A switch cabinet monitoring method, system, intelligent terminal and storage medium

Abstract

The application discloses a kind of switch cabinet monitoring method, system, intelligent terminal and storage medium, it is related to equipment state monitoring technical field, the method includes: when switching, vibration signal is monitored by vibration sensor, obtains vibration signal sequence;The history operation data sequence of switch cabinet is collected and analyzed, obtains aging leakage parameter and aging loose parameter;According to vibration signal sequence, real-time leakage parameter and loose parameter are obtained by analysis, according to aging leakage parameter, aging loose parameter configuration compensation correction coefficient, real-time leakage parameter and loose parameter are compensated and corrected and coupled superposition calculation processing, obtain leakage parameter and loose parameter;The aging proportion coefficient of aging leakage parameter and aging loose parameter is calculated, switch cabinet abnormality recognition precision is configured, according to leakage parameter and loose parameter, abnormal probability is obtained as monitoring result by identification.This application solves the technical problem of poor accuracy of switch cabinet monitoring in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of equipment state monitoring, and in particular to a switch cabinet monitoring method and system, an intelligent terminal and a storage medium. BACKGROUND

[0002] A switch cabinet is a commonly used electrical equipment, mainly used for opening and closing, controlling and protecting electrical equipment in the process of power generation, power transmission, power distribution and power conversion in a power system. As a key equipment of the power system, the internal arc-extinguishing medium leakage or mechanical component loosening of the switch cabinet can cause serious faults. The traditional monitoring method mainly collects real-time signals through smoke or arc sensors to make abnormality judgments. However, this kind of method has obvious deficiencies: on the one hand, the aging of the switch cabinet is a long-term cumulative process (such as slow gas leakage caused by deterioration of sealing materials, gradual loosening of mechanical connecting parts), and it is difficult to distinguish between transient interference and real aging trend by relying only on single real-time vibration signal analysis, and false judgments are easy to occur due to environmental noise or accidental impact; on the other hand, the performance of the sensor itself will decay with the use time, and the existing technology lacks a dynamic compensation mechanism for sensor errors, resulting in a decrease in the reliability of long-term monitoring data. In addition, since the historical operation data (such as the aging law of similar devices and historical fault characteristics) are not fused, the traditional monitoring method cannot quantify the influence of long-term aging on the current state, and it is difficult to predict the occurrence window of progressive faults, and often only triggers an alarm when the abnormality is obvious, which cannot predict the occurrence of electrical abnormalities in the early stage, resulting in the loss of the opportunity for proactive maintenance and causing safety problems. SUMMARY

[0003] The present application provides a switch cabinet monitoring method, system, intelligent terminal and storage medium, which is used to solve the technical problem that the existing technology cannot fuse historical aging trends and real-time monitoring data to achieve accurate switch cabinet monitoring.

[0004] In view of the above problems, the present application provides a switch cabinet monitoring method, system, intelligent terminal and storage medium.

[0005] In the first aspect, the present application provides a switch cabinet monitoring method, which comprises: during the operation of the switch cabinet, monitoring the vibration signal in the switch cabinet through a vibration sensor when the switch operates, to obtain a vibration signal sequence;

[0006] Collecting a historical operation data sequence of the switch cabinet, performing arc-extinguishing medium aging leakage analysis and component aging loosening analysis, and obtaining aging leakage parameters and aging loosening parameters;

[0007] According to the vibration signal sequence, real-time arc-extinguishing medium leakage analysis and real-time component loosening analysis are performed to obtain real-time leakage parameters and real-time loosening parameters; compensation correction coefficients are configured according to the aging leakage parameters and the aging loosening parameters; and the real-time leakage parameters and the real-time loosening parameters are compensated, corrected, and coupled and superimposed to obtain leakage parameters and loosening parameters.

[0008] According to the compensation correction coefficients and the aging proportion coefficients of the aging leakage parameters and the aging loosening parameters in the coupled superimposed calculation, switch cabinet abnormality recognition accuracy is configured; switch cabinet abnormality probability recognition is performed according to the leakage parameters and the loosening parameters; and an abnormality probability is obtained as a monitoring result.

