Analytical Methods for Power Monitoring and Control in an Underwater Observatory

Transcription

1 Analytcal Methods for Power Montorng and Control n an Underwater Observatory Tng Chan A dssertaton submtted n partal fulfllment of the requrements for the degree of Doctor of Phlosophy Unversty of Washngton 7 Program Authorzed to Offer Degree: Electrcal Engneerng

2 Unversty of Washngton raduate School Ths s to certfy that have examned ths copy of a doctoral dssertaton by Tng Chan and have found that t s complete and satsfactory n all respects, and that any and all revsons requred by the fnal examnng commttee have been made. Char of the Supervsory Commttee: Chen-Chng Lu Readng Commttee: Chen-Chng Lu Mark Damborg Bruce Howe Date:

3 n presentng ths dssertaton n partal fulfllment of the requrements for the doctoral degree at the Unversty of Washngton, agree that the Lbrary shall make ts copes freely avalable for nspecton. further agree that extensve copyng of the dssertaton s allowable only for scholarly purposes, consstent wth far use as prescrbed n the U.S. Copyrght Law. Requests for copyng or reproducton of ths dssertaton may be referred to Proquest nformaton and Learnng, North Zeeb Road, Ann Arbor, M 86-6, , to whom the author has granted the rght to reproduce and sell (a) copes of the manuscrpt n mcroform and/or (b) prnted copes of the manuscrpt made from mcroform. Sgnature Date

4 Unversty of Washngton Abstract Analytcal Methods for Power Montorng and Control n an Underwater Observatory Tng Chan Char of the Supervsory Commttee: Professor Chen-Chng Lu Department of Electrcal Engneerng The study of the undersea envronment requres the use of scentfc nstruments at the bottom of the ocean and n the water column to collect useful data. The tradtonal methods of conductng such studes by sendng shps or usng bottom nstruments and moorngs are not able to provde the necessary data over a long perod of tme due to weather and energy lmtatons. The objectve of the North Eastern Pacfc Tme- Seres Undersea Networked Experment (NEPTUNE) program s to construct an underwater cabled observatory on the seafloor of the northeast Pacfc Ocean off the coast of Washngton, Oregon, and Brtsh Columba, encompassng the Juan de Fuca Tectonc Plate. Ths system features over a few dozens of scence nodes for the connecton of scentfc nstruments that enhance our ablty to conduct contnuous ocean studes n ths regon. Ths dssertaton nvestgates mportant desgn and mplementaton ssues of the NEPTUNE power system. The power system assocated wth the proposed observatory s unlke conventonal terrestral power systems n many ways due to the unque operatng condtons of underwater cabled observatores ncludng the hgh relablty requrements and low observablty and controllablty. These unque aspects of the NEPTUNE system lead to the development of new hardware and software

5 applcatons that wll provde an essental and effcent operaton envronment. n ths dssertaton, the solutons to some of the techncal problems are proposed. The desgn of the Power Montorng and Control System (PMACS) allows PMACS to functon n a smlar way that Supervsory Control and Data Acquston (SCADA) and Energy Management Systems (EMS) are used to montor and control terrestral systems. A Fault Locaton algorthm s developed to dentfy a backbone cable fault by solvng nonlnear equatons wth only shore staton measurements. The same approach can be appled to underground power systems. n order to handle the request from the scence users to turn ther loads on and off, a Load Management algorthm s proposed based on nonlnear optmzaton. Ths algorthm takes nto account the dfferent prortes of scence node loads. The PMACS EMS modules requre dfferent parameters n the system model to provde accurate results. The effect of cable resstance varaton due to the temperature of the seawater s nvestgated and an algorthm s developed for PMACS to update the resstance values usng a quadratc programmng technque. The algorthms developed n ths research addresses challenges and dffcultes for an underwater observatory system. The results presented n ths dssertaton should be applcable to smlar underwater systems n the future.

6 Table of Contents Page Lst of Fgures... Lst of Tables...v Chapter : ntroducton.... Overvew of NEPTUNE.... Contrbutons of ths Dssertaton Power Montorng and Control System (PMACS) Fault locaton Load Management System Modelng.... Monterey Accelerated Research System..... MARS Fault Locaton Lab Test Results...5. Organzaton of Dssertaton...6 Chapter : Power Montorng and Control System...8. Overvew of Energy Management Systems...8. Overvew of PMACS Archtecture..... PMACS SCADA Functons..... PMACS EMS Functons...5. PMACS mplementaton Data Acquston and Control (DAC) Command Sequence enerator (CS) Alert Handler Data Collector Data Archvng Network Analyss and Control (NAC) Person Machne nterface (PM).... Summary... Chapter : NEPTUNE Fault Locaton.... Survey of Exstng Methods for Fault Locaton.... Fault Locaton Formulaton..... Fault Modelng Component Modelng System Modelng eneralzaton System Modelng for NEPTUNE.... Worst Case Analyss.... oltage Level Requrements...5

7 .5 Smulaton Results for the NEPTUNE System Software mplementaton of the Fault Locaton Module Summary...5 Chapter Neptune Load Management...5. ntroducton Operaton Modules of Load Management Securty Analyss System Restoraton On-Lne Operaton and Control Nonlnear Optmzaton Based Load Management Load Management for NEPTUNE...6. Test Scenaros and Numercal Results Scenaro Scenaro Scenaro NEPTUNE Power System User Contract Sample User Contract Summary...7 Chapter 5: Optmzaton Based Method to dentfy Cable Resstance Parameter dentfcaton for NEPTUNE Cable Resstance araton Optmzaton Based Cable Resstance dentfcaton Quadratc Programmng for Resstance Update Four Bus System Test Case aldaton of the Resstance dentfcaton Algorthm Smulaton Results for the NEPTUNE System Test Case Test Case Summary...89 Chapter 6: Concludng Remarks...9 Bblography...9

8 Lst of Fgures Fgure Number Page Fgure.: Essental elements of the NEPTUNE system... Fgure.: The NEPTUNE system... Fgure.: Connectons between the backbone cable and scence node...5 Fgure.: Recent desgn of NEPTUNE...6 Fgure.5: The MARS system... Fgure.6: MARS PMACS... Fgure.7: MARS PMACS external loads wndow... Fgure.: EMS archtecture 96's thru late 98's...9 Fgure.: Open archtecture of EMS... Fgure.: Overvew of PMACS... Fgure.: Power Montorng and Control System structure... Fgure.5: Master/remote staton nterconnectons... Fgure.6: Overvew of NEPTUNE PMACS components...6 Fgure.7: Data acquston and control node...7 Fgure.8: Network analyss and control... Fgure.9: PMACS console user nterface components... Fgure.: Fault model...6 Fgure.: Branchng unt...7 Fgure.: System topology wth node and lnk numbers...9 Fgure.: oltage requrements...7 Fgure.5: Fault locaton mplementaton for PMACS...5 Fgure.6: PMACS user nterface for fault locaton...5 Fgure.: oltage profle wth lmts...5 Fgure.: Load management modules...55 Fgure.: NEPTUNE system topology...6

9 Fgure 5.: -bus system v

10 Lst of Tables Table Number Page Table : Fault Locaton Lab Test Results...6 Table : Fault Locaton Results...8 Table : System parameter lmts...6 Table : Optmal Soluton for 8 Node System...66 Table : Soluton for Scenaro...67 Table : Soluton for Scenaro...69 Table : Smulaton Results...85 Table : Estmated Resstance for Test Case...88 Table X: Estmated Resstance for Test Case...89 v

