proposed solution
This paper presents an automated fault location algorithm for location of distribution shunt faults using synchronized voltage phasor measurement units (PMUs). PMUs have experienced rapid growth in transmission systems to almost 100 percent coverage, largely because of the range of applications they support[15][16], but in particular for observing angle stability and general operating state. A project conducted in partnership by the California Institute for Energy and the Environment, the University of California, Berkeley, Lawrence Berkeley National Labratory, and Power Standards Lab Inc. is developing a low cost, high precision phasor measurement unit specifically designed for distribution systems[17]. The GPS synchronized device has a frequency of 512 samples per cycle (31 kHz on a 60 Hz system) and is capable of measuring subcycle phase angle difference between devices. It utilizes the Simple Measurement and Actuation Profile (sMAP) technology for concentration of data, creating a foundation for application based software such as fault location[18]. It is expected that similar to transmission systems, PMU technology will present a compelling business case for the grid of the future by enabling many different applications through system awareness. A fault location approach based on PMU data differs from the specialized fault indicator networks and wave sensors described above by leveraging a technology with other applications and a readily available data stream; i.e. the cost of instrumentation need not be justified by fast fault location alone.
The proposed algorithm uses pre- and during-fault voltage magnitude and phase at the substation and remote PMUs, as well as current measurements at the substation and an imprecise system model in order to pinpoint a fault in less than a minute. The addition of PMU data to substation current measurements improves upon traditional impedance methods by measuring the behavior of the system on both sides of the fault, making the estimate more robust to the distance of the fault from the substation, variations in system conditions, and uncertainties in system models. In addition to the incorporation of remote measurements, a novel method of load aggregation is used that improves accuracy for higher impedance faults. The potential for using remote measurements of voltage sag data has been explored in Refs. 19 and 20, but these methods do not explore the potential benefit from phase angle difference. A method that uses PMU data is discussed in[21], but the paper focuses primarily on the related topic of PMU placement, rather than the response of the algorithm to variations in system conditions and uncertainty in system state.
Formulation of algorithm using PMU measurements
Scope of algorithm
The algorithm presented is designed for shunt faults occurring on sections of the feeder that are instrumented by a PMU on either end, such that the fault occurs between the two PMUs. In practice, this would mean faults that occur on the primary distribution system because each small lateral and customer connection would not be instrumented. Fast location of these faults is especially important because, unlike faults that can be isolated on laterals, these faults will affect the most downstream customers. Faults that occur on laterals that are not instrumented can be located to the branch point using the proposed algorithm in order to identify the faulted lateral; and then any of the established single end impedance measurement methods can be used to estimate the location of the fault on the lateral.
High level characterization
The proposed fault location algorithm is described by the following process:
Identify the type of fault and which phases are affected.
Select the PMU at the substation and one of the remote PMUs.
Starting at the substation, iterate over each line segment connecting the two PMUs. Assume a fault on that line segment and calculate the fault distance. If the calculated distance is greater than the line length, continue to the next line segment. If not, the fault is located according to these two PMUs.
Repeat steps 2 and 3 for each remote PMU.
Resolve the multiple location estimates from each PMU to a single estimate of the fault location.
Identification of fault type
Distribution shunt faults on three phase lines can be of various types. It is necessary to know the faulted phase(s) in order to accurately locate the fault. Ref. 22 describes a method to classify faults based on the rotation of the negative sequence voltage during a voltage sag. This method is employed at the substation to determine which phases to inspect for the fault.
Calculation of fault distance
The following procedure is used to calculate the location of the fault using pre- and during-fault substation measurements and a remote PMU at the end of one of the radial feeder paths. The consolidation of measurements from multiple PMUs is described in the next section. The fault is assumed to be on the line connecting nodes k and k+1. Pre-fault conditions are estimated. Then, during-fault load is estimated and a least squares algorithm is used to estimate fault distance d as a fraction of the line length. If the optimal value of d is one (implying that the fault is at or beyond the end of the line segment), then the next line section is considered.
There are PMUs on each phase at the substation (node S) and at the end of the feeder (node R). The substation current is also measured for each phase. All loads and laterals between nodes S and k are modeled as a single unknown load, as are all loads and laterals between nodes k+1and R, as shown by the dashed rectangles in Figure 1. Note that before the fault, If = 0.
