CN110705038B - High-voltage circuit breaker life cycle assessment and fault early warning method - Google Patents
High-voltage circuit breaker life cycle assessment and fault early warning method Download PDFInfo
- Publication number
- CN110705038B CN110705038B CN201910853174.5A CN201910853174A CN110705038B CN 110705038 B CN110705038 B CN 110705038B CN 201910853174 A CN201910853174 A CN 201910853174A CN 110705038 B CN110705038 B CN 110705038B
- Authority
- CN
- China
- Prior art keywords
- data
- detection data
- fault
- circuit breaker
- voltage circuit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention relates to the technical field of power equipment maintenance, in particular to a life cycle assessment and fault early warning method of a high-voltage circuit breaker, which comprises the following steps: a) Acquiring detection data; b, setting a fault source to enable the high-voltage circuit breaker to be continuously switched on and off until faults occur, wherein the switching on and off action times are N, and detecting periodically; c) Obtaining a life cycle evaluation result; d) Training a fault early warning neural network model; e) And inputting detection data of the high-voltage circuit breaker to be evaluated and early-warned into a fault early-warning neural network model, and if the fault early-warning neural network model outputs a fault type, sending out fault early-warning, wherein the corresponding fault is the fault type output by the fault early-warning neural network model. The invention has the following substantial effects: the service life of the high-voltage circuit breaker is characterized, so that the maintenance of the high-voltage circuit breaker is more targeted, and the working stability and reliability of the high-voltage circuit breaker are improved.
Description
Technical Field
The invention relates to the technical field of power equipment maintenance, in particular to a life cycle assessment and fault early warning method for a high-voltage circuit breaker.
Background
The high-voltage circuit breaker not only can cut off or close the no-load current and the load current in the high-voltage circuit, but also cuts off the overload current and the short-circuit current through the effect of the relay protection device when the system fails, and has quite perfect arc extinguishing structure and enough current breaking capability. High voltage circuit breakers are responsible for the dual tasks of control and protection in electrical power systems, with performance advantages directly related to the safe operation of the electrical power system. Thus, in the maintenance of the power grid, the detection and maintenance of the high-voltage circuit breaker is an important content. The high-voltage circuit breakers in the power grid are numerous, and the detection projects are numerous. The detection and maintenance of the high-voltage circuit breaker comprise a plurality of items such as secondary circuit detection, mechanical property detection, contact resistance detection and the like, and after detection is completed, detection data are required to be analyzed and judged, so that whether the detected high-voltage circuit breaker has potential safety hazards or not is determined, and the state of the high-voltage circuit breaker is evaluated. Wherein the mechanical characteristic parameter is one of the important parameters for judging the performance of the circuit breaker. According to the research results of CIGRE and China electric department, the mechanical fault accounts for approximately 37% of the faults of the switching equipment, so that the detection of the mechanical fault of the switching equipment is very necessary. The currently adopted overhaul mode of 'due overhaul' has serious defects. Such as temporary undercrown, overrown, blind repair or maintenance accidents caused by improper maintenance. The state maintenance based on the running state of the equipment is the most advanced equipment maintenance method at present. Therefore, the life cycle evaluation and fault early warning research on the high-voltage circuit breaker have important economic and technical significance.
For example, chinese patent CN105467309a, publication date 20156, month 4 and 6, a method for evaluating contact state of high voltage circuit breaker and maintenance strategy, the main technical characteristics are: connecting a dynamic resistance tester, a high-capacity storage battery, a speed sensor and a current sensor with a high-voltage circuit breaker contact to form a test loop and measuring the dynamic resistance of the high-voltage circuit breaker contact in the opening process; the dynamic resistance tester calculates and analyzes the measured speed, current and voltage signals to obtain a contact dynamic resistance-stroke curve in the brake separating process, and obtains a contact resistance value and a length value through the contact dynamic resistance-stroke curve; and comparing the acquired contact resistance value with a length value and a standard value, evaluating the contact state, and formulating an effective maintenance strategy. The technical scheme uses the contact resistance value and the length value to intuitively represent the contact state, and provides an important reference basis for evaluating the contact state of the circuit breaker and formulating the maintenance strategy. But it cannot provide technical guidance for detection of secondary circuits and mechanical characteristics, and cannot solve the problem of lack of life cycle evaluation technology of high-voltage circuit breakers.
Disclosure of Invention
The invention aims to solve the technical problems that: the detection of the current high-voltage circuit breaker cannot provide the technical problem of life cycle assessment of the high-voltage circuit breaker. A life cycle assessment and fault early warning method for a high-voltage circuit breaker based on big data technology is provided.
In order to solve the technical problems, the invention adopts the following technical scheme: a life cycle assessment and fault early warning method for a high-voltage circuit breaker comprises the following steps: a) Acquiring detection data of the same type of high-voltage circuit breaker in history maintenance; b) Acquiring a high-voltage circuit breaker with the same type, manually setting a fault source under laboratory conditions, enabling the high-voltage circuit breaker to continuously perform electrified switching-on and switching-off actions under the condition that the fault source exists until faults occur, detecting periodically while taking detection data as reference data, and correlating the detection data with the corresponding electrified switching-on and switching-off action times N; c) Comparing the detection data of the high-voltage circuit breaker with the reference data to obtain the number of times N corresponding to the reference data closest to the detection data, and taking N/N as a life cycle evaluation result of the high-voltage circuit breaker; d) Correlating the reference data with the type of the faults to form sample data, and training a fault early warning neural network model by using the sample data; e) Inputting the detection data of the high-voltage circuit breaker to be evaluated and pre-warned into the fault pre-warning neural network model obtained in the step D), and if the fault pre-warning neural network model outputs a fault type, sending out fault pre-warning, wherein the corresponding fault is the fault type output by the fault pre-warning neural network model. Through correlation between the switching-on times and the detection data, the equivalent switching-on times can be obtained by inputting the detection data, so that the representation of the life cycle of the high-voltage circuit breaker is obtained, and the maintenance of the high-voltage circuit breaker is more targeted; the fault early warning neural network model provides fault early warning information of the high-voltage circuit breaker, so that maintainers can maintain the high-voltage circuit breaker in a targeted manner, and the working stability and reliability of the high-voltage circuit breaker are improved.
Preferably, step D) further comprises: the detection data of M times before the occurrence of the fault is associated with the corresponding fault type to form fault sample data, and the fault sample data is used for training a fault judging neural network model; using detection data of the times (N-M) before occurrence of faults, correlating with the types of the faults, and forming a sample data training fault early warning neural network model; step E further comprises: and D) inputting the detection data of the high-voltage circuit breaker into the fault judging neural network model obtained in the step D), and giving out fault alarm if the fault judging neural network model outputs the fault type. The data near the occurrence of the fault can be used as sample data of the fault finding, thereby being used for training the fault finding neural network model.
