KR101148957B1 - Arc Fault Determination Method for the Breaking of the Arc Fault Current in the House - Google Patents

Abstract

The present invention can be easily applied to the arc failure generated in each line diverged from the distribution panel of the electric receiver, and the arc of the house can be judged on the basis of the discrete freeier transformation and the multilayer perceptron artificial neural network which can display a strong nonlinear relationship. Regarding the failure determination method,
That is, a main control device is installed through a single circuit breaker (MCCB) at a power supply unit of a distribution panel, and an earth leakage breaker (ELCB) is installed to cut a circuit when an overload or a short circuit occurs in each branch line. By configuring the single-phase power management device that collects and manages the power information from the single-phase power management device and the arc failure diagnosis device that determines the arc failure from the current, the single-phase power management device not only performs the power management task but also outputs the arc failure of the arc failure diagnosis device. If the circuit breakout command is not issued by the arc fault diagnosis device by periodically monitoring the pin, issue a circuit break command to the breaker (MCCB having an operating mechanism operated by an external command) or display the arc fault on the LCD and display the Internet. It is transmitted to the remote management system through the micro-controller provided in the arc fault diagnosis device The parser collects the sampled data to determine whether the arc has failed based on the arc fault determination technique, outputs the result to the arc output pin, issues a circuit breaker command to the circuit breaker, or checks the single phase power management device for arc failure. I would have to.

Description

Arc Fault Determination Method for the Breaking of the Arc Fault Current in the House}

The present invention is easy to apply the arc failure determination method for blocking the arc failure current of the house, more specifically, the arc failure generated in each line branched from the distribution panel of the electric customer, easy to apply a discrete non-linear relationship can be displayed The present invention relates to a method of determining arc failure of a house that can be determined based on transformation and multilayer perceptron artificial neural network.

In general, a distribution board is installed at an inlet of a system for supplying electricity to an electric customer, and a circuit breaker (MCCB) is provided at the power supply side of the distribution board to cut off the power supply when an overload occurs or work is performed. Each branch line branched from the distribution panel is provided with an earth leakage circuit breaker (ELCB) for the purpose of protecting an electric fire or a human body from leakage current when an electric leakage occurs (see FIG. 1).

However, since the circuit breakers or the ground fault circuit breakers operate based only on the root mean square (RMS) size of the load current or the image current, the arc failure, which is the main cause of the electric fire, cannot be determined. These circuit breakers have been exposed to the problem that the fault current cannot be blocked because the arc current is too small to detect faults in series and parallel arcs. Has caused a very serious problem.

According to a report from the United Ststes Fire Administration (USFA), 67,800 electric arc fires occur annually in homes in the United States, resulting in 485 victims, 2,300 injured people, and 680 billion won in property losses.

Therefore, it is urgently required to develop an arc failure determination technique that can significantly improve the safety of electricity use by accurately recognizing and blocking low voltage arc failure from the consumer perspective.

Until now, a number of studies have been conducted on the method for determining the arc failure of a house.However, the arc failure is judged by detecting changes in characteristics such as current size change, high frequency chattering phenomenon, shoulder phenomenon, intermittent steep rise rate and pulse width. Strategies were presented. However, since the change of these features is not consistent and the relationship between these features and the arc failure shows strong nonlinearity, the reliability of the arc failure judgment is greatly reduced.

The present invention is to solve the problem that can not properly detect the arc failure when the arc current is fine in the conventional electroconductor, the object of which is to determine the arc failure based on discrete freezer transformation and multilayer perceptron artificial neural network The present invention provides a method for determining arc failure for blocking current of arc faults in residential houses.

The present invention for achieving this purpose, the main circuit breaker (ELCB) is installed in the power supply unit of the distribution panel via a single circuit breaker (MCCB) as a medium and cut off the circuit when an overload or a short circuit occurs in each branch line. And the main controller comprises a single phase power management device for collecting and managing power information from a power line and an arc failure diagnosis device for determining an arc failure from a current. Periodically monitor the arc fault output pin of the arc fault diagnosis device, and if a circuit break command is not issued by the arc fault diagnosis device, issue a circuit break command to the breaker (MCCB having an operating mechanism operated by an external command), or It is displayed on the LCD and transmitted to the remote management system via the Internet. The microprocessor unit of the fault diagnosis device collects the sampled data to determine whether the arc has failed based on the arc fault determination technique, and outputs the result to the arc output pin, and gives a circuit breaker command to the circuit breaker or sends it to the single phase power management device. It is characterized in that it can be confirmed whether the arc failure.

