An Adaptive Routing Algorithm Over Packet Switching Networks for

882 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010

An Adaptive Routing Algorithm Over PacketSwitching Networks for Operation Monitoring of

Power Transmission SystemsCheng-Long Chuang, Student Member, IEEE, Yung-Chung Wang, Chien-Hsing Lee, Senior Member, IEEE,

Maw-Yang Liu, Member, IEEE, Ying-Tung Hsiao, Member, IEEE, and Joe-Air Jiang, Member, IEEE

Abstract—This paper reports on the development and sub-sequent use of a packet switching communication model formonitoring of power transmission line systems. Modern fault de-tection and fault location techniques for EHV/UHV transmissionnetworks usually work based on the data measured by a phasormeasurement unit (PMU). Digital cameras have also been widelyutilized to monitor the physical status of power transmission lineson electricity pylons. PMU measures voltage and current phasorswith synchronized time stamps, and then transmits the measureddata to a monitoring center for analysis. The transmission of thesedata is required to be very stable. For the sake of operation speedand system security, the development of an enhanced communica-tion infrastructure that guarantees the quality of service (QoS) forthe essential measured data is a crucial issue. We have developedan adaptive routing algorithm for packet switching networks toguarantee the QoS of important power system communicationsand have conducted computer simulations to demonstrate theeffectiveness of the proposed algorithm. The proposed algorithmcan minimize transmission delay and reduce the number ofredundant transmissions caused by loss of packets. Hence, thesimulation results show the feasibility of packet switching net-works on monitoring power systems.

Index Terms—Adaptive routing algorithm, network fault toler-ance, packet switching network, power transmission system, QoS.

I. INTRODUCTION

I N POWER systems, the importance of fault detection andlocation systems for power transmission lines has increased

dramatically in recently decades. EHV and UHV transmission

Manuscript received June 03, 2008; revised August 21, 2008. First publishedMay 19, 2009; current version published March 24, 2010. This work was sup-ported by the National Science Council of the Republic of China under ContractsNSC 94-2213-E-002-120, 95-2221-E-002-398, and NSC 96-2628-E-002-252-MY3. Paper no. TPWRD-00431-2008.

C.-L. Chuang is with the Department of Bio-Industrial Mechatronics Engi-neering and the Institute of Biomedical Engineering, National Taiwan Univer-sity, Taipei 106, Taiwan, R.O.C. (e-mail: [email protected]).

Y.-C. Wang is with the Department of Electrical Engineering, National TaipeiUniversity of Technology, Taipei 106, Taiwan, R.O.C. (e-mail: [email protected]).

C.-H. Lee is with the Department of Systems and Naval Mechatronic Engi-neering, National Cheng Kung University, Tainan 701, Taiwan, R.O.C. (e-mail:[email protected]).

M.-Y. Liu is with the Department of Electrical Engineering, National IlanUniversity, Ilan 260, Taiwan, R.O.C. (e-mail: [email protected]).

Y.-T. Hsiao is with the Department of Computer Science, National TaipeiUniversity of Education, Taipei 106, Taiwan, R.O.C. (e-mail: [email protected]).

J.-A. Jiang is with the Department of Bio-Industrial Mechatronics Engi-neering, National Taiwan University, Taipei 106, Taiwan, R.O.C. (e-mail:[email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TPWRD.2008.2008494

lines play essential roles in delivering electrical power frompower plants to local transformer substations [1]–[3]. Occur-rences of faults on the transmission line system may influencea large region of end-users and result in millions of dollarsof economic losses. Hence, power companies have the respon-sibility to ensure the quality and reliability of their electricalpower transmission networks. If a fault is happening on one ofthe power transmission lines, power companies have to isolatethe faulty region from the entire power transmission networkin order to protect the power systems from systematic collapse.Any delay within the fault detection and localization system [4],[5] to react may increase the chance of a major blackout. Thereare two primary reasons that may cause the entire or a part ofa power system to collapse, which are: 1) logical errors in theprotection algorithm and 2) delays of packets within the com-munication system. Accurate detection of the fault type and thefault location is crucial to inspection, maintenance, and repair oftransmission systems. Since these algorithms depend heavily onspeed and reliability of communication systems [6], this studyfocuses on improving the second reason mentioned above.

In order to monitor the status of the power system, powercompanies usually install a certain number of measurementunits in the power transmission network [4]. We can obtain twotypes of information from the power system: 1) transmissionline parameters, such as receiving end/sending end voltage andcurrent phasors measured by PMU and 2) video surveillance ofthe transmission lines on electricity pylons recorded by camera.The measurement units capture these data and then transmitthem to a monitoring center for analysis.