Preferably, in step B), the method for artificially setting the fault source includes the steps of: b11 Detecting the high-voltage circuit breaker for a plurality of times; b12 According to the maintenance requirement of the high-voltage circuit breaker, sequentially selecting one maintenance requirement to ensure that the maintenance requirement does not reach the standard, and detecting for a plurality of times after carrying out a plurality of electrified opening and closing actions; b13 Sequentially selecting two maintenance requirements to ensure that the maintenance requirements do not reach the standard, and detecting for a plurality of times after carrying out a plurality of charged opening and closing actions; b14 Using liquid nitrogen or dry ice to rapidly cool the high-voltage breaker, performing mechanical property tests for a plurality of times, and obtaining detection data of the mechanical property tests. And faults are actively generated, so that fault data are acquired, and the problem of insufficient fault data samples is effectively solved. The lubrication performance of the lubricant or the lubricating oil is reduced through liquid nitrogen or dry ice cooling, so that the state of jamming is simulated, after the test is finished, the lubrication performance of the lubricant or the lubricating oil is recovered, the type of the jamming failure of the mechanical part is simulated in a lossless mode, and the state data under the type of the jamming failure are obtained. The mechanical parts in nature become stuck because of poor lubrication or dust particles entering.
Preferably, in the step D, the method for establishing the fault diagnosis model of the high-voltage circuit breaker includes: d21 All detection data are obtained, and the detection data are associated with the corresponding fault types to be used as sample data; d22 Preprocessing, normalizing and training the neural network model, and taking the trained neural network model as a fault studying and judging model.
Preferably, in step D21), the method for associating the detection data with the corresponding fault type includes: d211 Obtaining the detection data in step B11) as historical detection data; d212 Comparing the detection data obtained by the detection in the step B12) and the step B13) with the historical detection data in sequence, and if the difference between the detection data and the historical detection data is larger than a preset threshold value, associating the detection data with the maintenance requirement which does not reach the standard; d213 Comparing the plurality of groups of detection data obtained by the plurality of mechanical characteristic tests in the step B14) with the historical detection data, and if the difference between the detection data and the historical detection data is larger than a preset threshold value, associating the group of detection data with the jamming fault of the mechanical part.
Preferably, the method for judging whether the difference between the detected data and the historical detected data is larger than a preset threshold value comprises the following steps: d31 Carrying out segmentation processing on the numerical value in the detection data and the historical detection data, and converting the numerical value into a state quantity by taking a segmentation interval as a name; d32 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d32 The Boolean amount of the processed historical detection data is regarded as a numerical value to be averaged, the average value is rounded to be an integer, the obtained integer is regarded as the Boolean amount again, and the processed detection data and the historical detection data are sequenced according to the setting to respectively form a detection vector and a historical detection vector; d33 Calculating the distance between the detection vector and the historical detection vector, comparing the distance with a preset distance threshold, if the distance is larger than the preset distance threshold, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is larger than the preset threshold, otherwise, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is not larger than the preset threshold. Boolean only eliminates the differences between the range of values of the data values.
Alternatively, the method for judging whether the difference between the detected data and the historical detected data is larger than a preset threshold value comprises the following steps: d41 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d42 Normalizing the detection data and the numerical value in the history detection data to obtain the minimum value and the maximum value of each item of the normalized history detection data respectively, arranging the processed Boolean quantity and the numerical value according to a set sequence, ordering the detection data to form a detection vector, ordering each minimum value of the history detection data to form a history detection left vector, and ordering each maximum value of the history detection data to form a history detection right vector; d43 The distance between the detection vector and the historical detection left vector as well as the distance between the detection vector and the historical detection right vector are calculated respectively, compared with a preset distance threshold, if the distance between the detection vector and the historical detection left vector as well as the distance between the detection data corresponding to the detection vector and the historical detection data are larger than the preset distance threshold, the distance between the detection data corresponding to the detection vector and the historical detection data is judged to be not larger than the preset threshold, otherwise, the distance between the detection data corresponding to the detection vector and the historical detection data is judged to be not larger than the preset threshold.
Alternatively, the method for judging whether the difference between the detected data and the historical detected data is larger than a preset threshold value comprises the following steps: d51 Carrying out segmentation processing on the numerical value in the detection data and the historical detection data, and converting the numerical value into a state quantity by taking a segmentation interval as a name; d52 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d52 Deleting the Boolean amounts with different values in the processed historical detection data, sorting the residual Boolean amounts of the historical detection data according to the setting to respectively form historical detection vectors, selecting the Boolean amounts corresponding to the historical detection vectors in the detection data, and sorting according to the setting to form detection vectors; d53 Calculating the distance between the detection vector and the historical detection vector, comparing the distance with a preset distance threshold, if the distance is larger than the preset distance threshold, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is larger than the preset threshold, otherwise, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is not larger than the preset threshold. The boolean component vector can eliminate the deviation caused by the numerical data. The Boolean amounts with different values are deleted, and only the Boolean amounts with consistent values are reserved, so that the influence caused by irrelevant factors can be eliminated, and the judgment result has a higher reference value.
Preferably, in step D51), the method for performing the segmentation processing on the numerical value amounts in the detection data and the history detection data includes: d511 Selecting a numerical value to obtain all numerical values of the numerical value in the historical detection data, and sequentially arranging the numerical values according to the numerical values, wherein the numerical values are marked as a set Ki, and the minimum value in the set Ki is k min And a maximum value of k max The method comprises the steps of carrying out a first treatment on the surface of the D512 To partition start point k s Giving an initial value of k min Partition endpoint k e Giving an initial value of k max Let the value k be examined m ＝k s +n×Δk, Δk is a step size manually set, n is a positive integer, and n is 1 as an initial value; d513 N is continuously added with 1, if the value k is examined m The following conditions are satisfied:
wherein the function N (x, y) represents the set Ki, and the number of data values in the numerical interval (x, y) is (2 k) m -k s ) As the interval dividing points and adding the dividing point set Km, the value of (2 k m -k s ) Assignment of the value of k to s Continuously adding 1 to n until k m ＞k max The method comprises the steps of carrying out a first treatment on the surface of the D514 To k) min And k max Adding a set Km, and dividing the numerical value data into numerical value intervals by using values in the Km as dividing points; d515 Selecting the next numerical value, and repeating the steps D511) to D514) until all the numerical values are divided into interval segments; d516 The detection data is divided into sections of a numerical value corresponding to the history detection data. According to the aggregation characteristics of the numerical values, segmentation is carried out, so that the segmentation is more close to different states of the numerical values.
Preferably, in step D51), the method of converting the numerical value into the state quantity by using the segment section as a name includes the steps of: d511 Dividing the numerical data into a plurality of intervals, [ n ] m(1) ，n m(2) ]，[n m(2) ，n m(3) ]...[n m(k-1) ，n m(k) ]Wherein n is m(1) And n m(k) Respectively a starting point and an end point of the numerical value interval, and nm (2) to nm (k-1) are intermediate dividing points of the numerical value intervalRespectively used as state names of corresponding numerical value intervals; d512 If the data of the historical detection numerical value falls into the interval [ n ] m(d) ，n m(d+1) ]，d∈[1，k-1]The state name is thenAs the value of the numerical quantity, conversion of the numerical quantity data into state quantity data is completed. The conversion of the numerical value quantity into the state quantity can be completed rapidly.
Preferably, in step D52), the method of converting the state quantity in the detection data and the history detection data into the boolean quantity includes the steps of: d521 Obtaining all state values of the state quantity data; d522 Splitting the state quantity field into a plurality of fields by taking the state value as a field name; d523 Setting the field with the same field name and state quantity data value as 1, setting the other split fields as 0, and completing the splitting of the state quantity data into Boolean quantity data. The state quantity is split into the Boolean quantity, so that the training efficiency of the neural network can be accelerated.
