IEEE TRANSACTIONS ON SMART GRID, VOL. 4, NO.

2, JUNE 2013

1067

Detecting and Locating Faulty Nodes in Smart Grids Based on High Frequency Signal Injection
Amir Mehdi Pasdar, Student Member, IEEE, Yilmaz Sozer, Member, IEEE, and Iqbal Husain, Fellow, IEEE
AbstractAn on-line method for detecting and locating a faulty node in the utility grid is proposed for smart grids. The method is based on injection of high frequency (A-Band) current signal into the grid that would impose voltages (less than 1V according to EN50065-1 standard) on the nodes to determine changes in the impedance characteristics. This detection is accomplished on-line without interrupting the power ow in the network. The developed algorithm has been implemented within an electrical power system model. This low voltage network model has been tested with different fault scenarios. The proposed procedure is able to detect the faulty nodes with high accuracy. Index TermsFaulty node detection, illegal electricity usage, signal injection to grid, smart grid services.

I. INTRODUCTION

AULT IN THE power grids is an unpredictable situation which could lead to an unbalance in the overall system operation. A Faulty node (FN) could appear in between the segments of the distribution network due to a variety of reasons. Illegal connection for electricity pilferage is one unwanted way of establishing a FN. This type of connection leads to an unpredictable technical as well as economical problems in the utility industry. From an economic point of view, this issue has a dramatic impact on the suppliers budget because of unaccounted electricity usage [1]. Early restoration of the fault improves the reliability and safety of the distribution network [2], [3]. A new method for the detection of illegal usage based on automatic meter reading has been suggested in [4]. The method is based on traditional automatic meter readers (AMR). The host controller can detect the existence of the FN in each feeder by comparing the sum of all the energy meters within the system with the host energy meter. The disadvantage of this method is its inability to localize the FN. This method also requires high quality calibrated energy meters. In [5], a FN locating method for a high voltage long range simple distribution network is presented. The fault is detected based on the difference between the

injected and reected signals. This method is not applicable for a short range or multilayer distribution network. In [6], a method to locate the position of the FN in the low voltage simple distribution network has been introduced. The method is based on determination of the power line characteristics by injecting the current signal into the grid to be measured by smart meters. The main disadvantage of this method is the requirement of powering down the system during the test, which is not acceptable in residential networks. In this study, a novel approach for detecting a FN, and the procedure for localizing the position of the FN in a three phase low voltage complex distribution network is presented. The approach is based on the information provided by the smart meters to the central station, and capturing the network parameters after the injection of the test signal into the network. The proposed method can be applied to a multilayer complex distribution network as opposed to other methods mentioned earlier. The localization of the FN can be accomplished quite accurately compared to most of the available technologies. Section I provides the introduction to the well-known procedures to detect and localize the FN; Section II provides an introduction to the smart grid structure and the description of the hardware needed for the new method; Section III describes the power network models at high frequency based on low frequency measurements and high frequency test signal injection; Section IV presents the proposed method for detection and localization of the FN. The test results for the example network topologies are presented in Section V; Section VI gives the summary of the research. II. SMART GRID STRUCTURE In modern times, smart grid infrastructure can contribute towards power network automation which would improve the capabilities of existing power networks signicantly. The primary focus is the improvements in the distribution network through communication and penetration of renewable energy systems [7], [8]. This modern power grid will integrate advanced information techniques, communication techniques, computer science and pre-existing physical grids [9]. In the smart grid infrastructure, it is possible to show when, where and how the energy is used through the on line bidirectional connectivity between users and utilities. It is also possible to manage the loads in smart grid networks to increase the energy delivery efciency [10]. The block diagram of a smart grid network in the low voltage side (LV) is shown in Fig. 1 where each individual smart meter communicates with the central unit. Two important elements in the smart grid network are the smart meter and the central unit. A smart meter is the data gateway which is installed at each customer node. The meter

Manuscript received April 15, 2011; revised September 02, 2011, December 19, 2011, May 18, 2012, and July 30, 2012; accepted September 18, 2012. Date of publication April 12, 2013; date of current version May 18, 2013. Paper no. TSG-00150-2011. A. M. Pasdar and Y. Sozer are with Electrical Engineering, University of Akron, Akron, OH 44325 USA. I. Husain was with Electrical Engineering, University of Akron, Akron, OH 44325 USA. He is now with North Carolina State University, Raleigh, NC 27606 USA. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TSG.2012.2221148

