DE102006019467A1 - Electrical network`s short-circuit detecting method for use in e.g. industry, involves determining actual values as intensity of current flowing in electrical network, and detecting short-circuit when threshold value exceeds threshold value - Google Patents

Abstract

The method involves determining three actual values as intensity of current flowing in an electrical network, and determining time dissipations. A respective threshold value for the respective actual values is determined based on a locus criterion. A short-circuit is detected when threshold value exceeds respective actual values. The respective threshold value is determined by theoretical derivatives from a model of the electrical network. An independent claim is also included for a device for detection of short-circuit in an electrical network.

Description

The The invention relates to a method and a device for early detection of short circuits in electrical networks.

each electrical network is generally to protect against short circuits in the network. in the The area of private house installation are safety devices intended. In the industry, especially in electrical systems or machines, this is achieved by circuit breakers, which via appropriate trip units be shut off in the event of a short circuit. Today will be here in particular Molded case circuit breaker with electronic trip unit used.

such Protection devices are characterized by their current-limiting effect out. This current limit protects both the electrical equipment (cables, tracks, etc.), as also the connected loads (machines etc.). By shortened response times in the event of a short circuit, this current limit can be further improved become. There are also positive effects on the switching capacity early Detection of the short circuit and thus a reduction in the Switching in the device reached energy reached.

• – Ansprechzeiten (Detektionszeit + Eigenzeit des Schalters) möglichst unter 0,5 ms
• – Impedanzerhöhungen von niedrigen Werten (< 1 mΩ) um ca. drei Größenordnungen im Millisekunden-Maßstab
• – Stabilität des erhöhten Widerstandes bis zum vollständigen Abschalten des Stromkreises (geringes Wärmedurchlassintegral)
• – Zerstörungsfestigkeit auch bei extremer Schaltarbeit
• – reversibles Schaltverhalten
• – von wesentlicher Bedeutung für die Strombegrenzung ist die Verzugszeit von Kurzschlusseintritt bis zu einer signifikanten Widerstandsänderung

The current limitation is characterized by a reduction of the dynamic (given by the current square I ² ) and thermal loads (given by ∫i ² dt) of an electrical system in the event of a short circuit. To realize a good current limit the following points are required:

• - Response times (detection time + proper time of the switch) if possible below 0.5 ms
• - Impedance increases of low values (<1 mΩ) by about three orders of magnitude in the millisecond scale
• - Stability of the increased resistance until complete shutdown of the circuit (low heat transfer integral)
• - Destruction resistance even with extreme switching work
• - reversible switching behavior
• - Of significant importance for the current limit is the delay time from short circuit onset to a significant change in resistance

The Own times of an electronic evaluation unit are fixed Threshold of the undelayed Short-circuit tripping dependent the type of short-circuit (single-phase, three-phase) and phase angle the current at short-circuit entry. Through the use of short-circuit detection algorithms and an adaptation of the Schalteraktorik these Eigenzeiten clearly be reduced.

The problem of long short-circuit detection times of electronic trip units is particularly evident from the characteristics of the circuit breaker. If one compares the switch-off characteristics of two circuit-breakers with identical mechanical design and different trip units (electronic and thermomagnetic), a significantly longer switch-off time of the circuit breaker with electronic trip unit is noticeable, especially at high short-circuit currents (> 30 · I _r ). Due to the long exposure time, this approximately twice as long period of time entails a high load on the system to be protected and, due to the high energy converted in the switch, also has a negative effect on the switching capacity of the circuit breaker.

to Reduction in response times, have already been in different Pamphlets solutions on Basis of a short circuit early detection presented. In addition to the classic overcurrent releases, which the Instantaneous value of the flowing Comparing current with a threshold is considered another criterion for short-circuit detection, a di / dt tripping in [FRANKEN70: "FRANKEN, H .: low-voltage circuit breaker, Springer-Verlag, Berlin / Heidelberg, 1970.]] This type of triggering speaks only on the steepness of the current curve but not on the reached amperage at. First this criterion has been shortened the switch-off times of DC high-speed switches only with direct current Since, in AC systems, the relationship between Current slew rate, short circuit current and voltage dependent on Short-circuit point and the power factor of the network. In In practice, the triggering be made dependent on the current rate of increase by the trigger winding placed parallel to an inductive shunt. At high current increases the voltage increases in proportion to the current increase.