Power companies prefer to use unified and simplified com-munication network structures in order to integrate the telecom-munication network with power system controls. The easiestand safest way is to use special-purpose communication lines,such as optical Ethernet, to connect directly all measurementunits with a central network hub in the monitoring center. Inthis way, the architecture of the communication network wouldbe a star-topology, and the power companies can minimize therisk of the entire network failing. However, the maintenance costof such a centralized network topology is high. In addition, thelink has no tolerance for any single link network failure; thus,essential data might be permanently lost on such an occurrence.

In this study, we present a packet switching network modelwith an adaptive routing algorithm and priority scheme thatcan transmit transmission line parameters and video surveil-lance streams with guaranteed quality of service (QoS) and pri-ority control. Packet switching techniques are used to achieve

0885-8977/$26.00 © 2010 IEEE

Authorized licensed use limited to: National Taiwan University. Downloaded on March 29,2010 at 12:35:28 EDT from IEEE Xplore. Restrictions apply.

CHUANG et al.: AN ADAPTIVE ROUTING ALGORITHM OVER PACKET SWITCHING NETWORKS 883

the maximal utility of the link capacity available in a telecom-munication network. It has proved very successful in wide-areanetwork cores. By taking advantage of packet switching tech-nology, we propose a new adaptive routing algorithm to enhancesystem fault tolerance and priority control, and to reduce systemdelay. The algorithm provides redundant links adjoined to thecommunication network without increasing the complexity ofthe network topology. With the proposed routing algorithm, thenetwork can transmit essential and general information aboutthe condition of power transmission lines with higher QoS. Fi-nally, we conducted a computer simulation based on the archi-tecture of the Taipower EHV/UHV transmission system.

This paper is organized into six sections, the first of which isthe introduction. In the Section II, we introduce the traffic modelincorporated with PMU, Global Positioning System (GPS), anddigital video cameras. Section III presents the packet-switchingbased communication model and the proposed adaptive routingalgorithm. In the Section IV, we introduce the Taipower 345kV power transmission network for performance evaluation.Section V presents the test results of the proposed communica-tion model and routing algorithm. Finally, we give conclusionsand a short discussion in the last section.

II. NETWORK TRAFFIC MODELS

A. Transmission Line Parameters at Power TransformerSubstations

In recent years, many previous studies have proved theeffectiveness of PMU in power protection systems [7]–[13].The distinct feature of PMU is that it is able to provide phasormeasurement of voltages and currents from widely dispersedlocations in a power transmission network. PMU measures thethree-phase voltages and currents, simultaneously. To avoidthe effect of delay in communication links, PMU requiresan external sampling synchronization device to become fullyfunctional [9]. There are two approaches to synchronize thetime on each PMU, GPS, and Synchronous Optical Networking(SONET). The GPS uses a constellation of Medium EarthOrbit synchronous satellites that transmit microwave signals toallow GPS receivers to determine their time. SONET uses twoclosely related multiplexing protocols that transfer multipledigital bit streams using laser diodes over the same optical fiber.GPS receiver is a portable device so it can be easily moved orreoriented. The accuracy of GPS time is 340 nanosecondsrelative to Coordinated Universal Time (UTC) on a singletime. SONET requires Ethernet links to operate. It is suitablefor power protection system if it starts with a highly accuratePrimary Reference Source (PRS) clock of Stratum-1 quality.However, installing a Stratum-1 PRS to the system is highlyexpansive and unsuitable for this study. In this study, sincethe pulse per second (PPS) signal in GPS has an accuracyranging from a few nanoseconds to a few microseconds, weintegrate a commercial GPS receiver with PMU to provide forthe sampling synchronization. Previous studies have verifiedthe validity and performance of sampling synchronizationof GPS-PMU configuration via field tests in substations ofTaipower 161-kV power transmission networks [8], [9]. Fig. 1

Fig. 1. One-line two-side simulated power transmission system that isequipped with GPS-PMU devices.

depicts a one-line two-side simulated power transmissionsystem that was equipped with GPS-PMU devices.