Estimation of pre-fault conditions
In this lumped load model with PMU measurements, the voltages and currents at nodes k and k+1are observable before the fault if the location of the equivalent loads are known, as expressed by the parameters α1 and α2. Z1 and Z2 are the known line impedance parameters between the source and node k and between node k+1and the remote PMU. Zf is the impedance of the assumed faulted line segment, and αi can be estimated using power flow simulations for the normal operation of the feeder. The advantage of this model is that it is robust to conincident load variations. This model is built off of the assumption that loads are not directly measured or known, and that as load varies, its spatial distribution remains roughly constant (this might not apply to all feeders, but is especially applicable to feeders with similar customers). Leading fault location algorithms proposed in Refs. 11 and 12 rely on the stronger assumptions of either knowing each load or placing all load at the end of the feeder, which can be relaxed because of synchronized remote end voltage measurements.
The voltage at the equivalent loads and at the nodes on either end of the assumed faulted section are given by Ohm’s law and KCL in Eqs. 1-5. These relationships hold both before and after the fault, though the PMU voltages and substation current will change as a result of the fault. Voltages and currents are considered to be 3×1 vectors representing phase quantities. Impedances are 3×3 matrices and αi is a 3×3 diagonal matrix representing the location of the equivalent load on each phase.
Before the fault, no current is lost on the line connecting nodes k and k+1, which allows us to write Eqs. 6 and 7 from KCL and Ohm’s law. The linear system of Eqs. 1-7 can be uniquely solved for the 7 unknown variables[VL1,VL2,Vk,Vk+1,Ik,IL1,IL2] to determine the pre-fault state.
Estimation of during-fault load currents
Following the fault, Eqs. 1-5 still hold, assuming that αi does not change (i.e., load currents may change after the fault, but their spatial distribution remains constant). At this point we must introduce a static load model to estimate the during-fault load currents. We use the ZIP model described in Ref. 14. This allows for the calculation of the during-fault load current where the parameters Z% +I% +P% = 1 describe the proportion of the load that is constant impedance, current and power, respectively. It is important to note that this approach does not assume knowledge of each individual load, but allows for a general adjustment of the aggregated load. The during-fault voltage VLi is given by Eqs. 1 and 2. The loads are assumed to be constant power factor as described in Eq. 9.
Eqs. 8 and 9 yield an estimate of the during-fault total load current IL0 i from during-fault measurements and the prefault estimates of loads from the previous section.
Calculation of fault distance
After the fault Eqs. 6 and 7 no longer hold, and Eqs. 10-12 are introduced. Note that in the previous section, pre-fault measurements were used, but now during-fault substation and PMU measurements are used.
We also include the estimates of during-fault load currents through Eqs. 13 and 14.
Forgetting about fault distance d for a moment, there are nine unknowns [VL1,VL2,Vk,Vf ,Vk+1,IL1,Ik,If ,IL2]. The system of equations given by Eqs. 1-5 and 10-14 gives 10 linearly independent complex equations with respect to those nine complex unknowns (for a three phase line it becomes 30 equations and 27 unknowns). If the fault is identified to be single phase or two phase, an additional equation (or two equations) is included that set the fault current on the healthy phases to zero. For a given d, this overdetermined system of equations can be solved by a least squares method, or weighted least squares if some measurements or assumptions are known to be more accurate than others. Because the system is overdetermined, there can exist some least squares error ε≥ 0. The problem of locating the fault can then be formulated as in Eq. 15, where ε(d) is the error from the least squares solution as a function of fault distance.
To solve Eq. 15, a simple search with a fixed step size can be used. The step size can be set to be the desired fault location accuracy (on the order of 1 m). The bounds on d and the relatively large acceptable error allow for a brute force search to be computationally tractable. If the optimal value of d is 1, then the fault is on the next line segment, and the preceding calculations are repeated for each line segment until the optimal d is less than 1.
Resolution of multiple estimates
The procedure described in the previous section yields an estimate according to a single remote end PMU. Consider the representation of a radial distribution feeder with a tree structure and multiple PMUs, as in Figure 2. The PMUs are located at the end of lines, and the arrows show the respective estimates of the fault. As shown, different remote PMUs will give different fault locations. This is because a fault occuring on the line connecting one PMU to the substation might appear as on a lateral for a different PMU, and the result from this PMU will report the fault as occuring at or near the junction point for the lateral. In Figure 2, PMUs 1-3 will estimate the fault near one junction, PMU 5 will provide another more accurate estimate near another junction (that is farther from the substation), and PMU 4 will provide the most accurate estimate as the fault occurs on the direct path to it from the substation. In this case, there are essentially three estimates, and the task is to choose the one closest to the fault. In some cases, the fault might occur on the direct path to k PMUs, and in this case not a single estimate, but an average of these k estimates should be taken.
The following recursive algorithm can be used to obtain a single estimate of the fault location. It relies on reasonable accuracy of the preceding calculations, and on the observation that incorrect estimates will be near a branch point that is closer to the substation than the actual fault location; implying that the best estimate (or cluster of estimates) will be the one farthest from the substation.