The detection of a short circuit with early detection of short - circuit from a combination of both criteria, instantaneous value of the current and current gradient, is recorded in DE 36 42 136 C2 presented. In every home For this purpose, a current sensor is used to determine the current gradients, eg Rogowski transducers. This is followed by a digital formation of pairs of values from current gradient and instantaneous value of the current, which are compared with predetermined threshold functions. This comparison is preferably carried out via a locus criterion. To trigger the switching device in the event of a short circuit, for example, a capacitor to the coil of a quick release (eg Thomson coil) are discharged. A disadvantage of the algorithm described in DE'136 is that it only allows evaluation of switching operations without bias current. Thus, changes in the power factor, for example, by switching on and off of large machines, and / or changes in the current already cause tripping, although they are still within the permissible range.

The algorithm according to [STEGE92, Stege, M .: short-circuit detection algorithms for current-limiting switching, dissertation, TU Braunschweig, 1992) is also based on the locus criterion shown in DE'136 and is intended to enable short-circuit detection within a few 100 μs Digital setups include the exact and easily reproducible setting options for the limit values as well as the possibility of remote monitoring.The sensor used here is a Rogowski coil, which measures the current gradient di (t) / dt.A numerical integration generates the temporal current profile from this signal i (t) superimposes this with the current gradients di (t) / dt occurring at the same time t and compares the sum of both with an extrapolation criterion according to FIG

An upstream low-pass filter reduces the influence of converter currents on the measuring signal. The value of the extrapolation criterion g _{extra is} calculated in a network with power factor cosφ _{r of} the maximum regular current I _or

Problematic the evaluation is carried out using such an algorithm of transients and harmonics in the current signal. Still must the power factor of the protected Network, i. known and rigid network conditions have to available.

In another method described in [RIVETTI 97 :, Rivetti, G., Teigland, J .: Intelligence in Low Voltage Circuit Breakers, ABB Technique, Issue 4, 1997], also becomes the envelope of a locus plot di (t) / dt i (t) characteristic as a triggering criterion used. The algorithm used there is essentially based on [STEGE92], has a detection time of approximately 300 μs and a for the Recognizing the short circuit necessary total detection time of less than 1 ms. The necessary sampling rate is 50-100 kHz per phase. For a trip of the switchgear are three consecutive samples outside the range defined as acceptable Area necessary. The algorithm became a self-powered protection device implemented, which allows protection in real time. Basic requirement for this are high-quality current sensors and adapted filtering high-frequency Signals. With the electronic trip unit PR223EF for molded case circuit breakers Starting at 250 A, the manufacturer offers ABB according to [ABB 04: ABB Leaflet, The new electronic releases PR223EF: more than just zone selectivity, ABB, 2004] the possibility the reduction of the tripping time at zone selectivity by an order of magnitude from a few hundred to a few milliseconds. Is realized this over an ASIC circuit, which the short circuit detection and the Locking the upstream devices simultaneously performs. In this electronic trip unit is the manufacturer's "EFDP" (Early Fault Detection and Prevention), which implements a Short early within 300 μs should be based on [STEGE 92] or [RIVETTI 97].

In [ROY 66: Roy, SC, extension of the triggering device of current-limiting switching devices to increase the tripping accuracy and achievement of selectivity, dissertation, TH Hannover, 1966] or DE 015 88 096 A a criterion for short-circuit detection was presented as the sum of the first and second derivative of the current.

The second derivative was generated by an LC circuit from the 1st derivative of the current, which is why the signal is associated with a balancing member. In addition, the signal is only very due to the LC circuit low performance, which makes his evaluation difficult. Also, no information is provided for determining the response value g and for determining the factors x and y.

The manufacturer ABB describes in DE 199 54 950 A1 a further method for detecting short circuits in distribution networks, in particular the medium voltage level, which consists of a conventional limit monitoring (eg current criterion, di (t) / dt-i (t) criterion) and an uncertainty area monitoring. This uncertainty improves the distinction between short circuits and short circuit-like events. Within this uncertainty range, discrete samples are examined for a present short circuit using artificial neural networks. In this case, the area "error" or "no error" of the classical method can be selected narrower, since a third area "perhaps error" is defined as an area of uncertainty, within which a decision is made by means of neural networks as to whether a short circuit exists Learning process, the neural network is able to classify the event correctly and to decide on a tripping of the switching device. Disadvantage of the method is the training process, which should include at least 500 circuits (including short circuits) and thus this type of early detection of short circuits, especially in the low voltage as little relevant to practice, even if a further independent adaptation of the algorithm to the present network conditions in this way feasible is.

DE 197 29 599 C1 describes a short circuit detection based on the locus criterion. In this case, an envelope of the loci is formed by polygons, whereby the switch-off criterion (tolerant locus criterion) becomes independent of the switching time, bias current and power factor of the network. The influence of harmonics can be compensated by widening the envelope, while at the same time worsening the detection properties. Current peaks occurring in the network are controlled by correspondingly high minimum sampling rates.

verwendet. Falls das Auslösekriterium erfüllt ist, wird das Auslösesignal an zwei antiseriell geschaltete J-FETs übertragen, welche den Stromkreis unterbrechen.In the DE 198 39 617 A1 the principle of short circuit tripping of a SiC-based semiconductor switch is shown. The trigger criterion is the product of instantaneous value of the current and the current current gradient

used. If the trigger criterion is met, the trip signal is transmitted to two anti-serial J-FETs which interrupt the circuit.

In DE 36 26 400 A1 and DE 41 28 618 A1 Further methods for fast detection of short circuits are presented. These are essentially based on a combined current and voltage measurement, from which the complex values for impedances and power of the network are formed. The resulting after a Fast Fourier transform FFT spectra are compared to detect short circuits with the thresholds stored for the corresponding network. Another method according to [FISCHER 95: Fischer, B., Method for rapid detection of state changes by means of short-term analysis in the audio-frequency domain, Dissertation, University of the Federal Armed Forces Hamburg, 1995] is also based on the method described above according to DE'400 and uses an analysis of Transient current and voltage signals in the audio frequency range during the short circuit. Here, real and imaginary part of the complex admittance of the network in the audio frequency range (f = 5 kHz) are detected and compared with thresholds. In addition, there is a discrete cross-correlation of current and voltage. The advantage of the algorithm is that the threshold values are independent of the prospective short-circuit current, the phase position of the current and voltage, and the time of the short-circuit occurrence. The prerequisite for this is that the spectral power densities of current and voltage, and thus the admittances or powers obtained therefrom, are known in the audio frequency range during normal operation.

task The invention is a method and an apparatus for improvement the current-limiting effect and the switching capacity of protective devices, in particular Compact circuit breakers with electronic trip unit for early detection of short circuits in electrical networks. A quick detection of short circuits in electrical networks should be so by electronic evaluation be realized by signals. By this shortening of the trip time should a significant improvement in current limitation compared to previous concepts are achieved.

The The object of the invention is achieved with respect to the method a method according to claim 1.

at the method according to claim Thus, 1 becomes a known locus criterion, which is a combined locus criterion Current and current gradient criterion (first and second actual value) represents, ie the current flowing in the network and whose first derivative uses the criterion of the second derivative (third actual value) of the current extended. Because the second derivative meets the other two criteria in the event of a short circuit, this additional criterion has an effect positive for a reduction of the short circuit detection times.

By the inventive method Therefore, the short circuit detection times are significantly reduced and thus the current-limiting properties of a corresponding Arrangement for significantly improved the electrical network.

Out the theoretical consideration of an equivalent circuit of the monitored Net regarding the second derivative of the flowing there Stromes or a simple network simulation can be done by overlaying loci Trip Levels be determined. The envelopes The loci are then used as a trigger criterion, ie recognition criterion for one Short circuit used.

Therefore Thus, the respective limit can be determined by theoretical derivation, Simulation or overlay of Ortskurvern be determined.

All Actual values occurring in a network can be displayed in a three-dimensional Space can be applied as a group of loci. The limits for short-circuit detection can from the envelope this locus can be determined. The realization of the method according to the invention gets through the approach the locus of the locus is simplified by an enveloping cuboid. This includes all regular ones conditions an ohmic-inductive system. A short circuit is always then detected when a three-dimensional value triple from first, second and third actual value leaves the cuboid.

bestimmt werden.The maximum value of the 2nd derivative, as the limit value of the third actual value, can be added

be determined.

und für die 1. Ableitung

The electricity continues to be the limit

and for the 1st derivative

The Actual values can when exceeded of a limit, activate a weighting factor. Only when the Sum of all weighting factors exceeds a limit, the short circuit recognized. Thus, e.g. a crossing an actual value alone does not trigger, if not Another actual value has exceeded its limit. This prevents false triggering become.

The Limit values or weighting factors or their sum can also stored in a queue. A short circuit is then recognized only when the sum of the values in the queue exceeds a limit. This brings an extra Avoidance of false triggering, caused by transient overshoot the limit values, with it, since the thresholds for the detection of a Short circuit exceeded several times Need to become.

A response value increases or decreases the response limits of the three individual criteria or limit values accordingly. For ohmic-inductive circuits, this threshold is eg I _i = 2. B. Machines with high inrush currents, the response value should be selected accordingly. Due to the multiplicative response value, the method according to the invention can also be adapted to machines and transformer starts as well as to the switching of capacitive loads, eg compensation systems, in addition to the entire spectrum of ohmic-inductive networks and thus offers an extended range of application.

As a sensor for recording the actual values or a variable for their derivation, a ro serve gowski coil.

When Actuator for unlocking the protection device can be used to further improve the current limit a high-speed release (preferably Thomsonantrieb or Piezoaktor) find use.

The Processing of the corresponding signals, ie the time course of the first, second and third actual value can be given in different ways respectively. The appropriate integration and / or differentiation can executed analogously become. The advantage here is a quick evaluation and thus Reaction time of the entire arrangement. A disadvantage of an analog Realization may be possible however, a high power requirement of the corresponding evaluation unit, susceptibility against EMC, insufficient temperature stability or aging.

alternative Therefore, the integration and / or differentiation can also numerically, so be digital. A digital realization is temperature stable, has a low power requirement and has hardly any aging. A simple implementation of changes is possible, what to flexibility or the possibility leads, environmental conditions to adapt. Could be problematic However, be a lower evaluation speed. Besides, they are for one exact discretization great Memory depths and high sampling rates necessary.

Regarding The device is achieved by a device according to claim 12. These and their benefits have already been linked with the method according to the invention explained.

For another Description of the invention will be to the embodiments of the drawings directed. Each shows a schematic schematic diagram:

1 the time course of a current and its first and second derivative,

2 the equivalent circuit diagram 2 a network to be protected,

3 a standardized representation of the second derivative of the current at the switch-on time of an ohmic-inductive circuit,

4 . 5 Loci of switch-on processes in ohmic-inductive networks with different switch-on phase angle of the voltage,

6 a three-dimensional border square,

7 an arrangement for carrying out the method according to the invention,

8th a flow chart for the inventive method.

1 shows over the time t plotted, the course of the current i (t) 100 whose first 102 and second 104 Discharge in the event of a short circuit. It can be seen as the second derivative 104 the other two curves ahead of time.

2 shows the equivalent circuit diagram 2 a protected network with a resistance R and an inductance L of the circle. For a rated voltage (effective value) U _N and a phase angle ψ of the voltage, the following differential equation for the switch-on operation of the single-phase ohmic-inductive circuit can be determined from this:

und einem Nennstrom (Effektivwert) I_N wird der Einschaltstrom im Zeitbereich durch folgende Gleichung beschrieben:

With a phase shift

and a rated current (RMS) I _N , the inrush current in the time domain is described by the following equation:

After two differentiations, the result for the second derivative of the current to be assessed is:

3 shows a normalized to ω ² I _N representation of the second derivative 106 of the current at the switch-on time of the ohmic-inductive circuit (t = 0), each having a lower power factor (eg cosφ _u = 0.1) and an upper power factor (eg cosφ _o = 0.9) over different phase angles ψ. Using theoretical considerations and simple network simulations, the tripping thresholds can be determined as a limit curve by superimposing loci within a defined power factor and switching angle range. The realization of an optimized for ohmic inductive networks short-circuit detection method thus takes place by using the envelope of the loci as a triggering criterion.

Examples of loci 24 and 26 of switch-on processes in ohmic-inductive networks with different switch-on phase angle of the voltage are in 4 and 5 shown.

In order to reduce the expense of implementing the algorithm in the protection device and to improve the disturbance behavior in use under real conditions, the tripping threshold values of the locus criterion become an enveloping cuboid 120 . 122 approximated. In the following, this simplified short-circuit detection algorithm is therefore referred to as a cuboid criterion.

beträgt der Maximalwert der 2. Ableitung für cosφ_o = 0,9 zum Einschaltzeitpunkt t = 0

At a switch-on phase angle of the voltage

is the maximum value of the 2nd derivative for cosφ _o = 0.9 at the switch-on time t = 0

With this maximum value of the second derivative as the triggering threshold, it is possible by extension with the limit values for current known from the literature and the current steepness for the application of the short-circuit detection algorithm to determine the in 6 set limit values for the cuboid criterion. 6 shows only the amounts of the three actual values current, 1st and 2nd derivative and thus only one eighth of the cuboid 124 in the three-dimensional positive and negative number space.

leaving the loci given by the short circuit detection criterion regular area, So the cuboid of the cuboid criterion, so an error is detected and the protection device (preferably circuit breaker) switched off.

und integriert durch einen Integrierer 48 (z. B. über Anwendung der Trapezregel)

7 shows a schematic structure of an apparatus for performing the method according to the invention on a network to be monitored 4 , As a sensor element is preferably a Rogowski coil 46 , Their signal is for further processing in the algorithm of an AD converter 44 digitized and differentiated by numerical methods by a differentiator 50 (eg via backward difference formation, central difference quotient or analog).

and integrated by an integrator 48 (eg via application of the trapezoidal rule)

In order to realize a comparison of the time-related values of current, current slope and 2nd derivative, the signal of the current gradient is via a delay element 58 each delayed by one sample.

An input filter 56 prevents false tripping caused by harmonics in the current signal, which are within the range of valid EMC standards (DIN EN 61000-3-2 and E DIN VDE 0838-2). As a result, the range of application of the algorithm, especially for use in industrial networks (eg Um- / inverter-powered loads), significantly expanded.

Subsequently, the signals are in the 8th shown algorithm for evaluation in an evaluation unit 60 fed. In this there is a simultaneous execution of the three short circuit detection criteria.

8th shows a flow chart for the inventive method.

About a weighting 30 . 32 . 34 the individual criteria (actual values 110 . 12 . 14 for current, current gradient, 2nd derivative ever solve when exceeding their limit 16 . 18 . 20 a weighting factor 30 . 32 . 34 off), false alarms should be avoided. The weighting of the signals is designed so that a triggering only by exceeding the limit 20 the second derivative is not possible. By the differentiation amplified high-frequency interference in the useful signal of the second derivative to be evaluated by an additional filter ( 52 ) damped. A high-pass filter compensates for the errors due to numerical integration of the signal by means of harmonics. The current instantaneous value signal generated in this way receives the strongest weighting because it is a comparatively safe criterion for a present short circuit. The weighting is done as a function of the network configuration or the consumers and equipment to be protected. For example, a weighting may be performed as follows: current criterion × 1, current slope × 0.5, and 2. derivative × 0.25.

A queue ( 38 ) is used for intermediate storage and evaluation of the last recorded weighted measured values. These are pushed through one digit at each sampling interval Δt of the A / D converter, so that old values fall out again. Exceeds the content of the queue 38 a set limit 42 , this leads to a triggering 22 of the protective device.

Since electrical networks, especially in the low voltage, can not be represented by a purely resistive-inductive network, has an adjustable response value 28 the instantaneous short-circuit release an adjustment of the criterion (limits 16 . 18 . 20 ) at switch-on processes of machines, transformers (inrush currents) and lines as well as capacitive loads (eg compensation systems). About this threshold 28 the response limits of the three individual criteria are defined according to the setting value. For ohmic-inductive circuits, this response value corresponds to I _i = 2. For other consumers z. B. Machines with high inrush currents, the response value should be selected accordingly.

The The method is therefore Ohmschinduktive in addition to the entire spectrum Networks also on machines and transformer starts as well as switching capacitive Loads, e.g. Compensation systems, customizable, thus offering a extended application area.

2

Equivalent circuit

4

network

10

Actual value (Current value)

12

Actual value (1st derivative)

14

Actual value (2nd derivative)

16

limit (Current value)

18

limit (1st derivative)

20

limit (2nd derivative)

22

short circuit

24 26

locus

28

pickup

30 32, 34

Weighting factor

38

queue

40

total value

42

limit

44

A / D converter

46

Rogowski coil

48

integrator

50

differentiator

52

lowpass

54

highpass

56

input filter

58

delay

60

evaluation

100

electricity

102

1. derivation

104 106

Second derivation

120 122, 124

cuboid

.delta.t

interval

R

Resitanz

L

inductance

Claims (24)