Preferably, in step D512), the setting method of the step Δk includes: and calculating the value data of the numerical value in the set Ki, removing the value data of the numerical value data of the set Ki from the value data of the numerical value data to obtain the value data of the numerical value data, performing absolute value operation on the residual value data, taking the minimum value as the step delta k, and participating in calculation.
Preferably, in step D521), the method for obtaining all state values of the state quantity data includes: if the state quantity data is the state of the breaker, all state values comprise all possible values of the state; if the state quantity data is the state quantity data converted from the numerical quantity data, all state values only comprise the values which appear in the history state.
Preferably, the detection data includes a closing time, a breaking time, a closing speed, a three-phase different period, an in-phase different period, a golden short time, a no-flow time, a moving contact maximum speed, a moving contact average speed, a moving contact action time, a bouncing number of times, a bouncing maximum amplitude, a breaking and closing stroke, a breaking and closing process current waveform curve, a time speed stroke dynamic curve in the breaking and closing stroke of the moving contact, an opening distance and a contact resistance. By acquiring various data of the high-voltage circuit breaker, the state data of the high-voltage circuit breaker is more comprehensive, the accuracy of fault research and judgment is improved, and meanwhile, conditions are provided for finding unobvious abnormal data.
Preferably, in step D, before training the fault analysis neural network model using the fault sample data, the normalization processing is performed on the fault sample data, including: d11 The method comprises the steps of) listing numerical data in fault sample data, obtaining a theoretical boundary value of the numerical data as a boundary value, if the theoretical boundary value does not exist, obtaining a historical boundary value of the data as the boundary value, wherein a left boundary value of the boundary value is regarded as 0, a right boundary value of the boundary value is regarded as 1, and dividing a difference of the numerical data minus the left boundary value by a difference of the right boundary value and the left boundary value to obtain a normalized value of the numerical data; d12 Splitting the state quantity data into a plurality of boolean data; d13 Boolean data into numerical values and normalized.
Preferably, in step D12, the method for splitting the state quantity data into a plurality of boolean data includes: d121 Obtaining all state values of the state quantity data; d122 Splitting the state quantity field into a plurality of fields by taking the state value as a field name; d123 Setting the field with the same field name and state quantity data value, setting the other split fields to zero, and completing the splitting of the state quantity data into Boolean quantity data.
Preferably, in step B), before the high-voltage circuit breaker is detected, a non-contact displacement sensor is mounted on each mechanical moving part of the high-voltage circuit breaker, and displacement data measured by the non-contact displacement sensor is added to the detection data of the high-voltage circuit breaker.
Preferably, in step E), each mechanical moving part of the high-voltage circuit breaker to be evaluated and pre-warned is provided with a non-contact displacement sensor, displacement data measured by the non-contact displacement sensor is obtained, the state data of the high-voltage circuit breaker to be evaluated and pre-warned and the displacement data measured by the non-contact displacement sensor are taken as input of a fault judging neural network model, the fault judging neural network model is trained, and fault judging is carried out.
Preferably, in the step B), a non-contact displacement sensor is installed on each mechanical moving part of a normal high-voltage circuit breaker, and under the condition of power failure, the switching-on and switching-off test is continuously repeated on the high-voltage circuit breaker until the mechanical parts of the high-voltage circuit breaker are damaged, and the switching-on and switching-off times K in the test process and the displacement data of each mechanical moving part in the switching-on and switching-off process are recorded as historical displacement data; in step E), if the fault judging result of the high-voltage circuit breaker to be judged is no fault, installing a non-contact displacement sensor on each mechanical moving part of the high-voltage circuit breaker to be judged, performing one-time opening and closing of the high-voltage circuit breaker to be judged to obtain displacement data measured by the non-contact displacement sensor, comparing the displacement data with the historical displacement data to obtain the opening and closing test times K corresponding to the closest historical displacement data, and taking (K-K) as the residual service life of the high-voltage circuit breaker to be judged.
The non-contact displacement sensor comprises a laser emitter, a current limiting resistor, a photoresistor, a power supply module, a reflection sticker, a voltage sensor and a communication module, wherein the laser emitter is fixedly arranged in a shell of the high-voltage circuit breaker, an alignment point on the outer surface of a mechanical moving part is aligned along the normal direction, an included angle is formed between emergent light of the laser emitter and the normal direction of the outer surface of the mechanical moving part, the alignment point of the laser emitter moves along the outer surface of the mechanical moving part in the stroke of the mechanical moving part to form a moving range, the reflection sticker is attached to the mechanical moving part and covers the moving range of the alignment point, the reflection sticker is provided with a plurality of high reflection areas which are arranged at equal intervals along the stroke of the mechanical moving part, a low reflection area is arranged between every two adjacent high reflection areas, the width of the high reflection area is equal to the width of the low reflection area, the spot diameter of the laser emitter is equal to the integral multiple of the interval width, the photoresistor is arranged on the other side of the laser emitter which is symmetrical to the outer surface of the mechanical moving part along the normal direction, one end of the photoresistor is grounded, the other end of the photoresistor is connected with the power supply module, the voltage sensor is connected with the voltage sensor through the current limiting resistor to the connecting point, and the voltage sensor is connected with the communication module.
The invention has the following substantial effects: through correlation between the switching-on times and the detection data, the equivalent switching-on times can be obtained by inputting the detection data, so that the representation of the life cycle of the high-voltage circuit breaker is obtained, and the maintenance of the high-voltage circuit breaker is more targeted; the fault early warning information of the high-voltage circuit breaker is provided through the fault early warning neural network model, so that maintenance personnel can maintain the high-voltage circuit breaker in a targeted manner, and the working stability and reliability of the high-voltage circuit breaker are improved; the technical problem that the number of the detection data samples under the fault is small can be solved by actively setting the fault source, the detection data and the fault are more relevant, the accuracy of fault analysis is improved, the convergence speed of the fault evaluation model can be increased by carrying out normalization processing on the sample data, the establishment efficiency of the fault evaluation model is increased, and the accuracy of the fault evaluation model is improved.
Drawings
FIG. 1 is a flow diagram of an embodiment.
FIG. 2 is a flow chart of a method for artificially setting a fault source according to an embodiment.
FIG. 3 is a flow chart of a method for associating detection data with a corresponding fault type according to an embodiment.
FIG. 4 is a flowchart of a method for determining whether a difference is greater than a predetermined threshold according to an embodiment.
Fig. 5 is a schematic structural diagram of a non-contact displacement sensor according to an embodiment.
Fig. 6 and 7 are schematic diagrams illustrating measurement of a non-contact displacement sensor according to an embodiment.
FIG. 8 is a flowchart of a method for determining whether the difference is greater than a predetermined threshold according to a second embodiment.
Fig. 9 is a flowchart of a method for determining whether the difference is greater than a preset threshold according to the third embodiment.
Wherein: 1. the device comprises a linear reflection sticker, 2, a laser emitter, 3, a cylindrical surface reflection sticker, 4, a cam, 5, a cylindrical end surface reflection sticker, 6, a moving part, 7, an alignment point track, 8, an arc reflection sticker, 100, a voltage sensor, 200 and a communication module.