[0009] In a second aspect, the present application provides a switch cabinet monitoring system, comprising:

[0010] A signal monitoring module is configured to monitor vibration signals in the switch cabinet through a vibration sensor when a switching action is performed during switch cabinet operation to obtain a vibration signal sequence.

[0011] An aging analysis and calculation module is configured to collect a historical operation data sequence of the switch cabinet, perform arc-extinguishing medium aging leakage analysis and component aging loosening analysis, and obtain aging leakage parameters and aging loosening parameters.

[0012] A real-time analysis and calculation module is configured to perform arc-extinguishing medium real-time leakage analysis and component real-time loosening analysis according to the vibration signal sequence to obtain real-time leakage parameters and real-time loosening parameters; compensation correction coefficients are configured according to the aging leakage parameters and the aging loosening parameters; and the real-time leakage parameters and the real-time loosening parameters are compensated, corrected, and coupled and superimposed to obtain leakage parameters and loosening parameters.

[0013] A monitoring result analysis module is configured to configure switch cabinet abnormality recognition accuracy according to the compensation correction coefficients and the aging proportion coefficients of the aging leakage parameters and the aging loosening parameters in the coupled superimposed calculation; perform switch cabinet abnormality probability recognition according to the leakage parameters and the loosening parameters; and obtain an abnormality probability as a monitoring result.

[0014] In a third aspect, the present application provides a switch cabinet monitoring intelligent terminal, comprising:

[0015] A memory is configured to store a first computer program.

[0016] A processor is configured to read and execute the first computer program, thereby realizing the switch cabinet monitoring method of the first aspect.

[0017] In a fourth aspect, the present application provides a non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores a second computer program, and the second computer program is executed by a processor to realize the switch cabinet monitoring method of the first aspect.

[0018] The one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0019] The application provides a switch cabinet monitoring method, system, intelligent terminal and storage medium. Through fusion analysis of historical operation data and real-time vibration signals, dynamic compensation correction coefficient configuration and abnormal recognition precision adjustment driven by aging proportion coefficient, the accuracy and early warning capability of switch cabinet fault recognition are significantly improved. Compared with the traditional method, the technical solution provided by the application uses the cloud-deployed aging analyzer and real-time analyzer to extract irreversible aging parameters from the historical vibration signals, and captures recoverable deformation parameters from the current vibration signals. Combined with the dynamic compensation correction of the sensing error table, the hardware error and the real fault feature are effectively separated. Based on the aging proportion coefficient, the number of abnormal recognition branches is dynamically adjusted, the calculation efficiency and recognition accuracy of the model are optimized, and the switch cabinet at different aging stages can adapt to the optimal monitoring strategy. The technology provided by the application significantly overcomes the single data dimension dependence problem and the sensor error accumulation problem. The integrated learning multi-branch BP neural network is used to predict the abnormal probability, the abnormal probability of multiple outputs is output, the progressive fault is actively predicted and the interference signal is intelligently filtered, and the accuracy and robustness of the switch cabinet abnormal monitoring are improved.

[0020] The application achieves the technical effect of fusing historical aging trend and real-time monitoring data to realize accurate, effective and timely switch cabinet monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 A flowchart of a switch cabinet monitoring method provided by an embodiment of the application is shown in the figure.

[0023] Figure 2 A structural diagram of a switch cabinet monitoring system provided by an embodiment of the application is shown in the figure.

[0024] Figure 3 A structural diagram of a switch cabinet monitoring intelligent terminal provided by an embodiment of the application is shown in the figure.

[0025] Figure 4 A structural diagram of a storage medium provided by an embodiment of the application is shown in the figure.

[0026] In the drawings, the components represented by the respective reference numerals are described as follows:

[0027] The signal monitoring module 100, the aging analysis calculation module 200, the real-time analysis calculation module 300, the monitoring result analysis module 400, the intelligent terminal 500, the memory 510, the processor 520, the first computer program 511, the computer readable storage medium 600, and the second computer program 611. DETAILED DESCRIPTION

[0028] The application provides a switch cabinet monitoring method and system, an intelligent terminal and a storage medium, which are used to solve the technical problem that the existing switch cabinet monitoring cannot combine historical aging trends with real-time monitoring data to achieve accurate monitoring.

