WO2023103027A1 - Fault arc detection apparatus - Google Patents

Abstract

A fault arc detection apparatus. A four-dimensional space M and a fundamental subspace thereof are constructed by taking the magnetic conductivity, signal amplitude, frequency and time of a group of multiple coils as axes, M space feature vectors of a signal, a TT space distance between two feature vectors, a weighting method for pre-determining and precisely determining the TT space distance, a weighted distance of a TT space, various types of TT space distance thresholds, a signal base and an arc base are designed, and the signal base and the arc base are constructed by means of multiple types of electric appliances. The implementation of a fault arc detection apparatus based on these parameters is provided, and a random process attribute of a fault arc is analyzed. By means of the cooperation of a hardware circuit, a processor and software, a fault arc detection apparatus is designed and finished.

Description

A fault arc detection device technical field

The invention relates to the field of fault arc detection and fire protection caused by a power frequency alternating current power supply system.

Background technique

According to statistics, fires caused by arc faults account for a large proportion of all fires, so there is an urgent need for a protective device to detect arc faults and disconnect the circuit in a short period of time before the arc fault ignites combustibles, thereby extinguishing the fault arc and prevent it from causing a fire. The main problem of existing arc fault detection products is that the accuracy rate of fault arc detection is relatively low, the false alarm rate and false negative rate of detection are relatively high, and fault arc false alarms and false negatives often occur. Therefore, it is necessary to design a fault arc detection device, which can collect, process and use the characteristic information of fault arc as much as possible, detect fault arc more accurately, improve the accuracy of fault arc detection, and reduce the cost of fault arc detection. False positive and false negative rates.

Contents of the invention

A fault arc detection device is designed, which can collect, process and use as much characteristic information of fault arc as possible, detect fault arc more accurately, improve the accuracy of fault arc detection, and reduce false alarms of fault arc detection rate and omission rate. Using the hardware circuit, processor and software to cooperate together, the fault arc detection device is designed and completed.

A 4-dimensional vector space is designed, defined as M vector space, and the basic subspace SPM(m, n, p, q) of M space is defined. When the signal arrives, by calculating the probability distribution of the signal falling into the basic subspace of each M space, the eigenvector of the M space that reflects the probability distribution characteristics of the signal on the M vector space is constructed, and the M space of the signal is The eigenvectors are compared to a known pre-acquired database of signal and arc bases to detect arc faults. Specifically, a group of coils with different relative magnetic permeability u of the magnetic core is used as the detection coil of the signal, and the L line of the power frequency alternating current passes through the center of the group of coils. There are a total of Nu in this group of coils, and the number axis is constructed in units of the logarithmic value of the relative magnetic permeability u, which is called the U axis. On the U-axis, take Nu points, these points are arranged from small to large, recorded as u(m), m is an integer variable, 0<m<Nu+1, and U-axis is one of the dimensions to build a multi-dimensional space M. The signal output by each coil is amplified by the amplifier, and the operational amplifier is used to realize the signal amplification function. The number of turns of each coil, the magnetic permeability of the skeleton material, and the gain of the signal amplifier jointly determine the amplitude of the signal at the output end of the amplifier and the overall gain. . The logarithm value of the amplitude value within the amplitude value range of the output signal of the amplifier corresponding to each coil is used as the second number axis, which is called the A axis. On the A-axis, take which point, these points are arranged from small to large, recorded as a(n), n is an integer variable, 0<n<Na+1, with the A-axis as the second dimension, construct this multi-dimensional space M. The logarithm value of the frequency value within the frequency value range of the output signal of the amplifier corresponding to each coil is the third number axis, which is called the F axis. On the F-axis, take Nf points, these points are arranged from small to large, recorded as f(p), p is an integer variable, 0<p<Nf+1, with the F-axis as the third dimension, construct this multi-dimensional space M. The cycle T of the power frequency AC is the maximum value of the coordinates, and the time t within a cycle is the coordinate point on the number axis to construct the fourth number axis, which is called the T axis. On the T-axis, take Nt points, these points are arranged from small to large, recorded as t(q), q is an integer variable, 0<q<Nt+1, take the T-axis as the fourth dimension to build this multi-dimensional Space M. The points on the number axes of all four dimensions can be selected uniformly or non-uniformly, but it must be ensured that in the same device, once selected, they remain consistent and cannot be changed. The currently constructed multi-dimensional space M is a four-dimensional space, select coordinate points {u(m), a(n), f(p), t(q)} and {u(m), a(n+1), f (p+1), t(q+1)}, the three-dimensional cube with these two coordinate points as diagonal vertices is recorded as the basic subspace SPM(m, n, p, q) of M space. The signal S1 in a power frequency AC cycle T will fall into the M space many times, the total number of times the signal falls into the entire M space is NM_S1(M), and the signal S1 falls into the basic subspace SPM(m, n, p , q) is NM_S1(m, n, p, q), then, the sum of NM_S1(m, n, p, q) in 4 dimensions is equal to NM_S1(M), and the signal S1 falls into the M space The probability of the basic subspace SPM (m, n, p, q) is recorded as POSB_S1 (m, n, p, q), then, POSB_S1 (m, n, p, q) = NM_S1 (m, n, p, q )/NM_S1(M). For any signal, define the eigenvector of a signal S1 in M space, called the M space eigenvector of the signal S1, expressed as S1_V{POSB_S1(m, n, p, q), //(m(0, Nu ], n(0, Na], p(0, Nf], q(0, Nt])}, it is composed of the probability values whose total number is TT throughout the M space, and it represents the sequence {POSB_S1(1 , 1, 1, 1), POSB_S1 (1, 1, 1, 2), ..., POSB_S1 (m, n, p, q), ..., POSB_S1 (Nu, Na, Nf, Nt)}, each probability value is called the coordinate of the vector S1_V, the symbol "_V" represents the vector, the same below, the vector S1_V is the vector of the TT dimension, and the space formed by the TT dimension coordinate is called the TT space. Signal S1 The M space eigenvector of is located in the TT space. The TT value is equal to Nu multiplied by Na multiplied by Nf multiplied by Nt, and the TT value is the total number of the basic subspace SPM (m, n, p, q) of the M space. Define two The TT space distance of the M space eigenvector of the signal is D(S1_V, S1_V), which is composed of two corresponding probability values of these two vectors in each basic subspace of the same M space, that is, the coordinate values of the vectors are respectively Corresponding subtraction, and then square them separately to get the square of the difference of each vector coordinate, and then sum the squares of all the vector coordinate differences to get the sum of the squares of the difference of all coordinates, and then take the square root to get the signal The TT space distance D(S1_V, S1_V) of the M space feature vectors of S1 and S2. Define a weighted vector C_V{c(m, n, p, q), //(m(0, Nu], n(0, Na], p(0, Nf], q(0, Nt])}, it represents the sequence {c(1,1,1,1), c(1,1,1,2),... ..., c(m, n, p, q), ..., c(Nu, Na, Nf, Nt)}, this vector is also a TT-dimensional vector. Use C_V to weight S1_V to obtain a weighted vector CS1_V, C_V&S1_V=CS1_V, "&" represents the weighted multiplication of two vectors, the process is as follows, use each coordinate in the vector C_V to multiply each corresponding coordinate in S1_V to get CS1_V{c(m,n , p, q)*POSB_S1(m, n, p, q), //(m(0,Nu],n(0,Na],p(0,Nf],q(0,Nt]))} .It represents the sequence {c(1,1,1,1)*POSB_S1(1,1,1,1), c(1,1,1,2)*POSB_S1(1,1,1,2) ,...,c(m,n,p,q)*POSB_S1(m,n,p,q),...,c(Nu,Na,Nf,Nt)*POSB_S1( Nu, Na, Nf, Nt)}, "*" represents the multiplication operation of the value, and the vector is also a TT-dimensional vector. This operation is defined as the weighted multiplication of two vectors, and the obtained operation result is still a coordinate dimension of TT dimension vector.

