High-voltage switch cabinet multi-source state fuzzy evaluation method based on double-layer normal cloud

Abstract

The invention discloses a high-voltage switch cabinet multi-source state fuzzy evaluation method based on double-layer normal cloud, and the method comprises the steps: carrying out the infrared thermal imaging and visible light imaging of a switch cabinet through double-spectrum infrared thermal imaging, so as to generate a temperature distribution and background image of the switch cabinet; the ultrasonic sensor is fixed on a shell of the switch cabinet to collect ultrasonic signals generated by partial discharge, and the transient voltage-to-ground sensor is attached to the switch cabinet to detect transient voltage-to-ground signals of the switch cabinet, measure and generate electrical parameters of an opening and closing coil and measure and generate vibration signals of a circuit breaker; two layers of evaluation models are established, in the first layer, the operation state membership degree of a data set is calculated based on a normal cloud model, the weight of the data set is distributed through an optimal weight algorithm to form an evidence body, the similarity coefficient and the similarity matrix between evidences are calculated, and a new evidence body is constructed; and redistributing the fused uncertain region through the original evidence by rule fusion to obtain an evaluation result and calculate an error.

Description

Technical Field

[0001] The invention relates to the technical field of high-voltage switch cabinet state monitoring, in particular to a fuzzy evaluation method for multi-source states of a high-voltage switch cabinet based on a double-layer normal cloud. Background Art

[0002] High-voltage switchgear is an important equipment in the substation. The switchgear is also one of the important equipment in the power system. The frequent occurrence of insulation flashover, abnormal heating and mechanism refusal to operate failures inside the high-voltage switchgear has a serious negative impact on the safety of the power grid. Its operating status directly affects the power quality and power supply reliability. Maintenance of the switchgear is a necessary means to ensure its safe and reliable operation. The existing online monitoring and live detection methods are limited to a single state quantity of the equipment, and fail to achieve state evaluation and fault diagnosis based on multi-source information. In addition, obtaining the operating status data of the switchgear without affecting the normal operation of the switchgear and using these monitoring data to evaluate the operating status of the switchgear are necessary prerequisites for state maintenance of the switchgear.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a multi-source state fuzzy evaluation method for high-voltage switchgear based on a double-layer normal cloud, which can realize multi-source information monitoring and evaluation of the operating status of the switchgear without affecting the normal operation of the switchgear, thereby significantly improving the evaluation accuracy.

[0005] The purpose of the present invention is achieved through the following technical solutions.

[0006] A fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud includes:

[0007] Dual-spectrum infrared thermal imaging is used to generate temperature distribution and background images of the switch cabinet through infrared thermal imaging and visible light imaging. The ultrasonic sensor is fixed on the shell of the switch cabinet to collect ultrasonic signals generated by partial discharge. The transient voltage-to-ground sensor is attached to the switch cabinet to detect the transient voltage-to-ground signal of the switch cabinet, measure and generate electrical parameters of the opening and closing coils, and measure and generate vibration signals of the circuit breaker.

[0008] Based on the normalization of temperature distribution, background image, ultrasonic signal, transient ground voltage signal, electrical parameters and vibration signal, a two-layer evaluation model is established. In the first layer, the running state membership of the data set is calculated based on the normal cloud model, and the weight of the data set is assigned through the optimal weight algorithm to form an evidence body. In the second layer, Θ is set as the recognition frame, m is the mapping of evidence to [0, 1], and A represents any subset of the recognition frame, which is denoted as And satisfy: m(A) is called the basic trust assignment function BPA of event A, where Reflects the empty set Without generating any credibility, the basic trust allocation function BPA of each piece of evidence in the evidence body is input, and the accuracy of the classifier is used as the weight to modify the basic trust allocation function BPA. The similarity coefficient and similarity matrix between the evidences are calculated, and a new body of evidence is constructed. Through rule fusion, the fused uncertain area is redistributed through the original evidence, and the evaluation result is obtained and the error is calculated.

[0009] In the multi-source state fuzzy assessment method of high-voltage switchgear based on double-layer normal cloud, dual-spectrum infrared thermal imaging adopts dual-channel imaging to synchronously image with dual bands, and the temperature distribution includes the temperature difference between the target object temperature and the surrounding objects based on the background image and the temperature difference between different areas on the surface of the target object and is added to the data set.

[0010] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, the range of the dual-spectrum infrared thermal imaging direction includes at least the busbar and wire sheath pipe, lightning arrester, capacitor, circuit breaker and disconnector of the switchgear.

[0011] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, an ultrasonic sensor is fixed on the outer shell of the switchgear to be detected, and a piezoelectric crystal is used as an acoustic-to-electric conversion element. When discharge occurs inside the switchgear, the sound wave signal generated by the local discharge is transmitted to the surface of the switchgear, and the ultrasonic sensor converts the ultrasonic signal into an electrical signal, which is added to the data set.

[0012] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, in the transient ground voltage sensor, polytetrafluoroethylene is used as the coupling capacitor plate material, the dielectric constant is 1, the thickness is 0.5mm, a thin copper sheet with a thickness of 25μm is placed on the front end of the plate, 50 ohm impedance is matched between the thin copper sheet and the coaxial cable core wire, and the sensor probe is 80mm above the surface of the switchgear.

[0013] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, the electrical parameters include closing time synchronization, opening time synchronization, closing coil current, contact resistance and circuit breaker accumulated electrical wear.

[0014] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, the vibration signal includes the average opening speed, the average closing speed and the closing spring state.

[0015] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, after the vibration signal of the circuit breaker is generated, the vibration signal is denoised based on SVD and the mutation point of the vibration signal is determined, the vibration signal is modally decomposed based on VMD, the amplitude, frequency and attenuation coefficient of each mode are determined and a vibration signal feature set is formed, and the vibration signal of the generated circuit breaker is added to the data set as a vibration signal.

[0016] In the described fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud, in the first layer, in the operation state membership of the data set calculated based on the normal cloud model, the data set is used as a variety of factors affecting the evaluated object U = {u1, u2, ..., um}, let V = {v1, v2, ..., vm} be n comments describing the state of each factor, m is determined by the number of evaluation indicators, n is the number of comments, ω is the weight of the i-th factor u, where ω ≥ 0, and Σω = l, the fuzzy set of the weight distribution of each factor on U is W, and the fuzzy matrix synthesis algorithm is applied to obtain the fuzzy comprehensive evaluation result B: B = WoR = {b1, b2, L, b n}, where o is a fuzzy operator, which is expressed as the synthesis algorithm adopted by the weighted fuzzy set and the evaluation matrix R. According to the maximum membership principle, b1, b2, L, b n The maximum value in is the evaluation result.