[0029] The technical solutions in the embodiments of the application will be clearly and completely described in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the application.

[0030] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those clearly listed steps or units, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0031] In one embodiment, as shown in the accompanying drawings, the application provides a switch cabinet monitoring method, which comprises the following steps: Figure 1

[0032] S10: When a switching action is performed in the running process of the switch cabinet, a vibration sensor is used to monitor the vibration signal in the switch cabinet to obtain a vibration signal sequence.

[0033] In the running process of the switch cabinet, the mechanical vibration generated has instantaneous and high dynamic characteristics. The traditional switch cabinet monitoring method only collects the vibration signal at a single time point when the switching action is performed, and cannot form continuous time sequence data, which leads to the inability to capture the dynamic change characteristics of the vibration signal, such as the vibration attenuation law in the process of opening and closing. In addition, the instantaneous signal is easily disturbed by environmental noise, lacks time sequence correlation analysis, and it is difficult to distinguish between normal mechanical action and abnormal vibration. In continuous monitoring, there is a problem of signal loss or waveform distortion. At the same time, electromagnetic interference in the running environment also causes signal baseline drift.

[0034] The step S10 in the method provided in the embodiments of the application comprises the following steps: ​

[0035] During the switching operation of the switch cabinet, the vibration signals in the switch cabinet are monitored through the vibration sensor;

[0036] The vibration signals obtained through monitoring are recorded in time sequence to obtain a vibration signal sequence.

[0037] Exemplarily, the vibration signals of the switch cabinet are monitored by using the optical fiber vibration sensor. When the switch cabinet is in the process of opening and closing, vibration will be generated. The optical fiber vibration sensor is used to quantitatively collect and characterize the vibration signals, and the vibration acceleration of the switch cabinet during the operation is collected, which is in units of m / s². The optical fiber vibration sensor is a vibration sensor that uses light waves as signal carriers and uses optical fibers as light wave transmission channels. It has the advantages of high sensitivity, strong safety, anti-electromagnetic interference, high insulation strength, and long-distance transmission.

[0038] The vibration signals obtained by the vibration sensor are recorded in time sequence. Specifically, each monitoring signal is given a time stamp, such as 2025.5.21.15:00. Then, the vibration signals are sorted according to the time stamp to obtain a vibration signal sequence.

[0039] The embodiment of the present application records the vibration signals in time sequence, completely retains the vibration characteristics of the whole process of switching, such as amplitude change and frequency distribution, and provides high-fidelity original data for subsequent analysis. At the same time, the use of optical fiber sensors can also prevent environmental electromagnetic interference from affecting signal collection. Sorting the vibration signals in time sequence can effectively suppress single-point sampling noise, verify the effectiveness of the signals through waveform continuity, avoid misjudgment caused by abnormal transient signal sampling, and provide accurate data support for subsequent analysis.

[0040] S20: Collecting a historical operation data sequence of the switch cabinet, performing arc-extinguishing medium aging leakage analysis and component aging loosening analysis, and obtaining aging leakage parameters and aging loosening parameters;

[0041] The vibration accumulated in the historical time will cause the leakage of arc-extinguishing media such as SF6 and the loosening of components (such as screws and sensors), which will further become the aging characteristics of the device. The aging characteristics have cumulative effect and nonlinear decay characteristics, and existing analysis methods are difficult to accurately separate irreversible aging caused by vibration and recoverable deformation. At the same time, the noise accumulation of long-term operation data will mask the real aging trend, and conventional filtering methods are easy to cause feature information loss. It is an urgent problem to extract key parameters representing material fatigue and structural degradation from multi-source heterogeneous historical data and establish an accurate aging quantification model.

[0042] The step S20 in the method provided by the embodiment of the present application includes:

[0043] According to the monitoring log of the switch cabinet, a vibration signal sequence of the switch cabinet in a historical time is collected, and a historical vibration signal sequence set is obtained;

[0044] According to the operation and maintenance data of the same type of switch cabinet, an aging analyzer is trained and configured in the cloud, wherein the aging analyzer includes an arc-extinguishing medium aging analysis branch and a component aging analysis branch, the input data of the arc-extinguishing medium aging analysis branch and the component aging analysis branch is the sample historical vibration signal sequence set, and the output data is respectively a sample aging leakage parameter and a sample aging loosening parameter;

[0045] The historical vibration signal sequence set is input into the trained aging analyzer, and an aging leakage parameter and an aging loosening parameter are obtained.