1949-3053/$31.00 2013 IEEE

1068

IEEE TRANSACTIONS ON SMART GRID, VOL. 4, NO. 2, JUNE 2013

Fig. 1. LV side smart grid topology.

measures the electricity usage information and sends it to the intermediate central unit. The intermediate unit collects the data from smart meters, and sends it to the utility central unit [11], [12], [15]. The connectivity between customers and the central unit could be power line carrier communication, ZigBee [13] or GSM [14]. This research will show that it is possible to detect and locate the FN in the distribution side before the meter. The existence of the FN in other areas is also detectable by the measurement devices. The central unit hardware needed to perform the proposed method has an extra injection unit. Each smart meter has a separate high frequency voltage measurement unit to measure the received signal voltage. A. Smart Meter The proposed method requires two units for proper operation. The rst unit is a common smart meter that should measure the data needed to calculate the impedance characteristics of each node; this unit also has the ability to store data into external memory and send it to the host via a physical layer. The meter transducer should have wide bandwidth to measure the current and voltage of the received signal. Another unit is the modem which sends data to a specied node. The smart meter unit is based on commercial energy meter chips [16], [17]; chips that have the ability to measure most of the energy usage parameters. Each meter sends six data to the host unit to calculate the impedance of the power line at low frequency: These are active power (P), reactive power (Q), apparent power (S), rms voltage , and rms current and line frequency (f). The power usage of each node is computed and transferred to the main unit. The modem unit could be designed based on several communication techniques such as power line carrier [18], [19] and wireless communication methods. B. Current Injection Unit The current injection unit is composed of a programmable reference signal generator, power amplier to generate the commanded current into the network and high voltage coupler [20]. In this research, a constant amount of current is injected into the network to determine the difference between the predicted and measured power line impedance model. A programmable current source is used for this purpose. The proposed current injection unit is considered to be installed in the central unit. The injected current signal at the main feeder imposes specic voltages on individual nodes [21], [22]. The EN50065-1 standard species the maximum peak to peak voltage which the central unit is allowed to impose on the

Fig. 2. EN50065-1 Standard for power line carrier communication.

Fig. 3. Test signal injection unit.

individual nodes. The maximum permissible peak-to-peak imposed voltage on each node is dened based on the frequency of the injected current signal; this denition is shown graphically in Fig. 2. Based on the desired frequency (A-Band) with respect to the EN50065-1 standard, the maximum peak-to-peak imposed voltage is 120 dB which is equal to 1 V. A constant value of the current is injected to maintain the limitation set by the standard. An accurate power line impedance estimation is needed to set the current value which is determined based on statistical data or on-line impedance measurements. The block diagram of the current injection unit is shown in Fig. 3. III. POWER NETWORK MODEL The steps for locating the FN are based on the capability of the hardware as well as on a developed model of the power network. The method relies on the load ow analysis of the power system at PLC frequency which requires the high frequency impedance of the loads. The high frequency load impedances are either directly measured at PLC frequency or rst measured at power transmission frequency and then transformed into PLC frequency by smart meters using load models. If the second method is chosen for the implementation, the model of individual loads helps to predict the high frequency impedance using the low frequency impedance can be programmed into the individual smart meters. After determining the load impedances, the smart meters transmit the information to the central unit periodically to be used in the high frequency load ow analysis. Since each measurement period can be xed, it is assumed that the load impedances are constant during these intervals. For better accuracy, the measurement periods can be shortened by the smart meters. Other information needed for this method is the network response characteristics to the injected signal. These steps are explained in detail in the following subsections.

PASDAR et al.: DETECTING AND LOCATING FAULTY NODES IN SMART GRIDS BASED ON HIGH FREQUENCY SIGNAL INJECTION

1069

Fig. 4. Smart grid impedance model in LV side.