[0017] In the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, multiple evidences of the identification frame Θ are represented as E1, E2, …, E k , and its corresponding basic trust allocation function is expressed as m1,m2,…,m k , each focal element is represented by A i (i=1,2,…,n), the accuracy of each classification of the classifier is used as the weight to modify the evidence source and construct the BPA of the evidence body:

[0018]

[0019] In the formula, R j It is in evidence E i After being processed by the classifier, focal element A i The accuracy of R jThe evidence body is weighted and the uncertainty is assigned to the recognition framework Θ to characterize its uncertainty region.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] The multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud of the present invention realizes multi-source information monitoring and evaluation of switchgear operation state without affecting the normal operation of switchgear. The temperature distribution and background image of switchgear are generated by dual-spectrum infrared thermal imaging toward infrared thermal imaging and visible light imaging of switchgear. The ultrasonic sensor is fixed on the shell of switchgear to collect ultrasonic signals generated by partial discharge. The transient voltage-to-ground sensor is attached to the switchgear to detect the transient voltage-to-ground signal of switchgear. The electrical parameters of the opening and closing coils are measured and generated, and the vibration signals of the circuit breaker are measured and generated to construct a real-time multi-state data set. The first layer is based on the subsystem evaluation of the normal cloud model, and the optimal weight algorithm is proposed to allocate the weights of each indicator. Considering the characteristics of randomness and model of switchgear state evaluation, the membership of the operation state of each indicator is calculated. Then, in order to eliminate the influence of subjective factors on the evaluation results, the DS evidence theory is used to perform the evaluation calculation of the second layer, and the evaluation results and calculation errors are obtained to significantly improve the evaluation accuracy. The multi-source parameter monitoring method of switchgear has the advantages of low cost, simple operation, high accuracy, excellent stability, etc., and is suitable for online monitoring and analysis of multi-state parameters of switchgear.

[0022] The above description is only an overview of the technical solution of the present invention. In order to make the technical means of the present invention clearer and to enable those skilled in the art to implement it according to the contents of the specification, and to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described below by way of example. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] By reading the detailed description of the preferred specific embodiments below, various other advantages and benefits of the present invention will become clear to those of ordinary skill in the art. The drawings in the specification are only for the purpose of illustrating the preferred embodiments and are not considered to be limitations of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative work. Moreover, the same reference numerals are used to represent the same components throughout the drawings.

[0024] In the attached picture:

[0025] Figure 1 It is a schematic diagram of the switch cabinet structure of a high-voltage switch cabinet multi-source state fuzzy evaluation method based on a double-layer normal cloud provided by an embodiment of the present disclosure;

[0026] Figure 2 It is a schematic diagram of opening current preprocessing of a high-voltage switchgear multi-source state fuzzy evaluation method based on a double-layer normal cloud provided by another embodiment of the present disclosure.

[0027] Figure 3 It is a schematic diagram of a vibration signal parameter identification process of a multi-source state fuzzy evaluation method of a high-voltage switchgear based on a double-layer normal cloud provided by another embodiment of the present disclosure.

[0028] Figure 4 It is a schematic diagram of a state model of a multi-source state fuzzy evaluation method of a high-voltage switchgear based on a double-layer normal cloud provided by another embodiment of the present disclosure.

[0029] The present invention is further explained below in conjunction with the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0030] The following will refer to the attached Figure 1 To Attachment Figure 4 Specific embodiments of the present invention are described in more detail. Although specific embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0031] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present invention, but the description is based on the general principles of the specification and is not intended to limit the scope of the present invention. The scope of protection of the present invention shall be determined by the attached claims.

[0032] To facilitate understanding of the embodiments of the present invention, further explanation will be given below by taking specific embodiments as examples in conjunction with the accompanying drawings, and each of the accompanying drawings does not constitute a limitation on the embodiments of the present invention.

[0033] For a better understanding, Figure 1-4 As shown, a fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud includes:

[0034] Dual-spectrum infrared thermal imaging is used to generate temperature distribution and background images of the switch cabinet through infrared thermal imaging and visible light imaging. The ultrasonic sensor is fixed on the shell of the switch cabinet to collect ultrasonic signals generated by partial discharge. The transient voltage-to-ground sensor is attached to the switch cabinet to detect the transient voltage-to-ground signal of the switch cabinet, measure and generate electrical parameters of the opening and closing coils, and measure and generate vibration signals of the circuit breaker.

[0035] Based on the normalization of temperature distribution, background image, ultrasonic signal, transient ground voltage signal, electrical parameters and vibration signal, a two-layer evaluation model is established. In the first layer, the running state membership of the data set is calculated based on the normal cloud model, and the weight of the data set is assigned through the optimal weight algorithm to form an evidence body. In the second layer, Θ is set as the recognition frame, m is the mapping of evidence to [0, 1], A Represents any subset of the recognition framework, denoted by And satisfy: m(A) is called the basic trust assignment function BPA of event A, where Reflects the empty set Without generating any credibility, the basic trust allocation function BPA of each piece of evidence in the evidence body is input, and the accuracy of the classifier is used as the weight to modify the basic trust allocation function BPA. The similarity coefficient and similarity matrix between the evidences are calculated, and a new body of evidence is constructed. Through rule fusion, the fused uncertain area is redistributed through the original evidence, and the evaluation result is obtained and the error is calculated.

[0036] In a preferred embodiment of the multi-source state fuzzy evaluation method of a high-voltage switchgear based on a double-layer normal cloud, dual-spectrum infrared thermal imaging adopts a dual-channel imaging method with dual-band synchronous imaging, and the temperature distribution includes the temperature difference between the target object temperature and the surrounding objects based on the background image and the temperature difference between different areas on the surface of the target object and is added to the data set.

[0037] In the preferred implementation of the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, the range of the dual-spectrum infrared thermal imaging direction includes at least the busbar and wire sheath pipe, lightning arrester, capacitor, circuit breaker and disconnector of the switchgear.

[0038] In the preferred implementation of the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, an ultrasonic sensor is fixed on the outer casing of the switchgear to be detected, and a piezoelectric crystal is used as an acoustic-to-electric conversion element. When discharge occurs inside the switchgear, the sound wave signal generated by the local discharge is transmitted to the surface of the switchgear, and the ultrasonic sensor converts the ultrasonic signal into an electrical signal, which is added to the data set.

[0039] In the preferred implementation of the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, in the transient ground voltage sensor, polytetrafluoroethylene is used as the coupling capacitor plate material with a dielectric constant of 1 and a thickness of 0.5 mm. A thin copper sheet with a thickness of 25 μm is placed on the front end of the plate. The impedance between the thin copper sheet and the coaxial cable core wire is matched with 50 ohms, and the sensor probe is 80 mm above the surface of the switchgear.