[0046] In the embodiment of the application, according to the monitoring log of the switch cabinet, a vibration signal sequence of the switch cabinet in a historical time is collected, and a historical vibration signal sequence set is obtained.

[0047] According to the historical operation and maintenance data of the same type of switch cabinet, an aging analyzer is trained. Exemplarily, a dual-branch time sequence convolutional network is used to construct an aging analyzer model. The dual-branch time sequence convolutional network (DB-TCN, Dual-Branch Temporal Convolutional Network) is a kind of convolutional neural network, which is designed by double branches, the branches are arc-extinguishing medium aging analysis branch and component aging analysis branch, can share bottom feature extraction layer, improve model efficiency. Exemplarily, the aging analyzer is constructed, including input layer, shared feature extraction layer, branch layer and output layer. Among them, the input layer is set to batch size 32, single channel 1000 point vibration signal, used to receive the input historical vibration signal sequence set; the shared feature extraction layer is divided into two layers, the first layer is causal convolutional layer, containing 64 convolutional kernels, using ReLU as the activation function, the second layer is dilated convolutional layer, the dilation factor is 2, the residual connection uses jump connection; the branch layer contains two branches, arc-extinguishing medium aging analysis branch and component aging analysis branch, wherein the arc-extinguishing medium aging analysis branch contains three layers, the global pooling layer output dimension is (32, 64), the number of neurons in the full connection layer is 32, and the activation function uses ReLU; the number of neurons in the output layer is 1, and the aging leakage parameter is output; the component aging analysis branch contains three layers, the global pooling layer output dimension is (32, 64), the number of neurons in the full connection layer is 32, and the activation function uses ReLU; the number of neurons in the output layer is 1, and the aging loosening parameter is output; NAdam optimizer is used for optimization.

[0048] The aging analyzer is trained by using a supervised training method, pre-training is performed first, and then the whole model is fine-tuned, exemplarily, a sample historical vibration signal sequence set is input, the output layer is first frozen, only the shared layer and the branch layer are trained to learn the general time sequence feature expression of the vibration signal, then all layers are unfrozen, the shared feature extraction layer and the branch layer parameters are jointly optimized to adapt to the specific aging parameter regression task, and finally the aging leakage parameter and the aging loosening parameter are output. The aging leakage parameter is the pressure drop caused by aging leakage in the running cycle, and the unit is KPa. The aging loosening parameter is the loosening displacement of the element aging loosening, and the unit is mm. The aging analyzer is supervised trained, the model parameters are adjusted until the model converges, for example, the accuracy of the aging leakage parameter and the aging loosening parameter output by the historical vibration signal sequence set is more than 95%, and the training of the aging analyzer is completed.

[0049] The trained aging analyzer is configured in the cloud and identified through Internet of Things transmission. The historical vibration signal sequence set is input, and the aging leakage parameter and the aging loosening parameter are output.

[0050] In the embodiment of the application, based on the historical vibration data, the aging analyzer model is trained, the long-term cumulative aging effect is converted into quantifiable parameters (aging leakage parameter and aging loosening parameter), and the contribution of the aging condition to the current state is determined. Through the training method of pre-training and then fine-tuning the whole model, a universal aging law model is established to provide a traceable benchmark reference for real-time monitoring. The model is deployed in the cloud, which saves the complexity of deploying operation equipment locally, and only simple sensor and single-chip microcomputer deployment is required for the switch cabinet, so that the switch cabinet monitoring method is more lightweight and convenient, and is suitable for diversified scenes.

[0051] S30: According to the vibration signal sequence, real-time leakage analysis and real-time loosening analysis of the arc extinguishing medium are performed to obtain real-time leakage parameters and real-time loosening parameters, compensation correction coefficients are configured according to the aging leakage parameters and the aging loosening parameters, and the real-time leakage parameters and the real-time loosening parameters are compensated, corrected and coupled and superimposed to obtain leakage parameters and loosening parameters.