To measure the three phase voltages and three phase currentswith higher phase angle precision in a power system that has afundamental frequency of 60 Hz, we suppose that the samplingrate of GPS-PMU is samples per second. Weuse MATLAB/Simulink to simulate the power system, and thentake the simulated voltage and current waveforms as the syn-chronized sampled data (three phase voltages and currents) frompower transformer substations. GPS-PMU organizes the mea-sured transmission line parameters into a single packet. The ar-rangement of the payload of the packet is as follows: 1) sequencenumber ( , 2 bytes); 2) time stamp ( , 7 bytes); 3)GPS time ( , 8 bytes); 4) GPS status ( , 1 byte);and 5) measured transmission line parameters ( , 48 bytes).Since the total size of measured data is 66 bytes per sample, ac-cording to the specification of packet-switching techniques [14],we need to split each packet into two smaller cells. The packetshave a 53-byte fixed-size format (including 5 bytes of headerinformation). Thus, the network traffic generated by each GPS-PMU is (106 cells) per second.Fig. 2 shows the format of cells generated by GPS-PMU.

B. Weather Monitoring and Video Surveillance on Pylons

In recent years, wireless sensor networks have became an ac-tive topic in the fields of wireless communication and measure-ment instrumentation. A network of compact surveillance sen-sors is now available on the commercial market. Each sensormodule is equipped with a tri-axial accelerometer, temperaturesensor, and humidity sensor. We can integrate a digital cameraand a wind speed meter with the sensor module to make wire-less video surveillance and wind speed monitoring possible.

In the simulation part of this study, we suppose that everyelectricity pylon has a sensor module attached to it. The sensormodules can spot potential threats such as bad weather orearthquake. Digital cameras on the sensor modules providevisual observation of the status of power transmission lines. Thesensor modules feed information on what is happening aroundthe pylon back to the monitoring center for further analysis.

Since the average distance between two pylons is about 300m, the wireless sensor module needs an external high-gain an-tenna for connectivity. The sensor module constantly measures



Fig. 2. Format of cells generated by GPS-PMU, where cell (a) carries time-stamp information (NC represents “no content”), and cell (b) carries the trans-mission line parameters measured by GPS-PMU.

the environmental conditions around the pylon, and then orga-nizes these data into packets. The arrangement of the payload ofthe packet is as follows: 1) sequence number ( , 3 bytes); 2)timestamp ( , 5 bytes); 3) three-axis accelerations ( ,

, , 6 bytes); 4) temperature ( , 2 bytes); 5) humidity( , 2 bytes); and 6) wind speed ( , 2 bytes). Fig. 3shows the format of a packet generated by the wireless sensormodule. According to the application note provided by MoteivCorporation, each packet has a 30-byte fixed-size format (in-cluding 10 bytes of header information). Since these environ-mental conditions do not change rapidly, we assume that thesampling rate of the sensor modules is one sample per minute.Thus, a constant data rate for each sensor module is 30 bytesper minute. Furthermore, if the monitoring center detects a faulton a transmission line, the fault localization algorithm will lo-cate the fault, and then activate the digital camera of the sensormodule nearest to the fault location. The digital camera videouses a resolution of 320 240 pixels, and frame rate of oneframe per second. With this video surveillance configuration,the bit rate of the video stream is 50 kbps according to the con-strained parameters bitstream (CPB) profile. The sensor modulewill dissemble the video streams into smaller packets and gen-erate 313 packets per second. Thus, the network traffic gener-ated by a sensor module is per second.

The destination of the video stream packet is the nearest trans-former substation, and then the packet assembler at the substa-tion will reassemble the payload of the video stream packet intothe format of a cell used in packet-switching networks. Thus, thesubstation sends the environmental data and video stream backto the monitoring center by generating one cell per minute and196 cells per second to the packet-switching network, which are

Fig. 3. Format of packets generated by wireless sensor modules.

equivalent to 53 bytes per minute and 10 388 bytes per second,respectively.

III. COMMUNICATION MODEL AND ADAPTIVE

ROUTING ALGORITHM

A. Communication Model

The packet generated by GPS-PMU is crucial for powersystem protection. Thus, it requires stringent traffic QoS. Theenvironmental conditions measured around pylons provide dataonly for long-term analysis and have no immediate causal effecton the security of power transmission lines. The monitoringcenter activates the video surveillance function of the wirelesssensor modules only when something of interest has happened.Thus, the packets that carry video surveillance streams havehigher traffic QoS than those for environmental conditions.

In this study, we assume that all switches in the substationsare shared memory type asynchronous transfer mode (ATM)[14] switches with priority control. According to a previouslyproposed study, shared buffer ATM switches provide the bestperformance compared to other types of switches. When anATM switch receives a cell, the switch stores the cell at a log-ical output port queue. The priority control immediately putsthe memory address of the cell into the address managementfirst-in first-out (FIFO) queue of corresponding priority class[15]. Next, based on its destination the priority control selectscells to be output by referring to the memory addresses in the ad-dress management FIFO queues. Fig. 4 depicts the architectureof the shared buffer ATM switches. In the simulation, we assumethat a time slot for a packet switching cell is equal to 2.83 s at155 Mbps link speed, the size of address management FIFO is5000 cells, and the total size of the logical output port queue is50 000 cells. We denote a set of priority classes by A, B, and C,which correspond, respectively, to measured transmission lineparameters, video surveillance streams, and environmental con-ditions around pylons. Among these priority classes, class A hasthe highest priority.

B. Adaptive Routing Algorithm for Packet-SwitchingCommunication Network

Although the priority control is able to guarantee QoS foressential information, the network’s ability to maintain an un-obstructed data flow is another important issue. Fault-tolerancemeans that the network can operate properly in the event of thefailure of some links. This goal can be attained by constructingredundant links. Load balancing is another crucial technique todistribute the network traffic evenly across the communication



Fig. 4. Architecture of the shared memory ATM switches.

network. Thus, no hot spot (i.e., the switch that works with max-imal capacity) is created. In order to attain these functions, net-work administrators need to constantly update the routing tableon every switch. In this section, we present an adaptive routingalgorithm that enables the switches to automatically update itsrouting table to attain an optimal path for packets. The routingalgorithm is a standalone algorithm that allows the switches towork independently without any assistance from network ad-ministrators or control computers in the monitoring center.

We suppose that we constructed a packet-switching commu-nication network , where is a set of ATM switches,and is a set of links. Let be a power transformer substation,and let be the monitoring center. The communication network

in fact is a graphic model as depicted in Fig. 5. The substationis equipped with a GPS-PMU. The GPS-PMU puts the values

of transmission line parameters in a packet and sends the packetto the monitoring center via the packet-switching network. Inaddition, the substation also relays packets received from othersubstations (e.g., , , etc.) and wireless sensor modules (e.g.,

, , etc.) to the monitoring center. We classify connectionsinto two categories, constant bit rate (CBR) connections and un-specified bit rate (UBR) connections.

The proposed routing algorithm consists of two stages, an ex-ploration stage, and an evolutionary stage. First, we connect allswitches in according to the predetermined links in . The firsttask for the switch at substation is to identify its neighboringswitches (forexample, at , , etc.) andconstructaprivatecon-nection matrix to register all possible outgoing links. It alsoinitializes a goodness of links matrix to register the goodnessof outgoing links for forwarding a cell to the destination . Then,the switch enters the first stage of the routing algorithm.

In this study, the mission of the packet-switching communica-tion network is to monitor and transfer the operation status of the

Fig. 5. Packet switching communication network for monitoring and trans-ferring the operational status of power transmission systems. The dashed bluelines are wireless links, and solid blue lines are hardwired links in the packetswitching network.

power transmission system. Thus, the final destination of all in-formation is the computers in the monitoring center. Under thisassumption, Fig. 6 shows the flow diagram of the first stage (ex-ploration stage) in the proposed routing algorithm. The switchthat connected to the network sends out a number of set-up re-quest cells to the switch in the monitoring center . Whenreceives these cells, it sends back ACK cells to , and thensends an empty packet switching cell with timestamp to . Fi-nally, sends ACK cells back to , again, to report transmissiondelay times of the cells. The switch uses the average transmis-sion delay on each outgoing link to calculate goodness of linksmatrix by

(1)



Fig. 6. Flowchart of the exploration stages. Switch � explores possible links toswitch � by sending many set-up request cells. Then they use timestamp cellsto measure the time delay of each found path.

where is the goodness of sending a cell to the monitoringcenter by forwarding the cell to the next switch , is theaverage transmission delay of cells that have been forwarded to

. After all switches have completed the first stage, the switchsends a setup request cell to the monitoring center via the

outgoing link with highest to construct a priority class-ACBR connection. This connection is for transmitting cells gen-erated by GPS-PMU. The remaining outgoing links withsmaller becomes redundant links and serve as backuplinks in case of network failure of outgoing link . The con-structed connection has the highest QoS, and is always valid atall times. The routing algorithm resets the connection channelwhen the QoS of the channel has degraded and becomes unableto satisfy the communication demands of the power transmis-sion system.