The key parameter for judging whether a fault arc occurs is to judge the probability distribution of the M-space eigenvector S1_V to the signal base STS_V(n) and arc The TT spatial distance of the base STA_V(m). Collect the M space eigenvector S1_V of the received signal S1, and establish the database of the M space eigenvector STS_V(n) of the standard normal signal sample and the M space eigenvector STA_V(m) of the standard fault arc signal sample in advance, n and m represent The sequence numbers of different data in the M-space standard feature vector database, respectively calculate the TT space distance D(S1_V, STS_V(n)) and D(S1_V, STA_V(m) of the vector S1_V to STS_V(n) and STA_V(m) ), the criterion Dth(sts) and Dth(sta) of the TT space distance of the preset vector, compare D(S1_V, STS_V(n)) and Dth(sts) and, D(S1_V, STA_V(m)) and Dth(sta), so as to detect the arc fault signal.

In order to complete the detection of the above-mentioned arc fault, the probability distribution of the standard normal signal samples and the standard arc fault signal samples in the basic subspaces of each M space must be collected in advance, and the M space eigenvector STS_V(n) of the standard signal samples and the standard The database of M-space eigenvectors STA_V(m) of arc fault signal samples, called signal base and arc base, respectively, are located in TT space. In order to simplify the complex work, the signal base and arc base are established in two steps. The first step is to select 10 common standard electrical appliances in daily life, as follows, power resistors, LED lights, halogen lights, air compressors, switching power supplies, air conditioners, electric drills, refrigerators, TV sets, washing machines, and build a foundation Signal base and base arc base. The second step is to add different electrical appliances in addition to the standard 10 electrical appliances, and establish a signal base and an arc base.

The 10 kinds of standard electrical appliances selected by establishing the basic signal base and the basic arc base are recorded as apl(h), 0<h<11, h is an integer, the first number axis is established, which is recorded as the APL axis, and apl(h ) as 10 points on the number axis, with apl(1) as the starting point, to sample a single electrical appliance, only these 10 situations need to be sampled. For the same type of electrical appliances, there are different rated power and rated current, the logarithm value of the minimum operating current of the designed fault arc detection device is li(1), and the rated current value of I different electrical appliances is selected as the sampling value , take li(1) as the starting point of the number axis, and take li(I) as the maximum value of the coordinates on the number axis, establish a second number axis, which is recorded as the LI axis, and the point on this axis is recorded as li(i), 0<i <I+1. A plane composed of APL and LI is denoted as APL-LI. On this plane, there are 10 times I points in total. According to the position of these 10 times I points, the number of times NM (m, n, p, q) that the acquisition signal falls into the basic subspace SPM (m, n, p, q) of M space, then according to the aforementioned formula, calculate The probability of a signal falling into SPM(m,n,p,q) POSB(m,n,p,q) to obtain the M-space feature vector APL_V(h,i), which is a vector of TT dimension. The next step is to find whether there is redundancy between APL_V(h, i) for different values of h and i, and calculate the TT space distance D(APL_V( h, i), APL_V(hx, ix)), where h is not equal to hx and i is not equal to ix. Define a TT space distance Dth(apl) threshold, when D(APL_V(h, i), APL_V(hx, ix))<Dth(apl), the data is considered redundant, and only one of them can be selected The data of the consumer is used as the basic signal base and the basic arc base, and another consumer is removed from these 10 standard consumer. Select another electrical appliance to enter the 10 standard electrical appliances, and repeat the above calculation process until 10 electrical appliances are found, and the TT space distance D(APL_V(h, i) between any two electrical appliances, APL_V(hx, ix))>Dth(apl) is always established, where h is not equal to hx, and i is not equal to ix. In the case of only the signal without arc fault, repeat the above measurement many times, and average the obtained data, so that 10 M space feature vectors BSTS_V(n), n(0, 10] are obtained, which are the basic signal base .Use the selected 10 kinds of standard electrical appliances, connect to the fault arc generator, and repeat the above-mentioned measurement several times on the basis of generating fault arc and superimposing the fault arc with the standard signal, and carry out the obtained data Average, and then get 10 M space eigenvectors BSTA_V(n), n(0, 10], they are the basic arc bases. These basic signal bases and basic arc bases are a subset of the entire basic signal base and basic arc bases , call them the basic basic signal base and the basic basic arc base. The Dth(apl) value is obtained in the following way, taking D(APL_V(h, i), APL_V(hx, ix)), h is not equal to hx, and i is not equal to ix, carry out arithmetic mean for all h(0,10]i(0,I], then select a ratio of the arithmetic mean, such as 5%, as Dth(apl) initial value, then according to the operating conditions of the device, Adjust this ratio, thereby expecting to obtain optimal value.Based on Dth (apl), select an appropriate ratio RT2, multiply RT2 with Dth (apl) to obtain Dth (apl2), Dth (apl2) is a later use threshold.

The 10 selected standard electrical appliances are combined in pairs, and the same standard electrical appliance is also combined in pairs. There are 100 times 1 situations in total. The signal superimposed after each pairwise combination is regarded as a signal , detect and calculate the M-space feature vector STS_V(n) of each such combined signal. Calculate the TT space distance D(APL_V(h, i) between any two combined signals, when D(APL_V(h, i), APL_V(hx, ix))<Dth(apl2), where h is not equal to hx , i is not equal to ix, the data is considered redundant, and one of the M space feature vectors STS_V(n) is discarded until D(APL_V(h, i), APL_V(hx, ix))>Dth(apl2), Among them, h is not equal to hx, i is not equal to ix, until the constant is established, the redundant M space feature vector STS_V(n) will be added to the basic signal base. The arc signal is added to the signal of the above pairwise combination, and the M space is obtained by detection and calculation. The eigenvector STA_V(m) uses the above-mentioned method of removing redundancy, and adds the M-space eigenvector STA_V(m) after removing redundancy to the basic signal base.

Then carry out every combination of 3 types of electrical appliances, repeat the above calculation process, you can choose to carry out until 5 types of electrical appliances are combined, detect and calculate the M space feature vectors STS_V(n) and STA_V(m), and perform redundancy removal processing. Add the data obtained after removing redundancy to the basic signal base and basic arc base. At this time, it is considered that a complete basic signal base and basic arc base have been established.

The establishment of the signal base and the arc base is based on the basic signal base and the basic arc base. Various electrical appliances are found. Using the establishment process of the basic basic signal base and the basic basic arc base, the normal signal and fault arc are sampled to obtain the The M-space eigenvectors of the i points of the consumer on the LI axis, and then calculate their TT space distances to the basic basic signal base and the basic basic arc base respectively. A distance discrimination threshold is set. If it is lower than the distance discrimination threshold, it will not be used. If it is higher than the distance discrimination threshold, the obtained M-space feature vector of the consumer will be added to the signal base and the arc base. Then make two, three, four, five combinations of the electrical appliance and the electrical appliance that obtained the signal base before, and also judge whether to add the signal base and the arc base according to the TT space distance from the M space feature vector to the signal base and the arc base . With continuous accumulation, this signal base and arc base will become more and more complete. This signal base and arc base are recorded as STS_V(n) and STA_V(m), n(0, Ns], m(0, Ma] , Ns and Ma are the total number of data in the signal-based and arc-based databases, respectively.