[0052] In real-time monitoring, real-time vibration signals are easily affected by sensor installation loosening such as screw micro-displacement and environmental interference such as electromagnetic noise, and the traditional method does not consider the error transmission of the real-time data caused by the state of the sensor itself. The real-time parameter fluctuation caused by elastic deformation has a time-varying coupling relationship with permanent aging, and existing compensation algorithms are also difficult to realize dynamic error correction. At the same time, the recoverable deformation (such as elastic vibration) and the irreversible aging (such as plastic deformation) are not distinguished, which may cause parameter aliasing, the difference in physical dimension between the historical cumulative parameters and the real-time parameters may cause system error when directly superimposed, and the normalization processing and fusion calculation of multiple source parameters need to be solved.

[0053] The step S30 in the method provided in the embodiment of the application comprises:

[0054] According to the operation and maintenance data of the switch cabinet of the same type, a real-time analyzer is trained and configured in the cloud, wherein the real-time analyzer comprises an arc-extinguishing medium real-time analysis branch and an element real-time analysis branch, input data of the arc-extinguishing medium real-time analysis branch and the element real-time analysis branch is a sample vibration signal sequence, and output data is respectively a sample real-time leakage parameter and a sample real-time loosening parameter;

[0055] The vibration signal sequence is input into the real-time analyzer, and a real-time leakage parameter and a real-time loosening parameter are obtained by output.

[0056] The aging loosening parameter is input into a sensing error table, a sensing error coefficient is obtained by mapping configuration, and a compensation correction coefficient is calculated, wherein the sensing error table is constructed by using a mapping relationship between a sample loosening parameter and a sample sensing error coefficient.

[0057] According to the compensation correction coefficient, an amplified compensation correction calculation is performed on the real-time leakage parameter and the real-time loosening parameter, and a corrected real-time leakage parameter and a corrected real-time loosening parameter are obtained.

[0058] According to the corrected real-time leakage parameter, the corrected real-time loosening parameter, and the aging leakage parameter and the aging loosening parameter, a coupling superposition calculation is performed, and a leakage parameter and a loosening parameter are obtained.

[0059] In the embodiment of the application, a double-branch time sequence convolution network is used to construct a real-time analyzer model, the real-time analyzer model is trained based on the same invention idea of step S20, and a real-time analyzer is constructed, which comprises an input layer, a shared feature extraction layer, a bifurcation layer, and an output layer. The input layer is set to batch size 32, single channel 1000 point vibration signal, and is used to receive an input historical vibration signal sequence set; the shared feature extraction layer is divided into two layers, the first layer is a causal convolution layer, contains 64 convolution kernels, and uses ReLU as an activation function, and the second layer is a dilated convolution layer, with a dilated factor of 2, and uses a jump connection for residual connection; the bifurcation layer contains two branches, an arc-extinguishing medium real-time analysis branch and an element real-time analysis branch, wherein the arc-extinguishing medium real-time analysis branch contains three layers, the output dimension of the global pooling layer is (32, 64), the number of neurons of the full connection layer is 32, and ReLU is used as the activation function; the number of neurons of the output layer is 1, and the real-time leakage parameter is output; the element real-time analysis branch contains three layers, the output dimension of the global pooling layer is (32, 64), the number of neurons of the full connection layer is 32, and ReLU is used as the activation function; the number of neurons of the output layer is 1, and the real-time loosening parameter is output; and an NAdam optimizer is used for optimization.

[0060] The real-time analyzer is trained by using a supervised training method, pre-training is performed for feature extraction, and then the whole model is fine-tuned, exemplarily, a sample historical vibration signal sequence set is input, the output layer is first frozen, only the shared layer and the branch layer are trained to learn the general time sequence feature expression of the vibration signal, then all layers are unfrozen, the shared feature extraction layer and the branch layer parameters are jointly optimized to adapt to the specific real-time parameter regression task, and finally the real-time leakage parameter and the real-time loosening parameter are output. The real-time leakage parameter is the pressure drop caused by real-time leakage in the running cycle, and the unit is KPa. The real-time loosening parameter is the loosening displacement of the element in real time, and the unit is mm. The real-time analyzer is supervised trained, the model parameters are adjusted until the model converges, for example, the accuracy of the real-time leakage parameter and the real-time loosening parameter output by the historical vibration signal sequence set is more than 95%, and the training of the real-time analyzer is completed.