Detailed Description
The following description of the embodiments of the present invention will be made with reference to the accompanying drawings.
Embodiment one:
a life cycle assessment and fault early warning method of a high-voltage circuit breaker is shown in fig. 1, and comprises the following steps: a) And acquiring detection data of the same type of high-voltage circuit breaker in history maintenance. The detection data comprise closing time, opening time, closing speed, opening speed, three-phase different periods, same-phase different periods, golden short time, no-flow time, moving contact maximum speed, moving contact average speed, moving contact action time, bouncing times, bouncing maximum amplitude, opening and closing stroke, opening and closing process current waveform curve, time speed and stroke dynamic curve in the opening and closing stroke of the moving contact, and contact resistance. By acquiring various data of the high-voltage circuit breaker, the state data of the high-voltage circuit breaker is more comprehensive, the accuracy of fault research and judgment is improved, and meanwhile, conditions are provided for finding unobvious abnormal data.
B) The method comprises the steps of obtaining a high-voltage circuit breaker of the same type, installing a non-contact displacement sensor on each mechanical moving part of the high-voltage circuit breaker, adding displacement data measured by the non-contact displacement sensor into detection data of the high-voltage circuit breaker, manually setting a fault source under laboratory conditions, enabling the high-voltage circuit breaker to continuously perform electrified switching-on and switching-off actions under the condition that the fault source exists until faults occur, detecting periodically until the times of the electrified switching-on and switching-off actions of the high-voltage circuit breaker before the faults are N, taking the detection data as reference data, and correlating with the times of the corresponding electrified switching-on and switching-off actions of a test. And each mechanical moving part of the normal high-voltage circuit breaker is provided with a non-contact displacement sensor, and under the power-off condition, the switching-on and switching-off test is continuously repeated on the high-voltage circuit breaker until the mechanical parts of the high-voltage circuit breaker are damaged, and the switching-on and switching-off times K in the test process and the displacement data of each mechanical moving part in the switching-on and switching-off process are recorded as historical displacement data.
As shown in fig. 2, the method for artificially setting the fault source includes the steps of: b11 Detecting the high-voltage circuit breaker for a plurality of times; b12 According to the maintenance requirement of the high-voltage circuit breaker, sequentially selecting one maintenance requirement to ensure that the maintenance requirement does not reach the standard, and detecting for a plurality of times after carrying out a plurality of electrified opening and closing actions; b13 Sequentially selecting two maintenance requirements to ensure that the maintenance requirements do not reach the standard, and detecting for a plurality of times after carrying out a plurality of charged opening and closing actions; b14 Using liquid nitrogen or dry ice to rapidly cool the high-voltage breaker, performing mechanical property tests for a plurality of times, and obtaining detection data of the mechanical property tests. And faults are actively generated, so that fault data are acquired, and the problem of insufficient fault data samples is effectively solved. The lubrication performance of the lubricant or the lubricating oil is reduced through liquid nitrogen or dry ice cooling, so that the state of jamming is simulated, after the test is finished, the lubrication performance of the lubricant or the lubricating oil is recovered, the type of the jamming failure of the mechanical part is simulated in a lossless mode, and the state data under the type of the jamming failure are obtained. The mechanical parts in nature become stuck because of poor lubrication or dust particles entering.
As shown in fig. 5, the non-contact displacement sensor comprises a laser emitter 2, a current limiting resistor, a photoresistor, a power supply module, a reflective sticker, a voltage sensor 100 and a communication module 200, wherein the laser emitter 2 is fixedly installed in a shell of a high-voltage circuit breaker, an alignment point of the outer surface of the mechanical moving part 6 is aligned along the normal direction, an included angle is formed between the emergent light of the laser emitter 2 and the normal direction of the outer surface of the mechanical moving part 6 by adjustment, in the stroke of the mechanical moving part 6, the alignment point of the laser emitter 2 moves along the outer surface of the mechanical moving part 6 to form a moving range, the reflective sticker is attached to the mechanical moving part 6 and covers the moving range of the alignment point, a plurality of high-reflection areas are arranged at equal intervals along the stroke of the mechanical moving part 6, a low-reflection area is arranged between every two adjacent high-reflection areas, the width of the high-reflection area is equal to the width of the low-reflection area, the diameter of a light spot of the laser emitter 2 is equal to the integral multiple of the interval width, one end of the photoresistor is grounded, the other end of the photoresistor is connected with the power supply module through the current limiting resistor, and the voltage sensor 100 is connected with the voltage sensor 100. Fig. 5 shows a linear reflection sticker 1, in which a mechanical moving part 6 to be detected moves in a linear direction, such as a moving contact, an unlocking lock, and the like. As shown in fig. 6, when the rotating member such as the shaft and the cam 4 is subjected to displacement non-contact displacement detection, a cylindrical surface reflection sticker 3 may be attached to the outer surface of the shaft or an equal radius circular arc portion of the cam 4, so as to avoid blurring of pictures, and the distance between the high reflection area and the low reflection area in the figure is distorted. When the equal radius arc portion of the cam 4 is also a working surface, a cylindrical end surface reflection sticker 5 may be attached to the end surface of the cam 4. As shown in fig. 7, when the detected moving part 6 has complex planar motion, that is, includes both translational motion and rotational motion, a suitable alignment point is selected on the detected moving part 6, so that the alignment point is always on the moving part 6 in the stroke of the moving part 6, the alignment point track 7 will be a segment of arc, the adaptive arc-shaped reflective sticker 8 is attached, and the arc-shaped reflective sticker 8 is arranged with high-reflection areas and low-reflection areas at intervals along the arc, so that the edges of the high-reflection areas and the low-reflection areas are perpendicular to the arcs at the corresponding positions. The present embodiment provides a non-contact displacement sensor implementation, in which a non-contact displacement sensor is known in the prior art for detecting vibration and displacement, and a person skilled in the art can design a non-contact displacement sensor in other forms to complete the detection of displacement.
C) Comparing the detection data of the high-voltage circuit breaker with the reference data, obtaining the times N corresponding to the reference data closest to the detection data, and taking N/N as a life cycle evaluation result of the high-voltage circuit breaker.
D) Associating the reference data with the type of the fault to form sample data, training a fault early warning neural network model by using the sample data, associating the detection data of M times before the fault with the corresponding fault type to form fault sample data, and training a fault studying and judging neural network model by using the fault sample data; and (3) using detection data of the times before occurrence of the fault (N-M), correlating with the type of the fault, and forming a sample data training fault early warning neural network model.
The method for establishing the high-voltage circuit breaker fault research model comprises the following steps: d21 All detection data are obtained, and the detection data are associated with the corresponding fault types to be used as sample data; d22 Preprocessing, normalizing and training the neural network model, and taking the trained neural network model as a fault studying and judging model. As shown in fig. 3, in step D21), the method of associating the detection data with the corresponding fault type includes: d211 Obtaining the detection data in step B11) as historical detection data; d212 Comparing the detection data obtained by the detection in the step B12) and the step B13) with the historical detection data in sequence, and if the difference between the detection data and the historical detection data is larger than a preset threshold value, associating the detection data with the maintenance requirement which does not reach the standard; d213 Comparing the plurality of groups of detection data obtained by the plurality of mechanical characteristic tests in the step B14) with the historical detection data, and if the difference between the detection data and the historical detection data is larger than a preset threshold value, associating the group of detection data with the jamming fault of the mechanical part.
