KR101510617B1 - A digital multi-function breaker for integrated blocking of arc, short circuit, leakage and overload current and the blocking method thereof - Google Patents

Abstract

The present invention relates to a digital multifunctional circuit breaker and a method thereof for integrally blocking an arc, a short circuit, a short circuit, and an overcurrent, and is capable of integrally detecting and interrupting an arc, a short circuit, a short circuit and an overcurrent through an artificial neural network, The structure of the neural network is designed to include an arc fault classifier for determining an arc fault and a special load classifier for classifying special loads that can cause malfunction, And an arc fault classifier based on a separate artificial neural network corresponding to an arc fault with respect to a special load is driven and verified to improve the accuracy of arc fault judgment, and a blocking method thereof.

Description

TECHNICAL FIELD [0001] The present invention relates to a digital multifunctional circuit breaker for integrally blocking an arc, a short circuit, a short circuit, and an overcurrent, and a blocking method thereof, and a digital multifunctional circuit breaker,

The present invention relates to a digital multifunctional circuit breaker for blocking an arc, a short circuit, a short circuit, and an overcurrent, and a method for blocking the same. More particularly, the artificial neural network structure is used for an arc fault classifier It is designed to include the special load classifier which can classify the special loads and it is possible to judge whether the arc fault and the special load type can be judged at the same time. In case of the arc fault to the special load, a separate artificial neural network based arc It is possible to improve the accuracy of the arc fault determination by performing the verification by driving the fault classifier.

In modern housing, electricity is a very important energy source that is indispensable for the convenience of life, but it is also a very dangerous thing that can lead to electric fire or electric shock, resulting in property loss or serious personal injury. Therefore, the Electricity Regulations require the installation of a molded circuit breaker (MCCB) or earth leakage circuit breaker (ELCB) at the entrance of houses or buildings. The overcurrent breaker is designed to protect the circuit from overloading and to prevent the resulting fire. On the other hand, when the leakage current from the circuit flows to the surface of the earth which is an unintended electric path, the leakage current causes the human body to constitute one electric path with the ground surface to prevent the electric shock to the human body Is designed. Such an earth leakage breaker prevents human shock as well as electric fire due to an arc or other failure between the ground surfaces. However, the cause of electric fire is caused not only by an overload but also by a short circuit as well as a large number of line arcs. Conventional circuit breakers or earth leakage breakers have a RMS (Root Mean Square) Could not be detected.

If the cord is stuck in a doorway or in furniture, the plug of the outlet may cause the plug to come loose when the wall is nailed or the staple is stuck to the table, , It occurs in a case where natural deterioration or a wire or a cord is exposed to a heat-radiating heat or sunlight, etc., as shown in FIG. 1 in terms of electrical and structural view. Such an arc current can be divided into a series arc occurring on the same voltage line, a parallel arc occurring between the voltage line and the neutral line, and a ground arc occurring between the voltage line and the ground line. That is, an arc current is a current generated for several microseconds (μs) by being electrically discharged by a voltage drop caused by contact of two electric wires such as an electric wire. Arc is a parallel arc that occurs between the voltage and neutral wires, a series arc that occurs when the voltage line is disconnected or loosely connected to an electric device, and a ground arc that occurs between the voltage line and the ground line. When an arc current is generated, joule heat is generated between the two electric wires. Joule heat generated by the arc current generates heat higher than the ignition point, which causes electric fire.

In this case, the series arc usually occurs in the contact portion or the disconnection portion of the voltage line due to the contact failure of the voltage line. In this case, the contact resistance is increased and the contact portion is overheated. A parallel arc occurs when a voltage line and a neutral line come into contact with each other, and a ground arc occurs when a so-called short-circuit occurs in which an arc is generated due to a decrease in insulation between a voltage line and a ground line. That is, a typical example of the occurrence of the series arc is a contact failure. A typical example of the occurrence of the parallel arc is a short, and a typical example in which a ground arc occurs is a short circuit.

These line arcs generate a high temperature that can be burned on wood, paper, and rugs, and thus cause electrical fires, so a quick and accurate arc judgment methodology is required.

However, as described above, the arc fault can not be protected by the existing protection circuit because it belongs to the failure detection area of the conventional circuit breaker. Therefore, a new arc detection methodology is required. Due to the nonlinearity and intermittency of the arc, There are many difficulties in establishing the In addition, there are various types of arc fault types according to the load, there is a fear of malfunction due to the similar waveforms of dimming waveform and arc fault, and existing arc fault breakers are developed separately from over current or earth leakage breaker Which is very inefficient in terms of cost and space.

Korean Patent Registration No. 0952222 (Apr. 02, 2010) discloses an arc current breaking apparatus and method for detecting an arc current having a low voltage and a low current characteristic, When the input current is subjected to signal processing, heat is generated in the wire, and when the resistance of the wire becomes irregular, it is judged that the current inputted to the load is the arc current. It is about what can be prevented. This prior art discloses only the arc current detection and the arc current interruption, but does not disclose an integrated mechanism capable of integrally detecting and interrupting the arc, short circuit, short circuit, and overcurrent of the present invention, It is not even mentioned.

Accordingly, the present invention proposes a multifunctional digital circuit breaker using an artificial neural network capable of improving efficiency in terms of cost and space by providing an integrated mechanism capable of integrally detecting and interrupting arc, short circuit, short circuit and overcurrent We aim. In addition, unlike the existing methods, the ARNT is designed to include arc fault classifiers for arc fault diagnosis and special load classifiers for classifying special loads that may cause malfunction, And the arc fault classifier based on a separate artificial neural network corresponding to the arc fault for the special load is driven and verified to improve the accuracy of the arc fault judgment.

The terms "arc current", "short circuit current", "leakage current", "overload current" and "fault current" are used in the present invention. It may also be called a term. That is, unless otherwise noted, the fault current is a concept that includes an arc current, a short circuit current, a leakage current, a overload (load) current, a fault current, or a combination thereof.

In the present invention, the arc is divided into a series arc, a parallel arc, and a ground arc. In the case of a short circuit, a short circuit, and an overload, an arc may or may not occur. , Short-circuit, short-circuit, and overload, the arc does not necessarily occur. Therefore, arc fault due to short circuit, short circuit and overload, as well as short circuit, short circuit and overload should be detected and interrupted.

It is an object of the present invention to provide a multifunctional digital circuit breaker using an artificial neural network capable of improving efficiency in cost and space by introducing an integrated mechanism capable of integrally detecting and interrupting an arc, a short circuit, a short circuit and an overcurrent It is for that purpose.

In addition, the present invention is designed to include an arc fault classifier for determining an arc fault and a special load classifier for classifying special loads that may cause malfunction, And to provide a digital multifunctional circuit breaker and a method of isolating the same.

Further, the present invention provides a digital multifunctional circuit breaker capable of improving the accuracy of arc fault determination by driving an artificial neural network-based arc fault classifier corresponding to an arc fault with respect to a special load It has its purpose.

The present invention relates to a digital multifunctional circuit breaker and a blocking method thereof for integrally blocking an arc, a short circuit, a short circuit, and an overcurrent, and more particularly, At least one current transformer for measuring the current; At least one analog-to-digital converter for converting an analog output of the current transformer into a digital signal; A main controller that integrally detects an arc current, a short-circuit current, a leakage current, an overcurrent, or a combination thereof using a current measurement value input from the analog-to-digital converter and outputs a trip command of the breaker according to the detected result; And at least one circuit breaker for shutting off power supply to all or a part of a house or a building according to the trip command. The input node of the artificial neural network may be an analog-to-digital conversion circuit, and the input node of the artificial neural network may be an analog-to-digital converter Wherein the output node of the artificial neural network is connected to at least one or more special load classifiers including an arc fault classifier and a dimming load, Or a combination thereof, the arc fault and the special load type can be determined and output simultaneously.

The present invention also provides a method for integrally detecting and blocking an arc current, a short circuit current, a leakage current, an overcurrent, or a combination thereof, comprising the steps of: detecting a current in a normal current region, a fault current region, a video current region, Measuring a current through at least one current transformer; Converting the analog output measured through the current transformer to digital through at least one analog-to-digital converter; Judging a short circuit through a signal inputted through a video current transformer and an analog-digital converter; Determining an overcurrent through a signal input through at least one current transformer independent of the image transformer and an analog-to-digital converter; And determining an arc fault through a signal input through the at least one current transformer independent of the image transformer and the analog-to-digital converter. The method of integrally detecting and interrupting the arc current, the short-circuit current, the leakage current, the overcurrent, or a combination thereof may further include: issuing a trip command to the breaker when the leakage current, the overcurrent, Determining the arc fault based on the artificial neural network and determining an arc fault based on the artificial neural network comprises the steps of analog-to-digital converting a signal output from the at least one current transformer, And applying a ratio of a fundamental wave and a harmonic wave to the input node of the artificial neural network. The arc fault determination based on the artificial neural network may be performed using a predetermined window Periodically judging an arc fault with a period of a half wave or a half wave according to the size of the arc And determining whether an arc fault occurs while moving the predetermined window. The arc fault analyzer may further comprise an arc fault classifier, and a controller for controlling the arc fault classifier, Further comprising the step of simultaneously determining whether the arc fault has occurred and the special load type through a classifier including at least one or more special load classifiers, or a combination thereof, and outputting the result through an output node of the artificial neural network, A method for integrally detecting and interrupting a short-circuit current, a leakage current, an overcurrent, or a combination thereof may include: driving an arc fault classifier based on a separate artificial neural network corresponding to an arc fault with respect to a special load; Further, it is possible to improve the accuracy of arc fault determination, The system is characterized by verifying an arc fault based on an artificial neural network.

The present invention relates to a digital multifunctional circuit breaker and a method thereof for integrally blocking an arc, a short circuit, a short circuit, and an overcurrent, and is capable of integrally detecting and interrupting an arc, a short circuit, a short circuit and an overcurrent through an artificial neural network, The structure of the neural network is designed to include an arc fault classifier for determining an arc fault and a special load classifier for classifying special loads that can cause malfunction, And an arc fault classifier based on a separate artificial neural network corresponding to the arc fault for the special load is driven and verified to improve the accuracy of arc fault judgment.

1 is an exemplary diagram for explaining a kind of line-to-line arc;
2 is a block diagram illustrating a structure of a digital multifunctional circuit breaker for integrally detecting and interrupting a fault current, a load current, an arc current, a short-circuit current, a leakage current, or a combination thereof according to an embodiment of the present invention. An example.
FIG. 3 is a block diagram illustrating a digital multifunctional breaker for detecting and interrupting a digital multifunctional breaker for integrally detecting and interrupting a fault current, a load current, an arc current, a short-circuit current, a leakage current or a combination thereof according to an embodiment of the present invention. A flowchart to illustrate the strategy.
4 is a flowchart for explaining an artificial neural network based arc current and load classifier learning method according to an embodiment of the present invention.
5 is a diagram illustrating an arc current and load classifier structure based on a neural network according to an embodiment of the present invention.
6 is a flowchart for explaining an arc current classifier learning method for a special load which is difficult to distinguish according to an embodiment of the present invention.
FIG. 7 is an exemplary view for explaining a structure of an arc current classifier for a special load which is difficult to discriminate according to an embodiment of the present invention; FIG.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a digital multifunctional circuit breaker for integrally detecting and interrupting a fault current, a load current, an arc current, a short-circuit current, a leakage current, or a combination thereof according to the present invention will be described with reference to the accompanying drawings. The embodiment will be described.

2 is a block diagram illustrating a structure of a digital multifunctional circuit breaker for integrally detecting and interrupting a fault current, a load current, an arc current, a short-circuit current, a leakage current, or a combination thereof according to an embodiment of the present invention. Fig.

As shown in FIG. 2, the present invention can be used to accurately and integrally detect and interrupt a fault current, a load current, an arc current, a short-circuit current, a leakage current, Digital multifunctional breaker. In particular, the multifunctional multi-function circuit breaker proposes a new strategy for verifying the arc fault determination result in order to increase the reliability of the arc fault judgment. The digital multifunctional circuit breaker includes a current transformer CT1 (10) for sensing a current value from an electric circuit, a current transformer CT2 (20), a current transformer ZCT (30) for sensing an image current, an analog current detected from the current transformers (Hereinafter referred to as ADC) 40-1, 40-2, and 40-3 for providing the digital main controller with the digital main controller, A main controller 50 based on a microcontroller 51 and a memory device 52 for detecting a short circuit current, a leakage current or a combination thereof and instructing an electric circuit breaker to cut off a circuit, And a circuit breaker (60) for breaking the power circuit. In order to avoid inaccuracies due to the saturation characteristics of the iron core, CT1 (10) for measuring the steady current region, CT2 (20) for measuring the current in the fault current region, and ZCT (30). The main controller 50 measures a normal current region and a fault current region through the CT1 10 and the CT2 20 to determine an overcurrent and a fault current based on the measured current and the fault current, It is possible to determine whether or not a short circuit is occurring.

If the current exceeds the limit in the fault current (overcurrent) range, the current transformer will saturate the iron core, and the error will increase rapidly, and the cross-sectional area of the iron core will be increased to improve the overcurrent characteristic. When a short-circuit current is generated in the power circuit, a large fault current flows in the primary winding of a current transformer connected to the main circuit, and the temperature of the primary winding rises due to a large fault current (overcurrent) The winding may be deformed by the electromagnetic force. For such an accident, the current transformer should be designed to withstand thermal and mechanical resistance. In the present invention, however, two current transformers CT1 and CT2 are used to stably measure the fault current.

FIG. 3 is a block diagram illustrating a digital multifunctional breaker for detecting and interrupting a digital multifunctional breaker for integrally detecting and interrupting a fault current, a load current, an arc current, a short-circuit current, a leakage current or a combination thereof according to an embodiment of the present invention. It is a flowchart for explaining the strategy.

3, the digital multifunctional circuit breaker according to the present invention performs an operation of making a breaker trip command through an arc fault determination algorithm based on an overcurrent judgment, a short circuit judgment and an artificial neural network . Here, the detection and blocking of the short-circuit current can be performed through the detection and interception of the over-current, so that the determination of the over-current can be regarded as a concept including the determination of the short-circuit current. If the breaker trip command is performed separately according to the overcurrent, the short circuit and the arc fault, the overcurrent, the short circuit and the arc fault can be judged separately. The digital multifunctional breaker of the present invention is efficiently designed to simultaneously determine the overcurrent, the short-circuit and the arc fault through the same current sampling data.

As described above, the digital multifunctional circuit breaker of the present invention includes an electrical leak judgment step, an overcurrent judgment step, and an arc fault judgment step. In the digital multifunctional blocking method, the ADCs firstly output currents inputted from CT1, CT2 and ZCT (S101), n = m = 0, E _n = P _m = CFCN = AFCFCN = 0 (S102).

For reference, E _n is the amount of image current accumulated up to the nth propagation waveform, and P _m is the amount of phase current accumulated up to the mth propagation waveform. That is, n and m represent the number of measurement times of the radio wave waveform measured by changing the image current and the phase current analog signal into a high-resolution digital signal, respectively. Also, the CFCN (Current Full-wave CouNter) represents the number of current propagation, and the AFCFCN (Arc Failure Current Full-wave CouNter) represents the number of current propagation in which an arc failure occurs.

On the other hand, the main controller collects the current sampling data from the ADCs at regular intervals and sequentially stores them in the arrays (S103), and checks whether half-waves are obtained based on the zero crossing every time sampling data is taken. If the radio wave is not complete, repeat the data collection until the radio wave is completed. On the other hand, if the propagation is completed, the process proceeds to the next step (S104). Through this process, data on the propagation waveform of the normal current can be constructed. Here, the zero crossing means when the sign of the current waveform is changed or when the sampling number is equal to a predetermined half wave number.

The main controller calculates the current value I ₀ of the current for the current ZCT propagating current and the current value I _a for the current CT propagating current at step S 105.

If the current I ₀ is less than the leakage current, n = 0, E _n = Is set to 0 and then turned over to the over-current determining step, the image current I ₀ is still the same or larger than the leakage determination process, a short circuit current operation (S106). The sequence current I ₀ is greater than or equal to the short-circuit operating current set the current (n-th) image current of the ZCT propagation current E _n a then calculated using the formula (1), n = n + 1 (S107) . If the next calculated E _n value is greater than the leakage current allowance, a breaker trip command is issued. If not, the next overcurrent judgment is performed. On the other hand, _{if the} value of E _n is smaller than the allowable leakage current limit, the operation proceeds to the overcurrent determining step (S108). That is, it is suspected that a short circuit has occurred only when the rms value of the image current measured at a specific region is larger than the rms value (or the leakage current operation current) of the normal current, and after the cumulative image current amount is measured, It is judged whether or not a short circuit has occurred.

For reference, the formula (1), △ T is the sampling period at, E _{n - 1} is the cumulative amount of current image to the previous (n-1-th) radio wave waveform.

If the phase current _Ia is smaller than the phase rated current, the main controller outputs m = 0 and _Pm = 0 and then proceeds to an arc fault determination step. If the phase current I _a is larger than the phase rated current, the overcurrent determination process is continued (S 201). Current (n-th) set the current image to the CT radio wave current to P _n and then calculated using the formula (2), m = m + 1 (S202). After calculating the overload ratio of the current I _a to the rated current, that is, the value for I _RMS / I _RATE, at the start of the program, corresponding to the ratio in the operating current table created based on the operating current characteristic curve stored in the memory in advance Is determined (S203). Next, if P _a is greater than the operation current allowable amount, the circuit breaker is operated, otherwise it is determined that there is no overcurrent, and the process goes to the arc fault determination step (S 204). Where I _RMS represents the RMS value of the current (I _a ), I _RATE (rated current) represents the rated current, and the overload ratio is calculated by the ratio.

In the equation (2),? T is the sampling period and Pm _-1 is the phase current amount accumulated up to the previous (m-1) th propagation waveform.

The main controller goes through a discrete Fourier transform (DFT) on the propagation waveform to calculate the magnitude components A (i), i = 0,. (1) divided by the size of the fundamental wave (A (1)), A (i) / A (1), i = 0, … (1), B (1), B (2), ...) as the input pattern of the learned multi-layer perceptron artificial neural network. , B (q)] (S302). Here, it is a well-known technique that a specific radio wave is subjected to discrete Fourier transform to divide the frequency component of the propagation waveform into frequency components up to the fundamental wave and the qth harmonic.

In step S303, the main controller determines whether the arc current propagates by using the arc current and load classification function of the artificial neural network by applying the IP to the multi-layer perceptron neural network input. If AFCFCN = 0 in the case where the inference result from the arc fault classifier node is a normal current propagation, the current sampling data is collected from the ADCs at regular intervals and stored in the arrays sequentially (S103), and AFCFCN = 0 Otherwise, CFCN = CFCN + 1 is set and the current electric wave number is incremented by 1 (S304).

For reference, a detailed description of the artificial neural network will be described in more detail with reference to FIGS. 4 and 5 below.

On the other hand, when the inference result from the arc fault classifier node is a fault current propagation, AFCFCN = AFCFCN + 1 is set and the value of hcn (AFCFCN) is stored as CFCN. Here, hcn (AFCFCN) is a memory space for storing the CFCN corresponding to the AFCFCN. Next, whether or not AFCFCN = 1 is checked. If AFCFCN = 1, set the window half-wave and set CFCN to 1. On the other hand, if AFCFCN is not 1, CFCN = CFCN + 1 is set (S305).

Next, the main controller determines whether or not AFCFCN ≥ 4 (that is, whether or not the arc propagation current is measured four times or more) (S306). If AFCFCN ≥ 4, it is based on UL1699 of an AFCI (Arc Fault Circuit Interrupter) After judging the arc fault, check the output from the load classifier node and check whether it is a special load that may cause malfunction. In case of normal load, it indicates breaker trip command. On the other hand, if it is a special load, the arc fault verification process for the special load is executed in order to improve the reliability of the arc fault judgment (S307). The arc fault validation process takes a discrete Fourier transform (DFT) on the propagation waveform determined as an arc current and calculates the magnitude component A (i), i = 0, ..., , q is obtained. (I) = A (i) / A (1) divided by the fundamental wave size, i = 0, ... (1), B (1), B (2), ...) as the input pattern of the learned multi-layer perceptron artificial neural network. , B (q)], and once again determines whether or not arc current is propagated by using the arc current and load classification function of the artificial neural network and determines the load type.

On the other hand, if AFCFCN ≥ 4 is not satisfied, it is determined that it is not an arc failure, and it is determined whether CFCN≥30. If CFCN? 30, CFCN with AFCFCN = 2 is determined as a new window start CFCN. Then, the current sampling data is collected at predetermined intervals from the ADCs and stored in the arrays sequentially (S307) .

 That is, the arc fault determination step sets 30 CFCNs to the arc propagation observation window range. If the AFCFCN is measured four times or more, it is determined that an arc fault has occurred, and then the arc fault is verified again. If the AFCFCN is less than 4, the arc fault determination process is repeated until the CFCN reaches 30, and when the CFCN is filled 30 times and the arc propagation observation window is completed, if the AFCFCN is less than 4, The next AFCFCN of the generated AFCFCN is set to the start CFCN of the new arc propagation observation window and the arc fault determination process is restarted so that it is possible to determine whether or not the arc fault is effectively caused even if the arc propagation occurs acyclic.

4 is a flow chart for explaining an artificial neural network based arc current and load classifier learning method according to an embodiment of the present invention.

As shown in FIG. 4, the structure, the neural network weight, and the bias threshold value of the multi-layer perceptron artificial neural network for judging the arc current and the load type in the digital multi-functional circuit breaker of the present invention are determined through an artificial neural network learning process .

First, normal waveforms and arc waveforms for each load are collected. At this time, the arc waveforms consist of series arc waveforms and parallel arc waveforms. In particular, arc waveforms for each load are collected for series arc waveforms. At this time, since the number of waveforms is large, clusters are determined as necessary for each of the normal waveform, the parallel arc waveform, and the serial arc waveform for each load (S401).

The clustering method is applied to each cluster, and the representative frequency spectrum A (i), i = 0, ... , q. Here, A (i) represents the magnitude of the i-th harmonic (S402).

Next, for all the half-wave waveforms, B (i) = A (i) / A (1), where i = 0, ..., , q are obtained (S403), and a learning pattern is generated for each normal half-wave waveform as follows.

P = [B (0), B (1), ... , B (q), 0, 0, 0]

P = [B (0), B (1), ... , B (q), 0, 1, 0]

P = [B (0), B (1), ... , B (q), 0, 0, 1]

On the other hand, the learning pattern for each half-wave half-wave is created as follows.

P = [B (0), B (1), ... , B (q), 1,0,0]

P = [B (0), B (1), ... , B (q), 1,1,0]

P = [B (0), B (1), ... , B (q), 1, 0, 1]

The learning pattern P = [B (0), B (1), ... , B (q), a, b, c] to B (0), ... B (q) represents the input value for the input node. On the other hand, a, b, and c denote output nodes, where a is an output node for arc current classification, b is an output node for dimming load classification, and c is an output node for other special load classification , The output node can be expanded according to the number of special loads that are difficult to classify the arc current (S404). That is, if [a, b, c] is [0,0,0], [steady current, dimming load (X), general load] ].

Next, by applying an error back propagation learning algorithm to all training patterns P and repeating learning until it reaches within the total tolerance (10 ^-4 ), the optimum structure of the multilayer perceptron, the optimal weight of the neural network and the bias threshold (S405)

The artificial neural network structure designed and learned at this time can be displayed as shown in FIG. 5 below.

5 is a diagram illustrating an arc current and load classifier structure based on a neural network according to an embodiment of the present invention.

5, the multi-layer Perceptron artificial neural network according to the present invention includes an input layer 100, hidden layers 200 and 200-1, and an output layer 300. The input layer 100, The number of input nodes of the hidden layer is determined according to the number of inputs influencing arc current or load type judgment, and the number of hidden layers and the number of nodes of the hidden layers are determined according to the result of learning. The output node 300 is composed of one output node for judging whether the propagating current is an arc current, two output nodes for determining whether the load is a general load or a special load capable of causing an error, Can be increased depending on the number of special loads that are difficult to classify.

6 is a flowchart for explaining an arc current classifier learning method for a special load which is difficult to discriminate according to an embodiment of the present invention.

As shown in FIG. 6, the multilayer perceptron artificial neural network for load type arc current determination is designed separately for each load type so as to have sufficient recognition capability. The structure of the artificial neural network for each load type, the weight of the neural network, and the threshold value are determined through the artificial neural network learning process.

First, after selecting a special load P _i that is difficult to distinguish, i is set in order starting from 1. Where P _i is the ith element of the set P of special loads that are difficult to classify. After creating a set W of waveforms with normal and series arc waveforms collected for the selected special load P _i , the waveform element W _i is selected as the selected waveform.

After selecting one representative radio wave for the selected waveform W _i , the frequency spectrum A (k), k = 0, ... , q. Here, A (k) represents the size of the kth harmonic (S501).

A (k) = A (k) / A (1), k = 0, ..., A (k) , q are obtained (S502). Then, based on the obtained results, the learning patterns for each stationary waves are T = [B (0), B (1), ... , B (q), 0] The learning pattern for each arc is T = [B (0), B (1), ... , B (q), 1]. The learning pattern T = [B (0), B (1), ... , B (q), o] to B (0), ... B (q) represents the input value for the input node. On the other hand, o indicates an output node, 0 indicates a normal current, and 1 indicates an arc current (S503).

Next, a set of all learning patterns TS is created, and it is repeatedly learned until all the elements of TS are within the total tolerance (10 ^-4 ) by applying an error back propagation learning algorithm to the structure of the multi-layer perceptron A neural network weight, and a bias threshold value are determined (S504).

If the designing and learning process of the artificial neural network for all the special loads is ended, that is, P = {}, the artificial neural network design and learning process is terminated. Otherwise, i = 1, i = i + 1 Then, the learning process is repeated for the remaining special loads, and a corresponding artificial neural network having the structure of FIG. 7 is designed for all the elements of P (S505).

At this time, the artificial neural network structure learned about P _i can be displayed as shown in Fig.

FIG. 7 is an exemplary view for explaining the structure of an arc current classifier for a special load which is difficult to discriminate according to an embodiment of the present invention.

As shown in FIG. 7, the artificial neural network structure learned for the special loads is configured as an input layer, hidden layers, and output layers as in the artificial neural network based arc current and load classifier structure do.

The number of input nodes of the input layer is determined by the number of inputs affecting arc current judgment for a particular load and the number of hidden layers and the number of nodes of hidden layers is determined according to the result of the learning. The output node has one output node for judging whether the propagating current is an arc current.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, I will understand the point. Accordingly, the technical scope of the present invention should be defined by the following claims.

10: CT1 20: CT2
30: ZCT 40: ADC
50: main controller 51: micro controller
52: memory 100: input layer
200: hidden layer 300: output layer

Claims (13)

At least one current transformer for measuring a current for a region including a steady current region, a fault current region, a video current region, or a combination thereof;
At least one analog-to-digital converter for converting an analog output of the current transformer into a digital signal;
A main controller for integrally detecting an arc fault, a short circuit current, a leakage current, an over current or a combination thereof by using a current measurement value input from the analog-to-digital converter and outputting a trip command of the circuit breaker according to the detected result; And
And at least one circuit breaker for shutting off power supply to all or a part of a house or a building according to the trip command,
In the case of the arc fault, the fault is detected through the artificial neural network, and the arc fault and the special load type are simultaneously determined through a classifier including an arc fault classifier, at least one special load classifier including a dimming load, or a combination thereof And outputs the digital circuit breaker. delete The method according to claim 1,
Wherein the input node of the artificial neural network performs a discrete Fourier transform (DFT) on an analog-to-digital converted signal, and then applies a ratio of each harmonic to a fundamental wave as an input signal thereof. delete A method for integrally detecting and shutting down an arc fault, a short-circuit current, a leakage current, an overcurrent, or a combination thereof,
Measuring a current for a region including a steady current region, a fault current region, a video current region, or a combination thereof through at least one current transformer;
Converting the analog output measured through the current transformer to digital through at least one analog-to-digital converter;
Determining a short circuit through a signal inputted through a video current transformer and an analog-to-digital converter as a current transformer for the image current region;
Determining an overcurrent through a signal input through at least one current transformer independent of the image transformer and an analog-to-digital converter; And
And determining an arc fault through a signal input through the at least one current transformer independent of the image transformer and the analog-to-digital converter,
The step of determining the arc fault is based on an artificial neural network. The arc fault is determined based on an arc fault classifier, at least one special load classifier including a dimming load, or a classifier including a combination thereof. The arc fault and the short-circuit current, the leakage current, the overcurrent, or a combination thereof are collectively detected and blocked. The method of claim 5,
And a step of issuing a trip command to the breaker when the leakage current, the overcurrent, or the arc failure is judged. The method of integrally detecting and blocking an arc fault, a short circuit current, a leakage current, an overcurrent or a combination thereof . delete The method of claim 5,
The arc fault determination is based on the artificial neural network,
And a step of performing a discrete Fourier transform on the result of analog-to-digital conversion of a signal output from at least one of the plurality of current converters and applying a ratio of a fundamental wave and a harmonic wave to an input node of the artificial neural network. And a short-circuit current, a leakage current, an overcurrent, or a combination thereof. The method of claim 8,
The arc fault determination is based on the artificial neural network,
The method of claim 1, further comprising: periodically determining an arc failure based on a predetermined window size, the arc fault and the short circuit current, How to. The method of claim 9,
The arc fault determination is based on the artificial neural network,
Further comprising the step of determining whether an arc fault occurs while moving the predetermined window. A method for integrally detecting and blocking an arc fault, a short circuit current, a leakage current, an over current, or a combination thereof. delete The method of claim 5,
And an arc fault classifier based on a separate artificial neural network corresponding to an arc fault with respect to the special load, thereby to improve the accuracy of the arc fault judgment. A short-circuit current, a leakage current, an overcurrent, or a combination thereof. The method of claim 12,
Wherein the verifying step verifies an arc fault based on an artificial neural network, wherein the arc fault and the short circuit current, the leakage current, the overcurrent, or a combination thereof are integrally detected and blocked.

Images