confirms that the electricity market can not be modeled as a perfect competitive market because the behavior of each company affects the profitmaximizing output of any other company and, most of all, the value of the correspondent marketclearing price. Conclusion:This letter represents an attempt to apply the oligopoly model features to the electricity market. For this purpose, a methodology has been developed, which is able to determine the profitmaximizing output for each supplier, taking into account the behavior of the companies that share the market. The effectiveness of the proposed model has been demonstrated on a test electricity industry containing three different suppliers. The results obtained clearly show how an oligopoly market model can better represent the electricity supply market. References: 1. S. Hunt, G. Shuttleworth, “Unlocking the Grid,” ZEEE Spectrum, July 1996. 2. D. Laider, S. Estrin, “Introduction to Microeconomics,” Philip Allan, 1989. 3. EL. Alvarado, “The Dynamics of a Power System Market,” PSERC publication, 9701. Copyright Statement: ISSN 02821724/99/$10.00 0 1999 IEEE. Manuscript received 24 February. This paper is published herein in its entirety.
IEEE PES Transactions Call for Papers The IEEE Power Engineering Society publishes three quarterly Transactions publications. /E€€ Transactions on Energy Conversion (TEC), /E€€ Transactions on Power Delivery (TPWRD), and /€E€ Transactions on Power Systems (TPWRS). Original contributions of lasting value to the profession are sought for publication in these Transactions. All contributions will be reviewed by a PES Technical Committee. Papers must be of unquestionably high quality, must be the original work of the authors, and must make a definite contribution to technical knowledge, particularly as it relates to the scope and interest of the Technical Committees. Locate author guidelines and submission requirements on the PES Web site, http://www. ieee.org/power, or request a copy of the revised Author’s Kit from PES Executive Office, IEEE Operations Center, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 088551331,U.S.A., (732)5623883,FAX (732)5623881,Email
[email protected]
June 1, 1999 Ekcironic Submission of TransamionsPapers Required
In support ofEEE’s trmsitmn t o fuuy elecfctllc pubbhtng on Januwg I , 2000, PES is rcqali auihors of T,ansacaons paperr to mbmt theu paperr as eletIcmi fles, beJune 1, 1999 Jure 1 was selected to allow the paperr to be rewewed and processed pnoi to IEEEs fanamon date Here
IS
genela1 information about the new procedures
A Fault Location Technique Using High Frequency Fault Clearing Transients Emmanouil Styvaktakis, Mathias H.J. Bollen, Irene Y.H. G u Author Affiliation: Dept. of Signals and Systems, Chalmers University of Technology, $412 96 Gothenburg, Sweden; Dept. of Electric Power Engineering, Chalmers University of Technology, S412 96 Gothenburg, Sweden Abstract: This letter suggests that a voltage recorder, placed next to a circuit breaker not as usual on the side of the substation, but on the side of the transmission line, may reveal the location of permanent faults. Two different approaches to estimating fault location are presented here: spectrum estimation and wavelet analysis. The methods are tested and compared with simulations of typical lransmission systems using the electromagnetic transients program (ITMTP). Keywords: Transmission lines, fault location, spectrum analysis, discrete wavelet transform. Introduction: Transmission lines suffer from fau Its that can damage the lines as well as equipment connected to them. Protection schemes are expected to operate and clear the fault b f opening the circuit breakers that connect the line with the source sicie of the network. The operation of some of these schemes is based on estimating the distance to the fault by measuring voltage and current at one end of the line. Different schemes are used for the fault clearing operation. These schemes involve either opening all three phases of the line or only those that suffered the fault. Fault clearing occurs at curreit zero. Usually a reclosing scheme is incorporated and closes the circuit breaker again. If the fault is temporary then the line will be back in operation. If the fault is not cleared then the circuit breaker has to open again. This may be repeated according to the reclosing scheme that is used. If all the reclosing attempts fail, the fault is permanent and the circuit breaker stays open. A repair crew should be sent to the fault poir t. For efficient dispatch of repair crews high accuracy is needed in determining fault location. Unlike protective relays, which need to detect whether a fault is in its zone of protection, fault locaion must be accurate to save time and expense of the crews, often searching in bad weather conditions or at places which are difficult I O reach. Different fault location algorithms have been proposed that use voltage and current data or techniques based on the detection of traveling waves [ l]. Proposed TechniquePrinciple: This letter cor siders two typical line configurations, shown in Figure 1. It shows that ~y measuring voltage on the lineside of a circuit breaker, the location of a permanent fault can be calculated using the transient caused by the fault clearing operation of the circuit breaker. In the case of a fault at a distance L, within the zcne of protection of the protective relay, circuit breaker 1 (CB1) receive:; a trip signal from the relay and opens to clear the fault. Usually operation time varies between a few cycles and more than one second. Due to the fault clearing operation of the circuit breaker, a surge is initiated and travels between an open circuit (circuit breaker open) and a short circuit (fault), if the fault is still present. When the surge hits the fault point, it is reflected with reversed sign and travels back to the open end of the line. Then it is reflected again from the open end but with the same sign and retums back to the fault point. Since the duration of this complete cycle is 42, (2 is the propagation time of the surge from the open er!d to the fault point) the main component of the voltage signal after the CBI opening has a frequency equal to [2]:
(Please see the Fehruaiy f!W v e r ~ o nof the PuLlrcaliaoGuide when auadabk far specrf,cs)
1 42
f =
Beginning June 1 , Tnwsaitions paper submttals must contain the following
* Ttie Transaction paper wbmiltal cover sheet  electronic :version and one p m e d copy * Two printed copies o f t h e payer (for reference and record piirpnsei) * The IEEE copyright form  hard copy only * The file far the o a o e i on a 3 5 f l o m v dink 01 L I D disk
58
The distance to the fault is:
IEEE Power Engineering Reoiew, May 1999
1
1
H ( z ) = = A(z)
1+
f:a$ k= I
L
V (b)
Figure 1. Single feed (@;doublefeed line (b)
1
I\
15
Faultstarts
1'
voltage involts
05
Y
0 0 51 '
Fault clcroing
15' 1 15
'
'
'
'
'
'
'
'
'

Figure 2. Voltageduring a fault clearing operation measured on the lineside of the circuit breaker for a double feed system
where c is the propagation speed of the surge. More frequency components are present due to other waves that propagate along the line, however, the frequency given by equation l is the dominant one. Simulations have been performed for the system shown in Figure la. The frequency dependent nonbalanced line model has been used in EMTP. Figure 2 shows the voltage on the lineside of CB1 during a fault, including the fault clearing operation (CB1 opens). Fault LocationFirst Approach, Spectrum Estimation: The estimation of the fault location related frequency in equation 2 is a task that should be considered separately because: Other frequencies, e.g., due to reflections of the initiated surge at elsewhere in the system, may be present making the estimation of the dominant frequency more difficult. This frequency must be estimated by using a small number of samples because the waves initiated by the circuit breaker operation are damped fast. During a circuit breaker operation, the three phases do not clear the current simultaneously. The current clears when it crosses through zero. When one phase is switched off, a transient is induced in the other phases posing an extra problem for the spectrum estimator. The accuracy of frequency estimation affects the accuracy of fault location so a method is needed that can provide high resolution. Three spectrum estimation methods [3] are considered and compared: Welsh's method: The power spectrum of a stationary signal can be estimated by using the discretetime Fourier transform and calculating the magnitude square of it. For a nonstationary signal the estimate is obtained by splitting the signal into segments, computing the spectrum for each segment and averaging these spectra. In Welsh's method the segments are overlapped and data in each segment is multiplied by a window function. Autoregressive (AR)Spectrum Estimation: This is a parametric method based on modeling the signal as the output of a linear system characterized by a rational system function: IEEE Power Engineering Revim, May 1999
where p is the order of the model. After ak are estimated, the power spectrum of the signal is computed from IH(o)l2.Different ways of estimating atexist. In this paper, Burge's approach was used, since the estimated AR spectrum has a high resolution in low noise data and a good spectral fidelity for short data records. 0 Multiple Signal Classification (MUSIC): This is a method of spectrum estimation based on eigenvalue analysis of the signal autocorrelation matrix. It categorizes the information based on the correlation of the data matrix, assigning information to either a signal subspace or a noise subspace. This method has been used for high resolution beamforming. CasesEMTP Simulations: The method has been tested for the above mentioned system configurations using the Electromagnetic Transients Program (EMTP). Several fault distances, fault types and fault resistances have been considered. The frequency dependent model (JMARTI's model) was used for modeling the 400kV transmission line. The length of the transmission line for both systems was 86 km, the propagation speed was calculated at 268.4 km/msec (using the line constants routines of EMTP) and a sampling frequency of 20 kHz was used in the simulations. The same length of data (200 samples) was used in spectrum estimation for all cases. The analysis window started at the data sample, which corresponded to the time instant of circuit breaker opening. For Welsh's method a Hamming window was used with 50% overlap. For Burge's approach on AR spectrum estimation order p=20 was used. For the MUSIC method the number of eigenvectors in the signal subspace was set to 8. Table 1 gives the results for several cases using the three spectrum estimation methods mentioned above. The results are acceptable and do not promote particularly any of these methods. The maximum error is 9 percent for short distance faults but less than 4 percent for most of the cases. Second Approach, Discrete Wavelet Analysis (DWT): DWT is a timefrequency signal analysis tool. It finds applications in different areas of engineering due to its ability to analyze the local discontinuities of signals. One way to interpret DWT is based on a filter bank shown in Figure 3. A filter bank consists of multilevel of highpass (H) and lowpass (L) filters that split the signal into the socalled details and approximations or different frequency bands. There are many different types of wavelet filters. In this letter Daubechie's wavelet of length 2
Doublefeed line, threephase to earth fault Doublefeed line, singlephase fault (fault resistance is 10 ohms) Doublefeed line, singlephase fault (fault ,resistance is 100 ohms)