The first step is to identify the point at which the feeder branches to different PMUs, or equivalently the last common node on the path from the substation to each remote PMU. If all of the fault location estimates are closer to the substation than that branch, then we can assume the fault occurs upstream of that branch, and the average of all of the estimates are taken. This is the first stopping criterion. If not, the remote PMUs are paritioned into groups based on which branch they are in. If we conceptualize the radial feeder as a tree, and each PMU is a leaf of the tree, then we are splitting the tree into subtrees at the first branch point (for simplicity assume there are two branches - a binary tree - recognizing that the approach can easily be extended if there are more than two subtrees at a branch point). Next, the mean fault location (distance from the substation) according to each of the two groups of PMUs is calculated. If both averages are within 50-m of each other, then there is no way to determine which is more correct, so the fault is assumed to be at the branch point. This is the second stopping criterion. However, if there is a significant distance, the group that estimates the fault farther away is chosen, and the same approach is applied recursively to that group until either of the two stopping criteria are met or there is only one PMU in the group.
For the case in Figure 2, the branch point is shown by the dotted circle. The arrows show the location of the estimate according to each PMU. PMUs 1 and 2 will be on one branch, and PMUs 3 and 4 will be on another. PMUs 1 and 2 will give estimates close to the branch point, but PMUs 3 and 4 will give estimates farther away, so the estimates from PMUs 1 and 2 are discarded. The next branch point is shown by the dashed lines, but because the estimates from PMUs 3 and 4 are close to each other, the branch point is taken to be the fault location. Had the estimate from PMU 4 been further away, then it would be taken to be the fault location.
Discussion
This paper proposes a novel fault location algorithm that leverages a synchronized voltage measurement stream in order to increase visibility of the feeder during the fault. The use of these accurate PMUs and the proposed algorithm has several advantages for fault location. When compared with other sensor based platforms, it drastically reduces the number of sensors needed to locate faults. When compared with traditional impedance based methods, it increases accuracy, especially at remote ends of the feeder and for higher impedance faults, which are traditionally the downfall of substation based impedance approaches. Travelling wave methods remain superior in accuracy and robustness because they are not affected by system loads (though uncertainties in line parameters can introduce error), but they require specialized equipment that does not serve any other purpose. The real power of the PMU based approach is that it leverages a data stream that is not specific to fault location. In the transmission system, PMUs have become ubiquitous because of their many applications to system diagnostics and control, and fault location is just one of many applications. This paper demonstrates that a PMU based approach has the potential to automatically and remotely locate faults on the order of meters or tens of meters, which is generally sufficient for dispatching maintenance crews or executing switching operations in the event of a fault.
The algorithm yields acceptable results for preliminary testing; however, there are multiple factors that need to be explored in simulation and physical environments. These areas of further research can broadly be classified as sources of error in the modeling of either loads, faults and measurement equipment. On the load side, one of the most important is the spatial distribution of the loads. Though the algorithm is robust to variations in the total load level, it is unclear what the effect of non-coincident load variation is. In addition, the response of loads under fault currents needs to be better modeled than the traditional ZIP model specified in the IEEE 123 bus system. In addition to load response during faults, the response of solid state voltage regulators needs to be modeled and accounted for in the fault location process.
It is important to note that for the purposes of the preliminary testing of the algorithm presented in this paper, only ideal resistive faults are modeled. Arc faults and other non-ideal faults, especially faults with time varying impedances, have the potential to introduce additional error. In addition, the simulation conducted for this paper assumes ideal measurements. While the voltage angle measurements from the PMUs have an accuracy of 0.01 degrees, there is a potential for fault conditions to saturate current transformers and limit the effectiveness of the method. The simulation in this paper shows that the method is robust to variations in fault impedance in the range of typical shunt faults[13]. In order to properly assess the limitations of the method, this variation in impedance needs to be combined with non-ideal fault modeling and error models in measurement equipment to determine under which conditions the algorithm yields unacceptable error. As faults create a system state that is outside of the bounds of normal operation and well developed models, there is no substitute for testing of a fault location algorithm on an operating feeder or physical test feeder. However, this research demonstrates that the use of high precision synchronized voltage measurement devices provides additional knowledge of the system state both before and during the fault that can be used for a more robust and accurate fault location algorithm than traditional impedance based methods.

This paper presents the development, simulation results, and field tests of an automated fault location system for primary distribution networks. This fault location system is able to identify the most probable fault locations in a fast and accurate way. It is based on measurements provided by intelligent electronic devices (IEDs) with built-in oscillography function, installed only at the substation level, and on a database that stores information about the network topology and its electrical parameters. Simulations evaluate the accuracy of the proposed system and the experimental results come from a prototype installation.