CN116125260A - Breaker electromechanical fault edge diagnosis method based on multi-element data fusion - Google Patents

Abstract

The invention discloses a breaker electromechanical fault edge diagnosis method based on multi-element data fusion, which comprises the following steps: collecting a switching-on/off coil current signal and a displacement signal of a circuit breaker to be tested; respectively extracting time domain features of the opening and closing coil current signals and the displacement signals, and respectively constructing a feature matrix and a feature map of the opening and closing coil current signals and the displacement signals based on the time domain features; and diagnosing whether the circuit breaker has electromechanical faults and the severity of the faults according to the characteristic matrix and the characteristic map.

Description

Breaker electromechanical fault edge diagnosis method based on multi-element data fusion

Technical Field

The disclosure belongs to the field of power equipment state diagnosis, and particularly relates to a breaker electromechanical fault edge diagnosis method based on multivariate data fusion.

Background

The circuit breaker is used as the switching equipment which is most widely applied in the power grid, and has the characteristics of strong arc extinguishing capability, high electrical service life, safe use, simple maintenance and the like. During the use process of the circuit breaker, the circuit breaker must be regularly subjected to power failure maintenance, but accidents caused by the circuit breaker are more than 60% of the total quantity of the circuit breaker in times or power failure duration, so that the service life loss of equipment and the power grid power failure loss are increased.

Typical electromechanical faults of circuit breakers include voltage variations in the storage capacitor, poor circuit contact, iron core air gap variations, iron core jamming, spring fatigue, etc., and primary monitoring quantities include main circuit current, displacement, vibration, and on (off) gate coil current. The conventional diagnosis method is to analyze the time domain and frequency domain characteristic quantity of the signals and evaluate the operation condition of the equipment by a threshold diagnosis mode. However, with the development of online monitoring, the above diagnostic methods have the following problems: firstly, accurate just (minute) point information is important to breaker fault diagnosis, but in actual working conditions, the breaker is in high-voltage operation, the just (minute) point always takes the passing time of the current of the last phase of the main loop, errors caused by pre-breakdown current exist, and therefore various characteristic quantities of a displacement curve are difficult to accurately obtain, and the curve utilization rate is not high; secondly, the vibration signal is easy to be interfered by electromagnetic interference, the diagnosis result is limited by the installation position of the sensor and the signal transmission path, and the universality of the diagnosis model is not high; in addition, experiments show that the closing (opening) gate current cannot accurately distinguish two similar faults of voltage change of the energy storage capacitor and poor loop contact. In summary, the above diagnosis method has the problems of difficult feature extraction, poor universality and accuracy in discrimination, low failure recognition rate and the like.

Meanwhile, with the vigorous development of the electric power internet of things, edge side devices (such as sensors, gateways and sink control terminals) are gradually intelligent, and have a certain calculation force. The monitoring network is not all dependent on the upper computer for fault diagnosis, and a part of diagnosis work can be put down to the edge side, so that the central computing load and the channel transmission pressure are reduced. However, the traditional breaker fault diagnosis model is mostly based on complex algorithms such as deep learning, neural network and the like, and multi-source and multi-parameter data are needed as fault diagnosis basis, which is difficult to realize in an environment with relatively scarce edge computing power. From the practical engineering point of view, the fault diagnosis should follow the law of the Olympic razor, namely, the complexity of the model is reduced as much as possible while the fault diagnosis rate is ensured, and the running state of the equipment is simply and effectively given.

Disclosure of Invention

Aiming at the defects in the prior art, the purpose of the present disclosure is to provide a breaker electromechanical fault edge diagnosis method based on multi-element data fusion, which constructs a corresponding identifiable space by constructing a high-coupling and strong-orthogonality ternary array matrix and expressing the characteristic track of the matrix through a ternary map.

In order to achieve the above object, the present disclosure provides the following technical solutions:

s100: collecting a switching-on/off coil current signal and a displacement signal of a circuit breaker to be tested;

s200: respectively extracting time domain features of the opening and closing coil current signals and the displacement signals, and respectively constructing a feature matrix and a feature map of the opening and closing coil current signals and the displacement signals based on the time domain features;

s300: and diagnosing whether the circuit breaker has electromechanical faults and the severity of the faults according to the characteristic matrix and the characteristic map.

