WO2001022104A1

WO2001022104A1 - Method for detection of high-impedance ground faults in a medium-voltage network

Info

Publication number: WO2001022104A1
Application number: PCT/FI2000/000816
Authority: WO
Inventors: Ari Nikander; Pertti JÄRVENTAUSTA
Original assignee: ABB Substation Automation Oy
Current assignee: ABB Substation Automation Oy
Priority date: 1999-09-23
Filing date: 2000-09-22
Publication date: 2001-03-29
Anticipated expiration: 2002-03-23
Also published as: FI112283B; AU7294200A; FI19992044A7

Abstract

The present invention relates to a method for detection of high-impedance ground faults in a medium-voltage network. According to the method, the degree of unsymmetry (|ki|) and/or the line-to-ground admittance (YtEi) as well as the zero-sequence voltage (U0) of each sending end (i) are determined. According to the invention, for the values of the line-to-ground admittance (YtEi) and the degree of unsymmetry (|ki|) of each sending end are determined a reference value (YtEiref and |kiref|) on the basis of measurement information obtained by means of an artificial deviation of the neutral voltage performed in a reference connection status, in a memory are stored as reference values the values of the line-to-ground admittance (YtEref) and the degree of unsymmetry (|kiref|) of each sending end (i), as well as the normal-connection status values of the zero-sequence voltage (U0ref) and the zero-sequence currents (I0refi) of the sending ends and the zero-sequence current (I0refs) of the feeding power source, the zero-sequence voltage (U0) is monitored at least essentially continuously and, if said zero-sequence voltage (U0) changes by more than a predetermined limit difference (ΔU0as), for each one of the sending ends (i) are computed new values of line-to-ground admittance (YtEi) and degree of unsymmetry (|ki|), the most recently computed values of the line-to-ground admittance (YtEi) are compared with the reference values (YtEiref) and from the comparison is determined whether the difference therebetween exceeds the inaccuracy (ΔYtEi) of the measurement technique used, whereby if the comparison gives a value greater than said measurement inaccuracy (ΔYtEi), it is checked for instance on the basis of the change in the entire network's summed line-to-ground admittance, which is computable from zero-sequence current of the feeding power source, whether a change has occurred in the connection status of the network end, if so, the most recently measured values of the line-to-ground admittance (YtEi) and degree of unsymmetry (|ki|) are stored as new reference values, while, if no change has occurred in the network connection status, a ground fault is indicated.

Description

Method for detection of high-impedance ground faults in a medium- voltage network

The invention relates to a method according to claim 1 for detection of high-impedance ground-faults in a medium-voltage network and identification of a faulted sending end.

The applications of the invention can be found in medium-voltage networks, parti cu- larly in continuous detection of high-impedance ground faults of the network, identification of a faulted sending end and monitoring of the network condition utilizing for this purpose the measurement information of, e.g., a numerical multifunction relay.

The magnitude of the ground fault current imposed on the medium- voltage sending end is dependent on the total line-to-ground capacitance 3C₀ of the line section connected to the sending end. The zero-sequence susceptance Bo of the sending end is computed based on the parameter values given by the line conductor manufacturers. One problem herein arises from the inaccuracy of the parameter values that must be used in computations. From experience it is known that the value of the zero- sequence susceptance B₀ used in computations may be erroneous by up to 10-20 % in regard to the theoretically correct value. However, it is most important to know the correct values of fault currents at the sending ends during any possible connection status of the network in order to assure the fulfillment of grounding voltage safety regulations and proper function of the network protective system. As known, a high- impedance ground fault changes the unsymmetry of the line-to-ground admittances at the sending end. Protective relays currently used are incapable of detecting the degree of line-to-ground unsymmetry at the sending end or changes thereof.

The detection of high-impedance ground faults and identification of the faulted sending end is problematic due to the small magnitude of the fault current. The most sensitive fault indicator directly measurable from a network under a fault situation is the zero-sequence voltage of the network. However, fault detection based on a change in the RMS value of the zero-sequence voltage or a change in the line-to- ground leakage resistance does not necessarily detect directly the faulted sending end that must be determined by other means. Hence, the method according to the inven- tion have been developed to provide independent status monitoring for each one of the sending ends. This goal is attained by way of determining the electrical insulation status and connection status of the network per sending end by way of continuous monitoring of line-to-ground admittances and level of line unsymmetry.

A significant benefit of the invention over conventional techniques is its realtime operation. This is based on up-to-date information available from the protective relays of the sending ends on the electrical length and insulation condition of the respective sending end. The protective relay of the sending end also has information on the electrical length of the entire network and its insulation condition. The method does not require information transmission between the protective relay and the overlying systems or protective relays.

Conventionally, computational methods for determination of ground fault current in medium-voltage networks are based on zero-sequence susceptance values published by line conductor manufacturers. The computation of fault currents is performed in conjunction the computation of the network general status.

Conventionally, attempts to indicate high-impedance ground faults have been based on monitoring the absolute value of the zero-sequence voltage of the network and changes occurring therein. When the zero-sequence voltage exceeds a threshold value, an alarm is triggered. This kind of method is incapable of giving any information on the location of the fault within the network.