A. Network Impedance Model at Low Frequency One of the most important data needed for detecting and locating the place of the FN is the power line network impedance model which consists of the nodes load impedance model and each feeder segments impedance model. Reactive as well as active power measured at each node determine the basic version of the node load impedance model at low frequency. The real and imaginary parts of the impedance are determined separately. The characteristics of these two elements are determined based on the apparent power as well as active power which are measured by the smart meters. The smart grid impedance model in the LV side is shown in Fig. 4. There are number of nodes in the network. Here is the feeder impedance in section is the cable impedance which connects feeder to the customer, is the grid frequency and is the load impedance which has resistive and inductive components. The parameters and are obtained from the energy consumption information as shown in (2) and (3). (1) (2)

Fig. 5. Heater impedance magnitude.

Fig. 6. Heater impedance phase. TABLE I POWER LINE CHARACTERISTICS AT 90 KHz

complex low frequency impedance general form for the is shown in (5)

to

. The

(3) and are the feeder segments, , and load real power, apparent power and RMS current. B. Network Impedance Model at High Frequency Smart meters in smart grids measure the apparent power as well as active power at the line frequency. The impedance measured of the power line frequency is converted to the current injection level at high frequency. It is possible to construct the high frequency impedance parameters based on the measured low frequency impedance. The polynomial mapped transfer function could be used for conversion. The transfer function varies for each load and changes with the operating frequency. The high frequency model denition is given as follows: (4) where is the complex impedance at the operating frequency is the low frequency impedance and is the polynomial mapped complex transfer function that converts are

(5) where and are the coefcients of the degree polynomial to represent . Polynomial coefcients are dened based on the impedance magnitude and phase variations of typical loads at home. As an example, the impedance magnitude and phase variation of the residential heater are presented in Figs. 5 and 6, respectively. Although it is possible to determine the conversion ratio for the load, it is more precise to measure the high frequency impedance directly. The current and voltage transducers should have wide bandwidth to measure the high frequency portion of the current and the voltage signals. Based on the high frequency voltage and current measurements the individual load impedances would be determined. The measured or estimated individual high frequency load impedances are communicated to the central unit by the individual smart meters. The central unit has predetermined network information which is basically the number of nodes and impedance between the individual nodes. The impedance between the nodes is based on the high frequency impedance of the power cable among the nodes. Table I shows the typical power line impedance characteristics at 90 KHz. The central unit constructs new admittance matrix every time it gathers high frequency load impedances from the smart meter. In the network architecture, there are

1070

IEEE TRANSACTIONS ON SMART GRID, VOL. 4, NO. 2, JUNE 2013

Fig. 7. Power and information ow diagram.

number of nodes including the ground and central unit nodes. The ground is considered as node number 0. The loads are treated as an impedance between the meter location and the ground. The admittance between two nodes (node i and k) can be found as: (6) where and are the real and imaginary parts of the impedance between the nodes. The admittance matrix in the network can be constructed as:

C. Response of the Network to High Frequency Test Signal A high frequency test signal can be injected to the starting point of the feeder to determine the real power line network characteristics. The FN is also considered since the test is conducted on the real network. The test signal is considered as a programmable current source. With respect to the EN50065-1 standard, the amount of current is limited due to the injected signal peak-to-peak voltage limitation at high frequency. Two different methods for adjusting the current magnitude are proposed. The statistical data showing the average of the total network impedance from the base station side for a reasonable time period could be used for setting the current. An alternative approach is the adaptive current adjustment. In this approach, the high frequency voltage is monitored continuously. The amplitude of the injected current is controlled based on measured voltages of the individual nodes. In the second method, although the control is more optimum than the rst one, the controller requires more processing time. The high frequency signal measurement unit is used to determine the characteristics of the received signal. The unit consists of the high voltage coupler, band pass lter and the high frequency RMS voltmeter. The high voltage coupler is used to separate the high frequency test signal from the high voltage power line. The structure of the coupler is based on LC band stop lter which has a high impedance within the range of the test signal. To mitigate the noise effects on the received signal a sharp band pass lter is used to pass only the test signal which has the frequency within the range of the EN50065-1 standard. The communication between devices can be based on power line carrier or wireless communication. Power and information ow diagram based on power line carrier communication method is shown in Fig. 7, where s represent the smart meters at the specied nodes. IV. FAULT DIAGNOSIS The array of the node voltages based on simulation of a predened model with respect to test signal injection is . The array of the measured high frequency received signal voltage at the nodes due to the injection of the test signal in the real world is . The detection unit runs the specic procedure at each iteration based on the two input matrices to detect the existence