Method for short-circuit detection in electrical networks, in which: - a first ( 10 ), second ( 12 ) and third ( 14 ) Actual value is determined as the current flowing in the network, as well as their first and second time derivative, - Based on a locus criterion a respective limit ( 16 . 18 . 20 ) for the respective actual value ( 10 . 12 . 14 ) is determined, - if at least one of the respective limit values is exceeded by the respective actual value, a short circuit ( 22 ) is recognized. Method according to Claim 1, in which the respective limit value ( 16 . 18 . 20 ) by - theoretical derivations from a model ( 22 ) of the electrical network, and / or - numerical simulation of the electrical network, and / or - superposition of loci ( 24 . 26 ) for predeterminable power factor and switching angle ranges of the electrical network. Method according to Claim 1 or 2, in which the respective limit value ( 16 . 18 . 20 ) by multiplication with a predefinable response value ( 28 ) is determined. Method according to one of the preceding claims, in which the exceeding of a respective limit value ( 16 . 18 . 20 ) a weighting factor ( 30 . 32 . 34 ), and a short circuit is not recognized until the sum of the weighting factors ( 30 . 32 . 34 ) is above a predefinable limit. Method according to one of the preceding claims, in which a short circuit ( 22 ) is recognized only when at least one of the respective limit values ( 16 . 18 . 20 ) is exceeded at several consecutive times (Δt). Method according to Claim 5, in which the exceeded limit values ( 16 . 18 . 20 ) of successive times (Δt) in a queue ( 38 ) and a short circuit ( 22 ) is recognized only when the sum value ( 40 ) of all limits ( 16 . 18 . 20 ) the queue ( 38 ) above a threshold ( 42 ) lies. Method according to one of the preceding claims, wherein the first derivative of the current intensity in the electrical network by a Rogowski coil ( 46 ) and the instantaneous current through integration ( 48 ), and the second derivative by differentiation ( 50 ) are determined from this. Method according to Claim 7, in which the differentiation ( 50 ) and / or integration ( 48 ) takes place analogously. Method according to Claim 7, in which the differentiation ( 50 ) and / or integration ( 48 ) is performed numerically after an A / D conversion. Method according to Claim 9, in which the first derivative of the current is a sampling interval (Δt) of the AD converter ( 44 ) is delayed. Method according to one of Claims 7 to 10, in which the differentiation ( 50 ) a low-pass filtering ( 52 ) and / or integration ( 48 ) a high-pass filtering ( 54 ). Method according to one of claims 7 to 11, wherein the Rogowski coil ( 46 ) an input filter ( 56 ) is connected downstream. Short-circuit detection device in an electrical network ( 4 ), comprising: - a determination unit ( 60 ) for determining a first, second and third actual value ( 10 . 12 . 14 ) than those in the network ( 4 ) flowing current, and their first and second time derivative, - and to determine a respective limit ( 16 . 18 . 20 ) for the respective actual value ( 10 . 12 . 14 ) using a locus criterion, - and an evaluation unit ( 60 ) for detecting a short-circuit when at least one of the respective limit values ( 16 . 18 . 20 ) by the respective actual value ( 10 . 12 . 14 ). Apparatus according to claim 13, wherein the determining unit ( 60 ) a determination unit for determining the respective limit value ( 16 . 18 . 20 ) by - theoretical derivations from a model of the electrical network ( 2 ), and / or - numerical simulation of the electrical network ( 2 ), and / or - superposition of loci ( 24 . 26 ) for specifiable power factor and switching angle ranges of the electrical network ( 2 ) is. Apparatus according to claim 13 or 14, wherein the determining unit ( 60 ) an input for a predefinable response value ( 28 ) for multiplication by the respective limit value ( 16 . 18 . 20 ) having. Device according to one of Claims 13 to 15, in which the evaluation unit ( 60 ) is formed in such a way that the exceeding of a respective limit value ( 16 . 18 . 20 ) a weighting factor ( 30 . 32 . 34 ), and a short circuit is not recognized until the sum of the weighting factors ( 30 . 32 . 34 ) is above a predefinable limit. Device according to one of Claims 13 to 16, in which the evaluation unit ( 60 ) is formed in such a way that a short circuit is not recognized until at least one of the respective limit values ( 16 . 18 . 20 ) is exceeded at several consecutive times (Δt). Device according to one of claims 13 to 17, wherein the evaluation unit ( 60 ) a queue ( 38 ) for storing exceeded limit values of successive points in time, wherein the evaluation unit ( 60 ) is designed in such a way that a short circuit is only recognized when the sum value of all limit values of the queue exceeds a limit value ( 42 ) lies. Device according to one of claims 13 to 18, with a Rogowski coil ( 46 ) for determining the first derivative of the current intensity in the electrical network, and with an integrator ( 48 ) to form the instantaneous current through integration, and a differentiator ( 50 ) to form the second derivative by differentiation from the first derivative of the current. Apparatus according to claim 19, wherein the differentiator and / or the integrator works analogously. Device according to Claim 19, in which the differentiator and / or the integrator operates numerically one and one AD converter ( 44 ) is connected downstream. Device according to Claim 21, having a delay element ( 58 ) at the time delay of the first derivative of the current by one sampling interval of the AD converter ( 44 ). Device according to one of Claims 19 to 22, with a forward-connectable low-pass filter ( 52 ) for the differentiator and / or a switchable high-pass filter ( 54 ) for the integrator. Device according to one of claims 19 to 23, with one of the Rogowski coil ( 46 ) switchable input filter ( 56 ).