From FI Pat. Appl. No. 973,520 is known a method for detecting a single-phase ground fault in a resonant-ground, three-phase network. According to this method, the symmetrical operation of the network is deviated artificially and thereby are measured a number of parameters, wherefrom the line-to-ground admittance and unsymmetπcal admittance component values are determined Due to its artificial implementation, the method is applicable to ground-resonant networks only. The artificially caused shift in the network neutral voltage is a risky procedure in terms of secure network operation and does not belong to the standard operatn e routines in the electrical power systems

In Fl-patent publication 100922 B is described a method for detection and location of a high-resistance ground fault Basically, the fault detection technique disclosed therein is based of a method outlined as follows At the distribution substation are measured the hne-to-neutral voltages with their phase angles and therefrom is computed the zero-sequence voltage Uo of the network as a vector sum of the measured hne-to-neutral voltages, the network zero-sequence impedance Z₀ is computed, the zero-sequence voltage U₀ is compared with the hne-to-neutral voltages U_v and the total zero-sequence impedance Z_o of the network. The thus obtained variables are then utilized for computing the fault impedances 2_ per each phase leg and the faulted phase leg is identified by being the leg with the largest real part of the fault impedance Z

In US Pat No 4,729,052 is described a method wherein the occurrence of a ground fault is detected by way of causing a change in the neutral point grounding impedance of the network and then measuπng the effect of the change on the neutral point voltage. This method is suitable for use only m resonant-ground systems that are equipped with automatic tuning of the arc-suppression reactor

The prior-art technique is handicapped among other things by the tendency of line conductor manufacturers to report the zero-sequence susceptance values on the "safe" side, whereby the ground-fault current values computed based thereon generally become larger that the actual values encountered Also m other aspects, the computed parameters of line conductors involve terms of inaccuracy For instance, no information is submitted to a protective relay on the change of the electrical line length that is caused by a change in the network connection status A fault location method based on monitoring the RMS value of the network zero- sequence voltage tells nothing about the location of the fault in the network. While a high-impedance fault may cause a larger change in the phase angle of the zero- sequence voltage than in the absolute value of the voltage, this information remains unutilized in said method. Weather phenomena (e.g., falling snow) and changes in the connection status of the network also cause deviations in the network neutral voltage. The latter type of deviations may be difficult to tell from the changes in the neutral voltage that are caused by a high-impedance ground fault. On the other hand, the method enjoys the benefits of simple use and relatively good sensitivity.

Prior- art detection methods of high-impedance faults do not facilitate network monitoring per sending end, because the direction of the fault must be determined after the fault detection that is performed on the basis of the changes measured in the zero- sequence currents of the energized sending ends. The method disclosed in FI patent publication no. 100922 B allows fault resistance monitoring only generally in regard to all the sending ends of the network. Moreover, the computation of the network zero-sequence impedance in said method is based on the zero-sequence susceptance values reported by the line conductor manufacturers that are accurate only to within a given error margin. The line-to-ground capacitances of the network can be measured by way of making ground fault tests, but such results are valid only for the connection status prevailing during the tests. Furthermore, the network zero-sequence impedance must be computed within a higher-level system, whereupon the information can be submitted to the protective relay. The benefits of the method include detection of the faulted phase leg, possibility of the fault direction determination and good sensitivity.

The method disclosed in FI Pat. Appl. No. 973,533 requires communications between the protective relays and the network control system.

It is an object of the present invention to overcome the problems of the prior-art techniques and to provide an entirely novel type of method for detection of high- impedance ground faults in a medium-voltage network. The goal of the invention is attained by way of computing for the line-to-ground admittance and degree of unsymmetry at each one of the sending ends a reference value under a reference connection status using measurement information obtained during an artificially caused deviation in the zero-sequence voltage or under an actual ground fault occurring in the network and then using these reference values for at least an essentially continuous realtime monitoring of the respective variable values in the network, whereby a possible fault situation can be detected on the basis of predetermined limit deviation values.

More specifically, the method according to the invention is characterized by what is stated in the characterizing part of claim 1.

The invention provides significant benefits.

One of the major novelty factors of the present method over prior-art methods and published techniques is related to the simplicity of the practical implementation and possibility of embedding the method as a part of other network automation. In practice, the application of the method does not need a priori knowledge on the electro- technical data of the network components (that is, the network data) nor realtime information on the concurrent connection status of the network. In fact, the present method can be appreciated to form a black box, whose input signals are the measured zero-sequence voltages and currents, while the output of the box is an indication of a ground fault that is sent as a unilateral event signal to a higher level system. The method can be implemented within a single compartmentalized cell terminal and the fault indication does not need communications with other protective relays or information processing systems, which makes the present method more robust and flexible than, e.g., the method disclosed FI patent publication no. 100922 B. Furthermore, a secondary relay/substation (REC) implemented according to the invention can be erected on a field disconnector station of the network. Inasmuch also this kind of application does not need for fault indication supplementary initial information or communications with the higher level relays or information systems, the practical implementation of the present method is substantially more straightforward to implement than the method described in cited FI patent publication no. 100922 B. The present method can detect in a self-contained manner changes in the network connection status and, hence, in contrast to the method disclosed in cited FI patent publication no. 100922 B, does not need realtime connection status information produced by another system, thus making the present system different from that described in cited FI patent publication 100922 B.