11 Acknowledgements would lke to express my sncere grattude to my advsor, Professor Chen-Chng Lu, for hs gudance, patence, support and nspraton throughout my studes n pursung my PhD. Durng my study, he has provded me valuable suggestons wth hs expertse on varous techncal problems. was gven a hgh degree of freedom to explore the problems and solutons. Hs encouragements and suggestons have helped me to understand new perspectve on the problems have faced. would also lke to thank Dr. Bruce Howe, Professor Mark Damborg, and Professor Steve Shen for ther gudance as members of the supervsory commttee. would lke the thank the engneers n the NEPTUNE power group, Dr. Harold Krkham, Dr. Bruce Howe, Professor Mohammad El-Sharkaw, Tm Mcnns, Chrs San and others. Workng wth them has been a good experence for me whch showed me how a large project can be conducted wth good teamwork and organzaton. The work on the NEPTUNE power system was funded by the Natonal Scence Foundaton grant OCE 675, "Development of a Power System for Cabled Ocean Observatores". Specal acknowledgement and thanks go to the members of the Appled Physcs Laboratory for the development of PMACS, Tm Mcnns and Chrs San for ther work on the PMACS hardware, Mke Kenney for hs work on the PMACS Server and NPC software, and John Ellott for hs suggestons on PMACS Console desgn and mplementaton. Durng the tme pursued my research, receved a lot of supports and assstance from the fellow graduate students. My thanks go to Dr. Juhwan Jung, Dr. Sung Kwan Joo, Dr. Hao L, Dr. Kevn Schneder, Dr. uang L, Mr. Andrew Hartono, and Mr. Chung- Ln for ther frendshp. would also lke to thank Phlp Plgrm, Phl Lancaster, Wayne Beckman, and Hongru Lu for ther valuable suggestons n ths research. v

12 Last but not least, would lke to thank my wfe, Sara Shum, for her constant support, encouragement, and understandng throughout the entre process of ths research. My deepest grattude goes to my parents for ther love and support. v

13 DEDCATON To My Famly v

14 Chapter : ntroducton The study of the undersea envronment requres the use of scentfc nstruments at the bottom of the ocean to collect useful data. The tradtonal method of conductng such studes s to send a shp to the locaton of nterest to collect data. Due to the lmtaton of weather condton, space and tme, the nformaton that can be collected s very lmted. Long-term ocean observatory systems need to be bult so that contnuous electrcal power can be suppled to scence users. Ths enables the scentfc nstruments to operate over a much longer perod of tme wthout nterrupton. There are currently two types of ocean observatory systems: moorngs and cabled observatory. The frst type requres the placng of buoys and juncton boxes on the surface and the bottom of the ocean. Ths nstrumentaton has typcally used batteres or a generator placed nsde the buoys for ts electrcal power requrements []. The data collected are transmtted back to shore va satellte telecommuncatons. The battery lfe and fuel capacty severely restrcts the duraton as well as the effcency wth whch the studes are conducted. The second type of ocean observatory s desgned by connectng multple scence nodes at the bottom of the ocean by submarne telecommuncaton cables from the shore. Contnuous power s suppled to the scence nodes from the shore. The communcaton capablty also enables the collecton and transmsson of real-tme data. Three dfferent ocean cabled observatores are beng developed over the world: NEPTUNE n North Amerca, ARENA n Japan [], and ESONET n Europe []. The desgns are fundamentally dfferent from one another whle amng at dfferent requrements and tradeoffs []. For example, earthquakes occur perodcally n Japan snce t s located near plate boundares. As a result, one of the prmary requrements of the ARENA power system desgn s the ablty to contnue system operaton when there s a dsturbance such as catastrophc earthquakes or shunt cable fault to montor

15 the sesmc actvty. ARENE therefore adopted a constant current power feedng desgn whch s robust aganst cable faults. On the other hand, the NEPTUNE power system s based on a constant voltage power feedng approach. Ths desgn allows the system to delver a larger amount of power to the scence nodes than the constant current approach. n ths dssertaton, several aspects of the NEPTUNE system are dscussed n detals.. Overvew of NEPTUNE The objectve of the North Eastern Pacfc Tme-Seres Undersea Networked Experment (NEPTUNE) program s to construct an underwater cabled observatory on the floor of the Pacfc Ocean, encompassng the Juan de Fuca Tectonc Plate. The underwater scentfc nstruments connected to the system can be operated for longterm sustaned measurements wth real-tme two-way communcaton usng the observatory s fber-optc/power cable, facltatng a host of new expermental capabltes [5]-[9]. The deployment of NEPTUNE wll provde a wde range of scentfc data of oceanographc, geologcal, and ecologcal processes. Real-tme and archved data collected by the nstruments connected to the system at the bottom of the ocean and n the water column can be provded to scentsts, engneers, educators, decson makers, and learners of all ages wth the nternet []. The essental elements of the NEPTUNE system are shown n Fgure..

16 Fgure.: Essental elements of the NEPTUNE system Tradtonal terrestral power systems are normally AC networked parallel confguratons whle underwater telecommuncaton systems are normally DC seres cabled systems. The proposed NEPTUNE power system dffers from both of them n that the NEPTUNE power system s a DC networked system. t s planned to have approxmately km of cables wth shore statons (ctora and Nedonna Beach) and up to 6 scence nodes, as llustrated n Fgure.. At each of the shore statons, a k DC power supply wll be used to provde power that serves the entre system.

17 Fgure.: The NEPTUNE system The cable connectng the scence nodes s called the backbone. At each of the node locatons, a branchng unt (BU) s used to connect the backbone cable wth the scence node through a spur cable. The connecton wth the backbone cable, BU, and the scence node s shown n Fgure.. n case of a backbone or spur cable fault, swtches n the BU wll be opened to solate the fault so that the rest of the system wll reman n operaton. The power supply for the swtches nsde the BU s based on Zener dodes. The operaton of a BU does not requre explct communcatons from the shore statons or scence nodes []; several dfferent voltage levels provde the

18 5 mnmal necessary mplct communcaton. The desgn of the BU and the operaton of the swtches s presented n []. oltages and currents on the backbone are not known to the operaton center snce no measurng devce s nstalled. n the current desgn, the length of the backbone cable between each of the BUs ranges from tens of klometers to over a hundred klometers. The lengths of spur cables would be several to tens of klometers. Fgure.: Connectons between the backbone cable and scence node At each of the scence nodes, a DC-DC power converter s used to convert the voltage level down to and 8 for scence users. The loads at the scence nodes have a constant power characterstc due to the nature of the DC-DC converters. Changes n power levels do not have a sgnfcant mpact on the loads.

19 6 The orgnal proposed NEPTUNE system as shown n Fgure. has been scaled back sgnfcantly. Fgure. shows a recent confguraton []. Ths new desgn has approxmately km of backbone cables and about a dozen scence nodes. The algorthms developed n ths study are based on the orgnal desgn and can be easly adapted to the new desgn. Fgure.: Recent desgn of NEPTUNE

20 7 Stage of ths new desgn s presently under constructon by NEPTUNE Canada []. Ths regonal cabled ocean observatory wll be the northern porton of NEPTUNE. t wll be a -node, 8 km loop termnated at Port Albern. t wll use a hybrd seres-parallel power system. The seres porton wll power optcal repeaters and the optcal supervsory system that wll, among other functons, control the BU breakers.. Contrbutons of ths Dssertaton The desgn of the NEPTUNE power system nvolves a number of engneerng challenges due to ts physcal locaton and the nature of a DC networked confguraton: The repar/replacement cost of a component for an underwater observatory system can be very hgh. Therefore, a crucal desgn crteron for the NEPTUNE system s a very hgh level of relablty. The NEPTUNE system has to provde relable power and communcatons to the scence nodes for a lfe span of years [7]. t s also our goal to desgn and mplement the system so that mnmum mantenance s requred over the lfe span. The relablty requrement results n a system that uses smple desgns for the system components whch leads to the low observablty of system status. Ths dssertaton addresses a number of ssues related to the desgn and mplementaton of an underwater observatory system by developng new montorng and control technologes and computatonal methods... Power Montorng and Control System (PMACS) The montorng and control of a conventonal terrestral power system s handled by the Supervsory Control and Data Acquston System (SCADA) system and Energy Management System (EMS) [5]-[7]. As an underwater observatory system,