(7) array has the individual nodes voltages and array represents the source currents into the individual nodes. The diagonal element of each node, known as the self-admittance, is the sum of all the admittances connected to it as given by: (8) The off-diagonal element, known as the mutual admittance is equal to the negative of the admittance between the nodes. This element is given by: (9) For a given high frequency injected current, load ow analysis has been carried out with the constructed impedance matrix using the Newton Raphson method [23] to nd the voltage of the individual nodes. In the high frequency domain, the only source that injects the signal is the programmable current source; therefore, all the node currents are zero except the rst node current . The individual node voltages are recorded in the array for the fault diagnosis presented in Section IV.

PASDAR et al.: DETECTING AND LOCATING FAULTY NODES IN SMART GRIDS BASED ON HIGH FREQUENCY SIGNAL INJECTION

1071

Fig. 8. 3 layer power line network model with 22 smart meters.

of the FN on the network. The difference between the measured and estimated voltage arrays is given as: (10) will be null. is used to show In the absence of a FN, the changes in the predened network impedance model. Each element in the array is compared with the threshold value. This procedure can be summarized as: (11) , and is the where is the set of elements in vector threshold voltage. An anomalous increase in matrix elements shows the area where a FN is present. The peak value of elements inside the faulty node area (FNA) shows the FN which is the closest node to the fault location in the network. The denition of the FN is given as: (12) The method allows the detection of the existence of the FN as well as the determination of the FN location. The ow diagram for the detection and localization method is given in Fig. 9. V. CASE STUDIES Different case studies are conducted to verify the validity of this novel method in experimental and simulated networks. In Section V-A, three different scenarios are simulated in the proposed three layer network to examine the validity and the performance of the concept. In Section V-B, the proposed method is tested in the real network with eight nodes. The effects of inaccurate measurements are also presented in simulated and experimental networks. A. Network Simulation A three layer network has been selected as the example network. The studied three layer network is shown in Fig. 8. This
Fig. 9. Detecting and localizing the FN procedure ow diagram.

model includes the most possible complicated topology for the low voltage power line network. All of the simulations have been performed on this model. In the rst case, there is no FN in the network. In the second and third cases, one FN is located at different locations for each one of the cases. In the simulated three layer network, there are 45 nodes (1,2, 45), and only 22 of them (24,25, 45) have a smart meter (sm24, sm25, sm45). The loads are selected based on the typical home load pattern. Inductive and resistive loads are both considered in this case study. In the model, each feeder and cable segments are modeled as a shunt medium transmission line. For modeling the feeder segments, the characteristics of the cable are obtained

1072

IEEE TRANSACTIONS ON SMART GRID, VOL. 4, NO. 2, JUNE 2013

Fig. 10.

for each node where FN is close to node 40.

Fig. 11.

for each node where FN is close to node 34.

from Table I based on the 25 feeder and cable are modeled as:

cable model. The

(13) is the feeder segment impedance, is the cable where impedance which connects the feeder to the load, and is the frequency of the injected signal. The feeder segment resistance is taken to be 100 and the feeder segment inductance is taken to be 16 . Since the length of the cable which connects the feeder to each load is very short, it is modeled with just a resistive element . has a range of (3 to 10 ) and has a range of (0 to 20 mH). The test signal frequency is selected as 90 KHz. Three different scenarios are considered for method validation. In the rst one, the network is running in the normal mode; is about zero where there is no illegal connection or FN in the network. In the second situation, a FN is considered close to node 40 with a 2 resistive load connection. Due to the absence of a meter at node 20, the vector has a peak in node 40 which is the closest node to the FN. For the third scenario, the FN is considered to be near node 16. The impedance of the FN is the same as the second scenario. As shown in Fig. 11, has the largest value near node 34. Since node 34 is the closest smart meter near the FN, we can conclude that the algorithm accurately detects the FN in the network. It can also be concluded that the FNA is depended on the value of the threshold voltage . The elements in the vector which are greater than the are considered as a FNA. The should be selected so that the FNA could be identied with a reasonable accuracy. Selecting as gives accurate identication of the FN in the third scenario. Many factors have an effect on the accuracy of the FN location estimation. The dominant factors that lead to miss identication of the FN are the truncation effects due to the analog-todigital converter resolution, the inaccurate model of the power

Fig. 12. Analog to digital converter resolution versus FNDR.

line impedance matrix, and the nonlinearity of the load estimation model. It is expected that the magnitude of the at the FN is much larger than the in any other node in the network. The faulty node detection ratio given in (14), is dened as the ratio of the magnitude at FN to the next largest value at any node in the network. The is a good measurement for the quality of the FN detection. (14) The analysis of the effect of the analog-to-digital converter resolution on the is presented in Fig. 12. should be grater than 1 for the possibility of the FN detection; the quality of the detection increases as increases. The high frequency network model is based on node power measurements and linear load conversion from low to high frequencies. Tolerance of power measurements causes inaccuracy in network model estimation. The tolerance is modeled by injecting a random error to the nal network model before performing the power ow analysis. Relation between the

PASDAR et al.: DETECTING AND LOCATING FAULTY NODES IN SMART GRIDS BASED ON HIGH FREQUENCY SIGNAL INJECTION

1073

Fig. 13. Measurement tolerance versus total error.

Fig. 14. Implemented experimental network. Fig. 15. Coupler/decoupler circuit.

tolerance percentage versus total error is shown in Fig. 13. The total error means the number of miss identication of the FN when the location of FN is changing from the rst node to the last. B. Experimental Network A one-layer eight-node network has been implemented in the laboratory to analyse the performance of the proposed method in a real system. Higher load impedance compared to the low impedance characteristics of the feeder segments increases the complexity of the FN detection procedure. In the experimental network, the impedance of the nodes have been selected in the range of 40 to 80 to reect the worst case scenarios. The implemented network is shown in Fig. 14. The injection unit has to inject a current into the utility network. Each node in the network has a decoupling network to measure the injected signal effect on that node. The coupling/decoupling circuit built in the lab is shown in Fig. 15. Coupling/decoupling circuits are made of a high pass lter that blocks the 60 Hz signal. Circuits protect the low voltage current injection and voltage measurment circuitry from higher power levels in the network. The EN50065-1 standard has been followed for current injection into the power network during the laboratory experiments. The amplitude of the current injected into the feeder is 142 mA. Amplitude of the injected current has been set based on maximum allowed peak-to-peak imposed voltage on each one of the nodes which is 1 V [21]. The voltage and the current of each node were used to compute the high frequency load impedances. Characteristics of the feeder segments are dened based on the length of each segment. The injected signal waveform is shown in Fig. 16. The test has been conducted with the FN connected to node 5. The voltage and current information for each node have been collected based on the 142 mA current injection. High frequency impedances for each node along with the predetermined cable impedance are combined to form the high frequency impedance network. The high frequency load ow analysis is performed for the injected current. The measured and the estimated node

Fig. 16. Injected test signal waveform. TABLE II SIMULATED AND MEASURED NODES VOLTAGE WITHOUT AND WITH FN

voltages from the load ow analysis are compared to calculate . Table II shows the measured and estimated node voltages. The voltage matrix elements are used in the FN detection algorithm. Based on the developed method as it is shown in Fig. 17, the FN has been identied to be close to where the fault occurred. The absolute value of peaks at node 4 which is very close to the FN (node 5). The measurement accuracy can affect the performance of the detection procedure. Measurement error has been added to the measured data to show the sensitivity of the performance with respect to measurement accuracy. The error tolerance is in the range of 0 to 20% of the total measurement. In Fig. 18, the minimum voltage difference of FN voltage compared to those of