In step D, before training the fault analysis neural network model by using the fault sample data, performing normalization processing on the fault sample data, including: d11 The method comprises the steps of) listing numerical data in fault sample data, obtaining a theoretical boundary value of the numerical data as a boundary value, if the theoretical boundary value does not exist, obtaining a historical boundary value of the data as the boundary value, wherein a left boundary value of the boundary value is regarded as 0, a right boundary value of the boundary value is regarded as 1, and dividing a difference of the numerical data minus the left boundary value by a difference of the right boundary value and the left boundary value to obtain a normalized value of the numerical data; d12 Splitting the state quantity data into a plurality of boolean data; d13 Boolean data into numerical values and normalized.
In step D12, the method for splitting the state quantity data into a plurality of boolean data includes: d121 Obtaining all state values of the state quantity data; d122 Splitting the state quantity field into a plurality of fields by taking the state value as a field name; d123 Setting the field with the same field name and state quantity data value, setting the other split fields to zero, and completing the splitting of the state quantity data into Boolean quantity data.
As shown in fig. 4, the method for determining whether the difference between the detected data and the historical detected data is greater than a preset threshold value includes: d31 Carrying out segmentation processing on the numerical value in the detection data and the historical detection data, and converting the numerical value into a state quantity by taking a segmentation interval as a name; d32 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d32 The Boolean amount of the processed historical detection data is regarded as a numerical value to be averaged, the average value is rounded to be an integer, the obtained integer is regarded as the Boolean amount again, and the processed detection data and the historical detection data are sequenced according to the setting to respectively form a detection vector and a historical detection vector; d33 Calculating the distance between the detection vector and the historical detection vector, comparing the distance with a preset distance threshold, if the distance is larger than the preset distance threshold, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is larger than the preset threshold, otherwise, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is not larger than the preset threshold. Boolean only eliminates the differences between the range of values of the data values.
E) And each mechanical moving part of the high-voltage circuit breaker to be evaluated and pre-warned is provided with a non-contact displacement sensor, displacement data measured by the non-contact displacement sensor are obtained, the state data of the high-voltage circuit breaker to be evaluated and pre-warned and the displacement data measured by the non-contact displacement sensor are taken as the input of a fault judging neural network model, the fault judging neural network model is trained, and the fault judging is carried out.
Inputting the detection data of the high-voltage circuit breaker to be evaluated and pre-warned into the fault pre-warning neural network model obtained in the step D), and if the fault pre-warning neural network model outputs a fault type, sending out fault pre-warning, wherein the corresponding fault is the fault type output by the fault pre-warning neural network model. Through correlation between the switching-on times and the detection data, the equivalent switching-on times can be obtained by inputting the detection data, so that the representation of the life cycle of the high-voltage circuit breaker is obtained, and the maintenance of the high-voltage circuit breaker is more targeted; the fault early warning neural network model provides fault early warning information of the high-voltage circuit breaker, so that maintainers can maintain the high-voltage circuit breaker in a targeted manner, and the working stability and reliability of the high-voltage circuit breaker are improved. And D) inputting the detection data of the high-voltage circuit breaker into the fault judging neural network model obtained in the step D), and giving out fault alarm if the fault judging neural network model outputs the fault type. The data near the occurrence of the fault can be used as sample data of the fault finding, thereby being used for training the fault finding neural network model.
Embodiment two:
the embodiment is further improved on the basis of the first embodiment, as shown in fig. 8, and the method for determining whether the difference between the detected data and the historical detected data is greater than a preset threshold value includes: d41 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d42 Normalizing the detection data and the numerical value in the history detection data to obtain the minimum value and the maximum value of each item of the normalized history detection data respectively, arranging the processed Boolean quantity and the numerical value according to a set sequence, ordering the detection data to form a detection vector, ordering each minimum value of the history detection data to form a history detection left vector, and ordering each maximum value of the history detection data to form a history detection right vector; d43 The distance between the detection vector and the historical detection left vector as well as the distance between the detection vector and the historical detection right vector are calculated respectively, compared with a preset distance threshold, if the distance between the detection vector and the historical detection left vector as well as the distance between the detection data corresponding to the detection vector and the historical detection data are larger than the preset distance threshold, the distance between the detection data corresponding to the detection vector and the historical detection data is judged to be not larger than the preset threshold, otherwise, the distance between the detection data corresponding to the detection vector and the historical detection data is judged to be not larger than the preset threshold.
Embodiment III:
the embodiment is further improved on the basis of the first embodiment, as shown in fig. 9, the method for determining whether the difference between the detected data and the historical detected data is greater than a preset threshold value includes: d51 Carrying out segmentation processing on the numerical value in the detection data and the historical detection data, and converting the numerical value into a state quantity by taking a segmentation interval as a name; d52 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d52 Deleting the Boolean amounts with different values in the processed historical detection data, sorting the residual Boolean amounts of the historical detection data according to the setting to respectively form historical detection vectors, selecting the Boolean amounts corresponding to the historical detection vectors in the detection data, and sorting according to the setting to form detection vectors; d53 Calculating the distance between the detection vector and the historical detection vector, comparing the distance with a preset distance threshold, if the distance is larger than the preset distance threshold, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is larger than the preset threshold, otherwise, judging that the distance between the detection data corresponding to the detection vector and the historical detection data is not larger than the preset threshold. The boolean component vector can eliminate the deviation caused by the numerical data. The Boolean amounts with different values are deleted, and only the Boolean amounts with consistent values are reserved, so that the influence caused by irrelevant factors can be eliminated, and the judgment result has a higher reference value.
In step D51), the method for performing the segmentation processing on the numerical value amounts in the detection data and the history detection data includes: d511 Selecting a numerical value to obtain all numerical values of the numerical value in the historical detection data, and sequentially arranging the numerical values according to the numerical values, wherein the numerical values are marked as a set Ki, and the minimum value in the set Ki is k min And a maximum value of k max The method comprises the steps of carrying out a first treatment on the surface of the D512 To partition start point k s Giving an initial value of k min Partition endpoint k e Giving an initial value of k max Let the value k be examined m ＝k s The method for setting the step size delta k comprises the following steps: and calculating the value data of the numerical value in the set Ki, removing the value data of the numerical value data of the set Ki from the value data of the numerical value data to obtain the value data of the numerical value data, performing absolute value operation on the residual value data, taking the minimum value as the step delta k, and participating in calculation.
The method comprises the steps of carrying out a first treatment on the surface of the D513 N is continuously added with 1, if the value k is examined m The following conditions are satisfied:
wherein the function N (x, y) represents the set Ki, and the number of data values in the numerical interval (x, y) is (2 k) m -k s ) As the interval dividing points and adding the dividing point set Km, the value of (2 k m -k s ) Assignment of the value of k to s Continuously adding 1 to n until k m ＞k max The method comprises the steps of carrying out a first treatment on the surface of the D514 To k) min And k max Adding a set Km, and dividing the numerical value data into numerical value intervals by using values in the Km as dividing points; d515 Selecting the next numerical value, and repeating the steps D511) to D514) until all the numerical values are divided into interval segments; d516 The detection data is divided into sections of a numerical value corresponding to the history detection data. According to the aggregation characteristics of the numerical values, segmentation is carried out, so that the segmentation is more close to different states of the numerical values.