The arc fault determination method for blocking the arc fault current of the house of the present invention is to sample the current signal flowing in each branch line passing through the distribution panel of the electric customer and collect the data to determine whether the arc fault caused by the overload According to the result, the circuit of the circuit breaker can be quickly disconnected or the arc fault can be checked. Thus, the problem that the arc current was too small and the failure detection was difficult due to the conventional arc occurrence can be minimized, thereby minimizing the fire caused by the arc failure. There is, there is an advantage to improve the safety of electricity use.

1 is a configuration diagram of a distribution panel installed at an inlet of a general electric customer;
Figure 2a and 2b is a block diagram showing a state in which the main control device is applied to the distribution panel and the main control device in detail for the arc fault current blocking of the house of the present invention,
3 is a waveform diagram illustrating an arc failure determination concept based on UL1699 proposed by the present invention;
4 is a flowchart showing a learning process based on a multilayer perceptron artificial neural network of the present invention;
5 is a flowchart illustrating an arc fault current determination method of current half-waves based on the artificial neural network of FIG. 3;
FIG. 6 is a flowchart illustrating an entire arc fault determination based on the discrete free Fourier transform and the multilayer perceptron artificial neural network proposed by the present invention.

Hereinafter, the arc failure determination method for blocking the arc fault current of the house of the present invention will be described with reference to the accompanying drawings.

Figure 2a is a block diagram showing a state in which the main control device is applied to the distribution panel for the arc fault current blocking of the house of the present invention, the main control device 100 is provided via a single circuit breaker (MCCB) in the power supply unit of the distribution panel An earth leakage breaker (ELCB) is installed in each branch line to cut off the circuit when an overload or a short circuit occurs in each branch line, and the MCCB and ELCB are operated by an external command and an Ethernet converter is designed to transmit information to a remote site. do.

FIG. 2B is a block diagram illustrating a main controller applied to the distribution panel in detail. The main controller 100 for diagnosing arc failure of the present invention includes a single-phase power management apparatus 10 for collecting and managing power information from a power line; It consists of the arc fault diagnosis apparatus 20 which judges an arc fault from an electric current.

The single phase power management IC provided in the single phase power management device 10 receives current and voltage signals from the current sensor CT and the voltage sensor circuit to record and manage voltage RMS, current RMS, active power, reactive power, and power consumption data. . The microprocessor periodically obtains power information from the single-phase power management IC, displays it on the LCD, transmits it to the home or remote management system via the Internet as needed, and also controls the electronic relay by the command of its own or remote management system. Control the branch line load on / off.

In particular, the single-phase power management device 10, in addition to performing a power management task, periodically monitors the arc failure output pin of the arc failure diagnosis device 20. If the value of the arc fault output pin changes from the steady state value 1 to 0, it is judged as an arc fault. If a circuit break command is not issued by the arc fault diagnosis device, the breaker (MCCB having an operating mechanism operated by an external command) It sends a circuit break command or displays an arc fault on the LCD and transmits it to the remote management system via the Internet.

The AD converter provided in the arc fault diagnosis apparatus 20 periodically records a current sample signal from the current sensor CT. The microprocessor unit collects the sampled data to determine whether the arc has failed based on the arc fault determination technique, outputs the result to the arc output pin, and sends a circuit break command to the breaker (MCCB having an operating mechanism operated by an external command). Lower or allow the single-phase power management device to check for arc failure. In addition, the test button is designed to check the self-operation performance, and when the test button is pressed, it is designed and implemented to output the arc failure to the output pin based on the arc failure determination technique based on the built-in arc failure waveform.