Design the pre-judgment weighted vector CY_V and the precise judgment weighted vector CJ_V, CY_V and CJ_V are both Nu times Na times Nf times Nt dimension vectors, that is, TT dimension vectors. The purpose of setting CY_V is to reduce the amount of calculation as much as possible and reduce the computational burden of the processor, at the cost of reducing the accuracy of detection. The purpose of setting CJ_V is to improve the accuracy of detection as much as possible, at the cost of a relatively large amount of calculation of the processor. The setting of these two vectors can be highlighted and adjusted by weighting the probability POSB_S1(m,n,p,q) of different signals S1 falling into the basic subspace SPM(m,n,p,q) of M-space Different basic subspace SPM(m, n, p, q) of M space influence degree on fault arc detection. If the coordinate value of the weight vector corresponding to this subspace is 0, then the basic subspace of the M space has no influence on the detection of arc faults at all. At this time, the weight vector plays the role of filtering the basic subspace of the M space. In order to reduce the amount of computation, the coordinate values of the weighting vector CY_V take as many 0s as possible. The fault arc detection process using weighted vectors is as follows. When a signal S1 enters the device, the device detects and calculates the M-space feature vector S1_V of S1. STS_V(n) and STA_V(m), n(0, Ns], m(0, Ma], are the signal base and arc base of the device, first use CY_V to vectorize S1_V, STS_V(n) and STA_V(m) The weighted multiplication of CY_V&S1_V, CY_V&STS_V(n), CY_V&STA_V(m), and then calculate the TT space distance of the vector D(CY_V&S1_V, CY_V&STS_V(n)), D(CY_V&S1_V, CY_V&STA_V(m), the obtained values are called The CY_V weighted TT space distance between the signal S1 and the nth signal base and the mth arc base of the device. The cy weighted judgment thresholds of the signal base and the arc base are Dth(cy-sts) and Dth(cy-sta ).

Set D(CY_V&STS_V(nx), CY_V&STS_V(n)), n(0, Ns], n is not equal to nx, D(CY_V&STA_V(mx), CY_V&STA_V(m)), m(0, Ma], m is not equal to mx , carry out the arithmetic mean for all n and m, and then select a proportion of the arithmetic mean, such as 5%, as the initial values of Dth(cy-sts) and Dth(cy-sta), and then adjust according to the operation of the device This ratio is expected to obtain the optimal value.

When there is at least one nx, nx(0, Ns] such that D(CY_V&S1_V, CY_V&STS_V(nx))<Dth(cy-sts), and for all m, m(0,Ma], D(CY_V&S1_V, CY_V&STA_V( m))>Dth(cy-sta) is always established, then it is judged that S1 is a normal signal, and the device outputs a normal signal indication through the pin.

When there is at least one mx, mx(0,Ma] such that D(CY_V&S1_V,CY_V&STA_V(mx))<Dth(cy-sta), and for all n,n(0,Ns], D(CY_V&S1_V,CY_V&STS_V( n))>Dth(cy-sts) is always established, then it is judged that S1 is the signal of fault arc, and the detection results of fault arcs in adjacent multiple M spaces are judged, and the principle of minority obeying the majority is adopted. If it is detected as a fault arc If the results are in the majority, the device outputs a high-level fault arc indication signal through the pin to control the tripping mechanism to trip or control the alarm mechanism to alarm.

When there is at least one nx, nx(0, Ns], making D(CY_V&S1_V, CY_V&STS_V(nx))<Dth(cy-sts), and there is at least one mx, mx(0, Ma], making D((CY_V&S1_V , CY_V&STA_V(mx))<Dth(cy-sta), or, for all n, n(0, Ns], D(CY_V&S1_V, CY_V&STS_V(n))>Dth(cy-sts) holds constant, and at the same time for All m, m(0, Ma], D(CY_V&S1_V, CY_V&STA_V(m))>Dth(cy-sta) are always established. At this time, it is considered that the weighting accuracy of the CY_V vector is not enough, and then the CJ_V vector is used for the M of the signal S1 The spatial feature vector S1_V is weighted and calculated. In this case, it may also be that the settings of Dth(cy-sts) and Dth(cy-sta) are unreasonable and should be adjusted again.

Using the CJ_V vector, calculate the weighted distances D(CJ_V&S1_V, CJ_V&STS_V(n)) and D(CJ_V&S1_V, CJ_V&STA_V(n)), and then compare them with the judgment thresholds Dth(cj-sts) and Dth( cj-sta) for a similar comparative judgment.

When there is at least one nx, nx(0,Ns] such that D(CJ_V&S1_V, CJ_V&STS_V(nx))<Dth(cy-sts), and for all m, m(0,Ma], D(CJ_V&S1_V, CJ_V&STA_V( m))>Dth(cy-sta) is always established, then it is judged that S1 is a normal signal, and the device outputs a normal signal indication through the pin.

When there is at least one mx, mx(0,Ma] such that D(CJ_V&S1_V, CJ_V&STA_V(mx))<Dth(cj-sta), and for all n, n(0,Ns], D(CJ_V&S1_V, CJ_V&STS_V( n))>Dth(cj-sts) is always established, then it is judged that S1 is an arc fault signal, and the detection results of arc faults in adjacent multiple M spaces are judged, and the principle of minority obeying the majority is adopted. If the result is in the majority, the device outputs a high-level fault arc indication signal through the pin to control the tripping mechanism to trip or control the alarm mechanism to alarm.

When there is at least one nx, nx(0, Ns], making D(CJ_V&S1_V, CJ_V&STS_V(nx))<Dth(cj-sts), and at least one mx, mx(0, Ma], making D(CJ_V&S1_V, CJ_V &STA_V(mx))<Dth(cj-sta), or, for all n, n(0, Ns], D(CJ_V&S1_V, CJ_V&STS_V(n))>Dth(cj-sts) holds constant, and at the same time for All m, m(0, Ma], D(CJ_V&S1_V, CJ_V&STA_V(m))>Dth(cj-sta) are always established, then the composition of the signal base and arc base of the device is unreasonable, or the judgment threshold Dth(cj -sts) and Dth(cj-sta) are unreasonably designed, and the signal base and arc base of the device or the judgment threshold Dth(cj-sts) and Dth(cj-sta) are reconstructed.

D(CJ_V&STS_V(nx), CJ_V&STS_V(n)), n(0, Ns], n is not equal to nx, perform arithmetic mean for all n, D(CJ_V&STA_V(mx), CJ_V&STA_V(m)), m( 0, Ma], m is not equal to mx, carry out the arithmetic mean for all m, and then select a proportion of these two arithmetic mean values, such as 5%, as the initial Dth(cj-sts) and Dth(cj-sta) value, the ratio can be adjusted later according to the operating conditions of the device, so as to obtain the optimal value.

The detection results of fault arcs in multiple adjacent M spaces are judged, and the principle of minority obeying the majority is adopted, that is, if the number of detected fault arcs is greater than half M/2 of the total number of M spaces, it is judged to exist arc fault. For example, in the 9 adjacent M spaces, if arc faults are detected 5 times, because 5 times are greater than half of the total number of 4.5 times, it is determined that there is an arc fault.

The device is completed by cooperating with hardware circuits, powerful processors and software.

Generally speaking, the more data sampling points used on each data dimension of the system, the more the number of basic subspaces SPM(m, n, p, q) in M space, and the greater the amount of information obtained, it is easier to Improve the accuracy of the judgment of the fault arc signal. However, the more sampled data points, the more basic subspaces SPM(m, n, p, q) in M space, the more complex the system is to implement, and the higher the cost of the system. With the development of integrated circuit technology, the ability to integrate complex circuits on integrated circuit chips is getting stronger and stronger. The contradiction between complexity and accuracy can be perfectly resolved through application-specific integrated circuits. Many circuit units of this device are similar. Therefore, it is very suitable for implementation on an integrated circuit chip.

Choose materials with different magnetic permeability as the skeleton of the coil. For the skeleton with higher magnetic permeability, you can choose ferrite with different magnetic permeability as the skeleton. For the bobbin with lower magnetic permeability, you can choose the one with similar magnetic permeability Special engineering plastics are used as the skeleton. Due to the limitation of volume and other factors, generally several coils are selected. If the volume limitation is not considered, many coils can be selected. This device has low requirements on the accuracy of the magnetic permeability of the skeleton material of the coil. In different implementations, the magnetic permeability of the material can vary within a certain range, which is convenient for selecting materials to realize the device, but in the same realization Among them, the requirements for the consistency of the magnetic permeability of the material are relatively high, which is relatively easy to achieve, and the coil can be made very thin. The signal output by each coil is amplified by the amplifier, and the operational amplifier is used to realize the signal amplification function. The number of turns of each coil, the magnetic permeability of the skeleton material, and the gain of the signal amplifier jointly determine the amplitude of the signal at the output end of the amplifier and the overall gain. . When using a 10A power resistor as a standard load, adjust the output signal amplitude of the amplifier to the ideal amplitude range by adjusting the number of turns of each coil, the magnetic permeability of the skeleton material, the gain of the signal amplifier, etc., usually 3.3 V. Each amplifier is followed by Nf filters. Considering the cost and volume, it is enough to select a few Nf. If the volume and cost allow, more than one can be selected. The edge of the frequency response curve of the filter is not required to be very steep, and there is no problem with a certain frequency aliasing between the filters. The filter can use a simple first-order resistance-capacitance filter composed of a capacitor and a resistor. But, in this way, this set of filter circuits is very easy to realize and easily integrated into integrated circuits, and this set of filter circuits basically does not need to be debugged, which is very convenient for mass production. Of course, high-order L-C filters can also be used. A group of comparators with different comparison levels can be connected behind each filter. The comparison levels of the comparators are arranged in order from small to large. According to the conditions, several to a dozen comparators are connected behind each filter. If the design To implement this device with a dedicated chip, more comparators can be used, which can improve the accuracy of fault arc detection. The power frequency alternating current is converted by AC/DC to power the whole device. The power frequency alternating current produces T-dimensional signals through the T-dimensional signal generation circuit. The circuit is composed as follows. The L line is directly connected to a diode. During the positive half cycle, the diode is turned on, and during the negative half cycle, the diode is cut off. The diode is then connected to a voltage divider circuit composed of two resistors. The voltage divider ratio is about 101:1, and the total resistance value is 2-3 megohms. The power frequency square wave that is completely synchronized with the power frequency alternating current is obtained, and the frequency multiplication of the square wave is Nt times, and the T-dimensional signal is obtained. The basic subspace SPM (m, n, p, q) signal acquisition circuit of M space is composed of a high-speed counter circuit, which is controlled by a T-dimensional signal. When the level of the T-dimensional signal changes from low to high, high-speed counting is started. When the level becomes low and then high again, the count value is read by the processor, the counter is cleared at the same time, and the new high-speed counting continues. The basic subspace SPM(m,n, p, q) signal acquisition. Calculation of signal base and arc base, management of database, calculation of TT space distance of different vectors, vector weighting, identification and detection of fault arc, and other tasks are all performed by high-speed high-performance processor and software running on the processor Done together. After the device detects the arc fault, it outputs a high-level arc fault indication signal through the pin to control the tripping mechanism to trip, thereby cutting off the power supply of the electrical equipment, thereby extinguishing the arc, and achieving the purpose of fire prevention. Or the fault arc indication signal notifies the alarm device to alarm. The device has signal output of running indicator light and arc fault indicator light.

Through a large number of experiments, it is found that the arc fault signal itself is a random process. Under the same conditions, the signal has strong randomness, and the arc fault signal is close to a stationary or quasi-stationary random process. Therefore, designing arc fault detection devices from the perspective of probability and stochastic process can more accurately approach the essence of physical phenomena, so higher detection accuracy can be obtained.

M space contains many basic subspaces SPM(m,n,p,q) of M space, especially when the number of data on each dimension increases, the basic subspace SPM(m,n,p , q), the number of eigenvectors in the M space, the dimensions of the signal base and the arc base increase sharply, and the TT space distance and the weighted TT space distance between the vectors also increase sharply, so that it is convenient to calculate according to the TT space distance Accurately identify the fault arc signal, because the increase of the TT space distance makes it easier and more accurate to detect the fault arc through the TT space distance, and can obtain a higher detection accuracy.

This device makes full use of the information related to fault arc and normal signal, so it can get higher detection accuracy. Practice has proved that this device can obtain higher accuracy of fault arc identification And lower false alarm and false alarm rate of arc fault.

This set of devices encapsulates the coil and the circuit well in one device. The shell of the device is made of flame-retardant plastic shell, which is potted with electronic potting glue, which fixes the relative position of each part well. Because the potting glue The insulation resistance is very high, and it also solves the problem that the creepage distance between the components of the small-volume fault arc detection device is too small.

Implementation example 1, select 3 coils, the first coil chooses a special engineering plastic with u<10 as the skeleton, and the second coil chooses NiZn (MgZn) ferrite with 1000<u<2000 as the skeleton. The third coil chooses 4000<u<5000 MnZn ferrite as the skeleton. The electric-magnetic-electric signal amplification and frequency response curves of the coils with different magnetic permeability materials are different. Considering the product Cost and volume, it is not advisable to choose too many coils. The signal output by each coil is amplified by the amplifier circuit, and each amplifier is followed by 3 filters. The filter uses a simple first-order resistance-capacitance filter composed of a capacitor and a resistor, and each filter is connected with 4 a comparator. In this example, the maximum value Nt of the T-dimensional signal is selected as 8, because the realization of the frequency multiplication circuit of an integer multiple of 2 is relatively easy, and the basic subspace SPM (m, n, p, q) signal acquisition circuit of the M space is obtained by High-speed counter configuration. Calculation of signal base and arc base, management of database, calculation of TT space distance of different vectors, vector weighting, identification and detection of fault arc, and other tasks are all performed by high-speed high-performance processor and software running on the processor Together, the device controls the arc fault indicator light and the running indicator light through the output pin.

Implementation example 2, in many application scenarios, the product is limited by cost and volume, and it is hoped that the volume and cost should be as small as possible, so a minimum working arc fault identification device is designed. In this device, one coil is selected, which uses a special engineering plastic with u<10 as the skeleton. Because the coil is large in size and cannot be made into an integrated circuit, only one coil is selected in the minimum system. The signal output by the coil is amplified by the amplifier circuit, and two filters are connected behind the amplifier, and the filter uses a simple first-order resistance-capacitance filter. In this example, in order to further reduce the size, the asymmetrical arrangement of the comparator group after the filter is adopted. The output of the filter 1 directly enters the high-speed high-performance processor, which is completed by the software after the analog-to-digital conversion by the processor. Compare function. The output of filter 2 enters a group of comparators composed of a total of 4 comparators, and the amplitude comparison function is completed by hardware. The power frequency alternating current is converted by AC/DC to power the whole device. In this example, the maximum value Nt of the T-dimensional signal is selected as 8, because it is relatively easy to realize frequency multiplication of an integer multiple of 2, and the function of generating the T-dimensional signal is completed by software in the processor chip. Basic subspace SPM (m, n, p, q) signal acquisition of M space, calculation of signal basis and arc basis and database management, calculation of distance in TT space of different vectors, vector weighting, identification and detection of fault arc The rest of the work is completed by the high-speed high-performance processor and the software running on the processor. The device controls the fault arc indicator light and the running indicator light through the output pin.

Implementation example 3, this device contains many signal amplifiers, filters, comparators and subspace SPM (m, n, p, q) signal acquisition function devices with similar functions but a large number, and the functions of each type of these devices are similar , a large number, very suitable for the use of integrated circuits to achieve. There is no way for the coil to be integrated into the chip, and it can only be connected outside the integrated circuit. This device uses 3 coils, the first coil chooses special engineering plastics with u<10 as the skeleton, the second coil chooses NiZn (MgZn) ferrite with 1000<u<2000 as the skeleton, and the third coil Choose 4000<u<5000 MnZn ferrite as the skeleton, and the electro-magnetic-electric signal amplification and frequency response curves of coils with different skeletons are different. Considering the problem of mutual interference between analog signals and digital signals, two integrated circuit chips (DIE) for analog signal chips (DIE) and digital signal chips (DIE) were designed to realize this device, and then multi-chip packaging (MCP) technology was adopted, The analog signal chip and the digital signal chip designed in this way are packaged together to form a chip. The signal amplifier, filter, and comparator are partly designed on one chip, which is called an analog signal chip. The signal amplifier needs to be authorized to use the IP in the existing integrated circuit design library. Each amplifier is followed by 3 filters. There are two ways to implement the filter. 1. Use a simple first-order resistance-capacitance filter composed of a capacitor and a resistor, because a large resistance value is integrated in the integrated circuit Resistors and large-capacity capacitors have certain difficulties, so the resistors and capacitors are completed by discrete resistor-capacitor components external to the chip, and a first-order circuit resistance-capacitance filter circuit is used. Therefore, corresponding to this realized device, a total of 12 external filter capacitors are required. And 12 filter resistors, 2, on the integrated circuit, the first-order active filter is realized through the operational amplifier, so that the capacitance of the capacitor can be very small, which is convenient for integration on the analog signal chip, and each filter is followed by 4 for comparison There are 36 comparators in total. These comparators need to be authorized to use the IP in the existing integrated circuit design library, and these comparators are integrated on the analog signal chip. The power frequency AC generates a T-dimensional signal through the T-dimensional signal generation circuit, and the maximum value Nt of the T-dimensional signal is selected as 8, because the realization of the frequency multiplication circuit with an integer multiple of 2 is relatively easy, and it is also more convenient to implement on a digital integrated circuit. The frequency circuit is realized on the digital signal chip. The basic subspace SPM (m, n, p, q) signal acquisition circuit of M space is composed of 16-bit high-speed counter, controlled by T-dimensional signal, the basic subspace SPM (m, n, p, q) signal of M space The acquisition circuit uses a high-speed counter, which is fully integrated on the digital signal chip, selects the high-performance fast processor IP in the integrated circuit design library, integrates it into the digital signal chip, and then uses multi-chip packaging (MCP) technology , the analog signal chip and the digital signal chip are packaged together to form a chip, and enough GPIO pins are reserved. Calculation of signal base and arc base, management of database, calculation of TT space distance of different vectors, vector weighting, identification and detection of fault arc, and other tasks are all performed by high-speed high-performance processor and software running on the processor Together, the device controls the arc fault indicator light and the running indicator light through the output pin.

Claims (6)

1. An arc fault detection device is characterized in that the arc fault detection device is designed and completed by using the cooperation of hardware circuit, processor and software; a 4-dimensional vector space is designed, which is defined as M vector space, and the basic sub-spaces of M space are defined Space SPM(m, n, p, q); when the signal arrives, by calculating the probability distribution of the signal falling into the basic subspace of each M space, the M that reflects the probability distribution characteristics of the signal on the M vector space is constructed The eigenvector of the space, the eigenvector of the M space of the signal is compared with the known pre-acquired signal base and arc base database, so as to detect the fault arc; as follows, the relative permeability u of the magnetic core is used A different group of coils is used as the detection coil of the signal, and the L line of the power frequency alternating current passes through the center of this group of coils; there are a total of Nu in this group of coils, and the number axis is constructed in units of the logarithmic value of the relative magnetic permeability u, which is called the U axis ;On the U-axis, take Nu points, these points are arranged from small to large, recorded as u(m), m is an integer variable, 0<m<Nu+1, take U-axis as one of the dimensions, and construct a multi-dimensional Space M; the signal output by each coil is amplified by the amplifier, and the operational amplifier is used to realize the signal amplification function. The number of turns of each coil, the magnetic permeability of the skeleton material, and the gain of the signal amplifier jointly determine the amplitude and The overall gain; the logarithmic value of the amplitude value within the range of the output signal amplitude value of the amplifier corresponding to each coil is the second axis, which is called the A axis; on the A axis, take which point, these points are from small to Large arrangement, recorded as a(n), n is an integer variable, 0<n<Na+1, with the A axis as the second dimension, construct this multi-dimensional space M; the output signal of the amplifier corresponding to each coil The logarithmic value of the frequency value within the frequency value range is the third number axis, which is called the F axis; on the F axis, Nf points are taken, and these points are arranged from small to large, recorded as f(p), p is an integer variable, 0<p<Nf+1, take the F axis as the third dimension to construct this multi-dimensional space M; take the period T of the power frequency alternating current as the maximum value of the coordinates, and take the time t within a period as the coordinate point on the number axis to construct the first The four number axes are called T-axis; on the T-axis, take Nt points, and these points are arranged from small to large, which is recorded as t(q), q is an integer variable, 0<q<Nt+1, and the T-axis is used as For the fourth dimension, construct this multi-dimensional space M; all the points on the number axis of these four dimensions can be selected evenly or non-uniformly, but it must be guaranteed in the same set of devices, once selected, Keep consistent and cannot be changed; the currently constructed multi-dimensional space M is a four-dimensional space, select coordinate points {u(m), a(n), f(p), t(q)} and {u(m), a( n+1), f(p+1), t(q+1)}, the three-dimensional cube with these two coordinate points as diagonal vertices is recorded as the basic subspace SPM(m, n, p, q ); the signal S1 in a power frequency AC cycle T will fall into the M space for many times, the total number of signals falling into the entire M space is NM_S1(M), and the signal S1 falling into the basic subspace SPM(m, n , p, q) is NM_S1(m, n, p, q), then the sum of NM_S1(m, n, p, q) in 4 dimensions is equal to NM_S1(M), and the signal S1 falls into M The probability of the basic subspace SPM (m, n, p, q) of the space is recorded as POSB_S1 (m, n, p, q), then, POSB_S1 (m, n, p, q) = NM_S1 (m, n, p , q)/NM_S1(M); For any signal, define the feature vector of a signal S1 in the M space, called the M space feature vector of the signal S1, expressed as S1_V{POSB_S1(m, n, p, q) , //(m(0,Nu],n(0,Na],p(0,Nf],q(0,Nt])}, which is composed of the probability values of the total number TT throughout the M space, It represents the sequence {POSB_S1(1, 1, 1, 1), POSB_S1(1, 1, 1, 2),..., POSB_S1(m, n, p, q),.... .., POSB_S1(Nu, Na, Nf, Nt)}, each probability value is called the coordinate of the vector S1_V, the symbol "_V" means a vector, the same below, the vector S1_V is a vector of TT dimension, composed of this TT dimension coordinate The space of is called TT space; the M space eigenvector of signal S1 is located in TT space; TT value is equal to Nu times Na times Nf times Nt, and TT value is the basic subspace SPM(m, n, p , the total number of q); define the TT space distance of the M-space eigenvectors of two signals as D(S1_V, S1_V), which is composed of these two vectors in each of the corresponding two basic subspaces of the same M-space The probability value, that is, the coordinate values of the vectors are subtracted correspondingly, and then squared separately to obtain the square of the coordinate difference of each vector, and then sum the squares of all the coordinate differences of the vectors to obtain the sum of the squares of the differences of all coordinates , and then take the square root to get the TT space distance D(S1_V, S1_V) of the M space feature vectors of the signals S1 and S2; define a weighted vector C_V{c(m, n, p, q), //(m( 0,Nu],n(0,Na],p(0,Nf],q(0,Nt])}, which represents the sequence {c(1,1,1,1), c(1,1 , 1, 2),..., c(m, n, p, q),..., c(Nu, Na, Nf, Nt)}, this vector is also a TT-dimensional vector; Use C_V to weight S1_V to obtain the weighted vector CS1_V, C_V&S1_V=CS1_V, "&" indicates the weighted multiplication of two vectors, the process is as follows, use each coordinate in the vector C_V to be multiplied by each corresponding coordinate in S1_V Got CS1_V{c(m,n,p,q)*POSB_S1(m,n,p,q), //(m(0,Nu],n(0,Na],p(0,Nf], q(0, Nt]))}; it represents the sequence {c(1,1,1,1)*POSB_S1(1,1,1,1), c(1,1,1,2)*POSB_S1 (1,1,1,2),...,c(m,n,p,q)*POSB_S1(m,n,p,q),...,c(Nu, Na, Nf, Nt)*POSB_S1(Nu, Na, Nf, Nt)}, "*" represents the multiplication operation of the value, and the vector is also a TT-dimensional vector. This operation is defined as the weighted multiplication of two vectors, and the obtained operation is The result is still a vector whose coordinate dimension is TT;

   The key parameter for judging whether a fault arc occurs is to judge the probability distribution of the M-space eigenvector S1_V to the signal base STS_V(n) and arc The TT space distance of the base STA_V(m); collect the M space feature vector S1_V of the received signal S1, and establish the M space feature vector STS_V(n) of the standard normal signal sample and the M space feature vector STA_V of the standard fault arc signal sample in advance (m) database, n and m represent the serial numbers of different data in the M space standard feature vector database, respectively calculate the TT space distance D(S1_V, STS_V(n) of the vector S1_V to STS_V(n) and STA_V(m) ) and D(S1_V, STA_V(m)), the criterion Dth(sts) and Dth(sta) of the TT space distance of the preset vector, compare D(S1_V, STS_V(n)) and Dth(sts) and, The relationship between D(S1_V, STA_V(m)) and Dth(sta), so as to detect the fault arc signal;

   In order to complete the detection of the above-mentioned arc fault, the probability distribution of the standard normal signal samples and the standard arc fault signal samples in the basic subspaces of each M space must be collected in advance, and the M space eigenvector STS_V(n) of the standard signal samples and the standard The database of M-space eigenvectors STA_V(m) of fault arc signal samples is called signal base and arc base respectively, and they are located in TT space; in order to simplify the complex work, the signal base and arc base are established in two steps; the first step , select 10 common standard electrical appliances in daily life, as follows, power resistors, LED lamps, halogen lamps, air compressors, switching power supplies, air conditioners, electric drills, refrigerators, TV sets, washing machines, and establish a basic signal base and Basic arc base; the second step is to add different electrical loads in addition to the standard 10 types of electrical loads, and establish signal bases and arc bases;

   The 10 kinds of standard electrical appliances selected by establishing the basic signal base and the basic arc base are recorded as apl(h), 0<h<11, h is an integer, the first number axis is established, which is recorded as the APL axis, and apl(h ) as 10 points on the number axis, with apl(1) as the starting point, to sample a single electrical appliance, only these 10 situations need to be sampled; for the same type of electrical appliance, there are different rated power and rated current, The logarithmic value of the minimum operating current of the designed arc fault detection device is li(1), select the rated current value of I different electrical appliances as the sampling value, take li(1) as the starting point of the number axis, and take li(I) For the maximum value of the coordinates on the number axis, establish a second number axis, which is denoted as the LI axis, and the point on this axis is denoted as li(i), 0<i<I+1; APL and LI together form a plane, denoted as , APL-LI, on this plane, there are 10 times I points in total; according to the positions of these 10 times I points, the number of times NM that the collected signal falls into the basic subspace SPM(m, n, p, q) of M space (m, n, p, q), and then calculate the probability POSB (m, n, p, q) that the signal falls into SPM (m, n, p, q) according to the above formula, so as to obtain the M space feature vector APL_V (h, i), which is a TT-dimensional vector; the next step is to find whether there is redundancy between APL_V(h, i) for different values of h and i, and calculate the vectors APL(h, i) and APL_V(hx , ix) between the TT space distance D(APL_V(h, i), APL_V(hx, ix)), where h is not equal to hx, i is not equal to ix; define a TT space distance Dth(apl) threshold, when D (APL_V(h, i), APL_V(hx, ix))<Dth(apl), it is considered that the data is redundant, and at this time only one of the consumer data can be selected as the basic signal base and basic arc base, Remove another electrical appliance from the 10 standard electrical appliances; select another electrical appliance to enter the 10 standard electrical appliances, and repeat the above calculation process until you find 10 electrical appliances, and any 2 electrical appliances The TT space distance between D(APL_V(h, i), APL_V(hx, ix))>Dth(apl) is constant until it is established, where h is not equal to hx, i is not equal to ix; in the case of only a signal without a fault arc Next, repeat the above measurement several times, and average the obtained data, so as to obtain 10 M space feature vectors BSTS_V(n), n(0,10], which are the basic signal bases; use the selected 10 kinds The standard electrical appliance is connected to the fault arc generator. On the basis of generating the fault arc and superimposing the fault arc and the standard signal, repeat the above measurement several times, average the obtained data, and then get 10 M The spatial eigenvectors BSTA_V(n), n(0, 10], they are the basic arc bases; these basic signal bases and basic arc bases are a subset of the entire basic signal base and basic arc bases, and they are called basic basic signal bases and the basic basic arc basis; the Dth(apl) value is obtained as follows, taking D(APL_V(h, i), APL_V(hx, ix)), h is not equal to hx, i is not equal to ix, for all h(0 , 10]i(0, I] carry out the arithmetic mean, then select a ratio of the arithmetic mean, such as 5%, as the initial value of Dth(apl), and then adjust the ratio according to the operation of the device, so as to expect to obtain the most Excellent value; based on Dth (apl), select an appropriate ratio RT2, multiply Dth (apl) by RT2 to obtain Dth (apl2), and Dth (apl2) is a threshold value used later;

   The 10 selected standard electrical appliances are combined in pairs, and the same standard electrical appliance is also combined in pairs. There are 100 times 1 situations in total. The signal superimposed after each pairwise combination is regarded as a signal , detect and calculate the M space eigenvector STS_V(n) of each such combined signal; calculate the TT space distance D(APL_V(h, i) between any two combined signals, when D(APL_V(h, i), When APL_V(hx,ix))<Dth(apl2), where h is not equal to hx and i is not equal to ix, the data is considered redundant, and one of the M space feature vectors STS_V(n) is discarded until D(APL_V (h, i), APL_V(hx, ix))>Dth(apl2), where h is not equal to hx, i is not equal to ix, until it is established, the redundant M space feature vector STS_V(n) will be added to the basic signal base; add the arc signal to the above pairwise combined signal, detect and calculate the M-space feature vector STA_V(m), use the above-mentioned method of removing redundancy, add the M-space feature vector STA_V(m) after removing redundancy base signal base;

   Then carry out every combination of 3 types of electrical appliances, repeat the above calculation process, you can choose to carry out until 5 types of electrical appliances are combined, detect and calculate the M space feature vectors STS_V(n) and STA_V(m), and perform redundancy removal processing. Add the data obtained after removing redundancy to the basic signal base and basic arc base. At this time, it is considered that a complete basic signal base and basic arc base have been established;

   The establishment of the signal base and the arc base is based on the basic signal base and the basic arc base. Various electrical appliances are found. Using the establishment process of the basic basic signal base and the basic basic arc base, the normal signal and fault arc are sampled to obtain the The M-space eigenvectors of the i points of the electrical appliances on the LI axis, and then calculate their TT space distances to the basic basic signal base and the basic basic arc base respectively; set a discrimination threshold of a distance, which is lower than the discrimination threshold of this distance If it is higher than the distance discrimination threshold, the obtained M-space eigenvector of the electrical appliance is added to the signal base and the arc base; Four or five combinations, also judge whether to add the signal base and the arc base according to the TT space distance from the M space eigenvector to the signal base and the arc base; with continuous accumulation, the signal base and the arc base will become more and more complete , this signal base and arc base are denoted as STS_V(n) and STA_V(m), n(0, Ns], m(0, Ma], Ns and Ma are the total number of data in the signal base and arc base database respectively;

   Design the pre-judgment weighted vector CY_V and the precise judgment weighted vector CJ_V, both CY_V and CJ_V are vectors of Nu times Na times Nf times Nt dimensions, that is, TT-dimensional vectors; the purpose of setting CY_V is to minimize the amount of calculation and ease the processing The cost is to reduce the detection accuracy; the purpose of setting CJ_V is to improve the detection accuracy as much as possible, but the cost is that the processor has a relatively large amount of calculation; the settings of these two vectors can be adjusted by different signal S1 The probability POSB_S1(m, n, p, q) of the basic subspace SPM(m, n, p, q) of the M space is weighted to highlight and adjust the basic subspace SPM(m, n, The degree of influence of p, q) on fault arc detection; if the coordinate value of the weight vector corresponding to this subspace is 0, then the basic subspace of this M space has no influence on the detection of fault arc at all. At this time, the weight vector starts from The basic subspace filtering function of M space; in order to reduce the amount of calculation, the coordinate value of the weighted vector CY_V should be as many as 0; the detection process of the fault arc using the weighted vector is as follows, when a signal S1 enters the device, the device detects Calculate and generate the M space feature vector S1_V of S1; STS_V(n) and STA_V(m), n(0, Ns], m(0, Ma], are the signal base and arc base of the device, first use CY_V to S1_V, STS_V (n) and STA_V(m) perform weighted multiplication of vectors, that is, CY_V&S1_V, CY_V&STS_V(n), CY_V&STA_V(m), and then calculate the TT space distance of the vector D(CY_V&S1_V, CY_V&STS_V(n)), D(CY_V&S1_V, CY_V&STA_V (m), the obtained numerical value is respectively referred to as signal S1 and the nth signal base of the device and the mth electric arc base using CY_V weighted TT space distance; the cy weighted judgment threshold of the signal base and the electric arc base is respectively Dth( cy-sts) and Dth (cy-sta);

   Set D(CY_V&STS_V(nx), CY_V&STS_V(n)), n(0, Ns], n is not equal to nx, D(CY_V&STA_V(mx), CY_V&STA_V(m)), m(0, Ma], m is not equal to mx , carry out the arithmetic mean for all n and m, and then select a proportion of the arithmetic mean, such as 5%, as the initial values of Dth(cy-sts) and Dth(cy-sta), and then adjust according to the operation of the device The ratio, thus expecting to obtain the optimal value;

   When there is at least one nx, nx(0, Ns] such that D(CY_V&S1_V, CY_V&STS_V(nx))<Dth(cy-sts), and for all m, m(0,Ma], D(CY_V&S1_V, CY_V&STA_V( m))>Dth(cy-sta) is always established, then it is judged that S1 is a normal signal, and the device outputs a normal signal indication through the pin;

   When there is at least one mx, mx(0,Ma] such that D(CY_V&S1_V,CY_V&STA_V(mx))<Dth(cy-sta), and for all n,n(0,Ns], D(CY_V&S1_V,CY_V&STS_V( n))>Dth(cy-sts) is always established, then it is judged that S1 is the signal of fault arc, and the detection results of fault arcs in adjacent multiple M spaces are judged, and the principle of minority obeying the majority is adopted. If it is detected as a fault arc If the results are in the majority, the device outputs a high-level fault arc indication signal through the pin to control the tripping mechanism to trip or control the alarm mechanism to alarm;

   When there is at least one nx, nx(0, Ns], making D(CY_V&S1_V, CY_V&STS_V(nx))<Dth(cy-sts), and there is at least one mx, mx(0, Ma], making D((CY_V&S1_V , CY_V&STA_V(mx))<Dth(cy-sta), or, for all n, n(0, Ns], D(CY_V&S1_V, CY_V&STS_V(n))>Dth(cy-sts) holds constant, and at the same time for All m, m(0, Ma], D(CY_V&S1_V, CY_V&STA_V(m))>Dth(cy-sta) are always established. At this time, it is considered that the weighting accuracy of the CY_V vector is not enough, and then the CJ_V vector is used for the M of the signal S1 The spatial feature vector S1_V is weighted and calculated. In this case, it may also be that the settings of Dth(cy-sts) and Dth(cy-sta) are unreasonable, so re-adjust;

   Using the CJ_V vector, calculate the weighted distances D(CJ_V&S1_V, CJ_V&STS_V(n)) and D(CJ_V&S1_V, CJ_V&STA_V(n)), and then compare them with the judgment thresholds Dth(cj-sts) and Dth( cj-sta) to perform a similar comparative judgment;

   When there is at least one nx, nx(0,Ns] such that D(CJ_V&S1_V, CJ_V&STS_V(nx))<Dth(cy-sts), and for all m, m(0,Ma], D(CJ_V&S1_V, CJ_V&STA_V( m))>Dth(cy-sta) is always established, then it is judged that S1 is a normal signal, and the device outputs a normal signal indication through the pin;

   When there is at least one mx, mx(0,Ma] such that D(CJ_V&S1_V, CJ_V&STA_V(mx))<Dth(cj-sta), and for all n, n(0,Ns], D(CJ_V&S1_V, CJ_V&STS_V( n))>Dth(cj-sts) is always established, then it is judged that S1 is an arc fault signal, and the detection results of arc faults in adjacent multiple M spaces are judged, and the principle of minority obeying the majority is adopted. If the result is in the majority, the device outputs a high-level fault arc indication signal through the pin to control the tripping mechanism to trip or control the alarm mechanism to alarm;

   When there is at least one nx, nx(0, Ns], making D(CJ_V&S1_V, CJ_V&STS_V(nx))<Dth(cj-sts), and at least one mx, mx(0, Ma], making D(CJ_V&S1_V, CJ_V&STA_V(mx))<Dth(cj-sta), or, for all n, n(0, Ns], D(CJ_V&S1_V, CJ_V&STS_V(n))>Dth(cj-sts) holds constant, and at the same time for all m, m(0, Ma], D(CJ_V&S1_V, CJ_V&STA_V(m))>Dth(cj-sta) is always established, then the composition of the signal base and arc base of the device is unreasonable, or the judgment threshold Dth(cj- sts) and Dth(cj-sta) are unreasonably designed, reconstruct the signal base and arc base of the device or the judgment threshold Dth(cj-sts) and Dth(cj-sta);

   D(CJ_V&STS_V(nx), CJ_V&STS_V(n)), n(0, Ns], n is not equal to nx, perform arithmetic mean for all n, D(CJ_V&STA_V(mx), CJ_V&STA_V(m)), m( 0, Ma], m is not equal to mx, carry out the arithmetic mean for all m, and then select a proportion of these two arithmetic mean values, such as 5%, as the initial Dth(cj-sts) and Dth(cj-sta) Value, the ratio can be adjusted according to the operation of the device in the future, so as to expect to obtain the optimal value;

   The detection results of fault arcs in multiple adjacent M spaces are judged, and the principle of minority obeying the majority is adopted, that is, if the number of detected fault arcs is greater than half M/2 of the total number of M spaces, it is judged to exist Arc fault; for example, if arc faults are detected 5 times in the 9 adjacent M spaces, because 5 times are greater than half of the total number of 4.5 times, it is determined that there is an arc fault;