In step D51), the method for converting the numerical value into the state quantity by using the segment section as a name includes the steps of: d511 Dividing the numerical data into a plurality of intervals, [ n ] m(1) ，n m(2) ]，[n m(2) ，n m(3) ]…[n m(k-1) ，n m(k) ]Wherein n is m(1) And n m(k) Respectively a starting point and an ending point of a numerical value interval, n m(2) ～n m(k-1) Dividing the middle of the numerical interval into pointsRespectively used as state names of corresponding numerical value intervals; d512 If the data of the historical detection numerical value falls into the interval [ n ] m(d) ，n m(d+1) ]，d∈[1，k-1]Then the status name->As the value of the numerical quantity, conversion of the numerical quantity data into state quantity data is completed. The conversion of the numerical value quantity into the state quantity can be completed rapidly.
The method for converting the state quantity in the detection data and the historical detection data into the Boolean quantity comprises the following steps: d521 All state values of the state quantity data are obtained, and if the state quantity data are the states of the circuit breaker, all state values comprise all possible values of the states; if the state quantity data is the state quantity data converted from the numerical quantity data, all state values only comprise the values which appear in the history state. The method comprises the steps of carrying out a first treatment on the surface of the D522 Splitting the state quantity field into a plurality of fields by taking the state value as a field name; d523 Setting the field with the same field name and state quantity data value as 1, setting the other split fields as 0, and completing the splitting of the state quantity data into Boolean quantity data. The state quantity is split into the Boolean quantity, so that the training efficiency of the neural network can be accelerated.
The above-described embodiment is only a preferred embodiment of the present invention, and is not limited in any way, and other variations and modifications may be made without departing from the technical aspects set forth in the claims.
Claims (15)
1. A life cycle assessment and fault early warning method for a high-voltage circuit breaker is characterized in that,
the method comprises the following steps:
a) Acquiring historical maintenance detection data of the high-voltage circuit breakers of the same type;
b) Acquiring a high-voltage circuit breaker of the same type, manually setting a fault source under laboratory conditions, enabling the high-voltage circuit breaker to continuously perform electrified switching-on and switching-off actions under the condition that the fault source exists until faults occur, detecting periodically while the times of the electrified switching-on and switching-off actions of the high-voltage circuit breaker before the faults are N, and taking periodic detection data as reference data and correlating the periodic detection data with the times of the electrified switching-on and switching-off actions corresponding to the tests;
c) Comparing the periodic detection data of the high-voltage circuit breaker with the reference data to obtain the number of times N corresponding to the reference data closest to the periodic detection data, and taking N/N as a life cycle evaluation result of the high-voltage circuit breaker;
d) Correlating the reference data with the type of the faults to form sample data, and training a fault early warning neural network model by using the sample data;
Step D) further comprises: the detection data of M times before the occurrence of the fault is associated with the corresponding fault type to form fault sample data, and the fault sample data is used for training a fault judging neural network model; using detection data of the times (N-M) before occurrence of faults, correlating with the types of the faults, and forming a sample data training fault early warning neural network model; step E further comprises: inputting the detection data of the high-voltage circuit breaker into the fault judging neural network model obtained in the step D), and giving out fault alarm if the fault judging neural network model outputs the fault type;
the method for establishing the high-voltage circuit breaker fault research model comprises the following steps:
d21 All detection data are obtained, and the detection data are associated with the corresponding fault types to be used as sample data;
d22 Preprocessing, normalizing and training the sample data, and taking the trained neural network model as a fault studying and judging model;
in step D21), the method of associating the detection data with the corresponding fault type includes:
d211 Obtaining the detection data in step B11) as historical detection data;
d212 Comparing the detection data obtained by the detection in the step B12) and the step B13) with the historical detection data in sequence, and if the difference between the detection data and the historical detection data is larger than a preset threshold value, associating the detection data with the maintenance requirement which does not reach the standard;
D213 Comparing the plurality of groups of detection data obtained by the plurality of mechanical characteristic tests in the step B14) with the historical detection data, and if the difference between the detection data and the historical detection data is larger than a preset threshold value, associating the group of detection data with the jamming fault of the mechanical part; the method for judging whether the difference between the detection data and the historical detection data is larger than a preset threshold value comprises the following steps:
d31 Carrying out segmentation processing on the numerical value in the detection data and the historical detection data, and converting the numerical value into a state quantity by taking a segmentation interval as a name;
d32 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d32 The Boolean amount of the processed historical detection data is regarded as a numerical value to be averaged, the average value is rounded to be an integer, the obtained integer is regarded as the Boolean amount again, and the processed detection data and the historical detection data are sequenced according to the setting to respectively form a detection vector and a historical detection vector;
d33 Calculating the distance between the detection vector and the historical detection vector, comparing the distance with a preset distance threshold, if the distance is larger than the preset distance threshold, judging that the difference between the detection data corresponding to the detection vector and the historical detection data is larger than the preset threshold, otherwise, judging that the difference between the detection data corresponding to the detection vector and the historical detection data is not larger than the preset threshold;
E) Inputting the detection data of the high-voltage circuit breaker to be evaluated and pre-warned into the fault pre-warning neural network model obtained in the step D), and if the fault pre-warning neural network model outputs a fault type, sending out fault pre-warning, wherein the corresponding fault is the fault type output by the fault pre-warning neural network model.
2. The method for evaluating the life cycle and pre-warning faults of a high voltage circuit breaker according to claim 1, wherein,
in the step B), the method for artificially setting the fault source comprises the following steps:
b11 Detecting the high-voltage circuit breaker for a plurality of times;
b12 According to the maintenance requirement of the high-voltage circuit breaker, sequentially selecting one maintenance requirement to ensure that the maintenance requirement does not reach the standard, and detecting for a plurality of times after carrying out a plurality of electrified opening and closing actions;
b13 Sequentially selecting two maintenance requirements to ensure that the maintenance requirements do not reach the standard, and detecting for a plurality of times after carrying out a plurality of charged opening and closing actions; b14 Using liquid nitrogen or dry ice to rapidly cool the high-voltage breaker, performing mechanical property tests for a plurality of times, and obtaining detection data of the mechanical property tests.
3. The method for evaluating the life cycle and pre-warning faults of a high voltage circuit breaker according to claim 1, wherein,
the method for judging whether the difference between the detection data and the historical detection data is larger than a preset threshold value comprises the following steps:
D41 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d42 Normalizing the detection data and the numerical value in the history detection data to obtain the minimum value and the maximum value of each item of the normalized history detection data respectively, arranging the processed Boolean quantity and the numerical value according to a set sequence, ordering the detection data to form a detection vector, ordering each minimum value of the history detection data to form a history detection left vector, and ordering each maximum value of the history detection data to form a history detection right vector;
d43 The distance between the detection vector and the historical detection left vector as well as the distance between the detection vector and the historical detection right vector are calculated respectively, compared with a preset distance threshold, if the distance between the detection vector and the historical detection left vector as well as the distance between the detection vector and the historical detection right vector are larger than the preset distance threshold, the difference between the detection data corresponding to the detection vector and the historical detection data is judged to be larger than the preset threshold, otherwise, the difference between the detection data corresponding to the detection vector and the historical detection data is judged to be not larger than the preset threshold.