  In addition to the power management task, the single-phase power management device is designed to monitor this output pin periodically and provide the result to the house or the remote management system if it is determined that the arc has failed. The arc fault judgment device of the house used the familiar AT91SAM7S256 as a microcontroller and the AD7688 of analog device as an AD converter.

According to the Arc Fault Circuit Interrupter (AFCI) standard UL1699, an arc fault circuit breaker is defined as having an arc fault current and detecting a circuit fault when eight half cycles (HF) are identified as arc fault currents within 0.5 seconds. Therefore, the signal processing window for determining the arc failure may be defined as 30 full cycles or 60 half cycles.

FIG. 3 is a waveform diagram illustrating an arc failure determination concept based on UL1699 proposed by the present invention, and black half waves show a shoulder phenomenon, which is a general phenomenon of arc failure, and thus an arc fault current half cycle (AFCHC). And the remaining half waves represent the steady current half waves. In the waveform diagram, the number corresponding to the arc fault current half wave indicates the number of the arc fault current half wave.

The arc fault determination mechanism begins by continuously monitoring the fault current. When monitoring the current continuously and discovering the first arc fault current half wave (AFCHC) that forms the black half wave, the starting point of the window for determining the arc fault is the starting point of the arc fault current half wave (point 1 in FIG. 3). The first window, window 1, is determined by recording 60 half-waves regardless of the steady current half-wave or arc fault current half-wave from the starting point.

At the same time, the number of arc fault current half-waves (AFCHCN) in Window 1 is counted, and if AFCHCN≥8, it is determined as an arc fault, otherwise it is determined to be normal. Because it is 7, it fails the arc fault judgment.

In this case, the window moves. The starting point of the second arc fault current half wave (AFCHC, point 2 in FIG. 3) is the starting point of the window, and the current half wave is continuously recorded so that the number of current half waves (CHCN) in the window regardless of the steady current half or arc fault current half wave. With 30), a new window, window 2, is defined.

At the same time, the number of arc fault current half cycle numbers (AFCHCN) in window 2 is counted. If AFCHCN≥8, it is determined as an arc fault, otherwise it is determined to be normal. In the drawing, since AFCHCN = 7, arc failure determination fails. Thus the window is moved again.

The starting point of the third arc fault current half wave (AFCHC, point 2 in FIG. 3) is the starting point of the window, and the current half wave is continuously recorded so that the number of current half waves (CHCN) in the window regardless of the steady current half or arc fault current half wave. With 30), a new window, window 3, is defined.

At the same time, the number of Arc Fault Current Half Cycle Numbers (AFCHCN) in Window 3 is counted. If AFCHCN≥8, it is determined as an arc fault, otherwise it is determined to be normal. In the drawing, since AFCHCN = 8, it is determined as an arc failure.

The remaining problem remains to determine whether each half wave is a steady current half wave or an arc fault current half wave. At this time, the determination of the arc fault current of the current half-wave is based on the multilayer perceptron artificial neural network described below.

That is, the arc fault determination method of the current half-wave is based on a multilayer perceptron artificial neural network, wherein the structure, neural weight and bias threshold (threshold voltage) of the multilayer perceptron artificial neural network are shown in FIG. 4, It is determined through the learning process of artificial neural network, and the learning process of the multi-layer perceptron artificial neural network is as follows.

First, collect steady and arc waveforms for each load. In this case, the arc waveforms are composed of series arc waveforms and parallel arc waveforms. In particular, the series arc waveforms collect arc waveforms for each load. At this time, since the number of waveforms is large, clusters are determined as necessary for the parallel arc waveform, the stationary waveform for each load, and the series arc waveforms for each load.

i번째 이산 프리에 변환결과 X[i]는 복소수이며, 실수부 Re(X[i])와 허수부 Im(X[i])로 구성된다.
상기 클러스터링 기법을 적용하여 각 클러스터 별, 각 반파 별로 대표적인 주파수 스펙트럼 A(i), i=1,...,n을 얻는다. 여기서, A(i)는 반파에 대한 이산 프리에 변환으로부터 얻은 i번째 고조파의 크기로서 다음 식(2)로 표시된다.