[0040] In the preferred implementation of the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, the electrical parameters include closing time synchronization, opening time synchronization, closing coil current, contact resistance and circuit breaker accumulated electrical wear.

[0041] In a preferred implementation of the method for fuzzy evaluation of multi-source states of a high-voltage switchgear based on a double-layer normal cloud, the vibration signal includes an average opening speed, an average closing speed and a closing spring state.

[0042] In a preferred implementation of the multi-source state fuzzy evaluation method for a high-voltage switchgear based on a double-layer normal cloud, after generating a vibration signal of a circuit breaker, the vibration signal is denoised based on SVD and the mutation point of the vibration signal is determined, the vibration signal is modally decomposed based on VMD, the amplitude, frequency and attenuation coefficient of each mode are determined and a vibration signal feature set is formed, and the vibration signal of the generated circuit breaker is added to the data set as a vibration signal.

[0043] In the preferred implementation of the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, in the first layer, in the operation state membership of the data set calculated based on the normal cloud model, the data set is used as a variety of factors affecting the evaluated object U = {u1, u2, ..., um}, and V = {v1, v2, ..., vm} is set as n comments describing the state of each factor, m is determined by the number of evaluation indicators, n is the number of comments, ω is the weight of the i-th factor u, where ω ≥ 0, and Σω = l, the fuzzy set of the weight distribution of each factor on U is W, and the fuzzy matrix synthesis algorithm is applied to obtain the fuzzy comprehensive evaluation result B:

[0044] B=WoR={b1,b2,L,b n}, where o is a fuzzy operator, which is expressed as the synthesis algorithm adopted by the weighted fuzzy set and the evaluation matrix R. According to the maximum membership principle, b1, b2, L, b n The maximum value in is the evaluation result.

[0045] In the preferred implementation of the multi-source state fuzzy evaluation method of high-voltage switchgear based on double-layer normal cloud, multiple evidences of the identification frame Θ are represented as E1, E2, ..., E k , and its corresponding basic trust allocation function is expressed as m1,m2,…,m k, each focal element is represented by A i (i=1,2,…,n), the accuracy of each classification of the classifier is used as the weight to modify the evidence source and construct the BPA of the evidence body:

[0046]

[0047] In the formula, R j It is in evidence E i After being processed by the classifier, focal element A i The accuracy of R j The evidence body is weighted and the uncertainty is assigned to Θ to represent its uncertainty region.

[0048] In one embodiment, Figure 1 As shown, the main electrical devices or components inside the switch cabinet are: primary main circuit devices, including busbars, disconnectors, circuit breakers, voltage and current transformers, and lightning arresters. Relays, voltage and current meters, signal lights, transfer switches, microcomputer integrated protection devices and other equipment of the secondary circuit, as well as other components such as pressure relief devices and interlocking devices to prevent misoperation that ensure the safe operation of the switch cabinet. Further, busbar room: the main busbar is fixed by the branch busbar and the static contact box and connected by a single splicing. The main busbar and the connecting busbar are rectangular copper bars, and two busbars are connected side by side when used for high current loads. The branch busbar is connected to the static contact box and the main busbar by bolts, and no other support is required. The busbars of adjacent cabinets are fixed with bushings, so that air will be retained between the connected busbars to buffer the arc. If an arc is generated inside, it can effectively prevent the busbar from melting through. The cabinet bushing can effectively limit the fault to the busbar compartment to prevent the fault from spreading to other places. Circuit breaker trolley room: Specific guide rails are installed inside to facilitate the sliding and working of the circuit breaker trolley. The trolley can slide between the working position and the experimental position. The partition of the static contact is installed on the rear wall of the trolley room, which is a locking device to ensure the five protections. In the process of the trolley sliding from the experimental position to the working position, the partition opens automatically, and the reverse sliding is completely compounded, ensuring the safety of the staff. Cable room: Since the switchgear adopts the central type, the cable room has a large space. The current transformer and the grounding switch are installed on the rear wall, and the lightning arrester is installed at the bottom. After the circuit breaker trolley and the pull-out partition are removed, the construction personnel can install and maintain from the front. The bottom of the cabinet for connecting the cables is equipped with a slotted, unloadable, non-magnetic metal cover or non-metal cover, which is convenient for construction and shortens the construction period. Low-voltage instrument room: Relay protection components, instruments, point detection indicators and some secondary equipment with special requirements are installed inside, and the control lines are laid in a wire trough with sufficient space and equipped with a metal cover to isolate the secondary lines from the high-voltage equipment. The low-voltage instrument room is mainly used to monitor and display some power parameters (such as voltage and current) of the main line of the high-voltage switchgear.

[0049] In one embodiment, in the case of busbar and wire sheath failure, in the long-term operation state, according to experience, the sheath is a defect-prone component, and the defect can be seen through abnormal temperature changes. The main defects of the sheath include poor contact of the conductive wire, lack of oil, surface dirt and dielectric loss.

[0050] In one embodiment, in the infrared detection of the arrester, the arrester can be divided into valve type arrester and zinc oxide arrester according to the composition structure. Obviously, most arresters have to work in an open air environment, and the failure is caused by the poor water insulation effect and internal moisture. Since the power loss of zinc oxide arrester is relatively large, it accelerates the aging of electrical equipment and the internal moisture causes the failure of oxide arrester and switch cabinet.

[0051] In one embodiment, during infrared detection of a capacitor, increasing dielectric loss of the capacitor, severe internal discharge, and lack of oil may cause the temperature of the capacitor to rise.

[0052] In one embodiment, in the infrared detection of the circuit breaker, the circuit breaker in the switch cabinet electrical equipment, that is, the automatic air switch, which is also a switch in the ordinary sense, is mainly divided into two types of faults: insufficient oil and excessive oil. Among them, the causes of the failure of the insufficient oil type circuit breaker are probably poor contact of components, lack of oil, poor contact of contacts, etc. These faults usually lead to temperature rise. The main causes of the failure of the excessive oil circuit breaker are internal moisture, poor contact of internal and external contacts, etc., which will eventually lead to increased internal overcurrent.

[0053] In one embodiment, in the infrared detection of the isolating switch, the direct cause of the overheating defect of the isolating switch is the poor contact between the connecting rod seat and the contact. The main reasons for the poor contact are dust, dirt and rust on the contact surface, as well as the oxidation chemical action of corrosive gas. These will produce impurity films on the contact surface, and ultimately cause the contact surface resistance to increase.

[0054] In one embodiment, when the electromagnetic waves inside the cabinet are transmitted to the surface of the equipment, an induced current will be generated on the surface of the switch cabinet equipment, and the frequency of the induced current is the same as the frequency of the electromagnetic waves that excite it. There is wave impedance on the surface of the equipment, and then a transient voltage to the ground is formed on the outer layer of the equipment, referred to as TEV. The local discharge signal occurring in the insulation structure of the switch cabinet can be regarded as a Gaussian pulse emitted by a point source, propagating in all directions in space, and transmitted to the surface of the cabinet through the discontinuous part of the shield layer to induce surface current and transient voltage to the ground. Since the electromagnetic waves excited by the Gaussian pulse in space are oscillating and attenuated, the surface current and transient voltage to the ground (TEV) excited by the electromagnetic wave on the surface of the conductor are also oscillating and attenuated.

[0055] The TEV sensor uses the capacitive coupling effect and places the sensor on the outer surface of the switch cabinet. At this time, the two form a flat plate capacitor, and the capacitance between the two is:

[0056]

[0057] Where: εr represents the relative dielectric constant of air, d represents the distance between the capacitor plates, and S represents the contact area between the two.

[0058] The sensor detects the TEV signal U as:

[0059]

[0060] Since the surface of the switch cabinet is a flat plate structure, the design structure of the sensor adopts a patch type. In order to simplify the design, polytetrafluoroethylene is selected as the coupling capacitor plate in this design, with a dielectric constant of 1 and a thickness of 0.5mm. A thin copper sheet with a thickness of 25μm is placed on the front end of the plate. 3~200MHz is the frequency band where the energy of the TEV partial discharge signal is mainly concentrated. In the signal conditioning module, the thin copper sheet and the coaxial cable core are matched with a 50 ohm impedance. Within the range of the partial discharge signal concentration, the designed capacitive reactance Xc is equivalent to the impedance of the back-end signal conditioning circuit, which plays a role in suppressing the reflection of high-frequency signal waves. Usually there are:

[0061]

[0062] Where f is the frequency of the original TEV signal. Considering the capacitive impedance matching at 200MHz, C is calculated to be about 15pF. A circular coupling capacitor plate with a radius of 30mm is used. In order to facilitate production, the radius of the coupling capacitor plate is increased to 32.5mm. In order to facilitate installation and operation for technicians, a strong magnet is placed between the bottom copper plate and the surface rubber. The strong magnet makes the sensor close to the surface of the switch cabinet, and the sensor probe is 80mm high.

[0063] In one embodiment, Figure 2 As shown, this method uses cubic spline interpolation to preprocess the coil current to eliminate the burrs contained in the signal collected by the instrument, so that the peaks and troughs of the signal can be reduced as much as possible while maintaining the original characteristics of the signal to the greatest extent, laying a solid foundation for the subsequent signal feature extraction.

[0064] According to the circuit breaker closing and opening coil current, the signal is smoothed by cubic spline interpolation, and the result is as follows: Figure 2 As shown in the figure, it can be seen that after preprocessing, the signal burrs are significantly reduced, which is helpful for the subsequent feature extraction.

[0065] In one embodiment, the local feature quantity of the closing and opening coil current; the selected global feature quantity is the signal mean, standard deviation, energy value, and kurtosis, and these two parts of feature quantity together reflect the working state of the circuit breaker. Extraction of local feature quantity,

[0066] For the extraction of typical time and current values, this method creates a multi-scale maximum value solution method. The solution of local feature quantities is divided into three parts: start time and end time, maximum value and corresponding time point, minimum value and corresponding time point. The solution process is solved according to these three parts: first, the start time and end time are solved. The principle is that the current value before the start time and after the end time fluctuates around zero. The solution process is: first find the maximum value of the opening and closing current signal, then solve all the time points corresponding to the maximum current value less than or equal to 0.01 times, and then perform differential processing on these time points, where the time point with the largest differential value and the time point after it are the start time and end time of the current signal. Then solve the maximum value and the corresponding time point: first solve the signal maximum value with a scale of 1, then transform the scale to solve the maximum value with the lowest height greater than 0.5 times the maximum current value, and then solve the maximum value by grouping the maximum value. Finally, solve the minimum value and the corresponding time point. First, flip the signal, that is, take a negative value for the signal, then the problem of solving the minimum value is transformed into solving the maximum value, and then the time point corresponding to the minimum value can be solved according to the steps of solving the maximum value. Among the global feature quantities, the mean μ can reflect the stability of the signal:

[0067] Where: xi is the value of the random variable.

[0068] The standard deviation σ can describe the degree of dispersion of the data:

[0069]

[0070] Energy W represents the work done by the opening and closing currents: Where: U is the control loop voltage; I is the coil current; T is time. Kurtosis K can reflect the signal distribution characteristics:

[0071]

[0072] When a high-voltage circuit breaker is performing closing and opening operations, the travel-time characteristics of its operating mechanism are important indicators for analyzing the operating status of the circuit breaker operating mechanism. By analyzing the travel-time characteristics, the wear, aging, deformation, etc. of the circuit breaker operating mechanism can be diagnosed. Its travel-time characteristics are also the basis for analyzing the speed of the moving contact during closing and opening operations, and the speed of the moving contact has an important influence on the electrical wear of the circuit breaker. The travel-time characteristics of its operating mechanism are generally extracted by measuring the travel-time signal of the moving contact. Therefore, when detecting the travel-time signal, the installation position of the displacement sensor is crucial to the accuracy of the measurement data.

[0073] Since the moving contact of the high-voltage circuit breaker is located in the arc extinguishing chamber, and the arc extinguishing chamber is sealed, the installation position of the displacement sensor must be carefully analyzed. After analysis, the stroke of the circuit breaker moving contact has a certain relationship with the rotation angle of the operating mechanism main shaft, so the stroke data of the moving contact can be obtained by measuring the rotation displacement of the main shaft after a certain conversion. Since the rotation angular displacement of the main shaft is small, the stroke of the moving contact and its angular displacement are approximately linear. The linear relationship can be expressed as:

[0074] L=Kθ, where L is the travel displacement of the moving contact, θ is the rotational angular displacement of the main shaft, and K is the relevant proportional coefficient, which is determined by the structure of the operating mechanism.

[0075] In one embodiment, the displacement sensor can be installed on the main shaft of the circuit breaker to detect the stroke-time characteristics of the operating mechanism. According to the obtained stroke-time, the average speed of the moving contact during closing and opening can be calculated. At the same time, the stroke-time data of each phase is detected, and the time difference of the moving contact of each phase moving the same displacement is compared by sampling the stroke-time data, and the maximum value of the difference is taken as the time difference of opening and closing. Using ADAMS as the experimental platform, the four states of the circuit breaker, namely normal, spring fatigue, handle jamming and four-link jamming, are simulated by simulation. Compared with the normal state, handle jamming and four-link jamming are equivalent to adding a damping to simulate the jamming state during the closing process, so the closing time will be extended and the closing speed will be reduced; spring fatigue reduces the stiffness coefficient of the contact spring, so although the contact closing start time is advanced, the closing time is extended. White noise with a signal-to-noise ratio of 20dB is added to the signal to simulate the noise and other interferences displayed by the actual signal in the acquisition, and then the signal is used for relevant information processing. The signal under normal state is processed by median smoothing and moving smoothing respectively, and its signal-to-noise ratio is improved to 35.56dB and 38.90dB respectively, that is, the result of moving smoothing is better, so moving smoothing is selected as the signal noise reduction processing.

[0076] In one embodiment, the stroke characteristics of the circuit breaker mainly include the following parts: opening and closing time, moving contact stroke, overstroke, opening and closing different phases, just opening speed, just closing speed, opening and closing maximum speed, opening and closing average speed, etc. It can be known from the physical meaning of overtravel that the time point corresponding to the maximum value of the speed curve is the overtravel time point. According to the stroke-time curve and the corresponding speed curve, this method extracts the characteristic parameters such as the average speed, just closing speed, closing time, stroke, opening distance, overtravel, and overtravel moment of the signal through the difference principle and the maximum value solution principle. The various parameters of the signal are different in the four states, among which the distinction between each parameter in the spring fatigue state and the other three states is very obvious; compared with the normal state, although the average speed and just closing speed are reduced and the closing time is extended when the handle is stuck, the distinction is not large, and the obvious distinguishing parameters are the opening distance and overtravel; when the four-link is stuck, the average speed, just closing speed, closing time, opening distance, overtravel, etc. all have very obvious changes.

[0077] In one embodiment, the vibration signal is a nonlinear, non-stationary superposition signal formed by a series of vibration events such as impacts of electromagnets, connecting rods, contacts, etc. generated during the opening and closing process of the circuit breaker. Therefore, the starting time of each vibration event will mutate on the original signal. The starting time of each signal can be extracted by identifying the mutation point, which is helpful for the analysis of the vibration signal. As an effective signal processing method for analyzing non-stationary and nonlinear signals, singular value decomposition (SVD) has a strong ability to extract transients. Through the reasonable setting of parameters, mutation point information can be obtained, and noise can be significantly reduced, highlighting the signal-to-noise ratio. Therefore, this method chooses to combine the Hankel matrix and singular value decomposition to identify the mutation points and reduce noise of the vibration signal. The specific steps are as follows:

[0078] (1) The vibration signal is represented by X = (x(0), x(1), ..., x(N-1)), and the Hankel matrix is ​​constructed based on the signal:

[0079]

[0080] Wherein: L is the beam parameter, and satisfies the relationship M+(L+1)=N, that is, M=NL-1.

[0081] (2) Perform singular value decomposition on the constructed matrix X1 as follows:

[0082] X1=UΛV T ,

[0083] In the formula: U=(u1,u2,…,u N-L) is an orthogonal matrix of (NL)×(NL); Λ is a diagonal matrix of (NL)×(L+1), where the main diagonal elements are the singular values ​​of X1 [σ1,σ2,…,σ L+1 ]; V = (v1, v2, ..., v L+1 ) is an orthogonal matrix of (L+1)×(L+1). Expanding the above formula gives:

[0084]

[0085] In the formula: m=min((NL),(L+1)) and σ1>σ2>…>σ m , at some point in the middle, the singular value suddenly changes.

[0086] (3) According to the above formula, the following matrix can be decomposed:

[0087]

[0088] Then Xi can be expressed as follows:

[0089]

[0090] (4) All elements in the first row of Xi and the L+1th column of the second row to the NLth row and the L+1th column form Pi:

[0091] P i =[x i.0 ,x i.1 ,…,x i.N-1 ],

[0092] That is, the length of the above formula Pi is N, and analysis shows that:

[0093]

[0094] After the above operation, we get [P1,P2,…,P m ] The vibration signal is divided into the main component and the noise signal, and the middle of the two is the mutation information. When the appropriate value is taken, the mutation signal can be separated from the two.

[0095] Signal separation is closely related to the selection of singular value thresholds. Common threshold processing methods include differential spectrum method, singular value median method and characteristic mean method. This method is based on the singular value norm:

[0096]

[0097] It is easy to know that Ki increases monotonically and gradually approaches 1. When Ki>μ, it means that the first i components contain the main signal of the vibration signal, followed by mutation information and noise information. When the row-column product of the Hankel matrix is ​​the largest, the denoising effect is the best, so the value of L for the vibration simulation signal is 1199. After a large number of simulation calculations, the threshold μ is 0.96. When L=1199, when i=123, Ki is 0.9600, that is, the signal before P123 is reconstructed into a denoised signal. After SVD denoising processing, the noise interference is reduced to a certain extent, the burrs of the signal are reduced, and the five main center frequencies of the signal are highlighted. After the signal is reconstructed, the signal-to-noise ratio increases from the original 10dB to 14.7227dB, which proves the feasibility and effectiveness of SVD denoising.

[0098] In the mutation point identification based on SVD, the mutation point identification of the signal after noise reduction is the key to the selection of the threshold. According to the simulation calculation verification, L is selected as 34 and the threshold μ is 0.995. When L=35 and i=21, Ki is 0.9951. Then, the starting time of the extracted mutation signal is determined by the extreme value method, and five mutation points are obtained. The time is 0.0152s, 0.0193s, 0.0242s, 0.0393s, and 0.0488s, respectively. The maximum error is 3.5%, and the recognition effect is good. In the vibration signal modal decomposition based on VMD, variational mode decomposition (VMD) can decompose the signal into a set of intrinsic mode function components with center frequency. Compared with EMD, which has defects such as mode aliasing and endpoint extension, VMD uses a different decomposition principle. It constructs a non-recursive variational model framework and uses the alternating direction multiplier method to continuously update each mode and the corresponding center frequency through iteration to search for the optimal solution of the model. The core of VMD signal decomposition is to decompose the input signal x into K inherent mode functions mk, each of which revolves around a corresponding center frequency ωk, while minimizing the sum of the estimated bandwidths of each mode. Therefore, it is necessary to construct a variational model for the signal and solve it. The constrained variational model is as follows:

[0099]

[0100] Where: {m k}={m1,m2,…,m K} is a set of K intrinsic mode functions; {m k}={m1,m2,…,m K} is the set of central frequencies of each intrinsic mode function; δ(t) is the impulse function, that is, the Dirac function; is its partial derivative with respect to t; * refers to the convolution operation. VMD introduces the quadratic penalty factor and the Lagrangian factor to construct an unconstrained variational model. η enables the signal to have good reconstruction accuracy, and α is used to ensure the strictness of the constraints. The problem after the introduction is as follows:

[0101]

[0102] In terms of signal processing, compared with EMD and wavelet transform, VMD has better results for signal modal decomposition and good noise robustness. However, VMD is a non-recursive variational model framework. It performs modal solution based on a predetermined number of modes to obtain each mode and the corresponding center frequency. The frequency range of the vibration signal of the high-voltage circuit breaker obtained under actual monitoring is very wide and contains complex components. When analyzing the vibration signal, not all vibration components are captured, but the analysis is carried out for the main vibration collision events. Therefore, the determination of the number of modes K seriously affects the decomposition effect. If K is too large, the mode will be over-decomposed. If K is too small, it will cause mode aliasing. It should be consistent with the number of main vibration events. If it is for the main event of the vibration signal, the mutation point identification mentioned above is to determine the starting time of the main event. The number of modes K can be determined according to the number of mutation points; if it is based on the main frequency displayed in the frequency domain, FFT can better reflect the change of signal frequency over time, and the number of modes can be determined according to the number of peaks in the FFT spectrum. According to the frequency domain graph of the vibration simulation signal and the corresponding mutation point identification mentioned above, it can be seen that the number of peaks displayed by FFT is five and there are five mutation points. The two methods support each other and determine that the number of modes K is 5, which is consistent with the theoretical results, proving the effectiveness of the two methods.

[0103] In one embodiment, the method uses the TLS-ESPRIT method based on the Hankel matrix to separate the signal subspace and the noise subspace of the signal to obtain the amplitude, frequency and attenuation coefficient of the signal. The specific calculation steps are as follows:

[0104] a) The original signal is represented by X = (x(0), x(1), ..., x(N-1)) and the Hankel matrix is ​​constructed based on the signal:

[0105]

[0106] Where: L is the beam parameter, and its size seriously affects the signal anti-noise effect.

[0107] b) Perform singular value decomposition on the constructed matrix X1 as follows:

[0108] X1=UΛV T ,

[0109] In the formula: U=(u1,u2,…,uN-L ) is an orthogonal matrix of (NL)×(NL); Λ is a diagonal matrix of (NL)×(L+1), where the main diagonal elements are the singular values ​​of X1 [σ1,σ2,…,σ L+1 ]; V = (v1, v2, ..., v L+1 ) is an orthogonal matrix of (L+1)×(L+1).

[0110] c) The main diagonal elements of Λ are singular values ​​of X1, and σ1≥σ2≥…≥σ L+1 ≥0, so it is necessary to divide the signal subspace V1 and the noise subspace V2 according to the size of the singular value. If the signal does not contain noise and other interference, the number of non-zero singular values ​​of X1 is i; if the signal contains noise, the singular values ​​are all greater than zero, and it is necessary to solve it according to the principle of singular value decomposition. Here, this method still selects the singular value norm mentioned above as the basis for division, and selects the threshold value as 0.995.

[0111] d) Introduce the total least squares method to solve. Delete the first and last rows of V1 respectively to obtain matrices V3 and V4, which are expressed as follows:

[0112] (V3+e1)=(V4+e2)Φ,

[0113] Where: e1 and e2 are disturbance matrices, representing the noise and other interference in the signal. Therefore, according to the TLS idea, it is necessary to find the optimal Φ to ensure that the overall error ||e1,e2|| is minimized. Therefore, the matrix [V3,V4] is constructed and its singular value decomposition is performed:

[0114] [V3,V4]=RΛZ T ,

[0115] Then Z can be divided into four p×p matrices:

[0116]

[0117] e) Let Φ TLS =Z 11 Z 21 -1 , solve its eigenvalue λ k (k=1,2,…,p) and construct the following matrix:

[0118]

[0119] f) The amplitude, frequency, attenuation coefficient, initial phase and other parameters of the exponentially decaying sinusoidal component can be obtained according to the following formula:

[0120] S=(λ T λ) -1 λ TX1

[0121]

[0122] A k =2|S|

[0123]

[0124] φ k =angle(S),

[0125] Where: fk, Ak, μk, φk are the frequency, amplitude, attenuation coefficient and initial phase of the kth exponentially decaying sinusoidal signal respectively.

[0126] The value of L is related to the effect of noise suppression. When the product of the number of rows NL and the number of columns L+1 of the Hankel matrix is ​​the largest, the main components of the signal and the noise interference will be isolated to the greatest extent, so the value of L should be around N / 2.

[0127] In one embodiment, the specific process of vibration signal feature extraction based on parameter recognition is as follows: Figure 3 As shown in the figure, the singular value (SVD) decomposition is performed, and then the main components of the signal are divided out according to the singular value norm and the corresponding threshold, and then the signal is reconstructed to achieve signal denoising and highlight the main center frequency of the signal. , Secondly, the corresponding Hankel matrix is ​​also constructed according to the vibration denoising signal, and the constructed matrix is ​​decomposed by singular value (SVD). Among them, the singular value norm and the corresponding threshold are very critical. Through a large number of simulation experiments and calculations, the appropriate values ​​are selected. When the signal is divided into the main components and the noise interference, the information mixed between the two is just a mutation signal, which can be used to represent the mutation point of the signal. Among them, a method is used to identify the mutation moment of the signal according to the mutation information, that is, to identify the starting moment of each major vibration event. Next, the denoised vibration signal is subjected to variational mode decomposition (VMD) to obtain k intrinsic mode functions with center frequencies, wherein the spectrum is analyzed by FFT, and it is mutually verified with the mutation moment to determine the number of main vibration events, that is, to determine the k value. Finally, TLS-ESPRIT is used to identify the main parameters of the modes, that is, the amplitude, frequency, and attenuation coefficient of the main vibration events, using each mode decomposed by VMD. The Hankel matrix of each decomposed mode is reconstructed and SVD is performed to divide it into signal subspace and noise subspace, and the signal subspace is used to identify and reconstruct related parameters.

[0128] In one embodiment, the operating state of the switch cabinet is reflected in the following aspects: insulation state, electrical state, mechanical state, state of other auxiliary components, and historical records. Insulation state: The insulation state of the high-voltage switch cabinet is mainly reflected in the busbar phase-to-phase insulation, insulation to the ground, and internal insulation to the switch cabinet housing. Before the insulation accident occurs, the insulation defect part of the switch cabinet often produces partial discharge phenomenon first. The insulation state of the switch cabinet can be detected by detecting the partial discharge phenomenon of the switch cabinet. Common partial discharge detection methods include transient ground voltage (TEV) detection, ultrasonic detection, and detection of the insulation resistance of the busbar through preventive experiments. Power state: The electrical state that reflects the operating state of the switch cabinet includes busbar current, voltage, closing periodicity, opening periodicity, closing motor coil current, contact resistance of contacts, and accumulated electrical wear of circuit breaker contacts. Mechanical state: The mechanical state of the switch cabinet is mainly reflected in the operating mechanism of the internal high-voltage circuit breaker. The average opening speed, average closing speed, closing spring state, and closing pressure of the high-voltage circuit breaker operating mechanism can be selected as evaluation indicators of its mechanical state. Historical records: According to industry standards, the switch cabinet has a certain operating life, and the internal circuit breaker also has a certain breaking number limit. With the extension of the operating life and the increase in the number of circuit breaker breaking times, the operating status of the high-voltage circuit breaker will also decline. Status of other auxiliary components: There are many auxiliary devices inside the switch cabinet, such as lightning arresters, interlocking devices, switch cabinet casings, etc., which will affect the operating status of the switch cabinet.

[0129] In one embodiment, a state assessment model with two levels is established, and the data indicator parameters are normalized and calculated. For the first-level subsystem assessment based on the normal cloud model, an optimal weight algorithm is proposed to allocate the weights of each indicator. Considering the randomness and model characteristics of switch cabinet state assessment, the normal cloud model is selected to calculate the membership of the operating state of each indicator. In one embodiment, Figure 4 As shown in the figure, the first layer is the subsystem evaluation, which includes insulation status, electrical status, mechanical status, historical records, and accessory status. This layer evaluates the status of related subsystems based on the detected index data; the second layer is the overall evaluation of the switchgear, which integrates the evaluation results of the subsystems to evaluate the overall operating status of the switchgear. Among them, H1, H2, H3, and H4 are the four operating status levels of the switchgear. In the first layer, let U = {u1, u2, ..., um} be the m factors affecting the evaluated object, that is, the n evaluation indicators. U is called the factor set of the evaluation object. Let V = {v1, v2, ..., vm} be the n comments describing the status of each factor, that is, n evaluation levels. m is generally determined by the number of evaluation indicators, and n is the number of comments, which is generally divided into 3 to 5 levels.

[0130] Since each factor has different influences on the evaluated object and different weights of each factor in the comprehensive evaluation need to be considered, it is necessary to assign weights to each influencing factor. Assume that ω is the weight of the i-th factor u, where ω≥0, and generally Σω=l. The fuzzy set of the weight distribution of each factor on U is W, that is:

[0131] W={ω1,ω2,L,ω n},

[0132] By applying the fuzzy matrix synthesis algorithm, the fuzzy comprehensive evaluation result B can be obtained:

[0133] B=WoR={b1,b2,L,b n},

[0134] Among them, o is a fuzzy operator, which is expressed as a weighted fuzzy set and the synthesis algorithm used by the evaluation matrix JR. Common synthesis algorithms include: weighted average algorithm, main factor highlighting algorithm, small upper bound and algorithm, etc. Each algorithm has its own scope of application, and it is necessary to select a suitable synthesis algorithm according to the evaluation object. According to the maximum membership principle, b1, b2, L, b n The maximum value in is the evaluation result. The normal cloud model is an uncertainty conversion model between qualitative concepts and quantitative representations. It constitutes a mapping between qualitative and quantitative, and considers two uncertainties of things in the objective world or concepts in human knowledge: fuzziness and randomness, and fully integrates the two. It takes into account both fuzziness and randomness in knowledge representation. This model conforms to the different operating conditions of different switchgears and the fact that the measurement error of state information is random.

[0135] In one embodiment, the method solves the similarity matrix based on the previously corrected evidence body, and uses this as the weight matrix to assign different weights to the evidence body. After normalization, a new evidence body is obtained and fused through the synthesis rule, that is, consistency and conflict are processed at the same time, the evidence source is modified, and the calculation is simple. Finally, in order to avoid the new evidence being too conservative and to improve the certainty of the diagnosis result, the original evidence with the greatest similarity to the new evidence is selected to redistribute the uncertainty area of ​​the new synthesis result. The evidence source obtained by processing the multi-source state monitoring signal of the switch cabinet needs to be corrected for the following three reasons: First, the multi-source state monitoring signals are all measured by sensors, and the external environment will inevitably have a certain impact on the signal; secondly, the same sensor will not measure completely the same signal at different times and under different conditions, and different sensors measure different signals, reflecting different switch cabinet state information, and the importance of the conclusion of the same proposition is also different. In one embodiment, the method uses the accuracy of the classifier classified after signal processing analysis as a weight to correct the evidence source. Assume that the recognition framework is Θ, and there are multiple pieces of evidence under this framework represented as E1, E2,…, Ek The basic trust distribution function corresponding to these evidences is expressed as m1,m2,…,m k , each focal element is represented by A i (i=1,2,…,n), the accuracy of each classification of the classifier is used as the weight to modify the evidence source and construct the BPA of the evidence body:

[0136]

[0137] In the formula, R j It is in evidence E i After being processed by the classifier, focal element A i The accuracy of is used to weight the evidence body and assign the uncertainty to Θ to characterize its uncertainty region, which is also in line with the actual situation that the recognition framework cannot contain all states.

[0138] After determining the basic probability allocation (BPA) based on training analysis, the following steps are used to construct a new body of evidence and further integrate it:

[0139] 1) Calculate the similarity coefficient between evidence bodies:

[0140]

[0141] Where: A k Evidence E i and Evidence E j The intersection of the focal elements

[0142] 2) Construct the similarity matrix between evidences based on the similarity coefficient:

[0143]

[0144] 3) Construct a new evidence matrix based on the similarity matrix:

[0145] m′=S·m,

[0146] Where:

[0147] m=(m i (A1),m i (A2)…m i (A n )) k×n ,i=1,2,…,k,

[0148] m is the original evidence matrix. To satisfy the requirement that the sum of the BPA of each piece of evidence is one, the new evidence matrix is ​​normalized row by row to obtain the evidence matrix m″. The new focal element is:

[0149]

[0150] 4) For the modified evidence matrix m″, Dempster synthesis rule is used to perform evidence fusion, and the obtained evidence fusion result is denoted by M.

[0151] 5) Modify the fusion result M based on the original evidence:

[0152]

[0153] Where: Ej is the evidence with the largest similarity coefficient between the original evidence body and the new evidence body, that is, the evidence is the most reasonable evidence, and mj is the corresponding basic probability distribution. Based on this, the uncertain area of ​​the fusion result is redistributed to improve the certainty of the fusion result.