4. The method for evaluating the life cycle and the fault pre-warning of the high voltage circuit breaker according to claim 1, wherein the method for judging whether the difference between the detected data and the historical detected data is larger than a preset threshold value comprises the following steps:
D51 Carrying out segmentation processing on the numerical value in the detection data and the historical detection data, and converting the numerical value into a state quantity by taking a segmentation interval as a name;
d52 State quantities in the detection data and the history detection data are converted into boolean quantities, and {0,1} is used to represent false and true, respectively; d52 Deleting the Boolean amounts with different values in the processed historical detection data, sorting the residual Boolean amounts of the historical detection data according to the setting to respectively form historical detection vectors, selecting the Boolean amounts corresponding to the historical detection vectors in the detection data, and sorting according to the setting to form detection vectors;
d53 Calculating the distance between the detection vector and the historical detection vector, comparing the distance with a preset distance threshold, if the distance is larger than the preset distance threshold, judging that the difference between the detection data corresponding to the detection vector and the historical detection data is larger than the preset threshold, otherwise, judging that the difference between the detection data corresponding to the detection vector and the historical detection data is not larger than the preset threshold.
5. The method for life cycle evaluation and fault pre-warning of high voltage circuit breaker according to claim 4, wherein,
in step D51), the method for performing the segmentation processing on the numerical value amounts in the detection data and the history detection data includes:
D511 Selecting a numerical value to obtain all numerical values of the numerical value in the historical detection data, and sequentially arranging the numerical values according to the numerical values, wherein the numerical values are marked as a set Ki, and the minimum value in the set Ki is k min And a maximum value of k max ；
D512 To partition start point k s Giving an initial value of k min Partition endpoint k e Giving an initial value of k max Let the value k be examined m ＝k s +n×Δk, Δk is a step size manually set, n is a positive integer, and n is 1 as an initial value;
d513 N is continuously added with 1, if the value k is examined m The following conditions are satisfied:
wherein the function N (x, y) represents the set Ki, and the number of data values in the numerical interval (x, y) is (2 k) m -k s ) As the interval dividing points and adding the dividing point set Km, the value of (2 k m -k s ) Assignment of the value of k to s Continuously adding 1 to n until k m >k max ；
D514 To k) min And k max Adding a set Km, and dividing the numerical value data into numerical value intervals by using values in the Km as dividing points;
d515 Selecting the next numerical value, and repeating the steps D511) to D514) until all the numerical values are divided into interval segments;
d516 The detection data is divided into sections of a numerical value corresponding to the history detection data.
6. The method for evaluating the life cycle and early warning faults of a high voltage circuit breaker according to claim 4 or 5, wherein in the step D51), the method for converting the numerical value into the state value by using the section as the name comprises the following steps:
D511 Dividing the numerical data into a plurality of intervals, [ n ] m(1) ,n m(2) ],[n m(2) ,n m(3) ]…[n m(k-1) ,n m(k) ]Wherein n is m(1) And n m(k) Respectively a starting point and an ending point of a numerical value interval, n m(2) ～n m(k-1) Dividing the middle of the numerical interval into pointsRespectively used as state names of corresponding numerical value intervals;
d512 If the data of the historical detection numerical value falls into the interval [ n ] m(d) ,n m(d+1) ],d∈[1,k-1]The state name is thenAs the value of the numerical quantity, conversion of the numerical quantity data into state quantity data is completed.
7. The method for evaluating life cycle and early warning faults of a high voltage circuit breaker according to claim 4 or 5, wherein in the step D52), the method for converting the state quantity in the detection data and the historical detection data into the boolean quantity comprises the following steps: d521 Obtaining all state values of the state quantity data;
d522 Splitting the state quantity field into a plurality of fields by taking the state value as a field name;
d523 Setting the field with the same field name and state quantity data value as 1, setting the other split fields as 0, and completing the splitting of the state quantity data into Boolean quantity data.
8. The method for evaluating the life cycle and early warning faults of a high voltage circuit breaker according to claim 4 or 5, wherein in the step D512), the step size Δk setting method comprises the steps of: and calculating the value data of the numerical value in the set Ki, removing the value data of the numerical value data of the set Ki from the value data of the numerical value data to obtain the value data of the numerical value data, performing absolute value operation on the residual value data, taking the minimum value as the step delta k, and participating in calculation.
9. The method for evaluating the life cycle and pre-warning faults of a high voltage circuit breaker according to claim 7, wherein,
in step D521), the method for obtaining all state values of the state quantity data includes: if the state quantity data is the state of the breaker, all state values comprise all possible values of the state; if the state quantity data is the state quantity data converted from the numerical quantity data, all state values only comprise the values which appear in the history state.
10. The method for evaluating the life cycle and pre-warning faults of a high voltage circuit breaker according to claim 1, wherein,
the detection data comprise closing time, opening time, closing speed, opening speed, three-phase different periods, same-phase different periods, golden short time, no-flow time, moving contact maximum speed, moving contact average speed, moving contact action time, bouncing times, bouncing maximum amplitude, opening and closing stroke, opening and closing process current waveform curve, time speed and stroke dynamic curve in the opening and closing stroke of the moving contact, and contact resistance.
11. The method for evaluating the life cycle of a high voltage circuit breaker and pre-warning faults according to claim 1 or 2, wherein in the step D, before training the fault diagnosis neural network model by using the fault sample data, the normalization processing is performed on the fault sample data, which comprises:
D11 The method comprises the steps of) listing numerical data in fault sample data, obtaining a theoretical boundary value of the numerical data as a boundary value, if the theoretical boundary value does not exist, obtaining a historical boundary value of the data as the boundary value, wherein a left boundary value of the boundary value is regarded as 0, a right boundary value of the boundary value is regarded as 1, and dividing a difference of the numerical data minus the left boundary value by a difference of the right boundary value and the left boundary value to obtain a normalized value of the numerical data;
d12 Splitting the state quantity data into a plurality of boolean data;
d13 Boolean data into numerical values and normalized.
12. The method for evaluating life cycle and early warning faults of a high voltage circuit breaker according to claim 11, wherein in the step D12), the method for splitting the state quantity data into a plurality of boolean data comprises:
d121 Obtaining all state values of the state quantity data;
d122 Splitting the state quantity field into a plurality of fields by taking the state value as a field name;
d123 Setting the field with the same field name and state quantity data value, setting the other split fields to zero, and completing the splitting of the state quantity data into Boolean quantity data.
13. A method for evaluating life cycle and early warning faults of a high voltage circuit breaker according to claim 2 or 3, wherein in the step B), before the high voltage circuit breaker is tested, a non-contact displacement sensor is installed on each mechanical moving part of the high voltage circuit breaker, and displacement data measured by the non-contact displacement sensor is added into the test data of the high voltage circuit breaker.