1074

IEEE TRANSACTIONS ON SMART GRID, VOL. 4, NO. 2, JUNE 2013

with an experimental network developed in the laboratory. Experimental results showed that the proposed method effectively detects and localizes the FN. REFERENCES
[1] M. T. Hagh, K. Razi, and H. Taghizadeh, Fault classication and location of power transmission lines using articial neural network, in Proc. 2007 IEEE Int. Conf. Power Eng., pp. 11091114. [2] M. Gilany, E. S. T. El Din, M. M. Abdel Aziz, and D. K. Ibrahim, An accurate scheme for fault location in combined overhead line with underground power cable, in Proc. 2005 IEEE Power Eng. Soc. Gen. Meet., pp. 25212527. [3] H. Zhang, Z. Tian, and E. Zhang, An improved algorithm for fault location in distribution network, in Proc. 2008 IEEE Int. Conf. Condition Monitoring and Diagnosis, pp. 727730. [4] I. H. Cavdar, A solution to remote detection of illegal electricity usage via power line communications, IEEE Trans. Power Del., vol. 19, no. 4, pp. 16631667, Oct. 2004. [5] M. da Silva, D. V. Coury, M. Oleskovicz, and E. C. Segatto, Combined solution for fault location in three-terminal lines based on wavelet transforms, IEEE Trans. Gener., Transm., Distrib., vol. 4, no. 1, pp. 94103, Jan. 2010. [6] A. Pasdar and S. Mirzakuchaki, A solution to remote detecting of illegal electricity usage based on smart metering, in Proc. 2007 IEEE Int. Workshop Soft Comput. Appl., pp. 163167. [7] C. Cecati, C. Citro, and P. Siano, Combined operations of renewable energy systems and responsive demand in a smart grid, IEEE Trans. Sustainable Energy, vol. 2, no. 4, pp. 468476, Oct. 2011. [8] K. Tanaka, K. Uchida, K. Ogimi, T. Goya, A. Yona, T. Senjy, T. Funabashi, and K. Chul-Hwan, Optimal operation by controllable loads based on smart grid topology considering insolation forecasted error, IEEE Trans. Smart Grid, vol. 2, no. 3, pp. 438444, Sep. 2011. [9] Z. Ruihua, D. Yumei, and Y. Liu, New challenges to power system planning and operation of smart grid development in China, in Proc. 2010 Int. Conf. Power Syst. Technol., pp. 18. [10] M. A. S. Masoum, P. S. Moses, and S. Deilami, Load management in smart grids considering harmonic distortion and transformer derating, in Proc. 2010 IEEE Innov. Smart Grid Technol., pp. 17. [11] S. Soergel, An economic smart metering pilot implementation using standards-based protocols, in Proc. 2010 IEEE Int. Conf. Innov. Technol. Efcient Reliable Electr. Supply, pp. 216219. [12] F. M. Cleveland, Cyber security issues for advanced metering infrastructure (AMI), in Proc. 2008 IEEE Power Energy Soc. Gen. Meet., pp. 15. [13] C. Tatsiopoulos and A. Ktena, A smart ZIGBEE based wireless sensor meter system, in Proc. 2009 IEEE Int. Conf. Syst., Signals, Image Process., pp. 14. [14] C. Hochgraf, R. Tripathi, and S. Herzberg, Smart grid charger for electric vehicles using existing cellular networks and SMS text messages, in Proc. 2010 IEEE Int. Conf. Smart Grid Commun., pp. 167172. [15] M. Venables, Smart meters make smart consumers [Analysis], IEEE Trans. Eng. Technol., vol. 2, no. 4, p. 23, Apr. 2007. [16] J. Jiang and L. Yu, Design of a new three-phase multi-rate watt-hour meter based on singlechip, in Proc. 2009 IEEE Int. Conf. Comput. Intell. Softw. Eng., pp. 14. [17] H. Ju-Yong and Y. Hu-Mei, Design of GPRS netwok watt-hour meter based on ADE7758, IEEE Trans. Microcomput. Inf., vol. 23, pp. 166167, Jan. 2007. [18] L. Dong and Z. BaoHuz, Design and emulation of high-speed narrowband PLC system for Smart meter reading, in Proc. 2010 IEEE Power Energy Eng. Conf., pp. 14. [19] A. Pasdar and S. Mirzakuchaki, Three phase power line balancing based on smart energy meters, in Proc. 2009 IEEE EUROCON, pp. 18761878. [20] T. D. Rybel, A. Singh, P. Pak, and J. R. Marti, Online signal injection through a bus-referenced current transformer, IEEE Trans. Power Del., vol. 25, no. 1, pp. 2734, Jan. 2010. [21] Signalling on Low-Voltage Electrical Installations in the Frequency Range 3 kHz to 148.5 kHz, CENELEC Standard EN 50065-1, 1991. [22] F. J. Cortes, E. M. Rubio, and A. Valdovinos, Embedded powerline DSP modem for domotic SNMP networking in European countries, IEEE Trans. Consum. Electron., vol. 48, no. 4, pp. 854862, Nov. 2002. [23] H. Saadat, Power System Analysis, 2nd ed. New York: McGraw-Hill, pp. 228233.