21 8 the NEPTUNE power system needs to be able to montor system parameters such as node voltages and swtch status and perform control operatons such as adjustng voltage outputs and swtchng of loads. Analytcal methods for varous system aspects need to be developed. For NEPTUNE, a Power Montorng and Control System (PMACS) s developed that combnes the functonaltes of SCADA and EMS. The purpose of PMACS s to provde computer, communcaton and software facltes for system operators to montor and control the system. As a result, a number of modules for PMACS are developed to provde functons such as fault locaton, state estmaton, load management and topology dentfcaton. The modules for state estmaton and topology dentfcaton are developed by Schneder [8]. The modules for fault locaton and load management are presented n later chapters of ths dssertaton. The desgn and archtecture of PMACS s dscussed n Chapter... Fault locaton One of the man challenges of the NEPTUNE power system s to dentfy the locaton of a backbone cable fault. Snce lmted resources are avalable for the development of a communcatons system of adequate relablty, t was decded that no communcaton would be avalable from shore statons to the branchng unts. As a result, voltages and currents on the backbone and the status of BU swtches are not known to the operaton center at the shore statons. Unlke tradtonal power systems, the fundamental assumpton s that no nstantaneous fault data wll be avalable snce no recordng devces and communcatons are avalable on the backbone cable. n the desgn of NEPTUNE, a sngle backbone cable fault would not cause a loss of any scence node n most locatons after the fault s solated. However, n case a secton of the backbone cable s mssng, the total power that can be delvered to scence nodes can be affected dependng on the fault locaton. Therefore, the faulted

22 9 cable has to be repared n order to allow the system to operate at full load. n case of a backbone cable fault, a repar shp s sent to the estmated locaton of the fault. Deep sea repars can be slow and costly; therefore, the Fault Locaton module of PMACS s ntended to locate a backbone cable fault to wthn ± km. Typcal terrestral power system technques requre the use of measurng devces such as Dgtal Fault Recorder and Phasor Measurement Unt [9]-[]. The analyss s done by observng the operatons of the crcut breakers and the transents of voltages. Due to the physcal sze of the branchng unts, measurng devces cannot be nstalled. The above technques cannot be appled for lack of these devces. The use of Tme Doman Reflectometry [] for submarne cable fault locaton cannot be used n ths system due to the networked confguraton and the sze of the system. The reflect sgnal s too small to be dstngushed from nose snce the fault mght be located thousand of klometers away from the source. n ths research, a fault locaton algorthm s developed based on the avalable measurements of voltage and current outputs from the shore statons durng a backbone cable fault. The algorthm uses knowledge of the system topology and solves for the locaton of the fault wth a set of nonlnear equatons. Whle typcal resstance estmaton method for underwater applcaton s done n a pont-to-pont manner, the algorthm developed n ths study generalzes the procedure by expendng the applcaton to a networked system. The development and mplementaton of the fault locaton algorthm s presented n Chapter... Load Management Load management s used n conventonal power systems to reduce the costs of operaton and ncrease relablty margn []. The methods are normally based on optmzaton technques by maxmzng the proft or mnmzng the cost. However,

23 for the NETPUNE system, the object of load management s to provde the maxmum amount of power to the scence node loads wthout volatng any system constrants. The total power that the NEPTUNE power system can supply to the scence nodes s lmted by the voltage outputs of the shore staton power supples and the current lmt of the backbone cable. The nomnal voltage output of the power supply at each shore staton s k. The backbone cables have a nomnal A current lmt. Therefore, the maxmum total power the system can provde at any gven tme s kw. Each ndvdual scence node can consume up to kw. t s clear that the system would not be able to smultaneously supply the maxmum load at every scence node. The loads at the scence nodes are categorzed nto dfferent prortes. Snce power s a lmted resource, t s to be reserved for loads wth hgh prortes n the case when all loads can not be served. A method for decdng the approprate acton s needed. For ths purpose, a Load Management module usng non-lnear optmzaton based technques s developed to determne the maxmum amount of power that the system can supply to the ndvdual scence nodes wthout volatng any of the system constrants. Ths algorthm takes nto account the load prortes at the scence nodes. The method ams at a dfferent objectve from exstng load management technques. The proposed algorthm s dscussed n Chapter... System Modelng System modelng and parameter dentfcaton are performed n power systems to ensure that the software modules of EMS provde a good representaton of the real system. To valdate the system models, real measured parameters of the system such as voltages and currents are compared wth the ones predcted by the model [] [5]. f a model fals to produce a good estmaton of the real system parameter, t needs to be tuned by adjustng the parameters of the model [6] [7]. The updatng

24 process of the model parameters are performed by varous numercal technques such as the auss-newton method [7] and enetc Algorthm [8]. The targetng models of these methods are usually used to predct the dynamc stablty and transents of the system. The EMS functons of PMACS requre a system model n order to provde accurate results that descrbe system operatng condtons. nstead of updatng the model to predct transents of the system, parameter dentfcaton s used to update the steady state system model. One of the parameters n the system model s the cable resstances of the backbone and spur cables. The cable resstance s consderably larger that the ones n conventonal terrestral power systems at about Ω/km. Due to the hgh resstance n combnaton of the physcal sze of the system, the voltage drop between the nodes s sgnfcant. A small varaton n per unt resstance can potentally lead to an unexpected scenaro f the system model does not reflect the change. The ncorrect system model can be dentfed by comparng the power outputs at the shore statons and the loads at the scence nodes wth the outputs from the power flow model wth nputs of voltages at the scence nodes. A cable resstance dentfcaton module based on the use of quadratc programmng s developed to solve ths problem. The detals are provded n Chapter 5.. Monterey Accelerated Research System The Monterey Accelerated Research System (MARS) project, headed by Monterey Bay Aquarum Research nsttute (MBAR), s near completon at the tme of ths wrtng and s scheduled for 7 nstallaton [9]. The purpose of MARS s to serve as a test bed for NETPUNE. The MARS system has one Shore Staton and one Scence Node as shown n Fgure.5. There are sea grounds at each end of the system,.e., the Shore Staton and the Scence Node). The cable s standard telecommuncatons cable (Alcatel OALC, 7

25 mm dameter core,.6 Ω/km) and the backbone communcatons technology s b/s Ethernet. The communcaton protocol s TCP/P. The prmary communcatons between the Node Controller and the PMACS uses the Ethernet provded by the Data Communcatons Subsystem (DCS). There s a secondary seral RS- communcatons channel for use durng operatons, n the case of a loss of the prmary communcatons system or for mantenance or troubleshootng. The Shore Staton contans a hgh voltage power supply from Unversal oltroncs (±5 k DC,. A) wth adjustable polarty, shore ground, the Power Supply Controller (PSC), and the PMACS server computer. The server s on the local area network, synchronzed by PS. Fgure.5: The MARS system

26 The mplementaton of the MARS PMACS s llustrated n Fgure.6. The same archtecture s used by NEPTUNE. PMACS s constructed wth a three-layer clentserver archtecture. At the lowest layer are the Node Power Controller (NPC) and Power Supply Controller (PSC), n the mddle s the PMACS Server, and on top are the PMACS Console and Clents. The NPC s conssted of one CPU board and four analog/dgtal /O boards. The PMACS Server s a HP ntel-based server wth RedHat Lnux. Fgure.6: MARS PMACS

27 The Console and Clent software s developed usng MS sual Basc.NET. The PMACS Console contans the man user nterface for the operator to nteract wth the actual system, and also provdes analytcal tools such as the Fault Locaton and State Estmaton module. The man wndow of the user nterface contans a one lne dagram of the MARS system. There are also multple wndows to show detals of the power subsystem components such as the external loads, nternal loads, engneerng sensor outputs, converter status, and power supply status. Many of these wndows also provde the ablty for the operator to control the status of the devces. Fgure.7 shows the external loads wndow for the MARS PMACS. There may be multple Clents but, at any gven tme, there must be one Console n communcaton wth the Server. The communcatons between the Console/Clent and the Server s usng Smple Object Access Protocol (SOAP). Fgure.7: MARS PMACS external loads wndow