[0154] This method uses the accuracy of information classification as weights to weight the classification results and assigns uncertainty to the recognition framework, which not only corrects the uncertainty of information processing, but also increases the uncertainty area in line with the actual situation that fails to contain all states. Next, the original source of evidence is corrected, that is, the similarity matrix between the evidence is calculated, and a new source of evidence is established based on this as a weight to solve the conflict between the evidence and make the new source of evidence more credible. In order to reduce the uncertainty interval of the corrected source of evidence, this method selects a group of original evidence with the largest similarity coefficient to redistribute the uncertainty area synthesized by the new evidence to improve the certainty of the evidence. This method not only corrects the errors generated by information processing and the conflict between evidence, but also increases the uncertainty area and improves the accuracy of the results. It is easy to calculate, and the results are more accurate and in line with the actual situation.

[0155] This method conducts a technical analysis of switchgear status assessment based on multi-source parameters, targeting built-in dual-spectrum infrared thermal imaging, built-in ultrasonic and TEV partial discharge, closing and opening coil current, vibration and acoustic monitoring data. According to the characteristics of different parameters, a suitable normalization algorithm is proposed for each indicator data. In addition, due to the complex system structure of the switchgear, the large amount of state information, inaccurate data measurement, and imperfect evaluation criteria, there is a great deal of uncertainty in the switchgear status assessment. On the basis of fully considering the uncertainty problems such as fuzziness and randomness in the evaluation process, a switchgear operation status assessment model with two layers of evaluation is established. The optimal weight method is designed to assign weights to each indicator, the normal cloud model is used to calculate the membership of each state indicator, and the fuzzy comprehensive evaluation algorithm is selected to evaluate the subsystem.

[0156] Industrial Applicability

[0157] The multi-source state fuzzy evaluation method of a high-voltage switch cabinet based on a double-layer normal cloud described in the present invention can be manufactured and used in the technical field of switch cabinet monitoring.

[0158] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0159] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud, comprising: Dual-spectrum infrared thermal imaging is used to generate temperature distribution and background images of the switch cabinet through infrared thermal imaging and visible light imaging. The ultrasonic sensor is fixed on the shell of the switch cabinet to collect ultrasonic signals generated by partial discharge. The transient voltage-to-ground sensor is attached to the switch cabinet to detect the transient voltage-to-ground signal of the switch cabinet, measure and generate electrical parameters of the opening and closing coils, and measure and generate vibration signals of the circuit breaker. Based on the normalization and data set construction of temperature distribution, background image, ultrasonic signal, transient ground voltage signal, electrical parameters and vibration signal, a two-layer evaluation model is established, in which: In the first layer, the running state membership of the data set is calculated based on the normal cloud model, and the weight of the data set is assigned through the optimal weight algorithm to form an evidence body. In the second layer, let Θ be the recognition framework, m is the mapping of evidence to [0, 1], and A represents any subset of the recognition framework, denoted as And satisfy: m(A) is called the basic trust assignment function BPA of event A, where Reflects the empty set Without generating any credibility, the basic trust allocation function BPA of each piece of evidence in the evidence body is input, and the accuracy of the classifier is used as the weight to modify the basic trust allocation function BPA. The similarity coefficient and similarity matrix between the evidences are calculated, and a new body of evidence is constructed. Through rule fusion, the fused uncertain area is redistributed through the original evidence, and the evaluation result is obtained and the error is calculated.

2. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: Dual-spectrum infrared thermal imaging adopts dual-channel imaging to synchronously image in dual bands. The temperature distribution includes the temperature difference between the target object temperature and the surrounding objects based on the background image and the temperature difference between different areas on the surface of the target object and is added to the data set.

3. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: The range of dual-spectrum infrared thermal imaging direction includes at least the busbar and wire sheath pipe of the switch cabinet, lightning arrester, capacitor, circuit breaker and disconnector.

4. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: The ultrasonic sensor is fixed on the outer shell of the switch cabinet to be tested, and uses a piezoelectric crystal as an acoustic-to-electric conversion element. When discharge occurs inside it, the sound wave signal generated by the local discharge is transmitted to the surface of the switch cabinet, and the ultrasonic sensor converts the ultrasonic signal into an electrical signal, which is added to the data set.

5. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: In the transient ground voltage sensor, polytetrafluoroethylene is used as the coupling capacitor plate with a dielectric constant of 1 and a thickness of 0.5 mm. A thin copper sheet with a thickness of 25 μm is placed on the front end of the plate. The impedance between the thin copper sheet and the coaxial cable core wire is matched with 50 ohms, and the sensor probe is 80 mm above the surface of the switch cabinet.

6. The method for fuzzy evaluation of multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: The electrical parameters include closing time synchronization, opening time synchronization, closing coil current, contact resistance and circuit breaker accumulated electrical wear.

7. The method for fuzzy evaluation of multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: The vibration signal includes an average opening speed, an average closing speed and a closing spring state.

8. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: After the vibration signal of the circuit breaker is generated, the vibration signal is denoised based on SVD and the mutation point of the vibration signal is determined. The vibration signal is modally decomposed based on VMD to determine the amplitude, frequency and attenuation coefficient of each mode and form a vibration signal feature set. The vibration signal of the generated circuit breaker is added to the data set as a vibration signal.

9. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: In the first layer, the running status membership of the data set is calculated based on the normal cloud model. The data set is a variety of factors that affect the evaluated object, U = {u1, u2, ..., um}, and V = {v1, v2, ..., vm} is the n comments describing the status of each factor, m is determined by the number of evaluation indicators, n is the number of comments, ω is the weight of the i-th factor u, where ω ≥ 0, and Σω = l, the fuzzy set of the weight of each factor on U is W, and the fuzzy matrix synthesis algorithm is applied to obtain the fuzzy comprehensive evaluation result B: B=WoR={b1,b2,L,b n }, where o is a fuzzy operator, which is expressed as the synthesis algorithm adopted by the weighted fuzzy set and the evaluation matrix R. According to the maximum membership principle, b1, b2, L, b n The maximum value in is the evaluation result.

10. The fuzzy evaluation method for multi-source state of high-voltage switchgear based on double-layer normal cloud according to claim 1, wherein: The multiple evidences of the recognition frame Θ are denoted as E1,E2,…,E k , the corresponding basic trust allocation function is represented by m1,m2,…,mk, and each focal element is represented by A i (i=1,2,…,n), the accuracy of each classification of the classifier is used as the weight to modify the evidence source and construct the BPA of the evidence body: In the formula, R j It is in evidence E i After being processed by the classifier, focal element A i The accuracy of R j The evidence body is weighted and the uncertainty is assigned to the recognition framework Θ to characterize its uncertainty region.

Images