Fig. 17.

for each node where FN is close to node 5.

Fig. 18. Minimum voltage deference of FN voltage to other node voltages versus measurement error tolerance.

other node voltages are presented with respect to the measurement error tolerance. In order to have a sufcient voltage difference for detecting the FN, the measurements and the low to high frequency conversion matrices should have a good accuracy. High error tolerance in the measurement or the conversion matrices makes inaccurate detection for FN. VI. CONCLUSION In this paper, a novel method for detecting and localizing a FN in low voltage power line networks is presented. The method can be run in a live network without any interference. The procedure is based on the power line network characteristics and the response of the network to a high frequency test signal. Simulations based on three different scenarios have been implemented to demonstrate the validity of the method. The truncation effects of analog-to-digital conversion on localizing the FN have been identied; the dependence of the method on the resolution of smart meters are simulated. The method has been validated

PASDAR et al.: DETECTING AND LOCATING FAULTY NODES IN SMART GRIDS BASED ON HIGH FREQUENCY SIGNAL INJECTION

1075

Amir Mehdi Pasdar (S10) received the B.Sc. degree in electrical engineering from the University of Mazandaran, Iran, in 2004 and the M.Sc. degree in electrical engineering from Iran University of Science and Technology in 2007. He is currently working toward the Ph.D. degree in electrical engineering at the University of Akron, Akron, OH. He was a Technical Manager of ABB in Iran. His main research interests include real time high frequency impedance modeling of the power network, smart grid sensor design, and next generation smart grid applications.

static power conversion systems that interface energy storage, and distributed generation sources with the electric utility and smart grid. Dr. Sozer has been involved in IEEE activities which support power electronics, electric machines and alternative energy systems. He is serving as an Associate Editor for the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS Society (IAS) Electrical Machine Committee and secretary for the IEEE IAS Sustainable and Renewable Energy Systems Committee.

Yilmaz Sozer (M05) received his B.S. degree in electrical engineering from the Middle East Technical University Ankara, Turkey, and his M.S. and Ph.D. degrees in electric power engineering from Rensselaer Polytechnic Institute, Troy, NY. He has worked at Advanced Energy Conversion, Schenectady, NY, and developed expertise in all aspects of electronic power conversion and its control. He joined the faculty of Electrical and Computer Engineering Department at the University of Akron, Akron, OH, in August 2009, where he is developing a research and teaching program on alternative energy systems. His research interests are in the areas of control and modeling of electrical drives, alternative energy systems, design of electric machines, large industrial

Iqbal Husain (S89M89SM99F09) received the B.Sc. degree from Bangladesh University of Engineering and Technology, Bangladesh, and the M.S. and Ph.D. degrees from Texas A&M University, College Station. He is currently the ABB Distinguished Professor in the department of Electrical & Computer Engineering at North Carolina State University, Raleigh, engaged in teaching and research. He is the Co-Director of the Advanced Transportation Energy Center (ATEC), and a faculty member at the NSF FREEDM Engineering Research Center at NC State. His research interests are in the areas of control and modeling of electrical drives, design of electric machines, development of power conditioning circuits, and design and modeling of electric and hybrid vehicle systems. Dr. Husain received the 2006 SAE Vincent Bendix Automotive Electronics Engineering Award, and the 2000 IEEE Third Millennium Medal and the 1998 IEEE-IAS Outstanding Young Member award.