In the second stage (evolutionary stage), we construct connec-tions for transmitting data generated by wireless sensor modulesto the monitoring center. Fig. 7 shows the configuration of thehybrid network that combines packet switching network and thewireless sensor networks. We take the switch in the substationas an example. It receives packets from wireless sensor modules,extracts the payload, and reassembles the payload into cells thatare compatible with the packet switching networks. When theswitch receives a packet that contains environmental conditionsor the starting packet of the video surveillance stream from thewireless sensor networks, the switch sends a setup request cellto the monitoring center . We initialize a proportion matrix ,and set the values in to . We define a function to calculatethe weights of outgoing links chosen to relay the information by

(2)

where is the proportion of the cell relayed to ,is bandwidth utility ratio of the outgoing link to , and and

are weighting factors that regulate the importance ofand during the calculation. The probability of choosing

as the next switch to relay the cell is

(3)

In (3), the proposed routing algorithm is able to achieve thefunction of load balancing because the higher , the lowerthe chance that has been chosen to be the next switch to relaythe cell. Therefore, the proposed routing algorithm is capableof preventing the sending of more cells to busy switches. When

Fig. 7. Configuration of the hybrid network that combines packet switchingnetwork and many wireless sensor networks.

the packet reaches the monitoring center, the monitoring centersends back the transmission delay time, denoted by , by anACK cell. is updated by

(4)

(5)

In order to keep the proportion matrix normalized, every timeafter it is updated, it must be normalized by

(6)

The value of is updated once the switch receives anACK cell of a setup request cell.

IV. TESTING SAMPLES OF POWER TRANSMISSION SYSTEMS

Fig. 8 depicts a real 345-kV power transmission networksystem encountered in Taiwan. In this study, we apply the pro-posed communication model and adaptive routing algorithm tothe power system in Fig. 8 using MATLAB/Simulink.

The power transmission system consists of three powerplants, thirty-nine lines, and twenty-four buses. We take ac-count of the parameters of the transmission lines during thesimulation. Table I summarizes the profile of the tested powertransmission system.

The monitoring system consists of four primary parts, GPS-PMU, the monitoring center, the packet switching network, andthe SQL database. In the simulation, we assume that each bus inthe power transmission system is equipped with a GPS-PMU re-gardless of the installation cost. Thus, the power system has 24GPS-PMUs attached to it. Fig. 9 shows the topology of the sim-ulated communication network. We can see that each switch hasat least two outgoing links to enhance the reliability of the com-munication system in case of network failures. Furthermore, thetotal length of the power transmission lines is 939.61 km. Wesuppose that the average distance between two pylons is 300m, and thus, 4705 pylons are required to hold the transmissionlines up in the air. According to this configuration, each pylonhas a wireless sensor module attached to it, and we integrate



Fig. 8. Topology of a real 345-kV power transmission network system encoun-tered in Taiwan.

the wireless sensor networks to the packet switching network asbranches connected to ATM switches.

To verify the performance of the proposed adaptive routing al-gorithm over the packet switching network, we have conductedextensive simulation studies for various network failures andnetwork traffic conditions. We discuss the simulation scenariosin the next section.

V. PERFORMANCE EVALUATION

In this section, we evaluate the performance of the pro-posed routing algorithm based on the traffic model and packetswitching network model described in previous sections. Thetransmission traffic consists of three types of data, transmissionline parameters, environmental conditions, and video surveil-lance streams. The GPS-PMU installed at each substationmeasure the three-phase voltages and currents on the bus,and sends the measurement results to the monitoring centervia a packet switching network. This type of data is essentialfor the protection system of power transmission network andrequires the highest priority traffic without excessive delaytime. The goal of this study is to test the robustness of the pro-posed routing algorithm when the network traffic encompassespackets from different kinds of networks.

To set up the simulation conditions, we refer to the specifica-tions of the ‘Cisco Catalyst 8540’ ATM switch. We assume thatthe ATM switch is equipped with one route processor and twoswitch processors, and the bandwidth of the shared memory is40 Gbps. The function of the two deployed switch processors

TABLE IPROFILE OF THE TESTING POWER SYSTEM

is to enhance the throughput. The maximum throughput of theATM switch is 24 million cells per second.