A half-wave discrete Fourier transform is attempted for each cluster. When the number of half-wave sample data collected is N and the time-domain sequence is x [n], the discrete Fourier transform is obtained from Equation (1).

The result of the i-th discrete free conversion, X [i], is a complex number and consists of a real part Re (X [i]) and an imaginary part Im (X [i]).
By applying the clustering technique, a representative frequency spectrum A (i), i = 1, ..., n is obtained for each cluster and each half wave. Here, A (i) is the magnitude of the i-th harmonic obtained from the discrete Fourier transform with respect to the half wave, and is represented by the following equation (2).

Next, for all half-wave waveforms, we obtain B (i) = A (i) / A (1), i.e., A (i) divided by the fundamental value, by dividing each harmonic's magnitude by the fundamental's magnitude, i = 1 , ..., 9. Next, learning patterns P = [B (0), B (2), .., B (9), 0] are generated for each of the standing wave waveforms. On the other hand, learning patterns P = [B (0), B (2), ..., B (9), 1] are generated for each of the arc half wave waveforms.

Next, the error back propagation learning algorithm is applied to all learning pattern Ps, and iterates and learns until it falls within the total tolerance (10 ^-4 ) to determine the structure, neural network weight, and bias threshold value of the multilayer perceptron. do.

In order to determine whether an arc of current half wave is broken using an artificial neural network, the AD converter collects half wave waveform data by sampling a current input from the CT at regular intervals. Once the half-wave data is obtained, first try discrete discrete conversion to obtain the frequency spectrum A (i), i = 1, ..., n.

Then, after obtaining B (i) = A (i) / A (1), i = 1, ..., 9 divided by the fundamental wave size, input pattern IP (Input Pattern) = [ B (0), B (2), ..., B (9)]. Finally, this input pattern (IP) is applied as an input to the multilayer perceptron artificial neural network learned through the process of 3.2.1 to determine whether the arc failure.

FIG. 5 illustrates an arc fault current determination technique of an artificial neural network based current half wave, and FIG. 6 shows an overall flow chart of an arc fault determination based on a discrete free-ier transform and a multilayer perceptron artificial neural network proposed in the present invention. This may be described as the following first to seventh steps.

Step 1: The AD converter samples the current input from the current sensor CT at regular intervals.

Step 2: The microprocessor collects current sampling data at regular intervals and stores them in an array through SPI communication with the AD converter. At the same time, it is checked whether or not all half-waves have been obtained on the basis of zero crossing at every sampling data. If the half wave is not complete, repeat step 2 and continue collecting data until the half wave is completed. On the other hand, if the half wave is completed, go to the third step. Here, the zero crossing means the moment when the sign of the current waveform is changed or when the number of sampling is equal to the number corresponding to the predetermined half wave.

Step 3: The microprocessor takes a discrete free transform on the half wave waveform to obtain the magnitude components A (i), i = 1, ..., n of 0 to n harmonics. Then, after obtaining B (i) = A (i) / A (1), i = 1, ..., 9 divided by the fundamental wave size, IP = [B (as the input pattern of the learned multilayer perceptron artificial neural network. 0), B (2), ..., B (9)] and go to the next step.

Step 4: The microprocessor determines whether it is an arc fault half wave or a steady current half wave by using a theory based on the artificial neural network of the multilayer perceptron structure. If the normal current half wave is AFCHCN (number of half fault currents of arc fault current) = 0, go to the second step and continue collecting sampling data. Otherwise, set CHCN (Current Half Cycle Number) as the number of current half-waves and the normal current half-wave number (arc fault current half-wave) = CHCN + 1, and increment the current half-waves by one, and then, in step 6, Going to On the other hand, if the arc fault current is half wave, go to the fifth step.

Step 5: The microprocessor sets AFCHCN = AFCHCN + 1. And stores the value of hcn (AFCHCN) as CHCN. Here, hcn is a memory space for storing the number of current half-waves (CHCN) corresponding to the AFCHCN-th arc fault current half-wave. Next, it is checked whether AFCHCN = 1. If AFCHCN = 1, set the window start half wave, set CHCN to 1, and then go to step 6. On the other hand, if AFCHCN is not 1, CHCN = CHCN + 1 is set, and the flow proceeds to step 6.