28 5 The PMACS Server and NPC are able to acqure accurate absolute tme-of-day from the DCS. Network Tme Protocol (NTP) s used for tme-of-day, wth an accuracy of approxmately msec. The NPC, PMACS Server and PMACS Console are on the same NTP Server to make sure they are all synchronzed. Although the MARS system s much smpler than the planned NEPTUNE system, wth only one shore staton and one scence node, the PMACS archtecture and operaton phlosophy s the same. NEPTUNE and MARS have the same State Estmaton and Fault Locaton modules... MARS Fault Locaton Lab Test Results The Fault Locaton module of the MARS PMACS uses the same method descrbed n Chapter to estmate the locaton of a fault on the backbone cable. The communcatons protocols, devces, and PMACS hardware are the same for both NEPTUNE and MARS. Real-tme voltage and current measurements are taken at the Shore Staton by the PSC. The PMACS Server sends the data to the PMACS Console for the Fault Locaton module to perform the calculaton. A lab test has been performed to verfy the proposed fault locaton algorthm usng the MARS PMACS software and hardware. nstead of the actual hgh voltage power supply, a low voltage power supply s used n the lab envronment. nstead of the Power Supply Controller, a Node Power Controller wth smlar functonalty and accuracy s used. Resstors are connected n seres to smulate the backbone cable. The Node Power Controller measures the nput voltage and current. These measurements are acqured by PMACS and processed by the Fault Locaton module to obtan the (estmated) resstance. The results are shown on the PMACS Console. The test s conducted at two dfferent voltage levels: 75 and 8. Two dfferent fault scenaros are tested by usng dfferent values of resstances: 9.9 Ω and 5. Ω. Dependng on the type of cable beng used n Neptune and MARS, these

29 6 values represent the locaton of the backbone cable fault from the Shore Staton. Measurements are taken over a second tme span whch ncludes samples. The lab test results are shown n Table. The results show that the estmated resstances are wthn Ω of the actual resstances. These results ndcate that the estmated fault locaton from the proposed algorthm s wthn km of the actual fault locaton. Table : Fault locaton lab test results for MARS nput oltage Measured oltage Measured Current Estmated Resstance Actual Resstance A 9.8Ω 9.9Ω A 9.7Ω 9.9Ω A 5.55Ω 5.Ω A.98Ω 5.Ω. Organzaton of Dssertaton n ths dssertaton, detaled solutons and algorthm are provded ncludng the general archtecture of PMACS and the EMS modules of Fault Locaton and Load Management. The cable resstance dentfcaton module s also ncluded. The chapters are organzed as follows: Chapter descrbes the functonaltes and features of SCADA system and EMS. The desgn and archtecture of the NEPTUNE s dscussed. Chapter descrbes the fault locaton requrement of NEPTUNE and provdes detals of the formulaton and development of the Fault Locaton Module. Chapter dscusses the ssues of management of scence node loads and gves a detaled descrpton of the Load Management module.

30 7 Chapter 5 deals wth the varaton of cable resstance and explans how that would lead to a change of system modelng. The problem formulaton and techncal method of resstance dentfcaton are provded. Chapter 6 summarzes the contrbuton of ths dssertaton.

31 8 Chapter : Power Montorng and Control System. Overvew of Energy Management Systems Snce the New York Blackout of 965 power system operators have realzed the necessty of coordnated power system operaton. A drect result of the 965 blackout was the emergence of computerzed Energy Management Systems (EMS). Due to the avalable technologes of the tme, these systems could perform only a lmted number of calculatons per second. n order to compensate for ths shortcomng n the hardware, software was hghly optmzed and talored specfcally to a gven hardware platform. The result of the software optmzaton was that the varous components of the EMS were so ntertwned that they effectvely became a sngle unt. Fgure. shows the seral manner n whch data was handled n the early EMS systems [5]. n addton, t was standard practce to have separate computers run dfferent components of the system. The prmary reason for ths was that whle the network analyss calculatons, power flow state estmaton, etc., requred a large number of floatng pont calculatons, the nputs from the power system meters requred a large number of nterrupts. These two requrements were ncompatble and lead to the use of separate sets of computers, one set for the floatng pont calculatons requred for network analyss, another set for SCADA functons capable of a hgh number of nterrupts, and possbly a thrd for data base functons. The use of varous sets of computers compounded the software problems snce the code had to be optmzed to work wth varous combnatons of computers. The result of these systems, often referred to as Legacy systems, was that pecemeal upgradng or replacement of components of the system was not possble. So by the late 8's, utltes found themselves wth outdated hardware that could not be upgraded, resultng n a total, and costly, replacement of the EMS

32 9 Fgure.: EMS archtecture 96's thru late 98's Utltes are accustomed to dealng wth components wth a lfe span on the order of a few decades. Ths was not the case wth the EMSs systems that would become outdated n less than a decade. Ths problem was further compounded by the rate at whch hardware and software developed durng the 8's and 9's. The soluton that was offered was an open archtecture system. An open archtecture system contans 5 key concepts that dffer radcally from the prevous closed archtecture systems; portablty, nteroperablty, expandablty, modularty, and scalablty. Fgure. shows an open archtecture of the EMS [6]. Portablty: Refers to ablty of the software to run on dfferent software and hardware platforms. nteroperablty: Refers to the ablty to run dfferent software and dfferent

33 hardware together n the same network. Expandablty: Refers to the ablty to ncrease the sze of the system as well as the scope of the software. Modularty: Refers to the ablty to add new software functons wthout adversely affectng the rest of the system. Scalablty: Refers to the ablty to apply the same software to systems of varous szes. Fgure.: Open archtecture of EMS. Overvew of PMACS Archtecture NEPTUNE s equvalent of a SCADA/EMS s called the Power Montorng and Control System (PMACS), whch conssts of the computer software and hardware that controls and montors the NEPTUNE power system n a real-tme envronment. n keepng wth the open archtecture concepts of Fgure., the top level archtecture of NEPTUNE PMACS s shown n Fgure.. The power source n the system s the shore staton converter located wthn a faclty sted at the Shore Staton. The remote locaton s then connected to Operaton Center. From the Operaton Center the system

34 s connected to the nternet va varous frewalls. The system can be operated through three dstnct locatons; the Shore Staton, Operaton Center, and the Unversty of Washngton (UW). Data from the scence node enters the remote ste through the data communcatons subsystem and s routed to the local PMACS computer and to the local RF transmtter/recever by the L/L swtches. The RF lnk as well as the frewalls wll be transparent to the PMACS systems for ease of use. Fgure.: Overvew of PMACS PMACS s constructed wth a -layer clent-server archtecture as shown n Fgure.. The frst layer has the Node Power Controller (NPC) and shore Power Supply Controller (PSC) that nteracts wth the hardware n the scence nodes and shore statons. The mddle layer s the PMACS Server, whch s responsble for collectng

35 the power system data from scence nodes and shore statons, as well as ssung control actons receved from the PMACS console; t s the centralzed bran of the system. The thrd layer s the PMACS Console and Clent that communcate wth the server to gather system parameters such as shore staton power supply status, external and nternal load status, current and voltage measurements at each bus, converter status and engneerng sensor measurements. The PMACS Console dsplays the system data and s used to perform control actons, such as turnng ON/OFF a specfc load, through the user nterfaces. Fgure.: Power Montorng and Control System structure

36 .. PMACS SCADA Functons PMACS has Supervsory Control And Data Acquston (SCADA) capabltes. (The SCADA system s a standard remote montorng and control system for electrc power systems.) However, PMACS does not have remote control capabltes of the branchng unt (BU) breakers snce all the protecton logcs are mplemented at the local BU. Accordng to [7], a SCADA system should have at least one master staton and one remote staton. t s also common for a system to have several remote statons. For NEPTUNE, there are two ways of mplementng the master and remote statons. The frst s to have both shore statons servng as sub-master statons and have a thrd centralzed locaton to be the master staton. The scence nodes wll be the remote statons. n ths case, both sub-master statons gather system data and transfer the data to the master staton. The communcaton between the two sub-master systems s also necessary. The master staton does not communcate wth the remote statons. Another way to mplement ths s to have both shore statons servng as master statons. One of them s set to be the prmary master staton and the other one as the secondary master staton. The remote statons are the scence nodes whch gather and transmt data to the masters. Both shore statons need to be able to communcate wth each other as well as the remote statons. n the case of a prmary master staton falure, the secondary master staton wll take over. The connectons between master and remote statons are shown n Fgure.5. The master staton(s) conssts of dual computer systems: the prmary computer system and the backup computer system. Both computers are connected to the remote statons through the communcatons nterface. n case of a prmary computer system falure, the complete computer system s swtched to the backup unt.

37 Sngle Master, Multple Sub-masters, Multple Remotes Dual Masters, Multple Remotes SUB-MASTER REMOTE MASTER REMOTE MASTER REMOTE REMOTE SUB-MASTER REMOTE MASTER REMOTE Fgure.5: Master/remote staton nterconnectons The man/machne nterface (MM) or person machne nterface (PM) s a standard component of any SCADA system. t s defned as the way an operator nteracts wth equpment [7]. The nterface should nclude nformaton dsplays and control capabltes. nformaton dsplays can be best represented by a wndows-based graphcal user nterface (U) whch has a -lne dagram of the system wth multple levels of detals. The user should be able to clck on a system component to vew detaled nformaton. System components should also be color-coded to ndcate the status. n case of an abnormal operatng condton of the system, a vsual and/or audble ndcaton should be used to alert the operator. The operator should be able to perform standard PMACS functons such as State Estmaton and Fault Locaton through the U. Control capabltes of the MM should nclude the use of standard wndows nput devces such as mouse and keyboard. The scence nodes are servng as the remote statons. Analog data such as voltage and current measurements as well as dgtal data such as converter status wll be sent to the master staton or the sub-master staton based on confguraton through the communcaton system every second.

38 5.. PMACS EMS Functons PMACS not only serves as a data acquston system but also an Energy Management System (EMS). Energy Management System functons are performed at the PMACS Console. Outputs wll be sent to the PMACS Server and to the MM for dsplay. The PMACS Energy Management System conssts of four separate modules: State Estmaton, Topology dentfcaton, Fault Locaton, and Load Management. The Fault Locaton module and Load Management module wll be dscussed n detal n later chapters of ths dssertaton.. PMACS mplementaton PMACS consst of dstnct functons; person machne nterface (PM), network analyss and control, data acquston and control (DAC) node, and the PMACS nternal Archve. Fgure.6 gves an overvew of the components of PMACS as well as ther assocated subcomponents... Data Acquston and Control (DAC) There are two paths for communcatons n the NEPTUNE system; the prmary and secondary systems. The secondary, out-of-band, s lttle more than a 96 baud trouble shootng and mantenance system. The prmary system s a ggabt Ethernet connecton that wll normally transmt and receve all of the PMACS telemetered data, as well as scence data. Whle there s only a sngle fber path there are conceptually three paths for data to flow nto and out of PMACS. Fgure.7 ndcates that all of the data must go through one of the three paths; the command sequence generator, the alert handler, or the data collector. Data can ether be

39 generated on the prmary or secondary communcatons system, n ether case the data s processed n the same manner. 6 Fgure.6: Overvew of NEPTUNE PMACS components

40 7 Fgure.7: Data acquston and control node... Command Sequence enerator (CS) The command sequence generator handles all the commands that are generated wthn the varous sectons of PMACS. Commands can be generated from the PM, or the network analyss module. Although the communcaton system s extremely capable and relable, t cannot be counted on to be avalable % of the tme. Therefore, the possblty must be allowed for that communcatons wll fal durng the tme a command s beng sent to an nstrument or a subsystem. The effect of such a falure may be to leave the nstrument or subsystem n an unknown or

41 8 undesred state. Whle there are ways to desgn systems to recover autonomously from loss of communcatons stuatons, t s better to guard aganst unpredctable behavor by sendng not commands, but command sequences to the node. The equpment n the node s then programmed not to act untl a complete sequence has been receved at the node, and ts acknowledgement receved at shore. A sequence generator s therefore nserted between the command requests and the communcaton system. Normally, commands wll lead drectly to command sequences, and these wll be sent forthwth va the communcaton system. By sendng a command sequence nstead of smple commands, communcatons errors can be avoded. The command generator encodes each command wth a header and footer and sends t to the approprate node controller. The node controller wll only execute commands that are accompaned by a complete header and footer, thus preventng the executon of partal commands. All commands that are sent from the command sequence generator are stored n the PMACS nternal archve. The command sequence generator s capable of asynchronous communcatons.... Alert Handler The alert handler s the communcatons buffer that collects all of the alarms, messages, and other alerts from the scence node(s). Ths data s then passed on to the PM as well as beng stored n the PMACS nternal archve. The alert handler s capable of asynchronous communcatons.

42 9... Data Collector The data collector s the communcatons buffer that collects the power system data from the scence node(s) and the shore staton converter. The data s then sent to the network analyss and control module as well as the PM. The data collector polls the data from the scence node(s) and shore staton converter at a rate that s determned by an nput from the user va the PM, tentatvely Hz. The data wll be tme stamped wth a resoluton of µs, and the precson s ms.... Data Archvng PMACS wll mantan an nternal data archve of all system status, parameters, commands, load, schedules, etc. The Node status updates wll be synchronously read at a Hz rate and all ths data wll be wrtten to the archve. n addton, all asynchronous data commands, alerts, schedule changes, etc. wll also be wrtten to the archve. PMACS wll mantan nternal data archves at dfferent locatons one at the PMACS Server and one at the PMACS Console. All data that s wrtten to the nternal archve wll also be transmtted over the network to an external archvng system that can easly be replcated else where. All data wll be archved n comma separated ASC format... Network Analyss and Control (NAC) The network analyss and control module s where most of the power system calculatons are performed as shown n Fgure.8. Raw data from the power system s suppled to the NAC va the DAC and operatonal parameters are nput va the PM.

43 Commands and system status are then generated and suppled to the DAC, PM, and the Load Management module as well as beng stored n the PMACS nternal archve. Raw data s suppled to the state estmaton block whch determnes the state of the system, ncludng any unknown values. The state varables nclude the measurements of voltage and current at the nodes. These values, along wth the raw data are then used to determne the correct topology of the system. Snce state estmaton requres knowledge of the correct system topology there wll be feedback loop between the state estmaton and topology dentfcaton block. n addton t s possble for the topology dentfcaton block to generate a varaton n the shore staton voltage, wth approval va the PM, n order to determne the correct topology of the system. Fgure.8: Network analyss and control

44 .. Person Machne nterface (PM) The PM s duplcated at each of the PMACS control statons; Shore Staton, Operaton Center, and UW as shown n Fgure.9. There are three classfcatons for the operatons of the PM; dsplayed values, nput values, and operator actons. Dsplayed values are tems such as scence node voltage, scence node current, low voltage breaker postons, low voltage connector currents, and ground fault ndcatons. nput values are tems such as shore staton voltage, backbone low voltage warnng level, backbone low voltage lmt, backbone low voltage emergency control lmt, and backbone hgh current thresholds. Operator actons are tems such as approval for adjustng shore staton voltage level for topology dentfcaton and approval for load sheddng. The PM also allows for the opton of performng off-lne system analyss such as power flow and dynamc smulaton.

45 Fgure.9: PMACS console user nterface components. Summary n ths chapter, the functonal descrpton of the Power Montorng and Control System s presented. The archtecture of PMACS conssts of the NPC/PSC, Server, and Console. A detaled descrpton s provded for the ndvdual components and modules such as command sequence generator and alert.

46 Chapter : NEPTUNE Fault Locaton. Survey of Exstng Methods for Fault Locaton Tradtonal power system fault locaton technques nvolve the use of dfferent protecton or recordng devces, such as Dgtal Fault Recorders [9], Phasor Measurement Unts [], Dgtal Relays [], and Sequence of Events Recorders []. Fault locatng methods are normally based on transents n voltages and currents measured by these devces. The usage of these types of devces s not feasble for NEPTUNE due to the physcal sze lmtaton of branchng unts and scence nodes. A common method for dentfyng submarne cable faults s Tme Doman Reflectometry (TDR) []. The same method s used n underground dstrbuton systems []. These applcatons are used for cable lengths from hundreds of meters to tens of klometers. Faults on the NEPTUNE power system can be stuated klometers from the shore. f TDR s used, the reflected sgnal from the fault would be very weak. Furthermore, the branchng unt swtches and Zener dodes wll also generate a large number of reflected sgnals whch further complcate the process of dstngushng the sgnal from the nose. Hence, t s determned that TDR s mpractcal for NEPTUNE due to the attenuaton and network confguraton of the cable system. For typcal submarne cables, fault locaton can be conducted by applyng voltage and current at one end of the cable nto the fault and estmatng the resstance of the cable []. Ths method would not work for the NEPTUNE system snce t s a networked confguraton. n ths research, a new fault locaton algorthm s derved and mplemented for a networked DC power system wth low observablty. The proposed algorthm makes use of the voltage and current measurements taken at the shore statons. The approach

47 estmated the locaton of the fault based on the resstance taken nto account the networked system topology Underwater power systems requre a hghly accurate fault locaton technque due to the hgh cost of repar. The algorthm developed n ths research does not requre extensve montorng devces to be nstalled at varous locatons on the system. Smlar methods may be appled to some terrestral power systems such as underground dstrbuton systems. Underground dstrbuton systems need to be hghly relable snce they are usually located n urban areas wth a hgher densty of load. The dffculty of locatng or reparng an underground cable fault s sgnfcantly hgher than overhead lnes. The method descrbed n ths paper s a good addton to the exstng fault locaton technques such as TDR. The Electrc Power Research nsttute (EPR) has a project that uses a smlar concept [] to locate cable faults for rural dstrbuton systems. The research s based on the method descrbed n [5] whch uses a Feeder Montorng System (FMS) to record the voltages and currents on a feeder. Ths method uses the recorded fault current to compare wth a default value stored n a database to estmate the locaton of a fault based on the feeder mpedance. mpedance-based fault locaton technques are used n power systems. The most common mpedance-based methods are one-end and two-end methods [6]. The applcatons are for a sngle lne of AC systems. The method proposed n ths paper uses a smlar method whch s desgned for a networked DC system.. Fault Locaton Formulaton For the NEPTUNE power system, a backbone cable fault causes the entre system to shut down because of voltage collapse. The system then restarts wth the shore staton voltage at a low postve voltage, 5. Durng ths tme, all swtches n the BU wll close onto the fault. Snce the DC-DC converters at the scence nodes requre

48 5 an operatng voltage < 5.9 k, none of the converters wll be turned on at the low voltage level of 5 and, as a result, there s no load or communcaton n the system. The only crcut carryng currents conssts of the backbone cable and the fault n the system. oltage and current measurements are taken at both shore statons. These measurements wll be used by PMACS to determne the fault locaton. After PMACS takes all the measurements, the polarty at the shore power supply wll be reversed ( 5 ), a sequence that causes the backbone swtches to open and solate the fault. Servce to all loads at the scence nodes s then restored []-[] by applyng the full k. n the fault locaton mode, all swtches wll be closed and the fault pont s drawng all the current,.e., there are no other loads. Snce communcaton s not avalable, the only accessble operatng condtons are the voltage and current outputs at the shore statons. PMACS uses these measurements to estmate the total cable resstance and the dstances between shore statons and the fault. To locate a backbone cable fault, several addtonal factors need to be taken nto account: ) the fault characterstcs, ) fault resstance, ) topology of the system, ) cable resstance, 5) voltage drop along the cable, 6) measurement errors... Fault Modelng A shunt fault on a submarne cable occurs when the cable s nsulaton deterorates, allowng sea water to contact the conductor. Typcal causes of shunt faults are: Cable s abraded or partally cut. Ths can occur f the cable s dragged along

49 6 the sea floor by a shp s anchor, fshng gear or ocean currents and t sustans cuts and abrason on the rocky seafloor or outcrops. Trawlng s the man cause for ths knd of cable fault. Cable has a manufacturng flaw such as a vod or an ncluson n the nsulaton. f the feld at that pont s hgh enough, delectrc breakdown can occur. n most cases, the cable remans a sngle pece connectng to the ground wth some resstance, nstead of completely breakng nto two separate peces []. The fault on a gven lnk between two Branchng Unts (BU) can therefore be modeled by the confguraton n Fgure.. Fgure.: Fault model n Fgure., R_AB s the resstance of the cable lnk between branchng unts A and B. R_fault s the unknown fault resstance, and n and m are the unknown fractonal dstances from each of the BUs to the fault locaton,.e., n m =. The fault resstance can vary over a range from a few ohms to tens of ohms dependng on the condton of the damaged cable and how much conductor s exposed to sea water. Ths range s based on the fndngs and experence over years by the author of [] usng a fall-of-potental test.

50 7.. Component Modelng t s known that the nomnal resstance of the cable s Ω/km or.6 Ω/km dependng on whch of the two types of cable s adopted. Each secton wll be precsely measured n factory durng assembly at a known temperature. However, dependng on the actual temperature of the sea water, t could be a few percent lower or hgher. There s a temperature coeffcent assocated wth the cable that can be used to calculate the actual resstance based on the temperature of the water (whch wll lkely be avalable from ndependent measurements). The estmaton of cable resstance can also be done by State Estmaton [8], [7]. Fgure.: Branchng unt Besdes the cable resstance, constant voltage drops along each secton of the cable need to be consdered whle locatng the fault. Across each repeater on the cable,

51 8 there s a voltage drop. Snce a BU ncludes seres Zener dodes, as shown n Fgure., there s also a constant voltage drop across each BU. Assumng a BU as shown n Fgure., the voltage drop across the BU s calculated as n (.): = Re (.) BU Zener verse ZenerForward Where: BU : voltage drop across a branchng unt ZenerReverse : reverse bas voltage of a zener dode ZenerForward : forward bas voltage of a zener dode The zener dode reverse bas voltage s 6.9, and the forward bas voltage s.7. Therefore, the voltage drop across one BU crcut s 5.. The voltage drop across a repeater s 7.6. The total voltage drop for each secton of the backbone s the sum of the repeater voltage drop and the BU voltage drop... System Modelng As mentoned, the mnmum operatng voltage for the DC/DC converters n the scence nodes s -5.9 k; therefore, there s no load n the system durng fault locaton except the fault tself. Snce all swtches wll be closed onto the fault, the topology of the entre system s known when takng the measurements. The fault s not solated untl all measurements are taken. Snce the system s a meshed network, currents converge to the fault pont through multple paths. Snce the system topology s known, the equatons for each path can be wrtten takng nto account the unknown currents, and known cable resstances and voltage drops. Fgure. shows the system topology wth node and lnk numbers.

52 9 Fgure.: System topology wth node and lnk numbers... eneralzaton For a system wth a meshed structure, each branch corresponds to an unknown current. A fault from a lne to ground s also modeled as a branch. For a system condton under whch no external load s connected, the fault current s known and t s the sum of all nput currents. For a system wth multple sources, multple equatons can be wrtten based on the crcut parameters and the current flowng through each path. For a Y-shape branch, there are a total of three currents, but one of them can be expressed as the sum or dfference of the other two. When there s no external load, / of the branch currents can be expressed n terms of a known current and another unknown current(s). Ths procedure reduces the total number of unknowns n the system to / of the number of

53 unknown currents plus the addton of the fault resstance and faulted secton cable resstance. For a fault on a system wth multple sources to be determned, the total number of paths from the sources to the fault should be larger than or equal to the total number of unknowns for any gven fault n the system. Based on the result from the prevous paragraph, the total number of unknowns for the NEPTUNE system s 7,.e., fault resstance, faulted cable resstance fracton n, and the number of unknown currents on dfferent paths. There are sources and the number of avalable paths from the sources to anywhere n the system s larger than 7. Therefore, all fault locatons on the NEPTUNE system are well specfed. The equatons for the paths can be wrtten n the followng general form (.): SS = p j ( p j R p j faulted _ lnk nr D pj ) faulted _ lnk D _ faulted f R f _ lnk (.) where: SS : oltage outputs of Shore Staton, =, P j : j th path from shore staton to the fault, =, pj : Currents on lnks of the path R pj : Cable resstance of the lnks of the path Dpj : oltage drop across lnks of the path D_faulted_lnk : oltage drop across the faulted lnk fault_lnk\ : Current on the faulted lnk n: Per unt dstance of the faulted lnk R fault_lnk : Cable resstance of the faulted lnk f : Fault current R f : Fault resstance

54 n the proposed formulaton, the current drecton s assumed to be from the shore staton toward the fault. The equatons needed are chosen based on the shortest dstance paths from each shore staton to the faulted lnk. Current drectons on the shortest path wll apply to the next paths dentfed for loop analyss. Snce the cable resstance s assocated wth an error, the shortest cable length would ntroduce the smallest error. n PMACS, the paths are dentfed automatcally by shortest-path search.... System Modelng for NEPTUNE Now suppose a backbone cable fault s present on cable lnk 9 between nodes and 5. The voltage and current measurements from both shore statons are gven. Snce the topology s known, the loop equaton from each shore staton to the fault can be wrtten. For the loop equatons, path P ncludes cable lnks,,,, 5, and 6 and path P ncludes cable lnks 5,,,,, and. Each equaton s nonlnear wth unknown currents as n (.) and (.). The non-lnearty s due to the nature of the Zener dodes n the system. SS SS where: = = R R 5 R R R D 9 5 D D 9 ' D nr D 5 D 9 R 5 R 5 R R D 9 R D 9 f D 5 D 9 " D mr SS : oltage outputs of Shore Staton, =, R 6 9 R 6 D R 9 R D 6 f D (.) (.)

55 k : Current on lnk k, k = 5 D k : oltage drop across lnk k, k = 5 9 : Current on lnk 9 from Node to fault 9 : Current on lnk 9 from Node 5 to fault m, n: Per unt dstance of Lnk 9 R f : Fault resstance The faulted lnk can be expressed n per unt length such that: nr R (.5) fault _ lnk mr fault _ lnk = fault _ lnk Addtonal non-lnear equatons need to be wrtten by loop analyss from the shore statons to the fault va the next shortest paths from shore statons and, respectvely, as shown n (.6) and (.7). SS SS = = R 7 R R R R 5 R 7 D 9 D D D D7 D 9" D 5 mr R R R R R D 9 6 R D 5 f D D 6 R 7 D R 7 R D 5 R D 7 5 D D5 ( R D ) ( 5 R5 D5 ) ( 6 R6 D6 ) ( 7 R7 D7 ) ( R D ) ( R D ) ( R ) ( R ) 6 R R 6 5 D 6 D D R 7 D 9 R 7 9 ' D nr D D R R R f D D (.6) (.7)

56 Note that all lnk currents are unknowns; however, snce there s no load durng fault locaton, shore staton currents are feedng the fault pont. Therefore: SS SS = 9 (.8) where: SS : Current outputs of Shore Staton, =, From the topology of the system, t can be seen that: SS = = = = = (.9) 5 SS = = (.) = 5 = SS (.) 7 = = = 6 5 = 6 = 7 = 9" = (.) Smlarly, the current of any other lnk can be wrtten as an expresson of the known currents SS, SS and some unknown current(s). Substtute (.9-.) nto (.), (.), (.6), and (.7), the number of unknowns n the equatons s reduced. The number of non-lnear equatons needed to solve a fault on a specfc lnk s dfferent for each lnk. The number of non-lnear equatons should be less than the number of unknowns snce there s an unknown fault resstance. However, wth the addton of (.5), there s an equal number of equatons and hence the soluton can be found by numercal technques. MATLAB s used to solve the non-lnear equatons for the values of m and n.

57 . Worst Case Analyss When takng the voltage and current measurements at the two shore statons, each of them s subject to error. Ths error affects the result of the estmated resstance and hence the estmated fault locaton. To reduce the error effect, multple ndependent measurements should be taken at shore statons. Assume that the lne resstance s Ω/km. Snce the goal s to locate the fault to wthn ± km, the error n terms of resstance should be wthn ±Ω. f the error n resstance for the worst case can be contaned wthn ±Ω, the error n fault dstance would be smaller than ± km for any other cases. n ths study, a worst case analyss s conducted to determne the maxmum allowable voltage and current measurement errors. Note that the worst case resstance as a random varable and ts varances are gven by: = = max max R R R R R R σ σ σ σ where: max max R R R R σ σ σ σ σ σ σ σ = = = =

58 5 Now assume a non-worst case R < R max, remans the same snce t s the shore staton output voltage and hence the non-worst case current >. ' = R σ < σ R R max σ R = σ ' σ ' ' ' ( ) ( ) σ Usng the values for the NEPTUNE system, t s found that the worst case s a fault on lnk 5 snce the cable resstance and voltage drops are both the largest among all fault scenaros. Ths analyss suggests that f the algorthm can locate a fault on lnk 5 wthn ± km, t should locate any other fault on the NEPTUNE system wthn better than ± km.. oltage Level Requrements As mentoned, when a fault occurs, the system shuts down and then restarts wth a postve voltage. The Zener dodes have a knee current of about 5 ma. n ths regon, the voltage drop s proportonal to the current (and hence s not constant). Due to the nature of the system, some currents on the branches mght be very small. Snce there s no communcaton durng the fault locaton mode, the currents on the branches are unknown. Therefore, voltage outputs at the shore statons need to reach a suffcent level to ensure that all currents on the branches are large enough so that the Zener dodes wll have constant voltage drops. The shore staton voltage requrements vary when a fault s located on dfferent lnks. For a fault on a specfc lnk, there s a requred mnmum voltage to locate the fault to wthn km. There s also a maxmum voltage level for each specfc scenaro snce the maxmum current allowed on a

59 6 backbone cable should not exceed A. f the voltage at the shore staton s hgher than the maxmum allowable level, the backbone current exceeds A somewhere n the system. Durng restartng, sometmes voltage and current measurements for fault locaton are taken before the system goes back to normal operaton. n ths case, the faulted lnk s not known at the pont when measurements are taken. Therefore, the voltage levels to apply at the shore statons cannot be determned. nstead, current outputs at the shore statons are rased untl the sum of the two currents s close to A. Ths ensures that Zener dodes are operatng n the saturated regon, and the constrant of A s not exceeded. f the system operator decdes to go back to normal operaton wthout takng fault measurements and come back for the measurements at a later tme, the faulted lnk can be dentfed before the measurements are taken. n ths case, the system can apply a voltage level that would guarantee the suffcent level of current n the branches wthout volatng the current lmt. Fgure. shows the mnmum and maxmum allowable voltage levels necessary to resolve a fault locaton on a gven lnk to the desred accuracy. Notce that lnk and lnk 5 are connected to the shore statons. f the fault s located close to the shore staton, even a small voltage mght result n a hgh current. Snce the true fault locaton s not known, the maxmum voltage level can not be used n order to avod a current that exceeds A. nstead, the voltage at the shore statons s ncreased untl the current reaches 5 A. The correspondng voltage and current measurements are then used to perform the fault locaton.

60 7 Fgure.: oltage requrements.5 Smulaton Results for the NEPTUNE System The frst step to estmate the fault locaton for the NEPTUNE power system s to formulate the set of non-lnear equatons smlar to (.) and (.) for the proposed topology shown n Fgure.. Port Albern s Shore Staton, and Nedonna Beach s Shore Staton. For a gven fault, the fault locaton algorthm constructs the nonlnear equatons based on the dscusson n Secton... The faulted lnk can be dentfed by the algorthm descrbed n [8]. Although the constant voltage drops on the cable sectons are not shown on the fgure, ther values are taken nto account when formulatng the equatons. The number of equatons requred to solve for the fault locaton depends on the specfc faulted lnk.

61 8 Table shows some results of smulated cable faults on dfferent lnks wth dfferent fault resstances. A normally dstrbuted random error of zero mean and.% standard devaton s added to the voltage and current shore staton measurements. The calculaton has been performed tmes smulatng sets of ndependent measurements. Table : Fault locaton results for NEPTUNE Faulted Lnk Faulted Locaton Fault Estmated Fault Locaton Resstance 9 km from Node Ω 9.5km from Node km from Node Ω 9.7km from Node 6 5km from Node Ω.km from Node 7km from Node 7 Ω 7.6km from Node 7 8 km from Node Ω.8km from Node 5 km from Node Ω 8.6km from Node km from Node Ω.8km from Node 5 5.8km from Node 7 Ω 9.9km from Node 7 Assume that a fault s presented at the far end of lnk 5 to represent the worst case scenaro. When both shore statons have a voltage output of, s.5 A and s. A. A normally dstrbuted random error of zero mean and.% standard devaton s added to these smulated voltage and current measurements. The calculaton was performed tmes smulatng sets of ndependent measurements. When solvng the non-lnear equatons, t yelds an average soluton of n =.9958 and m =.. Snce the lne segment s 5 km long, the error n estmatng the fault locaton s m tmes 5 or.9 km. Therefore, t shows that from both shore statons would be a suffcent voltage level to handle the worst case. For faults n dfferent locatons n the system, the voltage level does not exceed.

62 9 As shown n Table, the estmated fault locaton s very close to the actual locaton n most cases. The only case where the algorthm does not meet the km requrement s the cable fault on Lnk 5, wth an error of. km. Ths could be due to the fact that Lnk 5 s very far from both shore statons yeldng large errors n measurements and the fault resstance s larger than other cases..6 Software mplementaton of the Fault Locaton Module As shown n Fgure.5, the mplementaton of the proposed fault locaton algorthm for PMACS requres the followng nformaton as shown n Fgure.5: ) faulted lnk dentty, ) real-tme voltage and current measurements from both shore statons, and ) system topology. The faulted lnk wll be dentfed by the Topology dentfcaton module of PMACS. PMACS wll set the voltage levels for the shore statons and measure the current outputs. The topology s stored n a database. Once the NEPTUNE system restarted after a shutdown due to a fault, the PMACS Fault Locaton module wll apply the knowledge of the system topology combned wth the measurements from the shore staton and the algorthm descrbed n the prevous sectons to dentfy the locaton of the backbone cable fault. Once the fault s located, PMACS wll adjust the shore staton voltage outputs to -5 so that the fault wll be cleared by the swtches n the Bus and the system can be restored to normal operatons. Durng these operatons, the PMACS Console s only able to receve data and send command to the shore statons.

63 5 Fgure.5: Fault locaton mplementaton for PMACS n ths study, a software module has been developed for the fault locaton functon. Fgure.6 shows the PMACS user nterface for the Fault Locaton module for NEPTUNE. Currently, the shore staton measurements are generated by smulated data. The measurements are processed by the fault locaton algorthm software. The estmated fault locaton s dsplayed through the user nterface.

64 5 Fgure.6: PMACS user nterface for fault locaton.7 Summary The algorthm developed n ths chapter s a full scale verson of the resstance estmaton method that s used n pont-to-pont underwater applcatons. The algorthm apples the avalable voltage and current measurements from the shore staton to dentfy the locaton of a backbone cable fault. t has the ablty to locate a cable fault n a meshed confguraton and does not have the lmtaton of cable length as t does for the TDR method. The same algorthm may also be appled n underground cable systems or HDC systems.

65 5 Chapter Neptune Load Management. ntroducton For conventonal power systems, the goals of load management are normally to reduce the operatonal cost or ncrease relablty margn of the system [], [9]-[]. However, for the NEPTUNE power system, the purpose of the Load Management module of PMACS s to determne the maxmum amount of load the system can serve wthout volatng any system constrant. Even though current battery operated oceanographc equpment often requres less than W, t s expected (and hoped) that the nomnal kw at a node wll be quckly utlzed for lghtng (e.g., for hgh defnton vdeo), battery chargng (autonomous undersea vehcles), pumpng of water (hgh volume chemcal samplng), and acoustcs systems (navgaton, communcatons and tomography). Thus, managng the lmted resource of power gven the varous constrants wll lkely be an mmedate challenge. The total power the system can provde s lmted by the maxmum voltage output of the power supples as well as the current lmt on the backbone cables. The nomnal voltage output of the power supply at each shore staton s k. The backbone cables have a nomnal A current lmt. Therefore, the maxmum total power the system can provde at any gven tme s kw. Each ndvdual scence node can consume up to kw. A porton of the power s to be used to supply the communcatons devces at the scence nodes whch are the nternal loads. Assumng the power delvered to the nternal loads s W, the peak power delvered to scence users s 9 kw at each scence node. t s clear that the system would not be able to smultaneously supply the maxmum load at every scence node. n the NEPTUNE power system, the amount of power beng delvered to the external loads at scence nodes s defned by contracts wth the scence users. The user

66 5 contract provdes specfcatons of the load ncludng the nomnal amount of power the load s to consume as well as the prorty of the load. Should the user s equpment develop a fault so that the power demand exceeds the agreed amount, control should be taken to lmt the power. PMACS wll determne f power consumpton should be lmted usng the acqured data. A related effort s the dsconnecton of users n the event that the power, a lmted resource, s to be reserved for hgh-prorty applcatons. For example, t may be decded that prorty should be gven to lghts and removed from battery rechargng when some sudden underwater event such as an earthquake s detected. To dsconnect the users, PMACS dentfes the approprate swtches to open so that servce to other loads wll not be dsrupted. PMACS mantans a lst of prortes of the loads and uses the nformaton on power consumpton and the agreement between NEPTUNE and users to determne whether t s necessary to shed load. Based on the power avalable, PMACS dentfes load devces that need to be shed for the operatng condton. There are three levels of prortes: hgh, medum, and low. The system wll try to serve the loads at a hgher prorty before attemptng to serve loads at a lower prorty. t s assumed that the complextes of sensor networks (e.g., tree and mesh structures) beyond the prmary scence nodes are unmportant here. Note that over-current condtons resultng from a fault should be handled by the protecton system. The task of the Load Management module n PMACS s to handle hgh currents due to overloadng. n the current desgn of NEPTUNE, the length of the backbone cable between each of the BUs ranges from tens of klometers to over a hundred klometers. The lengths of spur cables would be several to tens of klometers. Typcal cable resstance s -.6 Ω/km. Snce the resstances of long cable sectons are not small, the voltage drops from the shore statons to the remote locatons wll be sgnfcant. f the current on a km backbone cable between nodes equals A, the voltage drop between the nodes would be f the cable resstance s Ω/km. Each DC-DC converter at the scence nodes has an nternal control loop to regulate ts own load voltage.

67 5 However, the DC-DC converters servng the loads can not operate f the nput voltage drops below (n absolute value) 5.9k. Hence, PMACS must be able to montor node voltages and determne f t s necessary to rase or drop source voltages and/or drop load so that the entre voltage profle along the cable system remans wthn an acceptable range at all tmes. A voltage profle s llustrated n Fgure.. One of the constrants of the Load Management algorthm s to ensure that voltage levels at all scence nodes are wthn range. Fgure.: oltage profle wth lmts n a smlar manner, PMACS must montor the current profle along the backbone. The current lmt used for the backbone cable s A. n a normal condton, overcurrent would not occur. However, f one of the shore statons s out of servce and all loads are served from the remanng shore staton, load currents wll ncrease sgnfcantly for some sectons of the backbone, partcularly those sectons close to the shore staton n servce. The Load Management module takes nto account the current lmt as one of the constrants and determnes the new optmal operatng condton when the system topology s changed.