64.16
63.00
63.90
61.49
64.16
63.90
63.90
61.33
64.16
70.00
71.38
74.33 59
m
I
I Aptoxmations
D e t d s at different levels
Slgnal
Figure 3. Filter bank interpretation of discrete wavelet transform ” 3”
Detmls of lowest level
1
Samples
I
Figure 4. Discrete wavelet transform of voltage (samplingfiequency is 40 kHz)
was used because of its good time localization properties. The time indexes associated with the local dominant peaks in the detailed signals are then picked up for the estimation of L in equation 2. For this case, a sampling frequency of 40 kHz is used. Consider the case of the double feed system and a singlephase fault at 64.16 km. The detailed signal from the lowest level of the DWT after circuit breaker opening is shown in Figure 4. Spike 1 corresponds to the instant that the circuit breaker opens and spike 2 corresponds to the instant that the reflected surge retums from the open end of the line. The time elapsed between spike 1 and spike 2 is twice the propagation time of the surge from the open end to the fault point. Knowing the speed of the propagation the distance to the fault can be estimated (63.75 km in this
2000 Modern Substations Conference and Exposition Call for Papers Abstract deadline: 1 September 1999 The PES Chapter of the IEEE Puerto Rico and CaribbeanSection will hold the Modern Substations Conference and Exposition in San Juan, Puerto Rico, 2428 April 2000. The event offers an excellent opportunity for presenting technical papers to one of the largest gatherings of intemational experts in the field. The wide range of topics will cover all aspects of power transmission and distribution substations. The conference theme is “Substation Technology for the Third Millenium.” Papers are encouraged on all topics related to new developments in substation technology. Authors wishing to make technical contributions are invited to submit papers. Title and abstract are due 1 August 1999, and the final manuscript is due 1 December 1999. Submit titledabstracts electro *ically (via email or on disk) to Ricardo L. Ramos, RG Engineering, Inc., 605 Condado Street, Suite 322 Santurce, P.R. 00907 USA. Phone: (787) 7234623, FAX (787) 7216678, Email r.l.ramosQieee.org. The file must include the paper tittle, abstract, author’s full name, address, phone number, FAX number, and Email address.
60
example). For detailed signals at higher levels these tM’o spikes are not so obvious as in the lowest level due to the lower :ime resolution. Therefore only detailed signals from the lowest level are used. The accuracy of this approach is mainly dependent on the sampling frequency of the voltage recorder. For a sampling freqLlency of 40 kHz and a propagation speed of 268.4 km/msec, one sample corresponds to 3.4 km (because the measured time corresponds to twice the distance in question). Therefore, the higher the sampling rate, the more accurate the fault location estimation. Conclusions: This letter proposes a new way of estimating the location of permanent faults in transmission lines. It requires a voltage recorder placed next to a circuit breaker on the side of the transmission line. The transient that results from the fault clearing operation of the circuit breaker is used to detect the position of the faLIt. Two different approaches have been described and tested using cases simulated in EMTP. The first approach is based on spectrum estimation. Three different spectrum estimation methods were used. The results are equally good for all three methods and not iromote any of them. The accuracy is dependent on the frequency resdution o f t c h spectrum estimation method. The second method utilixes the discrete wavelet transform to detect the discontinuity from the detailed signals. Its accuracy is highly dependent on the sampling frequency used. Acknowledgments: This work is supported by th Swedish National Board for Industrial and Technical Developmext (Nutek), Elforks, ABB Corporate research and ABB Network Partner. References: 1 . M.S. Sachdev (editor), Advances in Microprocessor Based Protection And Communications, IEEE Tutorial course, 9 7 TP 1200. 2. G.W. Swift, “The Spectra of Fault Induced Trarisients,” ZEEETransactions on Power Apparatus and Systems, vol. 98, no. 3, MayJune 1979, pp. 9407. 3. S.M. Kay, S.L. Marple, “Spectrum Analysis: A Modem Perspective,” Proceedings of the ZEEE, vol. 69, no. 11, hov. 1981, pp. 13801419. Copyright Statement: ISSN 02821724/99/$10.00 0 1999 IEEE. Manuscript received; 27 January1 999. This paper is pub1ished herein in its entirety.
1999 International Conference on Largie High Voltage Electrical Systems Colloquium on Turbogenerators, Hydmulic Generators, and Large Motors Lake Buena Vista, Florida, 810 September The ClGRE Rotating Machinery Study Committee 1 1, the IEEE PES Electric MachineryCommittee, and EPRI are organizing a joint colloquium on turbogenerators, hydraulic generators, and large motors. It will be held 810 September 1999, at the Buena Vista Palace Resort & Spa, Lake Buena Vista, Florida. The colloquium is open to members and nonmembers of these organizations. This 2.5day conference brings together users and equipment manufacturers (i.e., engineers from utilities and independent power producers, turbo and hydraulic generator and large motor manufacturers and designers). Attendees have the opportunity to share experiences on how to best use existing equipment, and how to best apply new rotating electric machinery in a deregulated environment. The language spoken at the colloquium is English. Papers will be posted on EPRlweb before the conference. To receive a registration packet, please contact: Michele Samoulides, conference manager, EPRI, 3412 Hillview Avenue, Palo Alto, CA 943041395, (650) 8552127, FAX (650) 8552166, Email msamouliQepri.com.
IEEE Power Engrneerzng Redrew, May 1999