Step 6: It is checked whether AFCHCN≥8. If AFCHCN≥8, it is determined as an arc failure based on UL1699, and outputs an arc failure state to an arc failure pin, and then stops inference. If not, continue with step 7 because the arc failure condition is not satisfied.

Step 7: It is checked whether CHCN≥30. If CHCN≥30, the failure of the arc failure is determined. Therefore, the half wave corresponding to AFCHCN = 2 is set as the start half wave of the new window. After setting AFCHCN = AFCHCN-1 and CHCN = hcn (AFCHCN), go to step 2 to continue data collection.

Therefore, the present invention as described above, by installing the intelligent main control unit 100 for diagnosing the arc failure by collecting the current signal data sampled through the circuit breaker (MCCB) installed in the distribution panel of the electric customer, overload during electricity use The main controller is to determine whether the arc failure is generated by the main controller and to quickly check the circuit breaker or the arc failure according to the result.

10: single phase power management device
20: arc failure diagnosis device
100: main controller

Claims (3)

A first step of sampling the current input from the current sensor CT at a predetermined period in the AD converter;
A second process of collecting current sampling data from the microprocessor at regular intervals through the SPI communication with the AD converter, and sequentially storing the collected current sampling data in an array and checking whether all half-waves have been obtained based on zero crossing; step;
When the half wave is completed based on the zero crossing, the microprocessor takes a discrete free transform on the half wave waveform to obtain the magnitude components A (i), i = 1, ..., n of 0 to n harmonics, and the fundamental After inputting B (i) = A (i) / A (1), i = 1, ..., 9 divided by wave size, the input pattern of the trained multilayer perceptron artificial neural network is IP = [B (0), B (2), ..., B (9)];
The microprocessor uses the theory based on the artificial neural network of the multi-layer perceptron structure to determine whether it is an arc fault current half wave or a steady current half wave, and as a result, if the normal current half wave is AFCHCN = 0, the sampling data is collected. Continue with, otherwise set CHCN = CHCN + 1, increment the current half wave by 1, and then skip to the sixth step described below;
If the current determined based on the theory is an arc fault current half wave, the microprocessor sets AFCHCN = AFCHCN + 1 and stores the value of hcn (AFCHCN) as CHCN, and then it is checked whether AFCHCN = 1 and if AFCHCN A fifth step of setting a window start half-wave when CH = 1 and setting CHCN to 1, while setting CHCN = CHCN + 1 when AFCHCN is not 1;
It is checked whether AFCHCN≥8, and if AFCHCN≥8, it is determined as an arc failure based on UL1699, outputs an arc failure state to an arc failure pin, and stops inferencing, otherwise, satisfies the arc failure condition. The sixth step, which continues since the next seventh step;
If CHCN≥30 is checked, and if CHCN≥30, the arc failure determination is failed, so the half wave corresponding to AFCHCN = 2 is set as the start half wave of a new window, and AFCHCN = AFCHCN-1, CHCN = hcn ( AFCHCN), and then, go to the second step and the seventh step of continuing data collection comprising the arc failure determination method for the current of the arc failure of the house. The method of claim 1, wherein in the second step of checking whether all half waves have been obtained based on the zero crossing, the zero crossing is performed when the sign of the current waveform is changed or when the number of samplings is equal to a number corresponding to a predetermined half wave. If the arc wave is present after the stationary wave or if the arc wave is present before the stationary wave, it means the moment when the sign of the waveform changes from the stationary wave first.If the half wave is not completed, it is until the half wave is completed. Repeating the second step of continuing to collect data, if the half wave is completed, the arc failure determination method for blocking the arc fault current of the house, characterized in that the third step. The memory of claim 1, wherein when the microprocessor sets AFCHCN = AFCHCN + 1 and stores the value of hcn (AFCHCN) as CHCN in the fourth step, hcn (AFCHCN) stores the CHCN corresponding to the AFCHCN th. Arc failure determination method for blocking the current of the arc failure of the house, characterized in that the space.

Images