Preferably, step S200 includes the steps of:

s201: acquiring time domain characteristic parameters of the opening and closing coil current signals and the displacement signals, and constructing an original sample matrix D based on the time domain characteristic parameters;

s202: carrying out standardization processing on the original sample matrix D to obtain a standardization matrix Z;

s203: calculating a covariance matrix R of the standardized matrix Z;

s204: calculating a eigenvalue lambda based on the covariance matrix R;

s205: screening the characteristic value lambda;

s206: constructing a ternary column matrix X based on the screened eigenvalue lambda;

s207: and constructing a characteristic map based on the ternary column matrix X.

Preferably, in step S300, the step of diagnosing whether the circuit breaker has an electromechanical fault according to the feature matrix and the feature map includes the following steps:

s301: converting the ternary column feature matrix to obtain a two-dimensional matrix;

s302: selecting boundary data points in the two-dimensional matrix, and intercepting data by taking the boundary data points as centers;

s303: carrying out weighted linear regression on the intercepted data to construct n regression curves;

s304: calculating the central value of each regression curve in the n regression curves;

s305: connecting the central values of each regression curve to obtain an optimal regression curve, wherein the optimal regression curve is a fault boundary equation of the circuit breaker to be tested;

s306: and judging whether the breaker fails or not according to whether any point on the characteristic map falls into the failure boundary.

Preferably, in step S300, diagnosing the fault severity of the electromechanical fault of the circuit breaker includes the steps of:

s3001: initializing K clustering centers based on a ternary column matrix;

s3002: calculating the distance from each point in the sample to the clustering center and dividing the distance to the clustering center with the smallest distance;

s3003: re-calculating the coordinates of the clustering centers, comparing whether the original k clustering centers change or not, if so, repeating the step S3002, otherwise, outputting the coordinates of the clustering centers;

s3004: and calculating Euclidean distances between the normal working condition and the data clustering centers with different fault levels to quantitatively represent the severity of the fault.

The present disclosure also provides a breaker electromechanical fault edge diagnosis device based on multi-metadata fusion, comprising:

the acquisition module is used for acquiring a switching-on/off coil current signal and a displacement signal of the circuit breaker to be tested;

the extraction module is used for respectively extracting time domain characteristics of the opening and closing coil current signal and the displacement signal;

the construction module is used for respectively constructing a characteristic matrix and a characteristic map of the opening and closing coil current signal and the displacement signal based on the time domain characteristics of the opening and closing coil current signal and the displacement signal;

and the diagnosis module is used for diagnosing whether the circuit breaker has electromechanical faults and the severity of the faults according to the characteristic matrix and the characteristic map.

Compared with the prior art, the beneficial effects that this disclosure brought are: and constructing a high-coupling and strong-orthogonality ternary column matrix by PCA, and constructing a corresponding identifiable space by expressing the characteristic track of the matrix by the ternary map. And determining a fault boundary equation by LWR, realizing fault type identification, and determining the severity of the fault by k-means clustering and Euclidean distance, thereby realizing accurate diagnosis of the electromechanical fault of the circuit breaker. Compared with the traditional diagnosis method, the on-site identification accuracy exceeds 97%, the on-site identification method does not need on-off point information, improves the coupling degree and the utilization rate of on-off gate current and displacement characteristics, and is more beneficial to on-line monitoring of equipment. Meanwhile, the method has the advantages of fewer signals, clear algorithm logic and less calculation force, is suitable for equipment edge diagnosis and environments with scarce calculation force (such as intelligent sensors, gateways and convergence control terminals), lowers an upper diagnosis algorithm to the edge side of a bottom layer, reduces central calculation load and channel transmission pressure, and is beneficial to pushing the construction of the electric power Internet of things.

Drawings

FIG. 1 is a flow chart of a method for breaker electromechanical fault edge diagnosis based on multi-metadata fusion provided in one embodiment of the present disclosure;

fig. 2 (a) and 2 (b) are schematic time domain characteristics (taking a closing process as an example);

fig. 3 (a) to 3 (f) are schematic diagrams of ternary matrix distribution trends under different working conditions;

FIG. 4 is a schematic diagram of a fault boundary equation;

FIG. 5 is a schematic diagram of determining whether an electromechanical failure of a circuit breaker occurs based on whether any point on the signature falls within the failure boundary equation;

FIG. 6 is a schematic diagram of a process for recalculating cluster center coordinates;

fig. 7 is a cluster center distribution for different fault severity.

Detailed Description

Specific embodiments of the present disclosure will be described in detail below with reference to fig. 1 to 7. While specific embodiments of the disclosure are shown in the drawings, it should be understood that the disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

It should be noted that certain terms are used throughout the description and claims to refer to particular components. Those of skill in the art will understand that a person may refer to the same component by different names. The specification and claims do not identify differences in terms of components, but rather differences in terms of the functionality of the components. As used throughout the specification and claims, the terms "include" and "comprise" are used in an open-ended fashion, and thus should be interpreted to mean "include, but not limited to. The description hereinafter sets forth the preferred embodiments for carrying out the present disclosure, but is not intended to limit the scope of the disclosure in general, as the description proceeds. The scope of the present disclosure is defined by the appended claims.

For the purposes of promoting an understanding of the embodiments of the disclosure, reference will now be made to the embodiments illustrated in the drawings and specific examples, without the intention of being limiting the embodiments of the disclosure.

In one embodiment, as shown in fig. 1, the present disclosure provides a method for diagnosing an edge of an electromechanical fault of a circuit breaker based on multi-data fusion, including the steps of:

s100: collecting a switching-on/off coil current signal and a displacement signal of a circuit breaker to be tested;

s200: respectively extracting time domain features of the opening and closing coil current signals and the displacement signals, and respectively constructing a feature matrix and a feature map of the opening and closing coil current signals and the displacement signals based on the time domain features;

s300: and diagnosing whether the circuit breaker has electromechanical faults and the severity of the faults according to the characteristic matrix and the characteristic map.

Compared with the traditional diagnosis method, the method does not need the information of the just-closed (opening) point, improves the coupling degree and the utilization rate of opening and closing current and displacement characteristics, and is more beneficial to the on-line monitoring of equipment. Meanwhile, the method has the advantages of fewer signals, clear algorithm logic and less calculation force, is suitable for equipment edge diagnosis and environments with scarce calculation force (such as intelligent sensors, gateways and convergence control terminals), lowers an upper diagnosis algorithm to the edge side of a bottom layer, reduces central calculation load and channel transmission pressure, and is beneficial to pushing the construction of the electric power Internet of things.

In another embodiment, step S200 includes the steps of:

s201: respectively acquiring time domain characteristic parameters of the opening and closing coil current signal and the displacement signal, and constructing a characteristic matrix D based on all the time domain characteristic parameters;

in this step, the time domain characteristic parameter of the current signal of the opening/closing coil is shown in fig. 2 (a), and specifically includes a first peak current I ₁ First valley current I ₂ Second peak current I ₃ Duration T of iron core touch ₀ Length of time of movement of iron coreT ₁ Time length T of contact opening and closing ₂ Duration T of current cut-off ₃ And total time length T of opening and closing current curve ₄ The method comprises the steps of carrying out a first treatment on the surface of the The time domain characteristic parameters of the displacement signal are shown in fig. 2 (b), and specifically comprise a movable contact initial potential U ₀ Maximum closing potential (minimum opening potential) U ₁ Final potential U of moving contact ₂ Time t when moving contact starts to move ₀ Time t of the highest switching-on potential (time t of the lowest switching-off potential) ₁ And the total switching-on and switching-off stroke d and the change rate v of the curve in the switching-on and switching-off process.

The feature matrix D constructed from all the time domain feature parameters described above is expressed as d= [ I ] ₁ ,I ₂ ,I ₃ ,T ₀ ,T ₁ ,T ₂ ,T ₃ ,T ₄ ,U ₀ ,U ₁ ,U ₂ ,t ₀ ,t ₁ ,d,v]。

S202: carrying out standardization processing on the feature matrix D to obtain a standardization matrix Z;

in this step, the feature matrix D is normalized by:

and is also provided with

Wherein N is the number of time domain characteristic parameters, i and j are the ith row and the jth column of the matrix D, the value range of i is 1 to N, the value range of j is 1 to 15 (the matrix D contains 15 time domain characteristics),

is the sample variance.

S203: calculating a covariance matrix R of the standardized matrix Z;

in this step, the covariance matrix R is calculated by:

wherein Z is ^T Is the transposed matrix of matrix Z, N is the number of time domain characteristic parameters.

S204: calculating a eigenvalue lambda based on the covariance matrix R;

in this step, the eigenvalue λ is calculated by:

|λE-R|＝0

wherein E is an identity matrix.

S205: screening the characteristic value lambda;

in this step, first, the cumulative contribution Con of the eigenvalue λ is calculated by the following equation:

/>

where the numerator part represents the lambda summation from the first lambda to the mth lambda and the denominator part represents the lambda summation from the first lambda to the 15 th lambda, m is required to be determined by a contribution of more than 90%, e.g. the first 3 lambda summations are more than 90% compared to the denominator, i.e. m is 3.

Then, selecting a characteristic value with the contribution rate larger than 90% as an element for constructing a ternary column matrix X subsequently, namely as follows:

s206: constructing a ternary column matrix X based on the screened eigenvalue lambda;

s207: and constructing a characteristic map based on the ternary column matrix X.

In the step, before constructing the characteristic map, the ternary column matrix X is normalized to X ₁ ，X ₂ ，X ₃ Drawing a corresponding ternary map by taking coordinate axes, and enabling the characteristic track of the ternary array matrix X to be represented by the ternary map so as to construct a corresponding identifiable space.

In another embodiment, in step S300, the diagnosing whether the circuit breaker has an electromechanical fault according to the feature matrix and the feature map includes the following steps:

s301: converting the ternary column matrix X to obtain a two-dimensional matrix X-y;

in this step, the ternary column matrix X is converted by:

s302: selecting boundary data points (x, y) in the two-dimensional matrix x-y, and taking the boundary data points (x, y) as the center, and intercepting a piece of data with the length of frac forwards and backwards;

s303: carrying out weighted linear regression on the intercepted data to construct n regression curves;

s304: calculating the central value of each regression curve in the n regression curves;

s305: connecting the central values of each regression curve to obtain an optimal regression curve, wherein the optimal regression curve is a fault boundary equation of the circuit breaker;

s306: and judging whether the breaker fails or not according to whether any point on the characteristic map falls in a failure boundary equation.

In another embodiment, in step S300, diagnosing the fault severity of the electromechanical fault of the circuit breaker comprises the steps of:

s3001: initializing K clustering centers based on a ternary column matrix;

s3002: the Euclidean distance from each point in the two-dimensional matrix x-y to the clustering center is calculated and is divided into the clustering center with the smallest distance;

s3003: re-calculating the coordinates of the clustering centers, comparing whether the original k clustering centers change or not, if so, repeating the step S3002, otherwise, outputting the coordinates of the clustering centers;

s3004: and calculating Euclidean distances between the normal working condition and the data clustering centers with different fault levels to quantitatively represent the severity of the fault.

In the step, the Euclidean distance between the normal working condition clustering center and the data clustering centers with different fault levels is calculated by the following formula:

wherein X is _f For the failure condition cluster center, X _n I=1, 2,3 for normal condition cluster center.

The present disclosure will now be described in detail with reference to specific data.

1. Collecting a switching-on/off coil current signal and a displacement signal of a circuit breaker to be tested;

2. respectively extracting time domain characteristic parameters of the opening and closing coil current signal and the displacement signal, and constructing a characteristic matrix and a characteristic map based on all the time domain characteristic parameters;

2.1, respectively obtaining time domain characteristic parameters of the opening and closing coil current signal and the displacement signal, and constructing a characteristic matrix D based on all the time domain characteristic parameters;

in this step, the matrix D is composed of 15 features of the opening and closing coil current and the displacement signal, that is, the number of rows of the matrix D is determined by the number of samples, the number of columns is determined by the number of time domain features, and the constructed matrix D is represented as follows:

D＝[I ₁ ,I ₂ ,I ₃ ,T ₀ ,T ₁ ,T ₂ ,T ₃ ,T ₄ ,U ₀ ,U ₁ ,U ₂ ,t ₀ ,t ₁ ,d,v]

for example, 10 combined gate coil currents and displacement signals were measured under normal conditions, corresponding to an original sample matrix as shown in table 1, wherein the number of rows is 10 and the number of columns is 15.

TABLE 1

2.2, carrying out standardization processing on the feature matrix D to obtain a standardization matrix Z;

in this step, the normalized matrix Z obtained according to the formula is shown in table 2:

TABLE 2

0.332885	0.211512	-0.15206	-920.106	-994.714	-1148.53	-579.286	-471.635	-155.172	147.541	345.1824	-386.311	-439.274	5.341876	-20.5423
															0.155311	0.211512	0.15681	-1601.67	-1427.2	-111.663	-1000.59	-373.785	620.6897	-344.262	-328.344	240.1392	-30.647	-5.75775	17.07201
-1.39133	-1.25853	-1.1299	2146.914	2465.161	-1387.81	-1316.56	-334.645	-155.172	147.541	-412.535	-125.29	684.4495	-5.44226	11.71517
															0.411959	0.429637	0.402368	783.794	-562.23	765.6859	368.6366	252.4517	1396.552	-16.3934	429.3732	240.1392	-132.804	4.280671	3.09765
0.708099	0.624422	0.164651	-920.106	-1859.68	-749.734	-1211.23	-549.914	-931.034	-508.197	-159.963	344.5476	-30.647	-0.93931	0.535684
															1.573807	1.356091	1.095037	-2964.79	-2724.65	606.168	-157.987	-178.086	620.6897	-180.328	260.9916	-20.8817	-132.804	2.37337	5.543163
-1.72541	-1.54739	-0.7262	3510.034	3762.614	1164.481	789.9356	800.4088	-1706.9	-508.197	-159.963	-73.0858	275.8229	-0.05019	-8.08184
															-009076	-002988	0228147	102234	-56223	1004963	3686366	2915915	6206897	3114754	8419083	8352668	1736663	-005019	2864744
-0.03239	-0.02203	0.03	102.234	-129.745	47.85537	1737.858	330.7313	-155.172	147.541	-159.963	-229.698	-30.647	-2.41639	-8.19829
															0.057814	0.024651	-0.06885	-238.546	2032.677	-191.421	1000.585	232.8818	-155.172	803.2787	176.8007	-73.0858	-337.117	2.660183	-4.00598

2.3 covariance matrix R obtained from the formula is shown in table 3:

TABLE 3 Table 3

0.91204	0.809153	0.530983	-1624.37	-1793.16	17.41443	-96.1975	-191.096	448.7063	3.355609	155.8761	57.19908	-210.985	1.512525	0.401253
															0.809153	0.720359	0.477578	-1437.36	-1607.21	36.1302	-81.8217	-164.764	419.5994	-0.28709	135.1579	58.95741	-184.079	1.270038	0.689615
0.530983	0.477578	0.366395	-902.876	-1042.05	219.0977	76.48329	-34.8448	307.9727	-15.7364	99.49807	48.74715	-117.074	0.930005	0.724543
															-162437	-143736	-902876	3407800	3388149	2210661	4366929	4848385	-775568	-22967	-176288	-687885	3670831	-12508	-286084
-1793.16	-1607.21	-1042.05	3388149	4324844	68222.46	645310.2	535750.6	-991731	186700.9	-197025	-158544	324486	-1342.06	-3554.59
															17.41443	36.1302	219.0977	221066.1	68222.46	797589.5	542304.9	304648	125139	-55787.7	66702.06	56164.37	-4707.68	749.692	511.2621
-96.1975	-81.8217	76.48329	436692.9	645310.2	542304.9	1053248	380290.4	-36318.9	164030.4	84240.1	-75144.7	-79501.7	1175.191	-4952.49
															-191.096	-164.764	-34.8448	484838.5	535750.6	304648	380290.4	195699	-69506.9	26342.72	12283.82	-13099.6	19925.31	266.0039	-1154.53
448.7063	419.5994	307.9727	-775568	-991731	125139	-36318.9	-69506.9	775862.1	110231.8	110319	50404.03	-66930.2	531.5915	4077.863
															3.355609	-0.28709	-15.7364	-22967	186700.9	-55787.7	164030.4	26342.72	110231.8	163934.4	36958.09	-40318	-30330.7	415.1984	-881.14
155.8761	135.1579	99.49807	-176288	-197025	66702.06	84240.1	12283.82	110319	36958.09	84190.83	-9083.23	-67562.9	1050.191	-1435.35
															57.19908	58.95741	48.74715	-68788.5	-158544	56164.37	-75144.7	-13099.6	50404.03	-40318	-9083.23	52204.18	6992.16	-194.397	1598.459
-210.985	-184.079	-117.074	367083.1	324486	-4707.68	-79501.7	19925.31	-66930.2	-30330.7	-67562.9	6992.16	102156.6	-843.915	1691.675
															1.512525	1.270038	0.930005	-1250.8	-1342.06	749.692	1175.191	266.0039	531.5915	415.1984	1050.191	-194.397	-843.915	14.34061	-26.2733
0.401253	0.689615	0.724543	-2860.84	-3554.59	511.2621	-4952.49	-1154.53	4077.863	-881.14	-1435.35	1598.459	1691.675	-26.2733	116.453

2.4, constructing a ternary column matrix X based on the filtered eigenvalue lambda;

in this step, matrix D is a 10×15 matrix, on which an X matrix is calculated, wherein D ^T For transposed matrix, i.e. a matrix of 15×10, the ternary column matrix X is specifically shown in table 4:

TABLE 4 Table 4

In the matrix shown in Table 4, the first behavior X1, the second behavior X2, and the third behavior X3 each have 10 sets of data for each principal component.

Fig. 3 (a) to 3 (f) are schematic diagrams of the trend of the ternary array matrix under different working conditions, wherein fig. 3 (a) is that three main components of the circuit breaker are distributed more intensively under normal working conditions, and the corresponding average value is [0.3952,0.2546,0.3502]; FIG. 3 (b) shows that the capacitance voltage decreases with a corresponding decrease in the first principal component, and the second and third principal components increase, with the third principal component increasing more slowly, under a change in the capacitance voltage; FIG. 3 (c) shows that the first principal component decreases and the second and third principal components increase when the loop resistance increases under poor loop contact, wherein the second principal component increases more slowly; FIG. 3 (d) shows a graph corresponding to the development of a curve toward the second principal component axis when the core air gap is reduced, i.e., the first and second principal components are increased and the third principal component is reduced, with the change in the core air gap; FIG. 3 (e) shows that under the condition of iron core jamming, as the original length of the spring increases, the curve is developed towards the first principal component axis, namely the first principal component and the third principal component are increased, and the second principal component is reduced; FIG. 3 (f) shows that the curve is developed toward the first principal component axis as the number of springs decreases, i.e., the first and third principal components increase, the second principal component decreases, and the various conditions are distributed more dispersed in clusters.

2.5, normalizing the ternary column matrix X, i.e. ensuring that each set x1+x2+x3=1, wherein a=x1/(x1+x2+x3), b=x2/(x1+x2+x3), c=x3/(x1+x2+x3); A. b, C is a normalized matrix, and the normalized ternary column matrix is shown in table 5:

TABLE 5

X1	0.564739	0.56903	0.56364	0.568696	0.567097	0.571073	0.565965	0.569069	0.567689	0.566451
											X2	0.419269	0.419775	0.420348	0.418089	0.420185	0.418671	0.41734	0.418101	0.419087	0.418908
X3	0.015992	0.011196	0.016013	0.013215	0.012717	0.010256	0.016694	0.01283	0.013224	0.014641

In table 5, each set of data x1+x2+x3=1.

3. And diagnosing the category of the electromechanical faults of the circuit breaker and the severity of the faults according to the characteristic matrix and the characteristic map.

3.1, converting the ternary column matrix X shown in the table 4, and obtaining a two-dimensional matrix X-y shown in the table 6:

TABLE 6

x	0.774374	0.778917	0.773813	0.777741	0.77719	0.780408	0.774636	0.778119	0.777232	0.775905
											y	0.363098	0.363535	0.364032	0.362076	0.363891	0.36258	0.361427	0.362086	0.36294	0.362785

3.2, selecting boundary data points in the two-dimensional matrix x-y, and taking the boundary data points (x, y) as the center, and intercepting a section of data with the length of frac forwards and backwards;

specifically, for example, there are five boundary data points, the coordinates of which are (1, 1), (2, 2), (3, 3), (4, 4), (5, 5), and, with (3, 3) as the center, frac=1 (the step size is 1, i.e. one coordinate point before and after taking), and the intercepted data are (2, 2), (3, 3), (4, 4); centering on (3, 3), frac=2, the truncated data are (1, 1), (2, 2), (3, 3), (4, 4), (5, 5).

3.3, carrying out weighted linear regression on the intercepted data to construct n regression curves, and calculating the central value of each regression curve;

in this step, for example, there are three data points (1, 0.9), (2, 2.1) and (3,3.2), the data with length 1 is cut back and forth with the point (2, 2.1) as the center, that is, curve fitting is performed on the points (1, 0.9), (2, 2.1) and (3,3.2), so as to obtain a curve fitting function of y=x, and then the regression curve is y=x, and the corresponding center value of the curve after regression is (2, 2).

And 3.4, connecting the central values of each regression curve to obtain an optimal regression curve, wherein the optimal regression curve is a breaker fault boundary equation, the fault boundary equation is shown in fig. 4, by taking a poor loop contact as an example, the curve is developed towards a third principal component axis, three principal components [ X1, X2 and X3] can be converted into [ X, y ] through a coordinate conversion formula, and the expressions of the upper boundary and the lower boundary of the fault are calculated through local weighted regression on the basis.

In the step, n boundary data points are provided, each boundary data point can obtain a corresponding regression curve and a regression curve center value after the steps, and the optimal regression curve can be obtained by connecting the n regression curve center values. For example, n is 5, that is, there are 5 boundary data points, and after the above steps, the 5 boundary data points all obtain corresponding regression curves and regression curve center values, the corresponding regression curve center values are (1, 1), (2, 2), (3, 3), (4, 4), (5, 5), and the optimal regression curve can be obtained by connecting the 5 center values.

And 3.5, judging whether the circuit breaker has electromechanical faults according to whether any point on the characteristic map falls in a fault boundary equation.

In this step, as shown in fig. 5, point a falls within the two boundaries of the fault and point B falls outside the boundaries of the fault.

4. Diagnosing the fault severity of the electromechanical faults of the circuit breaker.

4.1, randomly selecting K clustering centers and coordinates thereof, namely finishing initialization;

4.2, calculating the Euclidean distance from each point in the two-dimensional matrix x-y to the clustering center, and dividing the Euclidean distance to the clustering center with the smallest distance;

in this step, the Euclidean distance from each point in the two-dimensional matrix x-y to the cluster center is calculated by:

d＝[(x1-x2) ² +(y1-y2) ² ] ^1/2

wherein, [ x1, y1] is the cluster center coordinates, and [ x2, y2] is the sample coordinates.

4.3, recalculating the cluster center coordinates, comparing whether the original k cluster centers are changed or not, if yes, repeating the step S3002, otherwise outputting the cluster center coordinates; illustratively, as shown in fig. 6, 2 cluster centers are initially randomly selected, dividing the original sample matrix into 2 sets. On this basis, the average value of the two sets is obtained (namely, the new cluster center coordinates), if the newly determined center does not change (or the change is smaller than a set threshold value), the clustering reaches the expected result, otherwise, the re-iteration determination is needed.

4.4, calculating Euclidean distances between the normal working condition clustering center and the data clustering centers with different fault levels according to a formula, wherein the corresponding relation between the fault severity and the Euclidean distances is shown in a table 7:

TABLE 7

Fig. 7 shows the distribution of cluster centers under different fault severity, taking the case of poor loop contact as an example, as the loop resistance increases, the curve progresses toward the third principal component axis, and the farther from the normal working condition, i.e., the higher the severity.

5. Next, the present method is compared with the existing diagnostic method, and the comparison result is specifically shown in table 8:

TABLE 8

The above general description of the invention and the description of specific embodiments thereof referred to in this application should not be construed as limiting the scope of the invention. Those skilled in the art can add, subtract or combine the features disclosed in the foregoing general description and/or the detailed description (including examples) to form other technical solutions within the scope of the present application without departing from the disclosure of the present application.

Claims (5)

1. A breaker electromechanical fault edge diagnosis method based on multi-element data fusion comprises the following steps:

s100: collecting a switching-on/off coil current signal and a displacement signal of a circuit breaker to be tested;

s200: respectively extracting time domain features of the opening and closing coil current signals and the displacement signals, and respectively constructing a feature matrix and a feature map of the opening and closing coil current signals and the displacement signals based on the time domain features;

s300: and diagnosing whether the circuit breaker has electromechanical faults and the severity of the faults according to the characteristic matrix and the characteristic map.

2. The method according to claim 1, wherein preferably step S200 comprises the steps of:

s201: acquiring time domain characteristic parameters of the opening and closing coil current signals and the displacement signals, and constructing an original sample matrix D based on the time domain characteristic parameters;

s202: carrying out standardization processing on the original sample matrix D to obtain a standardization matrix Z;

s203: calculating a covariance matrix R of the standardized matrix Z;

s204: calculating a eigenvalue lambda based on the covariance matrix R;

s205: screening the characteristic value lambda;

s206: constructing a ternary column matrix X based on the screened eigenvalue lambda;

s207: and constructing a characteristic map based on the ternary column matrix X.

3. The method according to claim 2, wherein in step S300, the diagnosing whether the circuit breaker has an electromechanical fault according to the feature matrix and the feature map comprises the following steps:

s301: converting the ternary column feature matrix to obtain a two-dimensional matrix;

s302: selecting boundary data points in the two-dimensional matrix, and intercepting data by taking the boundary data points as centers;

s303: carrying out weighted linear regression on the intercepted data to construct n regression curves;

s304: calculating the central value of each regression curve in the n regression curves;

s305: connecting the central values of each regression curve to obtain an optimal regression curve, wherein the optimal regression curve is a fault boundary equation of the circuit breaker to be tested;

s306: and judging whether the breaker fails or not according to whether any point on the characteristic map falls in a failure boundary equation.

4. The method according to claim 1, wherein in step S300, diagnosing the fault severity of the electromechanical fault of the circuit breaker comprises the steps of:

s3001: initializing K clustering centers based on a ternary column matrix;

s3002: calculating the distance from each point in the sample to the clustering center and dividing the distance to the clustering center with the smallest distance;

s3003: re-calculating the coordinates of the clustering centers, comparing whether the original k clustering centers change or not, if so, repeating the step S3002, otherwise, outputting the coordinates of the clustering centers;

s3004: and calculating Euclidean distances between the normal working condition and the data clustering centers with different fault levels to quantitatively represent the severity of the fault.

5. A circuit breaker electromechanical fault edge diagnostic device based on metadata fusion, comprising:

the acquisition module is used for acquiring a switching-on/off coil current signal and a displacement signal of the circuit breaker to be tested;

the extraction module is used for respectively extracting time domain characteristics of the opening and closing coil current signal and the displacement signal;

the construction module is used for respectively constructing a characteristic matrix and a characteristic map of the opening and closing coil current signal and the displacement signal based on the time domain characteristics of the opening and closing coil current signal and the displacement signal;

and the diagnosis module is used for diagnosing whether the circuit breaker has electromechanical faults and the severity of the faults according to the characteristic matrix and the characteristic map.

Images