International Journal of Energy and Power Engineering 04; 3(3): 64 Published online May 30, 04 (http://www.sciencepublishinggroup.co/j/ijepe) doi: 0.648/j.ijepe.040303. A new schee based on correlation technique for generator stator fault detectionpart І R. Abd Allah, S. M. Mohaed, E. H. ShehabEldin, M. E. MASOUD Buraydah Colleges, Faculty of Engineering, Electrical Power Departent, Qassi Region, Kingdo of Saudi Arabia Helwan University, Faculty of engineering, Departent of Electrical Machines& Power Engineering, Cairo, Egypt Eail address: Mohandes_Ragab@yahoo.co (R. A. Allah), powervisionegypt@gail.co (S. M. Mohaed), shehab_eldin0@yahoo.co (E. H. ShehabEldin) To cite this article: R. Abd Allah, S. M. Mohaed, E. H. ShehabEldin, M. E. MASOUD. A New Schee Based on Correlation Technique for Generator Stator Fault DetectionPart І. International Journal of Energy and Power Engineering. Vol. 3, No. 3, 04, pp. 64. doi: 0.648/j.ijepe.040303. Abstract: In odern digital protection systes, a correlation technique is recently used for study and analysis. In this paper, a correlation technique is developed for fault detection, classification and direction discriination. The proposed technique is able accurately to identify the different types of faults that ay occur in synchronous generator stator winding. Three linecurrent easureents are required for both generator sides. The suggested technique perfors the fault classification task within halfcycle. Thus, the algorith is well suited for ipleentation in a digital generator protection schee. The used technique is applied for Elkurieat power station unit that produces 30 MVA. Alternative transient progra (ATP) and MATLAB progras are used to ipleent the proposed technique in this paper. Keywords: Power Syste, Generator Protection, Fault Classification, Correlation Coefficient, Internal Faults, ATPEMTP. Introduction Synchronous generator is the core of electrical power syste; once it is defected the network cannot continue working properly. Many types of faults will be occurring in the power syste and on the generator itself. So, there is a necessity to protect the generator fro those faults to liit the possible daage. Generator protection has dual protection objectives. Generator faults ay occur due to stator earth and phase faults, interturn faults, unbalanced stator currents, overheating, overvoltage, undervoltage, lossofexcitation, over excitation, over speed, generator otoring, rotor earth faults and other abnoral conditions, such as outofstep. The conventional protection schees for protecting large synchronous generators have any drawbacks, faults that ight take place near the neutral point of stator windings are not detected in case of high neutral grounding ipedance for generator, besides interturn faults are not usually discovered. In addition, the advent of large power stations and highly interconnected power systes akes rapid fault isolation to aintain syste stability. All faults associated with synchronous generators ay be classified as either insulation failures or abnoral running conditions []. An insulation failure in the stator winding will result in an interturn fault, a phase fault or a ground fault, but ost coonly the latter since ost insulation failures eventually bring the winding into direct contact with the core []. Differential relays, in particular the digital ones, are used to detect stator faults of generators. Electric power utilities and industrial plants traditionally use electroechanical and solidstate relays for protecting synchronous generators [3]. With the advent of digital technology, researchers and designers have ade significant progress in developing protection systes based on digital and icroprocessor techniques [45]. Several icroprocessor based algoriths for detecting stator winding faults have been proposed. Sachdev and Wind [6] developed an algorith that uses instantaneous differences between line and neutral end currents for detecting phase faults. Tao and Morrison [7] have used the discrete Fourier Transfor and Walsh functions to calculate the phasors of the fundaental frequency and third haronic voltages. An online digital coputer technique for protection of a generator against internal asyetrical faults is described by P. K. Dash and O. P. Malik [89] in which the
International Journal of Energy and Power Engineering 04; 3(3): 64 7 discriination against external faults is achieved by onitoring the direction of the negative sequence power flow at the achine terinals. This paper proposing a fault detection, classification and direction discriination schee based on correlation technique. The technique easures the currents at the both ends of stator windings and uses the calculated auto/crosscorrelation coefficients for the twoside currents of each phase, for aking relay trip or no trip decision. The suggested technique can operate accurately, during halfcycle period of the fundaental frequency.. ATP/EMTP Modeling of Synchronous Generator ATP library has any builtin odels including rotating achines, transforers, surge arrestors, switches, transission lines and cables [0]. The odel SM59 provides detailed dynaic odeling of synchronous achine. In addition to rated voltage, current and frequency, the odel needs d and q axis steady state, transient and subtransient reactance, value of oent of inertia, daping coefficient and nuber of poles. When dynaics of the achine is not required, sinusoidal voltage source odel Type4 can be used. Thus the package odel Type4 is preferred for odeling synchronous generator to analyze the internal faults located on stator windings; this is valid during the subtransient tie. The synchronous achine odel in ATPEMTP is used to do fault analysis to study the suggested protection schee based on correlation concept for fault detection. The proposed protection algorith is ipleented by using MATLAB package and is exained under different distances on stator windings and various types of internal, external faults that ay occur in the siulated power syste. 3. Proposed Technique In this paper, ATPEMTP software is used to get reliable siulation results during faults. The suggested protection schee is based on correlation concept in order to detect the faults on stator windings. Currents at both ends of each phase of the generator stator windings are easured and stored in a file where it is used for correlation estiation. On both sides of generator stator windings, the currents entering stator windings are first side as i a, i b and i c as shown in Fig.. The second side currents are i a, i b, and i c. Generated data easured by current eters at the both ends of stator windings is stored in a file; this data is in the discrete sapled for. These current saples of both sides are processed in MATLAB to estiate auto/crosscorrelation coefficients for twoside currents of each phase. This eans nine correlation coefficients are obtained for three phases, six autocorrelation coefficients and three crosscorrelation coefficients. 3.. Basic Principles Figure. Block diagra of proposed schee. This technique is based on auto/crosscorrelation coefficients estiation [] for each phase current signals ((i s, i s ). The linear correlation coefficient, in general r s, easures the strength and the direction of a linear relationship between two variables. It is soeties referred to as the Pearson product oent correlation coefficient in honor of its developer Karl Pearson, it is a diensionless quantity; that is, it does not depend on the units eployed. The value of correlation is such that < r s < +. The proposed differential protection algorith based on correlation technique has the following procedures: Read discrete sapled current of two sides (i s and i s ) for each phase s of stator windings (obtained fro ATP tool). Make digital soothing for sapled currents i s and i s for each phase s at neutral and load sides, respectively. The soothed sapled currents are shown in Eqs. (6). I s ( k ) = (5is ( k ) + is ( k ) is ( k + )) / 6 () I s ( k ) = ( is ( k ) + is ( k ) + is ( k + )) / 3 () I s ( k + ) = ( is ( k ) + is + 5is ( k + )) / 6 (3) I s (k), I s (k) and I s (k+): the soothed values of
8 R. Abd Allah et al.: A New Schee Based on Correlation Technique for Generator Stator Fault DetectionPart І sapled currents for phase s at neutral side corresponding to sapling instants k, k and k +, respectively. i s (k), i s (k) and i s (k+): the easured sapled currents for phase s at neutral side corresponding to sapling instants k, k and k +, respectively (Data obtained fro physical easureent often contains errors). I s ( k ) = (5is ( k ) + is ( k ) i s ( k + )) / 6 (4) I s = ( is ( k ) + is + is ( k + )) / 3 (5) I s ( k + ) = ( is ( k ) + is + 5is ( k + )) / 6 (6) I s (k), I s (k) and I s (k+): the soothed values of sapled currents for phase s at load side corresponding to sapling instants k, k and k +, respectively. i s (k), i s (k) and i s (k+): the easured sapled currents for phase s at load side corresponding to sapling instants k, k and k +, respectively. 3 Calculate autocorrelation coefficient (r is ) Autocorrelation coefficient (r is ) is calculated between two successive windows (each window has saples) shifted fro each other by one cycle interval for the instantaneous values of neutral current (i s ) of phase s. Eq. (7) shows the calculation of autocorrelation (r is ). is is ( k + N) i i ( k + N) s s k = k = r is = (7) ( ( is ) ( is ) ) ( ( is ( k + N)) ( i s( k + N)) ) r is : autocorrelation coefficient between two successive windows (each window has saples) shifted fro each other by one cycle interval for the instantaneous values of current (i s ) at first side, of phase s. N: the nuber of saples per cycle used in the siulation. 4 Calculate autocorrelation coefficient (r is ) The autocorrelation coefficient (r is ) is calculated between two successive windows (each window has saples) shifted fro each other by one cycle interval for the instantaneous values of load current (i s ) of phase s. Eq. (8) shows the calculation of autocorrelation (r is ). s s is is( k + N) is is ( k + N) r is = (8) ( ( i ) ( i ) ) ( ( is( k + N)) ( i ( k + N)) ) s r is : autocorrelation coefficient between two successive windows (each window has saples) shifted fro each other by one cycle interval for the instantaneous values of current (i s ) at (load) second side of phase s. 5 Calculate crosscorrelation coefficient (r is ) Crosscorrelation coefficient (r is ) is estiated between saples for a phase s current (i s ) signal at first side near neutral point and the corresponding saples for the sae phase s current (i s ) signal at second side near generator terinal. The selected saples are foursaples which are used as a correlated sall window in the differential protection function to obtain fast acting fault protection. The crosscorrelation coefficient (r is ) is calculated as shown in Eq. (9). i i s s s s r is = (9) ( ( is ) ( is ) ) ( ( is( k)) ( is( k)) ) r is : crosscorrelation coefficient for sapled currents between the two sides, neutral and load, of phase s for protected generator. s: the phase designation A, B or C. : the nuber of saples per window to be correlated used in the algorith (foursaples are selected in the algorith). i s (k): the sapled current values at instant k for neutral (first) side of phase s. i s (k): the sapled current values at instant k for load (second) side of phase s. 6 Tripping action of the algorith relies on the following conditions: (a) If r is and r is = (for all phases) & r is = (for all phases), then this case is noral operation condition. (b) If r is or r is < (for any phase) & r is = (for all phases), then this case is external fault condition. (c) If r is or r is < (for any phase) & r is < (for any phase) at the sae window and for the sae phase, then this case is internal fault condition. Tripping characteristic for the proposed differential protection is shown in Fig.. The characteristic illustrates two areas, blocking area in case of external fault condition and tripping area in case of internal fault condition. To confir internal fault occurrence, the crosscorrelation coefficient values are less than one with preset value (0.9 is selected as threshold value). The process of fault detection, faulted phase(s) identification and fault zone deterination either internal or external start altogether in parallel with a axiu execution tie of 0 sec. The crosscorrelation coefficient (r is ) deterines the faulted phase (s) and decides whether the fault is external or internal in case of r is <, whereas the autocorrelation coefficient (r is or r is ) deterines the fault condition in general, in case of r is < or r is <. But the later two coefficients do not decide the fault being external or internal and do not identify the faulted phase(s). The autocorrelation coefficient, r is or r is is used for fault occurrence confiration, discriination between healthy and external fault conditions and CT saturation detection besides the crosscoefficient r is. 7 Finally a tripping ode is selected for generation unit; i i
International Journal of Energy and Power Engineering 04; 3(3): 64 9 four ethods are used for isolating a generator once a fault has been detected [34]. They fall into four groups as follows: siultaneous tripping, generator tripping, unit separation and sequential tripping 3.. CT Saturation Detection The phenoenon of transient CT saturation during external faults causes false differential current in conventional differential relay; consequently a suggested algorith is introduced for CT saturation detection as follows: (a) If the auto/crosscorrelation coefficients (r is ) and (r is ) for phase s are suddenly decreased in the sae Generator condition Table. Auto/crosscorrelation coefficient ranges at different fault conditions. r is Crosscorrelation window (siultaneously), an internal fault is detected. (b) If the autocorrelation coefficient (r is ) is firstly dropping whereas the crosscorrelation coefficient (r is ) decreases later (shifting tie of T sat for phase s ), the fault is then an external fault with CT saturation. The interval (T sat ) is defined as the tietosaturation fro the beginning of fault detection, where it is supposed to have a iniu tie of.5 sec (/8 cycle) in case of severe CT saturation. Auto/crosscorrelation coefficient ranges (liits) at different fault conditions for synchronous generator are included in Table. r is or r is Autocorrelation Relay action. Noral operation for all 3phase for all 3phase Blocking. External fault for all 3phase < at least one phase Blocking 3. Internal fault < at least one phase < at least one phase Tripping 4. External fault with CT saturation (assue only one side CT saturated) (during the interval fro fault inception to CT saturation starting) at least one phase < at least one phase Blocking 4. Power Syste Description The single line diagra of power syste under study is shown in Fig. 3. The three phase wiring of the synchronous generator (Type4) used in siulation is shown in Fig. 4. The synchronous generator odel akes reference to the work described in article [57]. Their proposed siulation of internal fault includes the addition of two voltage sources of reverse polarities to each other in the faulted phase. Also each of the generator reactance in the EMTP reactance data card are reduced by a value of x which is equal to the subtransient reactance of the sound portion of the windings. Figure. Tripping Characteristic for the Proposed Differential Protection. Figure 3. Single Line Diagra for the Studied Power Syste.
0 R. Abd Allah et al.: A New Schee Based on Correlation Technique for Generator Stator Fault DetectionPart І Figure 4. Synchronous Machine Model The syste paraeters are obtained fro the Kurieat power station [8] of the Egyptian 500 KV unified network and are given in Appendix. 5. Siulation Results An internal singleline to ground fault (F ) was located on stator winding of the generator. Another singleline to ground fault (F ) was considered outside the generator protection zone as shown in Fig.4. The relay s CTs orientation is built for proper functioning of the proposed stator winding protection. The current signals, fro ATP EMTP software, generated at sapling rate of 00 saples per cycle, this gives a sapling frequency of 5 KHz. The total siulation tie is 0. Sec (i.e. the total nuber of saples is 500). The fault inception tie is 0.04 Sec (at saple 0) fro the beginning of siulation tie. 5.. Case : External Fault without CT Saturation Condition An external singleline to ground fault (F ) is located out of the generator protection zone at the priary side of stepuptransforer. The fault type is single phase Atoground assuing that short circuit is not resistive. The operating conditions of the siulated power syste are shown in Appendix. Figures 5 (ag) show the siulation results in case of external single linetoground fault. The Figures present the three phase secondary current signals at the two receiving and sending end sides, and their auto/crosscorrelation coefficients. In this case, it is noticed that the phase A secondary currents for neutral and load sides during the fault are identical and higher than the prefault currents; their values are nearly 4.8 ties the prefault currents as shown in Fig. 5(a). Phase B secondary currents for the two sides during the fault are identical and higher than the prefault currents; their values are approxiately. ties the prefault currents as shown in Fig. 5(b), however; phase C secondary currents for the two sides during the fault are identical and lower than the prefault current, their values are approxiately 0.9 ties the prefault currents as shown in Fig. 5(c). The calculated crosscorrelation coefficients (r ia, r ib, r ic ) are equal and close to unity before and during fault occurrence as presented in Fig. 5(d). The algorith calculates crosscorrelation coefficient between each two corresponding windows for secondary current signals at the two receiving and sending end sides. The calculated autocorrelation coefficients (r ia, r ib, r ic, r ia, r ib, r ic ) are equal and close to unity before fault occurrence, and they decrease with fault occurrence (see Figures 5(ef)). A trip flag depends on the crosscorrelation values, if their values are less than unity then a trip signal is sent for isolation generator CB. But this case is external fault condition and no tripping signal is issued as shown in Fig. 5(g). To assure that this case is external fault and not noral operation, autocorrelation coefficients is estiated between each two windows of load or neutral current signal shifted fro each other by one cycle. The autocorrelation values are closely to unity for neutral and load current signals in case of noral operation and their values are less than one in case of external faults. (a) The current signals i a, i a for case. (f) Autocorrelation coefficients r ia, r ib and r ic.
International Journal of Energy and Power Engineering 04; 3(3): 64 (b) The current signals i b, i b for case. (g) Autocorrelation coefficients r ia, r ib and r ic. (c) The current signals i c, i c for case. (h) No trip flag in case of external fault. (d) Crosscorrelation coefficients r ia, r ib and r ic. Figures 5 (ag). Three phase current signals and their auto/crosscorrelationcoefficients for case. 5.. Case : Internal Fault Condition In this case, all paraeters are kept as in case except the external fault type (F ) is changed to becoe internal fault type (F ). At operating power angle of generator δ is 0 degree, a singleline to ground fault (F ) is located on phase A for stator winding. Earth faults are siulated at different locations on stator winding such as %, 5%, 7%, 0%, 5%, 0%, 30%, 40%, 50% distance points fro the generator neutral. Figures 6(ah) show the siulation results for case in case of internal fault at 40% of stator winding. The Figures present the three phase secondary current signals at the two receiving and sending end sides and their auto/crosscorrelation coefficients (r ia, r ib, r ic, r ia, r ib, r ic ) and (r ia, r ib, r ic ). Appendix 3 shows the change in the Auto/crosscorrelation coefficients for current signals of each phase in cases SLG (ag) fault at different locations on stator winding (internal faults) at δ = 0 and R f = 0 oh. Fro the obtained results, it clearly appears that the proposed technique for differential algorith succeeded in differentiating between internal and external faults occurring on the generator stator besides identifying fault type and faulted phase(s). The technique has also the advantage of detecting ground faults occurring near the neutral point thus decreasing the dead zone to %. (a) The current signals i a, i a for case. (e) Crosscorrelation coefficients r ib and r ic.
R. Abd Allah et al.: A New Schee Based on Correlation Technique for Generator Stator Fault DetectionPart І (b) The current signals i b, i b for case. (f) Autocorrelation coefficients r ia, r ib and r ic. (c) The current signals i c, i c for case. (g) Autocorrelation coefficients r ia, r ib and r ic. (d) Crosscorrelation coefficient r ia. (h) Relay trip flag for case. Figures 6 (ah). Three phase current signals and their auto/crosscorrelationcoefficients for case. 6. Conclusions In this paper, a reliable and efficient technique has been presented for detecting generator stator winding internal and external faults by using correlation algorith. ATPEMTP software has been used for generating fault data and then processed in MATLAB to get auto/crosscorrelation coefficients of twoside currents for each phase of stator windings. These coefficients are used in the proposed algorith to ipleent relay logic. Results of case studies of single line to ground faults at different locations on stator winding and various fault types are presented. Case study results show that the technique used correctly discriinates between internal and external faults on the stator winding. The suggested algorith has the following features:. Introduced a new detection technique for internal faults on stator windings, the new technique has the advantages of avoiding CT saturation effect, detecting ground faults occurring near the neutral point thus decreasing the dead zone to % only, and avoid the effects of switching, DC coponents and haronics. Succeeded in identifying fault type, location and faulted phase(s), and produced a trip signal within 0 sec after fault occurrence. 3. Has the advantage of using the algorith of differential protection for CT saturation detection; whereas ost conventional differential protection algoriths require a separate algorith for detecting CT saturation. 4. The suggested auto/crosscorrelation technique is characterized by being siple, fast, reliable and accurate and can be ipleented practically, thus it can be used as a base for ipleenting a cheap and reliable digital protective relay for synchronous generators.
International Journal of Energy and Power Engineering 04; 3(3): 64 3 Appendix Appendix. Power syste paraeters data. Power syste paraeter Synchronous Generator (Sending source): Rated Voltapere / Rated line voltage / Rated frequency Nuber of poles /Neutral grounding ipedance (R n) Stepup Transforer: Rated Voltapere Transforation voltage ratio Connection priary/secondary Priary winding ipedance (Z p) Secondary winding ipedance (Z s) Vector group Z% Transission Lines: +ve sequence R Zero sequence R +ve sequence XL Zero sequence XL +ve sequence /Xc Zero sequence /Xc Transission line long (K) Main Load (load ): Load Voltapere Aux. Load (load ): Load Voltapere Power Network (Receiving source): Noinal line voltage Voltage phasor angle phase Noinal frequency Voltapere short circuit Current Transforer (CT): CTR Rated burden Class Voltage Transforer (VT): VTR Data 30 MVA / 9.57 kv / 50 Hz / 0.77 oh 340 MVA 9 kv /500 kv Delta/Star earthed neutral 0.007 + j0.84 oh 0.7708 + j 6.8 oh. YNd 5% 0.07oh /k 0.47 oh/k 0.30 oh/k 0.9 oh/k 3.96 icroho /k.94 icroho /k 00 K 5 GVA at PF = 0.85 lag 30 MVA at PF = 0.85 lag 500kV (pu) 0 0 50 Hz 5 GVA 000/ 30 VA 5p0 9570/00 V Appendix. Operating conditions of electrical coponents. Electrical coponent (operating condition) F operated Load (ain load) Load (aux. load) Generator operating power angle (δ ) Operating phase peak voltage of generator Generator grounding ipedance Data 50 Hz 8.5 + j 5.6 Oh 0.85 + j 6.7 Oh 0 Degree 6063 Volt 0.77 oh Appendix 3. Auto/crosscorrelation coefficients of each phase in cases of internal SLG (ag) faults, δ = 0, R f = 0 oh. Fault Type SLG (ag) Fault % Fault location on stator winding 50% 40% 30% 0% 5% 0% 7% 5% % Crosscorr. Between the two currents of neutral and load sides r ia prefault r ia duringfault + + 0.6 0.93 0.89 0.83 0.73 0.63 0.46 0.03 0.85 r ib prefault r ib duringfault r ic prefault r ic duringfault Autocorr. of current at neutral side r ia prefault r ia duringfault 0.90 0.87 0.8 0.7 0.6 0.46 0.6 + 0.03 + 0.84
4 R. Abd Allah et al.: A New Schee Based on Correlation Technique for Generator Stator Fault DetectionPart І Fault Type SLG (ag) Fault % Fault location on stator winding 50% 40% 30% 0% 5% 0% 7% 5% % Crosscorr. Between the two currents of neutral and load sides r ib prefault r ib duringfault 0.94 0.95 0.96 0.97 0.97 0.97 0.98 0.98 0.98 r ic prefault r ic duringfault 0.9 0.94 0.95 0.96 0.96 0.96 0.96 0.96 Autocorr. of current at load side r ia prefault r ia duringfault 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 r ib prefault r ib duringfault 0.94 0.95 0.96 0.97 0.97 0.97 0.98 0.98 0.98 r ic prefault r ic duringfault 0.9 0.94 0.95 0.96 0.96 0.96 0.96 0.96 0.96 References [] "Protective Relays Applications Guide," The English Electric Copany Liited, Relay Division, Stafford, 975. [] Mozina C. J., IEEE Tutorial on the Protection of Synchronous Generators, IEEE Tutorial Course, IEEE Power Engineering Society Special Publ., no. 95 TP0, 995. [3] Megahed A. I., Student Meber, IEEE, and O. P. Malik, Fellow, IEEE, "An Artificial neural network based digital differential schee for synchronous generator stator winding protection," IEEE Transactions on Power Delivery, vol 4, 999. [4] Sachdev M. S. (Coordinator) Microprocessor Relays and Protections Systes, IEEE Tutorial Course Text, Power Engineering Society Special Publ. No.88, EH069PWR, IEEE, Piscatway, NJ, USA, 988. [5] Gabriel Benouyal, Serge Barceloux, and Rolland Pelletier, "Field experience with a digital relay for synchronous generators," IEEE Transactions on Power Delivery, vol. 7, no. 4, 99. [6] Sachdev M. S. and D. W. Wind, "Generator differential protection using a hybrid coputer," IEEE Trans. Power Apparatus Syste, PAS9(973) 06307. [7] Tao H., Morrison I. F., "Digital winding protection for large generators," J. Electr. Electron. Eng. Aust., 3 (983), 36 3. [8] Dash P. K., Malik O. P., and Hope G. S., "Fast generator protection against internal asyetrical faults," IEEE Transactions on Power Apparatus and Systes, vol. PAS96, no. 5, 977. [9] Dash P. K., Malik O. P., and Hope G. S., "Digital differential protection of a generating unit schee and realtie test results," IEEE Transactions oxn Power Apparatus and Systes, vol. PAS96, no., 977. [0] EMTP Theory Book, Bonneville Power Adinistration, Portland, Oregon, USA 987. [] ATP version 3.5 for Windows 9x/NT/000/XP Users' Manual Preliinary Release No.. 00. [] W. Hauschild, and W. Mosch, Statistical Techniques for High Voltage Engineering, hand book, English edition published by peter pere grinus Ltd., London, United Kingdo, chapter, pp. 7879, 99. [3] Gabriel Benouyal Schweitzer, Engineering Laboratories, Lt., ''Power Syste Protection'' book, chapter, 005. [4] Leonard L. Grigsby, ''Electric Power Engineering Handbook'' book, Second Edition, chapter, 007. [5] A. I. Taalab, H.A. Darwish and T. A. Kawady, ANNBased Novel Fault Detector For Generator Windings Protection, IEEE Transaction on Power Delivery, Vol. 4, No 3, pp. 84 830, 999. [6] N.W.Kinhekar, Sangeeta Daingade and Ajayshree Kinhekar, Current Differential Protection of Alternator Stator Winding, Paper subitted to the International Conference on Power Systes Transients (IPST009) in Kyoto, Japan, 009. [7] K S Yeo and D T W Chan Application of Wavelet Analysis for Generator Stator Fault Detection, Eail: yks@pail.ntu.edu.sg. [8] Instruction Manual for Generator Electrical Equipent, Upper Egypt Electricity Production Copany, Elkureiat П 750 MW Cobined Cycle Project, Stea Turbine Generator & Auxiliaries (Generator Electrical Equipent), Hitachi, Ltd., Tokyo Jaban, 006.