A crucial novelty value of present method is in its self-contained continuous opera- tion as compared with, e.g., that described in reference publication [Lei94], whose method is chiefly related to monitoring on the basis of discrete measurements. The present invention in contrast utilizes inherent deviations of the zero-sequence voltage of the network. Accordingly, the computational routines of the method do not need tuning of the network compensation or connection changes in the network as some prior art methods do.

The determination of line-to-ground admittances at the sending ends by measurements improves the accuracy of fault current computations over the values obtainable by computations based on network data and, furthermore, facilitates providing the protective relay with the most recent information on the electrical length of the line connected to the monitored sending end. The method allows positive fault indication even if the fault should occur rapidly after a change in the sending end connection status.

Furthermore, the invention facilitates monitoring of the individual sending ends of the network as to their electrical condition. A change in the line leakage resistance not only tells about a fault but also indicates the faulted sending end and the faulted phase leg thereof. Prior- art methods for detection of high-impedance faults were chiefly based on monitoring the zero-sequence voltage of the network or determining the total leakage resistance of the entire network. Therefore, the faulted sending end must be identified by other techniques. One further advantage appreciated in the invention is that its implementation can be carried out without new equipment constructions or measurement arrangements in the network, but instead its computational algorithms are runnable on software that is programmed on existing numerical multifunction relays.

Since the line-to-ground admittance computation at a sending end is based on a change in the zero-sequence current and zero-sequence voltage instead of the absolute values of these variables, the error factor caused by the capacitance unsymmetry of the sending end is eliminated. At low values of the zero-sequence voltage, the zero-sequence current component due to the capacitance unsymmetry may play a significant part. Also a portion of the measurement error in the zero-sequence voltage can be eliminated.

In the following, the invention will be described, among other aspects, with the help of an exemplifying embodiment illustrated in the appended drawings of FIG. 1.

Chiefly due to the unsymmetry of line-to-ground capacitances in a network, a small neutral voltage always occurs even under normal conditions in a network grounded via a large reactance (resonant-ground system) or in an isolated-neutral network. In a resonant-ground network, the neutral voltage can be altered by changing the tuning of the arc-suppression reactor. A change in the connection status also causes a change in the neutral voltage in both a resonant-ground and an isolated-neutral network. The neutral voltage under a normal condition may also be changed by means of an artificially generated capacitance unsymmetry. Herein, a capacitor is connected to one phase leg of the network, whereby the neutral voltage of an operative network rises. Also current injection to the neutral of the network can be used for changing the neutral voltage. This can be implemented in a relatively simple manner by connecting a 230 V voltage to an auxiliary winding (500 V) or measurement winding of the arc-suppression reactor. The current injection system can be isolated from the low-voltage circuit with the help of a suitable isolation transformer. A capacitor can be used for current limiting. The injection current may be smaller than 1 A and the duration of current injection need not be longer than a few seconds. These techniques are well known in the art. The inventiveness of the present invention is appreciated in the application of the measured neutral voltage change.

When the neutral voltage of the network changes, also the zero-sequence currents of the sending ends change. The line-to-ground admittance of the sending end can be computed with the help of the changes in the zero-sequence current and voltage. The following description defines the parameters and computational techniques needed in this computing procedure. Equations 1-4 used in the invention for computing the line-to-ground admittances and degree of network unsymmetry are known to those versed in the art. These equations are described in, e.g. cited publications [Lei97] and [Lei94]. The inventiveness of the present invention is based on the novel and continued development and application of these computational techniques into a network protection method utilizing the capabilities of a numerical multifunction relay without any need to provide additional information to the relay from higher-level automa- tion systems of the network.

The definitions use the following subindexes:

i = sending end subindex, i = 1,2,3,... v = phase subindex 1,2,3 t = sum of all the three phases

E = ground point

Ui = line-to -neutral voltage a = -1/2 + jV3/2 (phase-shift operator)

The summed line-to-ground admittance of all the three phases is predominantly capa- citive. Denoting the angular frequency corresponding to a nominal frequency of

50 Hz with symbol ω (ω = 2πf), the line-to-ground admittance is

^κ,ε According to the conventions of a resonant-ground network, the degree of network unsymmetry is defined as

The degree of unsymmetry can be defined for both the entire network and a single sending end. For a single sending end, the admittances are the line-to-ground admittances for the sending end concerned.

The line-to-ground admittance of a sending end can be computed from the change of the zero-sequence current and voltage (Eq. 3). Herein, subindexes a and b refer to two separate measurements of the zero-sequence voltage and current. The zero- sequence voltage and current phasors are compared with such a reference phasor that does not change during a ground fault. Such an applicable reference phasor is one of the line-to-line voltages. The use of the change of the zero-sequence voltage and current in the computation of the line-to-ground admittance eliminates the effect of zero-sequence component due to the capacitance unsymmetry on the end result of the computation. For low-resistance ground faults, the sending-end line-to-ground admit- tances may also be computed directly from the zero-sequence current and voltage, because herein the capacitance unsymmetry effect is negligible. This is because the zero-sequence current component related to the capacitance unsymmetry is not dependent on the zero-sequence voltage, whereby

Respectively, the degree of unsymmetry for the entire network or a single sending end can be computed from Eq. 4 written as:

The basic concept of the invention is that the line-to-ground admittances and degrees of unsymmetry of the sending ends are computed each time the zero-sequence volt- age of the network changes. In this manner it is assured that the line-to-ground admittances based on the measurement data obtained at the sending ends and the fault current values derived therefrom are at any time representative of instantaneous connection status of the network. The goal of the method is to indicate ground faults having a fault resistance in the range of 100 - 200 kohm. For ground fault resistances smaller than 10 kohm, the faulted sending end can be detected directly on the basis of the change in its line-to-ground admittance. Herein, the computation is supportive to the normal protection given by directional relays. The method is in the same manner applicable to an isolated-neutral network as to a resonant-ground network and does not need a change of the relay operating characteristics if the network-compensating reactor is disconnected from the network.

Now referring to FIG. 1, the steps are described elucidating the operation principles of the invention in the determination of the electrical length of a line, indication of high-impedance ground faults and determination of a faulted sending end:

Step 1

A reference value of the initial status (that is, the reference connection status, where- from computations begin) is computed for the line-to-ground admittance Y_tEl of each sending end and for the degree of unsymmetry Ik,!, whereby an artificial shift of the neutral voltage or a change thereof caused by a switching status change or a ground fault is utilized for this purpose. Herein, the computation follows the techniques disclosed in cited publication [Lei97]. In the case of a resonant ground network, additional computations are made to determine the values of line-to-ground admittance and the degree of unsymmetry for the entire network utilizing for this purpose the zero-sequence current computed from the line-to-neutral currents of the sending end or, alternatively the zero-sequence current of the reactor branch if such a measurement value is available. The zero-sequence voltage is computed as a phasor sum of the line-to-neutral voltages or, alternatively, is measured from the open-delta winding of the voltage transformers. Serving as the line-to-ground capacitance C_tε in the computation of the degree of unsymmetry at each sending end is used a reference capacitance whose value is selected to be equal to the summed line-to-ground capacitance of all the three phase legs of the entire network in its reference connection status. The summed line-to-ground capacitance of the entire network can be computed from the zero-sequence current change of each sending end using Eqs. 1 and 3. The values of line-to-ground capacitance per sending end are required for determining the electrical length of the line and indication of high-impedance ground faults.

Step 2

As reference values are stored the values of line-to-ground admittance Y_tE_ιref and |k,_ref| of each sending end denoted by subindex i. Also the normal-connection status values of the zero-sequence voltage, the zero-sequence currents of the sending ends and the zero-sequence current of the feeding power source are stored as reference values Uo_ref, Ioreti and I_0refs-

Step 3

When a change in the network connection status or a ground fault causes in the network a zero-sequence voltage change ΔUo that exceeds a preset limit difference value ΔUoas, new values of line-to-ground admittance Y_tE, and degree of unsymmetry Ik, I are computed for each sending end. As a second set of reference values are stored the zero-sequence voltage and sending-end current values denoted as Uo_ref, lo_refi- The zero-sequence current change at the faulted sending end is not used as such for fault indication, but instead the fault-situation zero-sequence voltage, sending-end zero- sequence current and precomputed value of the line-to-ground admittance are utilized for computing a parameter k which is representative to the unsymmetry of the sending end and whose change can be used for fault indication. The computation is performed simultaneously for each one of the sending ends, whereby the faulted sending end is found directly from the result of the computation and there is no need to identify the fault at the network level as is suggested in FI patent publication 100922 B. This approach gives higher sensitivity in fault location than what can be attained by detecting the fault merely from the change in the line-to-ground admit- tance of the faulted sending end.

Step 4

The computed line-to-ground admittance values of the sending ends are compared with the reference values Y_t&ref- If the sending-end line-to-ground admittance does not differ from the reference value Yt_Bref by more than a preset error tolerance ΔY_tE, caused by computational and measurement inaccuracy errors, the status of the sending end is assumed to be unchanged and the sending end obviously has no ground fault. In this case the stored reference value is not changed. If the deviation of the sending-end line-to-ground admittance value from the reference value Yt&r_ef is greater than ΔY_tEι, it is plausible to assume that the sending-end connection status has been changed or that the sending end is affected by a ground fault. A change in the connection status may also be inferred from the summed line-to-ground admittance change of the entire network, which is computed from the zero-sequence current of the feeding power source. However, this approach is applicable only when the change in the electrical line length exceeds the possible deviation due to measurement inaccuracy in the computed value of the entire network's admittance. When a change occurs in the connection status, new reference values are stored.

Step 5

The computed degrees of unsymmetry at the sending ends are compared with the stored reference values |k_lref). If the computed degree of unsymmetry is greater than

|k_ιrefj and the change thereof is greater than |Δk,|, it is plausible to assume that the sending end is affected by a high-impedance ground fault. The values of degree of unsymmetry can be computed from Eq. 4 without the need for a new shift of the zero-sequence voltage when the line-to-ground admittances of the sending ends are known in the prevailing connection status.

Step 6

An alternative method of monitoring the degree of unsymmetry is to monitor the ground-leakage resistances of the different phase legs at the sending ends. The moni- toring of the leakage resistances may be arranged to take place in parallel with the monitoring of the degree of unsymmetry. If the degree of unsymmetry at any sending end changes, it is possible to infer from the leakage resistance R_f whether a ground fault or a change in the connection status has occurred. This inference also gives such information on the faulted phase leg that may be utilized in the location of faults not identifiable by visual inspection. For instance, it is possible in conjunction with lightning arresters to determine which one of three phases has a defective arrester. Arc arresters are typically used for overvoltage protection of substations and underground cables. The physical location of phase-leg conductors at substations and underground cable terminations are generally identified. The leakage resistances of the different phases can be computed from Eq. 5 written as:

R_f = — (5)

L_0l + jωC_tEl U₀ where

U_v = is voltage of phase v lo, = zero-sequence current lo of sending end i C_tE_! = line-to-ground capacitance of sending end i in normal connection status Uo = zero-sequence voltage.

The zero-sequence current at the sending end comprises a zero-sequence-voltage- dependent component Quo) and an unsymmetry-dependent component (I_k) in accordance with Eq. 6 below. In practice, the line-to-ground capacitances per phase are not exactly equal.

Lo, = Lυo + = Y_tE, U₀ + (Y_0λ U + Y₀₂ U₂ + Y₀₃ U₃) (6)

where

Y,_Eι = summed line-to-ground admittance of three phases at sending end i

XQ_{I J} ¥.02. _O3 ⁼ line-to-ground admittances of phases 1, 2 and 3

Ui, Lb, Lb = symmetrical voltages of phases 1, 2 and 3.

The effect of the error caused by capacitance unsymmetry can be reduced in the following manner. After the zero-sequence susceptance B_tE at the sending end has been determined by the admittance computation, the capacitance-unsymmetry-dependent component of the zero-sequence current can be determined with the help of Eq. 6, whereupon its effect on the leakage resistance may thus be eliminated. Thus, the leakage resistances at each phase leg can be monitored on a continuous basis without allowing the capacitance unsymmetry to have any effect on the computational result. Changes in the zero-sequence voltage do not affect the capacitance-unsymmetry- dependent component of the zero-sequence current. In this manner, the status monitoring can be carried out on a continuous basis without a need for deviations to occur in the zero-sequence voltage.

The unsymmetry-dependent component of the zero-sequence current may also be used as a direct fault indicator according to Eq. 6. As can be seen from the equation, a ground fault causes a change equal to the fault current in the unsymmetry-depen- dent component of the zero-sequence current.

where

I_OKI = unsymmetry-dependent current component of sending end i in a normal condition Ioκfι ⁼ unsymmetry-dependent current component of sending end i in a fault condition.

When the zero-sequence voltage changes due to, e.g., a high-impedance fault (R_f = 10 - 200 kohm), the fault resistance may also be determined from Eq. 8 [Lei97]. In this case, the computation is carried out replacing the absolute values of zero- sequence current and voltage by the changes of these variables:

u vE

R 7r (8)

Δ/_0l -Z,_fiΔC/₀ where U_VE = voltage of phase v during a fault (high-impedance ground fault)

ΔI_0l = (zero-sequence current I₀ of sending end i during fault) - (zero-sequence current I₀ of sending end i before fault)

Yo_tot ⁼ summed line-to-ground admittance of three phases of sending end i in a normal condition ΔU₀ = change in zero-sequence voltage due to fault (= during fault - Lb before fault)

Indication of a high-impedance fault and detection of a faulted sending end from the degree of unsymmetry or by means of Eqs. 5, 7 or 8 require the line-to-ground admit- tances of the sending ends to be known a priori for a situation prevailing before the fault. If the change in the zero-sequence voltage is caused by a ground fault, the line- to-ground admittances can be computed for all sound sending ends. In contrast, if the change in the zero-sequence voltage is caused by a change in the connection status, the line-to-ground admittances can be computed for all those sending ends that maintain their connection status unchanged. In the following is described a method for indication of a change in the connection status and determination of a new line- to-ground admittance value after a change in the connection status.