   The device is completed by cooperating with hardware circuits, powerful processors and software.
2. A fault arc detection device according to claim 1, characterized in that materials with different magnetic permeability are selected as the skeleton of the coil, and for a skeleton with higher magnetic permeability, ferrite with different magnetic permeability can be selected as the coil skeleton. Skeleton, for the skeleton with lower magnetic permeability, you can choose special engineering plastics with close magnetic permeability as the skeleton; because of the limitation of volume and other factors, generally choose a few coils, if you do not consider the volume limit, you can choose many coils ; This device has lower requirements on the accuracy of the magnetic permeability of the skeleton material of the coil. In different implementations, the magnetic permeability of the material can vary within a certain range, which is convenient for selecting materials to realize the device, but in the same In the implementation, the requirements for the consistency of the magnetic permeability of the material are relatively high, which is relatively easy to achieve, and the coil can be made very thin; the signal output by each coil is amplified by the amplifier, and the operational amplifier is used to realize the signal amplification function. The number of turns of each coil, the magnetic permeability of the skeleton material, and the gain of the signal amplifier jointly determine the amplitude of the signal at the output end of the amplifier and the overall gain; when using a 10A power resistor as a standard load, by adjusting the number of turns of each coil , the magnetic permeability of the skeleton material, the gain of the signal amplifier, etc., adjust the amplitude of the output signal of the amplifier to an ideal range, usually 3.3V; each amplifier is followed by Nf filters, considering the cost and volume, Nf can be selected as a few, if the size and cost allow, you can choose multiple; the edge of the frequency response curve of the filter is not required to be very steep, and there is no problem with a certain frequency aliasing between the filters. The filter can use a simple first-order resistance-capacitance filter composed of a capacitor and a resistor. In this way, this filter circuit is very easy to implement and easily integrated into an integrated circuit, and this filter circuit basically does not need to be debugged. It is very convenient for mass production. Of course, high-order L-C filters can also be used; each filter can be followed by a group of comparators with different comparison levels. The comparison levels of the comparators are arranged in order from small to large. According to conditions, each Several to more than a dozen comparators are connected behind each filter. If a dedicated chip is designed to realize this device, more comparators can be used, which can improve the accuracy of fault arc detection; power frequency alternating current is converted by AC/DC , to supply power to the whole device; the power frequency alternating current produces T-dimensional signals through the T-dimensional signal generating circuit, the circuit is composed as follows, the L line is directly connected to a diode, during the positive half cycle, the diode is turned on, and during the negative half cycle, the diode is cut off; The diode is then connected to a voltage divider circuit composed of two resistors. The voltage divider ratio is about 101:1, and the total resistance value is 2-3 megohms. The power frequency square wave that is completely synchronized with the power frequency alternating current is obtained, and the square wave is multiplied by Nt times to obtain a T-dimensional signal; the basic subspace SPM (m, n, p, q) signal acquisition of M space The circuit is composed of a high-speed counter circuit, which is controlled by a T-dimensional signal. When the level of the T-dimensional signal changes from low to high, high-speed counting starts. When the level becomes low and then high again, the count is read by the processor. At the same time, the counter is cleared, and the new high-speed counting is continued, and the basic subspace SPM (m, n, p, q) signal acquisition of the M space can also be realized by software; the calculation of the signal base and the arc base and the management of the database , Calculation of distance in TT space of different vectors, weighting of vectors, identification and detection of arc faults, and other work are all completed by high-speed high-performance processors and software running on the processors; after the device detects arc faults, through The pin outputs a high-level fault arc indication signal to control the tripping mechanism to trip, thereby cutting off the power supply of the electrical equipment, thereby extinguishing the arc and achieving the purpose of fire prevention; or the fault arc indication signal notifies the alarm equipment to alarm; the device has Signal output of running indicator light and arc fault indicator light.
3. A fault arc detection device according to claims 1 and 2, characterized in that 3 coils are selected, the first coil selects a special engineering plastic with u<10 as the skeleton, and the second coil selects 1000<u< 2000 NiZn (MgZn) ferrite is used as the skeleton; the third coil chooses 4000<u<5000 MnZn ferrite as the skeleton, and materials with different magnetic permeability are used as the skeleton coil to amplify the electric-magnetic-electric signal The functions and frequency response curves are different. Considering the cost and volume of the product, it is not advisable to choose too many coils; the signal output by each coil is amplified by the amplifier circuit, and each amplifier is followed by 3 filters, which are easy to use A first-order resistance-capacitance filter composed of a capacitor and a resistor, each filter is followed by 4 comparators; in this example, the maximum value Nt of the T-dimensional signal is selected as 8, because the multiple of integer multiples of 2 The realization of the frequency circuit is relatively easy. The basic subspace SPM (m, n, p, q) signal acquisition circuit of the M space is composed of a high-speed counter; The rest of the calculation, vector weighting, identification and detection of fault arc are all completed by the high-speed high-performance processor and the software running on the processor. The device controls the fault arc indicator light and running indicator light through the output pin.
4. An arc fault detection device according to claims 1 and 2, characterized in that, in many application scenarios, the product is limited by cost and volume, and it is desired that the volume be as small as possible and the cost as low as possible, so the design A working minimum arc fault identification device; in this device, select 1 coil, which uses special engineering plastics with u<10 as the skeleton, because the coil volume is large, and there is no way to make an integrated circuit, so in the minimum In the system, only one coil is selected; the signal output by the coil is amplified by the amplifier circuit, and two filters are connected behind the amplifier, and the filter uses a simple first-order resistance-capacitance filter; in this example, in order to further reduce the volume , the asymmetric arrangement of the comparator group after the filter is adopted, the output of filter 1 directly enters the high-speed high-performance processor, and after the analog-to-digital conversion by the processor, the comparison function is completed by software; the output of filter 2 enters a The comparator group consists of a total of 4 comparators, and the amplitude comparison function is completed by hardware; the power frequency alternating current is converted through AC/DC to supply power to the entire device; in this example, the maximum value Nt of the T-dimensional signal is selected as 8, Because the realization of frequency multiplication of integer multiples of 2 is relatively easy, the function of generating the T-dimensional signal is completed by software in the processor chip; the basic subspace SPM (m, n, p, q) signal acquisition and signal base Calculation of arc base and database management, calculation of TT spatial distance of different vectors, vector weighting, identification and detection of arc faults and other work are all completed by high-speed high-performance processor and software running on the processor , the device controls the arc fault indicator light and the running indicator light through the output pin.
5. According to claim 1, a kind of arc fault detection device described in 2 is characterized in that, contains many signal amplifiers, filters, comparators and subspace SPM (m, n, p, q) Devices with signal acquisition functions. Each of these devices has similar functions and a large number, and is very suitable for realization by integrated circuits; the coil cannot be integrated into the chip and can only be connected outside the integrated circuit; the device used in this device 3 coils, the first coil chooses special engineering plastics with u<10 as the skeleton, the second coil chooses NiZn (MgZn) ferrite with 1000<u<2000 as the skeleton, and the third coil chooses 4000<u <5000 MnZn ferrite is used as the skeleton, and the electro-magnetic-electric signal amplification and frequency response curves of coils with different skeletons are different; considering the mutual interference between analog signals and digital signals, an analog signal chip is designed (DIE) and digital signal chip (DIE) two integrated circuit chips (DIE) to realize this device, and then use multi-chip packaging (MCP) technology to package the analog signal chip and digital signal chip designed in this way together to form a Chip; the signal amplifier, filter, and comparator are partly designed on one chip, called an analog signal chip, and the signal amplifier needs to be authorized to use the IP in the existing integrated circuit design library; each amplifier is followed by 3 filters, There are two ways to realize the filter. 1. Use a simple first-order resistance-capacitance filter composed of a capacitor and a resistor, because it is difficult to integrate a large-value resistor and a large-capacity capacitor in an integrated circuit. Therefore, the resistance and capacitance are completed by the discrete resistance-capacitance components connected to the chip, and a first-order circuit resistance-capacitance filter circuit is used. Therefore, corresponding to this realization device, a total of 12 filter capacitors and 12 filter resistors need to be connected externally, 2. In the integrated circuit Above, the first-order active filter is realized through an operational amplifier, so that the capacitance of the capacitor can be very small, which is convenient for integration on an analog signal chip. Each filter is followed by 4 comparators, and there are 36 comparators in total. These comparators It is necessary to authorize the use of IP in the existing integrated circuit design library, and integrate these comparators on the analog signal chip; the power frequency alternating current generates a T-dimensional signal through the T-dimensional signal generation circuit, and the maximum value Nt of the T-dimensional signal is selected as 8, Because the realization of the frequency multiplication circuit of an integer multiple of 2 is relatively easy, it is also more convenient to implement on a digital integrated circuit, and the frequency multiplication circuit is realized on a digital signal chip; the basic subspace SPM (m, n, p, q) of M space The signal acquisition circuit is composed of a 16-bit high-speed counter, controlled by T-dimensional signals. The acquisition circuit of the basic subspace SPM (m, n, p, q) signal of M space uses a high-speed counter, which is fully integrated on the digital signal chip. Select the IP of the high-performance and fast processor in the integrated circuit design library, design and integrate it into the digital signal chip, and then use the multi-chip packaging (MCP) technology to package the analog signal chip and the digital signal chip together to form a chip. Sufficient GPIO pins are reserved; the calculation of signal base and arc base, database management, calculation of TT space distance of different vectors, vector weighting, identification and detection of arc faults, etc. are all performed by high-speed high-performance processors It is completed together with the software running on the processor, and the device controls the fault arc indicator light and the running indicator light through the output pin.
6. An arc fault detection device according to claims 1 and 2, characterized in that the device encapsulates the coil and the circuit well in one device, the outer shell of the device is made of flame-retardant plastic shell, and the electronic pot is used to Sealant potting can well fix the relative position of each part, because the insulation resistance of the potting compound is very high, it also solves the problem that the creepage distance between the components of the small-volume fault arc detection device is too small.

Images