In order to reduce the cost of constructing the telecommuni-cation network, we build 22 short distance ATM transmissionlinks between 17 power transformer substations in the testedsample network. Each substation has no more than threetransmission links connected with other substations. Thus, thenetwork topology is simply a mesh network. The transmissiondelay time for transmitting a cell depends heavily on the speedof the ATM fabric and is calculated by

(7)

where the maximum delay time on the ATM fabric is 41.6 ns(switching 24 000 000 cells per second). The routing delay de-pends on the operating frequency of the route processor. In thissimulation, we assume that the routing delay is 10 ns. For ashared memory switch with 155 Mbps transmission speed, thequeueing delay is 2.83 s times the number of cells in the queueof the output port controller. The miscellaneous delay time isextra delay time caused by other factors, e.g., speed of electricsignals (speed of light). The channel release time is set to 1 min.Each switch can maintain a maximum of 65 535 channels.



Fig. 9. Topology of the real power transmission network (solid black lines) andsimulated communication network (dashed blue lines).

First, we evaluate the performance of the priority control.All substations transmission channels to the seventh substation(Panchiao substation and monitoring center). In the explorationstage, each node sends 250 set-up request cells in order to findthe possible paths to the Panchiao substation. The transmissionrate of each channel is set to 2 Mbps. The discovery result of theexploration stage is

The average delay times listed above are most likely the sameall the time since they carry essential information that requireshighest priority traffic. Traffic of priority classes B and C are notable to influence the QoS of priority class A cells. Thus, the av-erage delay times listed above are the delay times of the priorityclass A cells. The average delay times may change when theconnection path is released due to network failure, QoS degra-dation, or other factors. We will discuss this issue later.

The goodness of links matrix constructed in the explorationstage then becomes the proportion matrix in the evolutionarystage. The proportion matrix is the fundamental guideline totransmit priority classes B and C cells. The priority classesB and C cells are mainly from wireless sensor networks.Since these networks are based on wireless communication,the transmission rate of communication links between twowireless sensor modules is relatively limited if we compareit with the packet switching network. The delay time fortransmitting a packet is about 30 microseconds with 1 Mbpstransmission speed.

The video surveillance streams recorded by cameras fittedwith wireless sensor modules are transmitted by priority classB cells in the packet switching network. Since the bit rate ofthe video surveillance stream is ,the nearest connected ATM switch needs to create a connec-tion channel to the monitoring center with a transmission rateof 96 kbps. In this part of the simulation, we randomly activatecameras installed on the pylons. Most of transmission delay iscaused by the wireless sensor network because of the long dis-tances between the activated cameras and their nearest substa-tions. Since all wireless sensor modules need to report environ-mental conditions around the pylons in a fixed interval of time,each wireless sensor module generates a 30-byte priority classC packet, and transmits the packet to the nearest substation.The transmission delay time of the priority class C packet in-creases if administrators have activated a large number of cam-eras. However, once these wireless packets reach a substation,the packet switching network is capable of transmitting them tothe monitoring center with minimal delay time. The path of con-nections created for transmitting priority classes B and C cellsmay be different depending on the traffic condition in the packetswitching network. Fig. 10 shows the box-plot of average delaytime versus the number of activated cameras obtained in 1000repeats of simulations. An inspection of the figure clearly showsthat the transmission time delay of the proposed communica-tion model for transmitting video surveillance streams mightbe up to around 70 s when the number of activated cameras isaround 200. This is because the bandwidth of the WSN node isvery small. So, once we activate too many cameras, it will causea large transmission delay. However, once these WSN packetsreach the nearest substation, the ATM network will transmit thepackets to the monitoring center within microseconds.

Another problem is network failure. The network instrumentsutilized for power protection systems need to be very stable andreliable. However, in the case of network failure, we need extrabackup links to provide fault tolerance. Thus, each substationhas two or three communication links connected with other sub-stations. In Fig. 9, the 8th substation obviously is a hot spot inthe communication network.



Fig. 10. Box-plot of average delay time versus number of activated cameras obtained in 1000 repeats of simulations. In this figure, thin red lines represent thesamples that fall in the inter-quartile range, and thin black lines represent the samples that fall in the range of non-outlier observations. Red crosses represent outlierobservations, mostly caused by cameras at some distance from substations.

We suppose that all communication links connected with the8th substation are entirely failed during the simulation. The sub-stations that depend on the 8th substation to relay network traffic(6th, 9th, 11th, 12th, 14th, and 15th substations) send out 250setup request cells to find out an alternative path to transmit cellsto monitoring center. The experimental results are

where “ ” represents a modification of the connection path. Wecan see that the network failure occurring at the 8th substa-tion slightly affected the QoS of other communication channels.However, the time needed for the network to recover stabilityis lesser than 2 ms. Fig. 11 shows the delay time variances ofeach communication channel before and after the 8th substa-tion failed.

VI. CONCLUSIONS

In this study, we proposed a communication model for apower transmission protection system and an adaptive routingalgorithm for packet switching networks. By cooperatingwith ATM switches, the communication model encompassesa complete monitoring configuration for power transmissionswitching networks and wireless sensor networks to form asystem. In order to produce precise simulation results, the

Fig. 11. Delay time variances of each communication channel before and afterthe 8th substation failed.

detailed traffic model and network architectures are given todefine the entire system.

A real Taipower 345-kV transmission network was modeledfor the simulation. The simulation results have shown that theproposed algorithm is able to automatically discover the bestpath to transmit data to the desired destination, to collect datafrom different platforms, and is robust against network failure.Also, the priority control helps the ATM switches guarantee theQoS of essential transmission line parameters. Finally, the pro-posed algorithm and communication model have shown theireffectiveness for adaptive and priority controls.

ACKNOWLEDGMENT

The authors would also like to thank Prof. R. R. Bailey for hisproofreading of the manuscript and making a number of helpfulsuggestions.

REFERENCES

[1] B. Lian and M. M. A. Salama, “An overview of digital fault loca-tion algorithm for power transmission lines using transient waveforms,”Elect. Power Syst. Res., vol. 29, pp. 17–25, 1994.



[2] J. A. S. B. Jayasinghe, R. K. Aggarwal, A. T. Johns, and Z. Q. Bo,“A novel nonunit protection for series compensated EHV transmissionlines based on fault generated high frequency voltage signals,” IEEETrans. Power Del., vol. 13, no. 2, pp. 405–413, Apr. 1998.

[3] A. K. S. Chaudhary, K. S. Tam, and A. G. Phadke, “Protection systemrepresentation in the electromagnetic transient program,” IEEE Trans.Power Del., vol. 9, no. 2, pp. 700–711, Apr. 1994.

[4] K. P. Lien, C. W. Liu, C. S. Yu, and J. A. Jiang, “Transmission net-work fault location observability with minimal PMU placement,” IEEETrans. Power Del., vol. 21, no. 3, pp. 1128–1136, Jul. 2006.

[5] C. L. Chuang, J. A. Jiang, Y. C. Wang, C. P. Chen, and Y. T. Hsiao,“An adaptive PMU-Based fault location estimation system with a fault-tolerance and load-balancing communication network,” in Proc. IEEEPower Engineering Society—Power Tech 2007, Lausanne, Switzerland,Jul. 1–5, 2007.

[6] J. A. Jiang, C. S. Chen, and C. W. Liu, “Closure to discussion of a newprotection scheme for fault detection, direction discrimination, classi-fication, and location in transmission lines,” IEEE Trans. Power Del.,vol. 18, no. 2, pp. 653–655, Apr. 2003.

[7] J. A. Jiang, J. Z. Yang, Y. H. Lin, C. W. Liu, and J. C. Ma, “An adaptivePMU based fault detection/location technique for transmission lines,Part I: Theory and algorithms,” IEEE Trans. Power Del., vol. 15, no. 2,pp. 486–493, Apr. 2000.

[8] J. A. Jiang, Y. H. Lin, J. Z. Yang, T. M. Too, and C. W. Liu, “Anadaptive PMU based fault detection/location technique for transmis-sion lines, Part II: PMU implementation and performance evaluation,”IEEE Trans. Power Del., vol. 15, no. 4, pp. 1136–1146, Oct. 2000.

[9] T. M. Too, J. A. Jiang, J. Z. Yang, Y. H. Lin, and C. W. Liu, “De-sign, implementation and field tests for performance evaluations ofa brand-new phasor measurement unit based on global positioningsystem,” J. Chin. Insti. Electr. Eng., vol. 8, pp. 103–118, Apr. 2001.

[10] C. S. Yu, C. W. Liu, S. L. Yu, and J. A. Jiang, “A new PMU-based faultlocation algorithm for series compensated lines,” IEEE Trans. PowerDel., vol. 17, no. 1, pp. 33–46, Jan. 2002.

[11] C. S. Chen, C. W. Liu, and J. A. Jiang, “A new adaptive PMU basedprotection scheme for transposed/untransposed parallel transmissionlines,” IEEE Trans. Power Del., vol. 17, no. 2, pp. 395–404, Apr. 2002.

[12] J. A. Jiang, C. W. Liu, and C. S. Chen, “A novel adaptive PMU basedtransmission line relay—Design and EMTP simulation results,” IEEETrans. Power Del., vol. 17, no. 4, pp. 930–937, Oct. 2002.

[13] J. A. Jiang, C. S. Chen, and C. W. Liu, “A new protection scheme forfault detection, direction discrimination, classification, and location intransmission lines,” IEEE Trans. Power Del., vol. 18, no. 1, pp. 34–42,Jan. 2003.

[14] R. O. Onvural, Asynchronous Transfer Model Networks. Boston,MA: Artech House, 1995, pp. 112–117.

[15] L. Kleinrock, Queueing Systems. New York: Wiley, 1976, vol. II, pp.119–126.

Cheng-Long Chuang (S’04) received the B.S. degree in electrical engineeringand computer science and the B.S. degree in information engineering in 2003and the M.S. degree in electrical engineering in 2005, all from from TamkangUniversity, Taipei, Taiwan, R.O.C. He is currently pursuing the Ph.D. degreein bio-industrial mechatronics engineering and bio-medical engineering at Na-tional Taiwan University, Taipei.

He is an Adjunct Lecturer of Computer Science at National Taipei Universityof Education. His research interests are in the area of evolutionary computing,power systems, cognitive psychology, artificial intelligence, bioinformatics, andneuroscience.

Yung-Chung Wang received the M.S. and Ph.D. degrees in electrical engi-neering from National Tsing Hua University, Hsinchu, Taiwan, R.O.C., in 1990and 2000, respectively.

From 1990 to 2001, he was a Research Engineer with the Chung-HwaTelecommunication Laboratory, where he was engaged in research on thedevelopment of ATM switching systems and IP switch router systems. Since2001, he has been with the Department of Electrical Engineering, NationalTaipei University of Technology (NTUT), Taipei, Taiwan, where he is currentlyan Associate Professor. His research interests include wireless networks, opticalnetworks, queueing theory, and performance evaluation of communicationnetworks.

Chien-Hsing Lee (S’93–M’98–SM’06) was born in Pingtung, Taiwan, R.O.C.,on June 13, 1967. He graduated from National Kaohsiung Institute of Tech-nology, Taiwan and received the B.S. degree in electrical engineering from Ari-zona State University, Tempe, in 1993 and the M.S.E.E. and Ph.D. degrees fromthe Georgia Institute of Technology, Atlanta, in 1995 and 1998, respectively.

He is currently an Associate Professor at National Cheng Kung University,Taipei, Taiwan. R.O.C. His research interests are power system groundinganalysis, power system transient modeling, power quality, and applications ofwavelet theory in power systems.

Maw-Yang Liu (M’08) was born in Hualien, Taiwan, R.O.C., in 1970. He re-ceived the Ph.D. degree in electrical engineering and computer science fromNational Taiwan University, Taipei, Taiwan, in 2001.

From 1993 to 2006, he was with National Taiwan Police TelecommunicationAgency, where he worked on digital microwave radio, ATM, digital transmis-sion systems, and WLAN. He joined the faculty of the Department of ElectricalEngineering, National Ilan University, Ilan, Taiwan, in 2007. His main researchinterests include optical communication systems, spread spectrum techniques,and networking.

Ying-Tung Hsiao (M’92) received the B.S. degree in electrical engineeringfrom National Taiwan Institute of Technology, Taipei, Taiwan, R.O.C., in 1986and the M.S. and Ph.D. degrees in electrical engineering from National TaiwanUniversity in 1989 and 1993, respectively.

Subsequently, he joined the faculty of St. John’s and St. Mary’s Institute ofTechnology and was a Professor of Electrical Engineering at Tamkang Univer-sity, Taiwan. He is currently a Professor in the Department of Computer Scienceand Information, Nation Taipei University of Education. His research interestsinclude power system analysis, optimal theory, and motor control.

Joe-Air Jiang (M’01) was born in Taipei, Taiwan, R.O.C., in 1963. He receivedthe M.S. and Ph.D. degrees in electrical engineering from National Taiwan Uni-versity, Taipei, Taiwan, in 1990 and 1999, respectively.

From 1990 to 2001, he was with Kuang-Wu Institute of Technology, Taipei.Currently, he is a Professor with the Department of Bio-Industrial MechatronicsEngineering, National Taiwan University, Taipei. His research interests includecomputer relaying, mechatronics engineering, wireless sensor network, and bio-effects of electromagnetic wave.


Documents

An Adaptive Routing Algorithm Over Packet Switching Networks for