A change in the connection status of a sending end can be indicated and distinguished from a line fault with the help of the changes occurring in the zero-sequence voltage and current. The basic principle of the method is to measure at sending end i the zero-sequence voltage change ΔUo = Lbi - Lb₂ and, respectively, the zero- sequence current change ΔI₀₎ — loi - 1₀2 that are detected from measurements 1 and 2. Herein, the sending-end line-to-ground admittance Y_0l is already known in the initial condition. The change in the sending-end line-to-ground admittance is denoted as ΔYo. When the zero-sequence voltage of the network changes, the zero-sequence current change at sending end i can be written as:

Δ ₀₍ = Y_0l U_m ^~ (lo, + ΔF₀) /₀₂ (9) Herefrom, the change in line-to-ground admittance can be solved:

Δio - Zo, (10)

The formulation of Eq. 9 is based on the assumption that the zero-sequence current caused by the capacitance unsymmetry is not changed due to the change in the connection status. Eq. 9 is exactly valid for those sending ends on which the electrical line length does not change, that is, ΔY₀ = 0. Hence, the detection of connection status change can be performed in a reliable manner. If the sending end has no change in its connection status nor a ground fault, the change ΔY₀ in the line-to- ground admittance remains zero. Simultaneously, the method makes it possible to compute an estimate for the change in the electrical length of the line. Then, the line- to-ground admittance of the sending end can be corrected to a value corresponding to its new connection status. The ground fault detection is implemented by means of computing the leakage resistance per sending end with the help of Eqs. 5 and 8.

After the sending end has undergone a change in its connection status, whereby its line-to-ground admittance has changed, the new value of line-to-ground admittance can be determined using one of the methods described below:

Method a

The changed situation is allowed to continue until a new change in the zero-sequence voltage of the network occurs due to a ground fault or connection operation, whereby the new value of line-to-ground admittance can be computed.

Method b

In a resonant-ground network, the neutral voltage can be altered artificially, e.g., by changing the tuning of the arc-suppression reactor. Another possibility is to change the electrical length or capacitance unsymmetry of the network. If the substation is equipped with a centralized compensation system, the neutral voltage can be shifted by means of connecting an adjustable or constant voltage supply to an auxiliary winding (e.g., to a 500 V winding) of the arc-suppression reactor. These techniques for shifting the neutral voltage are known in the art. Typically, they have been used for determination of network parameters and tuning of the arc-suppression reactor.

Method c

The zero-sequence susceptance of the network is determined with the help of the network data. A novel feature of the present invention is that the error margin of the zero-sequence susceptances can be reduced by virtue of assuming each line section to contribute by a given percent proportion to the total zero-sequence susceptance Bo at the sending end. Then, the computations are made using the value prevailing before the change in the connection status as the total zero-sequence susceptance value. Resultingly, the upper level control system can compute the percent change in the sending end zero-sequence susceptance, whereby the new value of line-to-ground admittance is obtained. In this manner it is possible to eliminate the systematic error possibly biasing the zero-sequence susceptance values given by the line conductor manufacturers. According to actual ground fault tests, the theoretically computed zero-sequence susceptance values are generally slightly biased on the higher side.

Method a of the above-described alternatives is not useful in cases where the ground fault occurs rapidly after a change in the network connection status, whereby a new line-to-ground admittance value has not yet been computed. In this situation, the degree of unsymmetry or line-to-ground leakage resistance computed on the basis of the ground admittance value valid before the connection status change is not physi- cally correct, while it may be sufficiently close to indicate the fault.

If the sending end connection status has changed and the sending end is faulted before the new value of the line-to-ground admittance has been determined, the computation of the sending end line-to-ground admittance and the degree of unsymmetry can be based on the reference value Ioreti and the values of the zero-sequence voltage and current that were valid during the fault. Since the value of I_0refi was measured during the earlier connection status, the values of line-to-ground admittance and degree of unsymmetry computed from Eqs. 3 and 4 are not physically correct. However, the changes occurring in these parameter values make it possible to detect a ground fault on the monitored sending end. Even if the sending end would be found to be in an "uncertain" state after the connection status change, the fault indication is yet entire feasible.

One possibility of determining the line-to-ground admittance of the monitored send- ing end after a change in the connection status is to use first method c and, after the network has undergone a change in its neutral voltage, use the alternative method a for computing the line-to-ground admittance.

Knowing the degree of compensation in the network is important, since this param- eter has a significant effect on the suppression conditions of the arc occurring during a ground fault and, thus, on the short-duration outages on customer services. Such medium-voltage networks that have the ground fault compensation implemented using fixed arc-suppression reactors or tunable reactors having a stepwise adjustable tap selection, the determination of network degree of compensation is rather inaccurate on the basis of network data and the nominal specifications of the reactor.

In some cases the arc-suppression reactors may also be located in a distributed manner around the network. Then, a change in the connection status of the network also changes the network degree of compensation. Modeling techniques to be described later facilitate computational determination of network degree of compensation for each different connection status. Herein, the network degree of compensation can be monitored in the same fashion as is described for the line-to-ground admit- tances and degree of capacitance unsymmetry. The computation of degree of compensation also permits automatic tuning control of the arc-suppression reactor after a change in the connection status.