14. The method for evaluating the life cycle and early warning faults of the high-voltage circuit breaker according to claim 13, wherein in the step E), a non-contact displacement sensor is installed on each mechanical moving part of the high-voltage circuit breaker to be evaluated and early warned, displacement data measured by the non-contact displacement sensor are obtained, the state data of the high-voltage circuit breaker to be evaluated and early warned and the displacement data measured by the non-contact displacement sensor are taken as inputs of a fault judging neural network model, the fault judging neural network model is trained, and fault judging is carried out.
15. The method for evaluating the life cycle and early warning faults of the high-voltage circuit breaker according to claim 13, wherein in the step B), a non-contact displacement sensor is arranged on each mechanical moving part of the normal high-voltage circuit breaker, the switching-on and switching-off test is continuously repeated on the high-voltage circuit breaker under the power-off condition until the mechanical parts of the high-voltage circuit breaker are damaged, the switching-on and switching-off times K in the test process and the displacement data of each mechanical moving part in the switching-on and switching-off process are recorded as historical displacement data;
in step E), if the fault judging result of the high-voltage circuit breaker to be judged is no fault, installing a non-contact displacement sensor on each mechanical moving part of the high-voltage circuit breaker to be judged, performing one-time opening and closing of the high-voltage circuit breaker to be judged to obtain displacement data measured by the non-contact displacement sensor, comparing the displacement data with the historical displacement data to obtain the opening and closing test times K corresponding to the closest historical displacement data, and taking (K-K) as the residual service life of the high-voltage circuit breaker to be judged.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910853174.5A CN110705038B (en) | 2019-09-10 | 2019-09-10 | High-voltage circuit breaker life cycle assessment and fault early warning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910853174.5A CN110705038B (en) | 2019-09-10 | 2019-09-10 | High-voltage circuit breaker life cycle assessment and fault early warning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110705038A CN110705038A (en) | 2020-01-17 |
| CN110705038B true CN110705038B (en) | 2023-09-26 |
Family
ID=69196133
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910853174.5A Active CN110705038B (en) | 2019-09-10 | 2019-09-10 | High-voltage circuit breaker life cycle assessment and fault early warning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110705038B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113567845A (en) * | 2021-06-11 | 2021-10-29 | 国网浙江省电力有限公司嘉兴供电公司 | High-voltage circuit breaker life cycle intelligent evaluation method based on deep belief network |
| CN113960465B (en) * | 2021-09-29 | 2023-01-17 | 北汽福田汽车股份有限公司 | Relay detection method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1049761A (en) * | 1989-08-25 | 1991-03-06 | 株式会社日立制作所 | Determine the method and the device of deterioration of electric power line |
| CN106199412A (en) * | 2016-07-01 | 2016-12-07 | 太原理工大学 | A kind of permanent magnet mechanism high-pressure vacuum breaker method of fault pattern recognition |
| CN106291351A (en) * | 2016-09-20 | 2017-01-04 | 西安工程大学 | Primary cut-out fault detection method based on convolutional neural networks algorithm |
| CN109034413A (en) * | 2018-07-11 | 2018-12-18 | 广东人励智能工程有限公司 | Intelligence manufacture equipment fault prediction technique and system based on neural network model |
| CN109470944A (en) * | 2018-04-10 | 2019-03-15 | 国网浙江省电力有限公司嘉兴供电公司 | Protect electric equipment fault early warning method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107884707B (en) * | 2017-09-30 | 2018-12-21 | 杨启蓓 | High-voltage circuit-breaker switching on-off time on-Line Monitor Device and method |
-
2019
- 2019-09-10 CN CN201910853174.5A patent/CN110705038B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1049761A (en) * | 1989-08-25 | 1991-03-06 | 株式会社日立制作所 | Determine the method and the device of deterioration of electric power line |
| CN106199412A (en) * | 2016-07-01 | 2016-12-07 | 太原理工大学 | A kind of permanent magnet mechanism high-pressure vacuum breaker method of fault pattern recognition |
| CN106291351A (en) * | 2016-09-20 | 2017-01-04 | 西安工程大学 | Primary cut-out fault detection method based on convolutional neural networks algorithm |
| CN109470944A (en) * | 2018-04-10 | 2019-03-15 | 国网浙江省电力有限公司嘉兴供电公司 | Protect electric equipment fault early warning method |
| CN109034413A (en) * | 2018-07-11 | 2018-12-18 | 广东人励智能工程有限公司 | Intelligence manufacture equipment fault prediction technique and system based on neural network model |
Non-Patent Citations (2)
| Title |
|---|
| 改进量子神经网络高压断路器故障诊断方法研究;张莲等;《重庆理工大学学报(自然科学)》;20180915(第09期);全文 * |
| 高压断路器机械参数监测与寿命预估方法研究;兰建军等;《东北电力大学学报》;20110215(第01期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110705038A (en) | 2020-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110648315B (en) | High-voltage circuit breaker state evaluation method based on big data technology | |
| CN110705038B (en) | High-voltage circuit breaker life cycle assessment and fault early warning method | |
| CN103292989B (en) | Breaker mechanical characteristic performance testing method and system | |
| CN102590683B (en) | Online diagnosing and early warning method for current-carrying faults of power equipment | |
| Kang et al. | Condition monitoring of power transformer on-load tap-changers. Part 2: Detection of ageing from vibration signatures | |
| US20210278464A1 (en) | Condition assessment method and device for an outdoor post-mounted vacuum switch | |
| CN104081301B (en) | Monitoring of primary devices in a power system | |
| CN107728509A (en) | A kind of breaker mechanic property on-line expert diagnostic system based on Multidimensional Data Model | |
| CN110673024B (en) | High-voltage circuit breaker fault early warning method based on power Internet of things | |
| Oleg et al. | Fault diagnosis of power transformer using method of graphic images | |
| CN116664015A (en) | Intelligent charging pile management system and method thereof | |
| CN113567845A (en) | High-voltage circuit breaker life cycle intelligent evaluation method based on deep belief network | |
| CN110716133B (en) | High-voltage circuit breaker fault studying and judging method based on Internet of things and big data technology | |
| CN111950822B (en) | Performance detection method and device for vehicle traction transformer cooling system, storage medium and terminal | |
| CN115575863A (en) | OLTC (on-line test) voiceprint signal identification method based on phase space trajectory diagram principle | |
| CN110133408A (en) | A kind of High Voltage Circuit Breaker Contacts life appraisal device and method | |
| CN110133459A (en) | A kind of line insulator operating condition ground detection method | |
| CN105466685B (en) | A kind of detection method of breaker actuating spring energy-storage state | |
| Ni et al. | A Real-time Fault Diagnosis Method for Grid-Side Overcurrent in Train Traction System using Signal Time Series Feature Pattern Recognition | |
| CN110703075B (en) | High-voltage circuit breaker quick overhaul method based on big data technology | |
| Lee et al. | Failure prediction system of water treeing using 3D VLF tandelta of MV cable | |
| CN113933757A (en) | Traction transformer overcurrent diagnosis protection device and method | |
| Lee et al. | A condition indexing for assessment management in MV underground power cables | |
| CN112241511B (en) | Performance detection method and device for auxiliary electrical system of vehicle and terminal | |
| Hussain et al. | Early fault identification for operating circuit breaker based on classifier model system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |