International Journal of Energy and Power Engineering 4; 3(): 93- Publihed online May, 4 (http://www.ciencepublihinggroup.com/j/ijepe) doi:.648/j.ijepe.43.8 Protection cheme for tranmiion line baed on correlation coefficient R. Abd Allah Buraydah College, Faculty of Engineering, Electrical Power Department, Qaim Region, Kingdom of Saudi Arabia Email addre: Mohande_Ragab@yahoo.com To cite thi article: R. Abd Allah. Protection Scheme for Tranmiion Line Baed on Correlation Coefficient. International Journal of Energy and Power Engineering. Vol. 3, No., 4, pp. 93-. doi:.648/j.ijepe.43.8 Abtract: In modern digital power ytem protection ytem, tatitical coefficient technique i recently ued for fault analyi. A correlation technique i developed for fault detection and dicrimination. The propoed technique i able to accurately identify the condition of phae() involved in all ten type of hunt fault that may occur in extra high-voltage tranmiion line under different fault reitance, inception angle and loading level. The propoed technique doe not need any extra equipment a it depend only on the three line-current meaurement which are motly available at the relay location. Thi technique i able to perform the fault detection, type and phae election in about a half-cycle period. Thu, the propoed technique i well uited for implementation in digital protection cheme. The uggeted cheme i applied for a part of 5 Kv Egyptian network. Alternative tranient program (ATP) and MATLAB program are ued to implement the propoed technique. Keyword: Power Sytem, Fault Detection, Fault Claification, Correlation Coefficient. Introduction In modern power ytem protection, accurate, fat and reliable fault claification technique i an important operational requirement. On one hand, correct information of the type of fault i readily needed for fault location algorithm []. On the other hand, in digital ditance protection cheme, for proper operation of the protective relay, correct determination of the fault type i a prerequiite []. A ignificant amount of reearch work ha been directed to addre the problem of an accurate fault claification cheme. Among the variou technique reported for fault claification in tranmiion ytem, Artificial-neural-network (ANN) approach i the mot widely ued technique. Although the neural-networkbaed approache have been quite ucceful in determining the correct fault type, the main diadvantage of ANN i that it require a coniderable amount of training effort for good performance, epecially under a wide variation of operating condition uch a, ytem loading level, fault reitance, fault inception intance [3-9]. Similarly, Expert ytem-baed approach effectivene depend largely on the domain knowledge of the expert, which i often quite time-conuming to be obtained [-]. Fuzzy and fuzzy neural-network-baed approache alo require extenive training of the ANN [3-4]. Recently, fault claification technique uitable for a fault recorder ha been propoed in [5], which can identify all ten type of hort-circuit fault. Some tatitical coefficient technique were ued for fault analyi uch a an alienation technique wa developed for fault detection and dicrimination. The paper [6] preented protection cheme for tranmiion line baed on alienation coefficient for current ignal; the cheme ued another algorithm beide alienation coefficient to ditinguih between double phae and double phae-to-ground fault. In thi paper, a fault detection and election baed on correlation technique i propoed which i able to determine accurately, during one cycle period of fundamental frequency, all fault type and phae election. Alo thi technique take into conideration the wide variation of operating condition uch a witching, pre-fault power level, fault reitance and fault inception angle.. Fault Claification Strategy The fault election algorithm i baed on the autocorrelation technique of two half ucceive cycle with the ame polarity. For tranmiion line protection, thi
94 R. Abd Allah: Protection Scheme for Tranmiion Line Baed on Correlation Coefficient method need only three line-current meaurement available at the relay location (i a, i b, i c )... Correlation Coefficient Calculation The auto-correlation coefficient i etimated a follow for any two dependant variable, y (x) and y (x) [7]. The auto-correlation coefficient (r) calculated a follow: r = ( N x = y ( x) y ( x) N N N y ( x) N ( y) )( x = x = y y y ( x) N ( y ) ) Where, N = the number of ample per cycle ued in the imulation r: empirical correlation coefficient of y (x), and y (x). y (x): i the initial intantaneou value of the current at time t. y (x): i the intantaneou value of the current at next cycle. y : arithmetic mean of y (x) and y (x), repectively. y, N y = N y ) x= ( x N y = y ( x) N x = The trength of linear aociation between two variable i quantified by the correlation coefficient (r), it value lie between - and + [7]... Fault Detection and Faulty Phae Selection To implement our technique, three tak are tarting in parallel: fault detection, fault confirmation, and faulty phae election a follow:... Fault Detection (Initiation) A tranition i detected if: I > % I n, where I n i the line nominal current.... Faulty Phae Selection - Fault confirmation and faulty phae election are done according to the following equence. Three-phae current correlation coefficient value are calculated. If fault i detected, phae current correlation value are orted in acending order and compared. - If fault i detected, phae current correlation value are orted into acending order and compared. The poible fault cae are: (a) If the three-phae correlation coefficient are nearly equal and their value are le than.7, then the fault i three-phae fau - If r a r b r c <.7, the fault i three-phae (a-b-c fault) (b) If the two-phae correlation coefficient are equal and their value are nearly, while the third phae () () (3) correlation coefficient i le than.7, the fault i inglephae to ground fault. - If r a <.7, r b, r c, the fault i ingle phae- toground fault (a-g fault) - If r b <.7, r a, r c, the fault i ingle phae- toground fault (b-g fault) - If r c <.7, r a, r b, the fault i ingle phae- toground fault (c-g fault) (c) If the two-phae correlation coefficient are equal and their value are le than.7, while the third phae correlation coefficient i nearly, the fault i double phaeto-ground fault. - If r a r b <.7, r c the fault i double phae-toground fault (a-b-g fault) - If r b r c <.7, r a the fault i double phae-toground fault (b-c-g fault) - If r a r c <.7, r b the fault i double phae-toground fault (a-c-g fault) (d) If the three-phae correlation coefficient are not equal and their value: one phae i le than.3, econd phae i le than.7, while the third phae alienation coefficient i nearly, the fault i phae-to-phae fault. - If r a <.7, r b <.7, r c the fault i phae-to-phae fault (a-b fault) - If r b <.7, r c <.7, r a the fault i phae-to-phae fault (b-c fault) - If r a <.7, r c <.7, r b the fault i phae-to-phae fault (a-c fault) - To make ure of ditinguihing between double phae and double phae-to-ground fault, the cro-correlation between the two phae current of the faulted phae i calculated. If the value of cro-correlation i nearly -, the fault i double phae fault. - If r ab - the fault i phae-to-phae (a-b fault) otherwie the fault i double phae-to-ground (a-b-g fault). - If r bc - the fault i phae-to-phae (b-c fault) otherwie the fault i double phae-to-ground (b-c-g fault). - If r ac - the fault i phae-to-phae (a-c fault) otherwie the fault i double phae-to-ground (a-c-g fault). 3. Cae Study Power Sytem The propoed technique i applied on the power ytem hown in Fig.. The ytem parameter are obtained from one-generation unit in El-KURIEMAT power tation that produce 3 MVA [8]. The parameter of the elected ytem are a follow: Figure. The imulated power ytem.
International Journal of Energy and Power Engineering 4; 3(): 93-95 Machine (Sending Source): Rated line voltage i 9 kv, Volt-Ampere rating i 3 MVA, Frequency i 5 Hz, Voltage phae angle i and number of pole i. Machine (Receiving Source): Machine ha the ame parameter of Machine except the teady-tate voltage phae angle i. Main Tranformer: At each ide there i a tep up tranformer 34 MVA, 9.57/5 kv (Delta/Star earthed neutral), it primary impedance i.7 + j.84 Ω, it econdary impedance i.778 + j 6.8 Ω. Aux. Tranformer: At each ide there i an auxiliary tranformer 3 MVA, 9.57/6.3/6.3 kv (Delta/Star/Star earthed neutral), it primary impedance i.978 + j.4894 Ω, it econdary impedance i.39 + j.6 Ω. Line Tranmiion line (T.L.) impedance i.7 + j.3 Ω /Km with Km length for each circuit. Load Each load i repreented a impedance of value. + j6. Ω. 4. Simulation Reult A fault (F) wa conidered at the middle of one circuit of the tranmiion line auming that hort circuit i temporary and not reitive. The developed technique wa applied by calculating the auto-correlation coefficient (r x ) between two ucceive half cycle with the ame polarity of current ignal at the ending-end where relay would normally be intalled. The propoed technique i capable to dicriminate between the two type of two -phae fault either earthed or iolated. To confirm the two faulted phae type, a cro-correlation coefficient i calculated for the faulted phae. To implement the preent technique, the tudied power ytem configuration wa imulated by uing ATP oftware [9]. The generated and meaured three phae line current ignal are taken from the tranmiion line terminal at '' BB '' ide. Five imulation cae tudie are done to dicriminate between the faulty phae, by uing the propoed method, for fault claification. Thee cae are done under effect of different pre-fault power level, fault reitance, and fault inception angle located on the imulated power ytem. The current meaured ignal ampling rate i 5 ample per cycle, which mean ampling time of.4 msec. The total imulation time i Sec (i.e. the total number of ample i 5). The fault inception time i.4 Sec and the fault clearing time i.5 Sec from the beginning of the imulation time. 4.. Three Phae-to-Ground Fault (Cae ) Thi cae tudie the performance of the propoed technique during the three phae-to-ground fault condition on the propoed technique. The operating power angle of generator (δ ) i Degree. Figure how the imulation reult for cae ''''. Figure (a-c) preent the intantaneou value for the three phae current. In thi cae, it i noticed that the three phae current during the fault are higher than the pre-fault current; their value are nearly five time the pre-fault current. The correlation coefficient (r x ) are calculated between two ucceive half-cycle with the ame polarity for the three-phae current ignal. The three-phae current correlation coefficient r ia, r ib and r ic are hown in Fig. (d-f).the value of r ia, r ib, r ic are equal and cloe to one before fault inception and after fault clearing. At fault tart they are equal and le than.4 while they are le than.4 at fault clearing. From the above reult, it i clear that the correlation coefficient value at fault initiation are good detector to determine the faulted phae; Their value are cloely to one for current ignal in cae of normal operation and they are le than one in cae of fault condition. Summary of the correlation coefficient for the meaured current ignal at the different period are hown in Table. Table. Correlation coefficient in cae of three phae-to-ground fault (δ = degree, three phae-to-ground fault (a-b-c-g) with R f = ohm). Fault type Signal (r) pre-fault (r) at fault tart (r) during fault (r) at fault clearing (r) pot-fault Three line-toground fault ia ria =.4. ib rib =..35 ic ric =.3.4 4.. Single Line-To-Ground Fault (Cae ) All parameter are kept a in cae '''', except the fault type i changed from three line-to-ground fault (a-b-c-g) to ingle line-to-ground fault (a-g). Figure 3 (a-c) preent the intantaneou value for three-phae current. The faulty phae current value i nearly ix time the pre-fault current, while both the healthy phae current are equal and nearly.35 time the pre-fault value. The three phae current correlation coefficient r ia, r ib, r ic are hown in Fig. 3 (d-f). Their value are equal and cloe to one before fault inception and after fault clearing. At fault tart r ia i le than.74 while r ib and r ic are nearly one. At fault clearing, r ib i equal one while r ia, r ic are le than.74. From the
96 R. Abd Allah: Protection Scheme for Tranmiion Line Baed on Correlation Coefficient above reult, it i clear that the correlation coefficient value at fault initiation i good indicator to determine the faulted phae A. Summary of the correlation coefficient for the meaured ignal at the different period are hown in Table. Table. Correlation coefficient in cae of ingle phae-to-ground fault (δ = degree, ingle phae-to-ground fault (a-g) with R f = ohm). Fault type Signal (r) pre-fault (r) at fault tart (r) during fault (r) at fault clearing (r) pot-fault Single line-toground fault ia ria = -..5- ib rib = ic ric =.74.74 4.3. Double Line-To-Ground Fault (Cae 3) Thi cae tudie the effect of double line-to-ground fault condition on the performance of the propoed algorithm. Therefore, all parameter are kept a in cae '''', except the fault type i changed to double line-to-ground fault (a-c-g). Figure 4(a-f) how the imulation reult for three-phae current a the two faulty phae current are nearly ten time the pre-fault value while the healthy phae current i nearly.35 time it pre-fault current value. The value of correlation coefficient r ia, r ib and r ic are equal and cloe to one before fault inception and after fault clearing. At fault tart, r ia and r ic are le than the value of.7, wherea r ib i nearly one. At fault clearing, r ia and r ic are le than.78 wherea r ib i nearly one. From the above reult, it i clear that the correlation coefficient value at fault initiation can define the double line-to-ground fault. The cro-correlation coefficient r iac, calculated between the two faulted phae (A and C), at fault tart ha a value of -.85; thi value confirm the type of fault i double line-to-ground fault. Before fault inception, the value of cro-correlation coefficient r iac i equal to co( ) =.5, which i conidered a normal value. Summary of the correlation coefficient for the meaured ignal at the different period i hown in Table 3. From the obtained reult, it i clear that the correlation coefficient value at fault initiation i good detector to determine the faulted phae (A and C). Table 3.Correlation coefficient in cae of double phae-to-ground fault (δ = degree, double phae-to-ground fault (b-c-g) with R f = ohm). Fault Type Signal (r) pre-fault (r) at fault tart (r) during fault (r) at fault clearing (r) pot-fault Double line-toground fault 4.4. Double Line Fault (Cae 4) ia ria = -.5 -.7 ib rib =.7.78 ic ric =.7.38 ib & ic ribc = -.5 -.85 -.5 -.58 -.5 Thi cae tudie the effect of double line fault Condition on the performance of the propoed algorithm. Therefore all parameter are kept a in cae '''', except that the fault type i changed to double line fault (a-c). Figure 5(a-f) how the imulation reult for the three-phae current a the two faulty phae current are nearly ten time the pre-fault value, wherea the healthy phae current i nearly.35 it pre-fault value. The value of correlation coefficient r ia, r ib and r ic are equal and cloe to one before fault inception and after fault clearing. At fault tart, r ia, r ib and r ic have different value with value of -.55,.95 and.7, repectively. While at fault clearing, r ia, r ib, r ic have value of -.57,.95 and.38, repectively. The cro-correlation coefficient r ibc, calculated between the two faulted phae (A and C), at fault tart ha a value of, wherea it value i equal to.5 before fault inception. The value of cro-correlation coefficient, at fault inception, confirm that the fault type i phae-tophae fault. Conequently, our technique can determine the fault type whether double phae-to-ground or phae-tophae fault by calculating the cro-correlation between the two faulted phae. If r ac -, at fault tart, the fault type i phae-to-phae otherwie it i double phae-to-ground. Summary of the correlation coefficient for the meaured ignal at different period i hown in Table 4. Thee reult how that the correlation coefficient value at fault initiation i good detector to determine the faulted phae and ditinguih between phae phae iolated and grounded faulty without adding any extra meauring equipment. Table 4. Correlation coefficient in cae of double phae fault (δ = degree, double phae fault (b-c-g) with R f = ohm). Fault type Signal (r) pre-fault (r) at fault tart (r) during fault (r) at fault clearing (r) pot-fault ia ria = -.55 -.57 Double line fault ib rib =.95.95 ic ric =.7.38 ib & ic ribc = -.5 - - -.6 -.5
International Journal of Energy and Power Engineering 4; 3(): 93-97 4.5. Three Phae-to-Ground Fault with High Reitance (Cae 5) Thi cae tudie the effect of three-phae-to-ground fault with high fault reitance on the propoed technique. In thi cae, the applied fault reitance (R f ) i ohm. The operating power angle of generator (δ ) i degree. Figure 6 how the imulation reult for cae 5. Figure 6 (a-c) preent the intantaneou value for the three-phae current. In thi cae, it i noticed that the three-phae current during the fault are higher than the pre-fault current; their value are nearly four time the pre-fault current (thee value are le than that are in cae ''''). The auto-correlation coefficient (r x ) are calculated between two ucceive half-cycle for the three-phae current ignal. The three-phae current correlation coefficient r ia, r ib, r ic are hown in Fig. 6 (d-f).the value of r ia, r ib, r ic are equal and cloe to one before fault inception and after fault clearing. At fault tart they are equal and le than.65 while they are le than.74 at fault clearing. From the above reult, it i clear that the auto-correlation coefficient value at fault initiation are good detector to determine the faulted phae with high reitance. Alo, the correlation coefficient value at fault initiation are higher than that are in cae ''''.The correlation coefficient value are cloely to one for current ignal of healthy phae and they are le than one for faulty phae. Summary of the correlation coefficient for the meaured current ignal at the different period are hown in Table 5. Table 5. Correlation coefficient in cae of three phae-to-ground fault (δ = degree, three-phae to ground fault (a-b-c-g) with R f = ohm). Fault type Signal (r) pre-fault (r) at fault tart (r) during fault (r) at fault clearing (r) pot-fault Three line-toground fault with high reitance ia ria =.58.65 ib rib =.65.47 ic ric =.54.74 4 current ignal ia correlation coefficient(ria) -.8 ia(amp) -4-6 -8 corr(ria).6.4 -. - -4 6 4 (a) The current ia for cae. (d) ria for the current ia for cae. correlation coefficient(rib) 8 6 4 - corr(rib).8.6.4. -4 (b) The current ib for cae. (e) rib for the current ib for cae.
98 R. Abd Allah: Protection Scheme for Tranmiion Line Baed on Correlation Coefficient 8 current ignal ic correlation coefficient(ric) 6 4.8 ic(amp) - -4 corr(ric).6.4-6. -8 - (c) The current ic for cae. (f) ric for the current ic for cae. Figure (a-f). The imulation reult for cae, δ = degree, Three-phae to ground fault (a-b-c-g) 3 current ignal ia. correlation coefficient(ra)).8 ia(amp) - - -3-4 -5 corr(ra)).6.4. -6 -. (a) The current ia for cae. (d) ria for the current ia for cae. correlation coefficient(rb)) 5 5.8 corr(rb)).6-5.4 - -5. - 8 (b) The current ib for cae. (e) rib for the current ib for cae. current ignal ic. correlation coefficient(rc)) ic(amp) 6 4 - -4 corr(rc))..9.8.7.6.5.4.3-6. (c) The current ic for cae. (f) ric for the current ic for cae. Figure 3(a-f). The imulation reult for cae, δ = degree, ingle line-to-ground fault (a-g).
International Journal of Energy and Power Engineering 4; 3(): 93-99 3 current ignal ia. correlation coefficient(ra)).8 ia(amp) - - -3-4 -5 corr(ra)).6.4. -6 -. (a) The current ia for cae 3. (d) ria for the current ia for cae 3. correlation coefficient(rb)) 8 6 4 corr(rb))..9.8.7 -.6.5.4-4 8.3 (b) The current ib for cae 3. (e) rib for the current ib for cae 3. current ignal ic. correlation coefficient(rc)) ic(amp) 6 4 - -4-6 -8 corr(rc))..9.8.7.6.5.4.3 -. (c) The current ic for cae 3. (f) ric for the current ic for cae 3. -.4 correlation coefficient(ribc) -.45 -.5 -.55 -.6 corr(ribc) -.65 -.7 -.75 -.8 -.85 -.9 (g) riac for the current ia & ic for cae 3. Figure 4(a-g). The imulation reult for cae 3, δ = degree, double line-to-ground fault (a-c-g).
R. Abd Allah: Protection Scheme for Tranmiion Line Baed on Correlation Coefficient 4 current ignal ia. correlation coefficient(ria) 3.8.6 ia(amp) corr(ria).4. - -. - -.4-3 6 -.6 (a) The current ia for cae 4. (d) ria for the current ia for cae 4. 6 4 4 - - -4-4 -6 8-6 (b) The current ib for cae 4. (e) rib for the current ib for cae 4. current ignal ic. correlation coefficient(ric) 6. ic(amp) 4 - -4-6 -8 corr(ric).9.8.7.6.5.4.3 -. (c) The current ic for cae 4. (f) ric for the current ic for cae 4. -.4 correlation coefficient(ribc) -.5 -.6 corr(ribc) -.7 -.8 -.9 - -. (g) riac for the current ia & ic for cae 4. Figure 5(a-g). The imulation reult for cae 4, δ = degree, double line fault (a-c).
International Journal of Energy and Power Engineering 4; 3(): 93-8 current ignal ia. correlation coefficient(ria) 6 4.9 ia(amp) - -4 corr(ria).8-6.7-8 -.6 - (a) The current ia for cae 5 (d) ria for the current ia for cae 5. correlation coefficient(rib) 8 6.9 4 - -4-6 corr(rib).8.7.6.5-8 8.4 (b) The current ib for cae 5 (e) rib for the current ib for cae 5 current ignal ic. correlation coefficient(ric) 6 4.9 ic(amp) corr(ric).8 -.7-4 -6.6-8.5 (c) The current ic for cae 5 (f) ric for the current ic for cae 5 Figure 6(a-f). The imulation reult for cae 5, δ = degree, Three-phae to ground fault (a-b-c-g) with fault reitance (R f) = ohm. From the comparion between cae '''' and cae ''5'' we deduce the following finding: (a) The greater the applied fault reitance (R f ) the lower the DC component in the fault current (becaue of the lower primary time contant of the power ytem, (τ p =X L /ωr). (b) The greater the applied fault reitance (R f ) the greater the correlation coefficient value at fault initiation. (c) The greater the applied fault reitance (R f ) the greater the degree of power ytem tability. 5. Simulation Reult and Technique Evaluation From the different cae tudie, we ummarize the following: () Fault caue tranient of the tranmiion line current and a a reult caue collape for power ytem voltage magnitude. () The propoed technique i baed on the two type of algorithm for fault detection: Algorithm : Fault detection uing uperimpoed quantitie (delta algorithm) Algorithm : Fault detection uing auto-correlation coefficient value (3) Three tak are tarting in parallel for fault detection, fault claification and faulty phae election. (4) Auto-correlation coefficient between two ucceive half-cycle with the ame polarity for each phae current ignal can be ued to identify the faulted phae tatu. The cro-correlation coefficient between the two faulted
R. Abd Allah: Protection Scheme for Tranmiion Line Baed on Correlation Coefficient phae i ueful for determining the type of fault either double phae-to-ground or phae-to-phae fault. (5) An extenive imulation tudie are done to tudy the effect of different loading level, fault reitance (R f ) and fault inception angle on the performance of the propoed technique. The reult how that our technique ha high accuracy and efficiency for fault detection and claification with a wide range of fault reitance from zero value up to 5 Ω. 6. Concluion In thi paper, a correlation technique of tranmiion line for fault identification and faulty phae election ha been propoed. The main achievement of thi work are a follow:. Three line current meaurement are ufficient to implement thi technique.. It i accurate to identify all ten type of hort-circuit fault. 3. It i efficient to ditinguih between phae phae iolated and grounded faulty without needing any extra equipment. 4. The reliability of the propoed method i quite high. 5. It i quite effective over a wide range of a pre-fault power level, fault reitance, and fault inception angle. 6. Fat and imple method, a the time taken by thi method i about m (for a 5-Hz power ytem). 7. The effect of DC component and harmonic are eliminated with etimation of correlation coefficient. 8. The faulted phae can be determined by uing correlation technique. 9. The technique doe not ue the data of power ytem element but it need only three phae current meaurement available at the relay location. Reference [] M. M. Saha et al., A new accurate fault location algorithm for erie compenated line, IEEE Tran. Power Delivery, vol. 4, pp. 789 797, July 999. [] A. G. Phadke, Computer Relaying for Power Sytem. New York: Wiley, 988. [3] R. K. Aggrawal, Q. Y. Xuan, R. W. Dunn, and A. Bennett, A novel fault claification technique for double-circuit line baed on a combined unupervied/upervied neural network, IEEE Tran. Power Delivery, vol. 4, pp. 5 5, Oct. 999. [4] W.-M. Lin, C.-D. Yang, and J. H. Lin, A fault claification method by RBF neural network with OLS learning procedure, IEEE Tran. Power Delivery, vol. 6, pp. 473 477, Oct.. [5] T. Daltein and B. Kulicke, Neural network approach to fault claification for high peed protective relaying, IEEE Tran. Power Delivery, vol., pp., Apr. 995. [6] D. K. Ranaweera, Comparion of neural network model for fault diagnoi of power ytem, Elect. Power Syt. Re., pp. 99 4, 994. [7] K. H. Kim and J. K. Park, Application of hierarchical neural network to fault diagnoi of power ytem, Int. J. Elect. Power Energy Syt., vol. 5, no., pp. 65 7, 993. [8] A. L. O. Fernandez and N. K. I. Ghonaim, A novel approach uing a FIRANN for fault detection and direction etimation for high voltage tranmiion line, IEEE Tran. Power Delivery, vol. 7, pp. 894 9, Oct.. [9] A. Poeltl and K. Frohich, Two new method for fat fault type detection by mean of parameter fitting and artificial neural network, IEEE Tran. Power Delivery, vol. 4, pp. 69 75, Oct. 999. [] A. A. Girgi and M. B. John, Ahybrid expert ytem for faulted ection identification, fault type claification and election of fault location algorithm, IEEE Tran. Power Delivery, vol. 4, pp. 978 985, Apr. 989. [] C. A. Protopapa, K. P. Patira, and A. V. Machia, An expert ytem for ubtation fault diagnoi and alarm proceing, IEEE Tran. Power Delivery, vol. 6, pp. 648 655, Apr. 99. [] H. T. Yang, W. Y. Chang, and C. L. Huang, On line fault diagnoi of power ubtation uing connectionit expert ytem, IEEE Tran. Power Syt., vol., pp. 33 33, Feb. 995. [3] A. Ferrero, S. Sangiovanni, and E. Zapitelli, A fuzzy et approach to fault type identification in digital relaying, IEEE Tran. Power Delivery, vol., pp. 69 75, Jan. 995. [4] H.Wang andw.w. L. Keerthipala, Fuzzy neuro approach to fault claification for tranmiion line protection, IEEE Tran. Power Delivery, vol. 3, pp. 93 4, Oct. 998. [5] T. Adu, An accurate fault claification technique for power ytem monitoring device, IEEE Tran. Power Delivery, vol. 7, pp. 684 69, July. [6] M.E. Maoud, M.M.A. Mahfouz, Protection cheme for tranmiion line baed on alienation coefficient for current ignal, IET Gener. Tranm. Ditrib., Vol. 4, I., pp. 36 44. March. [7] W. Hauchild, and W. Moch, Statitical Technique for High Voltage Engineering, hand book, Englih edition publihed by peter pere grinu Ltd., London, United Kingdom, chapter, pp. 78-79, 99. [8] Intruction Manual for Generator Electrical Equipment, Upper Egypt Electricity Production Company, Elkureimat П 75 MW Combined Cycle Project, Steam Turbine Generator & Auxiliarie (Generator Electrical Equipment), Hitachi, Ltd., Tokyo Jaban. [9] ATP - verion 3.5 for Window 9x/NT//XP - Uer' Manual Preliminary Releae No.. - October.