In general, the network degree of compensation can be computed from Eq. 11 below: &>C_(£ _{y =} ^!L ^{~ J} ₌ __E ₌ _ _χ (11)

I_c ωC,_ε ω²C_lEL_0E

The summed ground capacitance _E of the three phase legs of the network can be determined from the change of the zero-sequence current at the sending end with the help of Eqs. 1 and 3. If the network includes distributed arc-suppression elements, the susceptances of the separate reactors connected to the sending ends must be subtracted from the term CDQ_E and added to the term l/ωL₀ε which represents the susceptance of a reactor connected to the star point of the main transformer at the substation or, respectively, a star point formed with the help of a grounding transformer. The term L_OE can be determined with the help of the zero-sequence current and zero-sequence voltage of the sending end or, alternatively the zero-sequence current and zero-sequence voltage of the reactor branch.

When the compensation is implemented in a distributed manner so that the arc-suppression units are situated at the sending ends, an overcompensation status of a given sending end can be identified by a measurement of its line-to-ground admittance. In the case that a sending end is overcompensated, a protective system based on the measurement of the losinφ value (which is a common practice in networks compensated with distributed arc-suppression elements) operates erroneously when the ground fault occurs elsewhere in the network.

If the steady-state zero-sequence voltage of the network is insufficient for the com- putation of network parameters and an artificial elevation of the network neutral voltage is undesirable, the zero-sequence susceptances of the lines connected to the sending ends can be computed during ground faults. It must be noted that networks having overhead lines are particularly subject to a relatively high rate of temporary ground faults that elevate the neutral voltage of the network. Then, the zero-sequence susceptances of the sound sending ends can be computed directly from the reactive component (I₀sinφ) of the sending-end zero-sequence current and the zero-sequence voltage of the network. On the other hand, the protective relay of the faulted sending end measures the ground fault current invoked by the rest of the network. Therefore, the portion of the ground fault current produced by the faulted sending end does not contribute to this measurement.

In the case a change occurs in the connection status of the sending end, the above- described method c must be used. The zero-sequence susceptance of the sending end can be measured at a higher accuracy when the neutral voltage rises the next time. In practice, method c gives a sufficiently accurate estimate of the zero-sequence susceptance of the sending end after a connection change.

Today, there is in field used a relatively small quantity of the newest-generation multifunction relays that are capable of synchronizing the phase of the sending end zero-sequence currents with regard to the line-to-line voltage phase and of converting the measured zero-sequence currents into the complex format required in the computations. This is because the useful life to the older-generation numeric relays appears to be rather long. If a substation is equipped with numeric protective relays that do not give directly the phase relationship between the zero-sequence voltage and the zero-sequence current, the synchronization of the zero-sequence currents can be carried out as follows.

Generally, the registers of the protective relay store information on the zero-sequence current absolute value and its real or reactive component (depending on the operating characteristic of the relay). If the absolute value and, e.g., the reactive component of the zero-sequence current are known, they can be used for computing the phase angle between the zero-sequence voltage and the zero-sequence current as follows: ^ = arcsin^ (12)

In this case, it is sufficient to provide a substation with only one multifunction relay that synchronizes the zero-sequence voltage phasor in regard to the phase angle of one of the line-to-line voltages. After the phase angle between the zero-sequence voltage and the line-to-line voltage phasors has been resolved, it is also immediately possible to synchronize the zero-sequence current phasors in regard to the line-to-line voltage phasor, since the phase angles between the zero-sequence currents and zero- sequence voltage are known (see Eq. 12). Hence, the system can be implemented without replacing the protective relays of the sending ends.

As to its overall cost, the method disclosed herein is cost-efficient inasmuch it dispenses with the installation of new measurement techniques or equipment. The only operation required to implement the method is the programming of a new functional software routine for the protective relay. In practice, an implementation of the present method when carried out by way of placing only one multifunction relay at a substation with communications to the older-generation numeric relays is the most cost-efficient alternative.

In compensated networks, the neutral voltage is generally sufficiently high for the computation of the parameters needed. A disadvantage that may be encountered is that in ground-isolated overhead networks with properly transposed conductors and in underground cable networks the steady-state zero-sequence voltage and, hence, the zero-sequence currents of the sending ends may be too small for reliable compu- tation of the line-to-ground admittances. However, even here the line-to-ground admittances can be computed in conjunction with ground faults occurring in the network or, alternatively, by shifting the neutral voltage of the network in an artificial manner.

The method needs at least one multifunction relay per each substation. The measurement data required include one line-to-line voltage, the neutral voltage, the zero- sequence current of the sending end or, alternatively, the zero-sequence current of the arc-suppression reactor branch and the zero-sequence currents of the sending ends. In practice all of these measurement values are available at essentially any substation.

References

(Lei97) : Leitloff V. et.al. : Detection of resistive single-phase earth fault in a compensated power-distribution system. ETEP Vol. 7, No 1, January/February 1997.

(Lei94) : Leitloff V. et.al. : Messung der Parameter eines kompensierten Netzes durch Injektion eines Stromes in den Sternpunkt. Elektrizitatswirtschaft, Jg. 93 (1994), Heft 22.

(Sch94) : Schafer H.-D. : Erhδhung der Verlagerungsspannung in Mittelspannungs- Kabelnetze mit Erdschlusskompensation. Elektrizitatswirtschaft, Jg. 93 (1994), Heft 21.

Claims

What is claimed is:

1. Method for detection of high-impedance ground faults in a medium-voltage network, said medium-voltage network including at least one sending end (i) for feeding electrical power into said network, whereby in the method the line-to-ground admittance (Yt_Ei) and the zero-sequence voltage (Uo) of each one of said sending ends (i) is determined,

characterized by the steps of

determining the degree of unsymmetry (jk,!) of each sending end (i), computing for the values of the line-to-ground admittance (Y_tEι) and the degree of unsymmetry (|k,|) of each sending end a reference value (Yt&_ref and |k,_ref|) on the basis of measurement information obtained by means of an artificial devia- tion of the neutral voltage performed in a reference connection status or, alternatively, caused by a ground fault occurring in the network in a natural manner, storing in a memory as reference values the values of the line-to ground admittance (Y_tEi_ref) and the degree of unsymmetry (|k,_refj) of each sending end (i), as well as the normal-connection status values of the zero-sequence voltage (Uo_ref) and the zero-sequence currents (Io_refi ) of the sending ends and the zero- sequence current (Io_refs) of the feeding power source, monitoring the zero-sequence voltage (Uo) at least essentially continuously and, if said zero-sequence voltage (Uo) changes by more than a predetermined limit difference (ΔUo_as), computing for each one of the sending ends (i) new values of line-to-ground admittance (Y_tEι) and degree of unsymmetry (|k,|), comparing the most recently computed values of the line-to-ground admittance (Y_tEi) with the reference values (Yt_Ei_re_f) and determining whether the difference therebetween exceeds the inaccuracy (ΔY_tEι ) of the measurement technique used, - if the comparison gives a value greater than said measurement inaccuracy (ΔYtEi ), checking for instance on the basis of the change in the entire network's summed line-to-ground admittance, which is computable from zero- sequence current of the feeding power source, whether a change has occurred in the connection status of the network and, if so, storing the most recently measured values of the line-to-ground admittance (Y_tEi) and degree of unsymmetry (|k,|) as new reference values, while, if no change has occurred in the network connection status, indicating a ground fault, and when necessary, comparing the computed degrees of unsymmetry of the sending ends (i) with the stored reference values |k_ιref| and, if the change of the computed degree of unsymmetry exceeds said predetermined limit |Δk,| of change, defining the monitored sending end to be affected by a high- impedance, single-phase ground fault.

2. Method according to claim 1, characterized in that the values of line-to-ground admittance (Y_tEι) and degree of unsymmetry (|k,|) are computed in a compensated network for the entire network from the zero-sequence current that is computed from the phase leg currents of the sending end in concern.

3. Method according to claim 1, characterized in that the values of line-to-ground admittance (Y_tEι) and degree of unsymmetry (|k_ι|) are computed in a compensated network for the entire network from the zero-sequence current of the arc-suppression reactor branch with the provision that such measurement facility is available.

4. Method according to claim 1 for detection of high-impedance ground faults in a medium -voltage network, characterized in that the degree of unsymmetry at the sending ends (i) is monitored in said method.

5. Method according to claim 1 for detection of high-impedance ground faults in a medium-voltage network, characterized in that the insulation status of sending ends (i) is monitored by way of monitoring the line-to-ground leakage resistances of the phase legs of the sending ends (i), as well as the degree of unsymmetry or the unsymmetrical current component at the sending ends (i), so that at the occurrence of a change in the degree of unsymmetry at a given sending end, from the computed fault resistance (R_f) is made a distinction between a change in the connection status and a ground fault, the latter result also giving such information on the faulted phase leg that can be later utilized in the location of faults not identifiable by visual inspection.

6. Method according to any one of foregoing claims for detection of high-impedance ground faults in a medium-voltage network, characterized in that a change in the connection status is indicated from the consequences in the network zero-sequence voltage caused by a randomly occurring ground fault or change in the connection status.

7. Method according to any one of foregoing claims for detection of high-impedance ground faults in a medium-voltage network, characterized in that a change in the connection status is indicated by way of artificially causing a change in the zero- sequence voltage through, e.g., altering the tuning of the arc-suppression reactor or making a change in the electrical length or capacitance unsymmetry of the network.

8. Method according to claim 7 for use on a substation having a centralized compensation system, characterized in that to an auxiliary winding (e.g., to a 500 V winding) of the arc-suppression reactor is connected either an adjustable or a constant voltage supply for the purpose of shifting the neutral voltage.

9. Method according to any one of foregoing claims for detection of high-impedance ground faults in a medium-voltage network, characterized in that a change in the connection status is indicated by way of resolving with the help of the network data the zero-sequence susceptance of the sending end in concern.

10. Method according to any one of foregoing claim, characterized in that the unsymmetry current component (I_OK is used in lieu of the values of degree of unsymmetry (|k,|) and its reference values (|k_iref|).