[0061] The trained real-time analyzer is configured in the cloud and identified through Internet of Things transmission. The vibration signal sequence is input, and the real-time leakage parameter and the real-time loosening parameter are output.

[0062] A sensing error table is constructed. The sensing error table is a kind of pre-constructed mapping relationship table, which can associate the aging loosening parameter with the sensor measurement error. Through experiments or historical data accumulation, the sensor measurement error corresponding to different loosening degrees is obtained, the mapping relationship between the sample loosening parameter and the sample sensor error coefficient is used to construct the sensing error table, and the loosening parameter is positively correlated with the sensor error. Exemplarily, the sensor error coefficient corresponding to the aging loosening parameter of 5 mm is set to ±10%, and the sensor error coefficient corresponding to 10 mm is set to ±20%. According to the sensor error coefficient, a compensation correction coefficient is configured, and the compensation correction coefficient is equal to the absolute value of the sensor error coefficient. Exemplarily, the sensor error is ±10%, and the compensation correction coefficient = |sensor error coefficient| = |±10%| = 10%.

[0063] According to the compensation correction coefficient, the real-time leakage parameter and the real-time loosening parameter are amplified and compensated for correction calculation, the leakage and loosening are redundantly compensated, and the possible electrical abnormality probability can be found in time. The corrected real-time leakage parameter = (1 + compensation correction coefficient) x real-time leakage parameter, and the corrected real-time loosening parameter = (1 + compensation correction coefficient) x real-time loosening parameter. Exemplarily, the compensation correction coefficient is 10%, the real-time leakage parameter is 0.5 KPa, and the real-time loosening parameter is 10 mm. Then the corrected real-time leakage parameter = (1 + 10%) x 0.5 = 0.55 KPa, and the corrected real-time loosening parameter = (1 + 10%) x 10 = 11 mm.

[0064] According to the corrected real-time leakage parameter, the corrected real-time looseness parameter, the aging leakage parameter and the aging looseness parameter, coupling superposition calculation is performed to obtain the leakage parameter and the looseness parameter. The corrected real-time leakage parameter and the aging leakage parameter are added to obtain the leakage parameter. For example, the corrected real-time leakage parameter is 0.55 KPa, the aging leakage parameter is 0.45 KPa, and then the leakage parameter = corrected real-time leakage parameter + aging leakage parameter = 0.55 + 0.45 = 1 KPa. The corrected real-time looseness parameter and the aging looseness parameter are added to obtain the looseness parameter. For example, the corrected real-time looseness parameter is 11 mm, the aging looseness parameter is 9 mm, and then the looseness parameter = corrected real-time looseness parameter + aging looseness parameter = 11 + 9 = 20 mm.

[0065] The embodiments of the present application compensate for the performance degradation of the sensor dynamically by constructing the sensing error table, eliminate the interference of hardware errors on real-time parameters, perform redundant compensation on leakage and looseness, amplify data correction interference, improve abnormal detection sensitivity, and facilitate timely discovery of possible electrical abnormality probability. The corrected real-time parameter reflects recoverable elastic deformation, the aging parameter reflects irreversible cumulative degradation, coupling superposition calculation recovers deformation and irreversible aging influence, and can realize precise characterization of device state in both instantaneous and long-term dimensions.

[0066] S40: According to the compensation correction coefficient and the aging proportion coefficient of the aging leakage parameter and the aging looseness parameter in coupling superposition calculation, the switching cabinet abnormality recognition accuracy is configured, the switching cabinet abnormality probability recognition is performed according to the leakage parameter and the looseness parameter, and the abnormality probability is obtained as a monitoring result.

[0067] The traditional abnormality recognition model adopts a fixed threshold strategy and cannot adapt to state feature changes in different aging stages. The dynamic relationship between the compensation correction coefficient and the aging proportion is not quantitatively modeled, resulting in a contradiction between recognition accuracy and false alarm rate. The existing probability recognition method does not consider the spatiotemporal correlation characteristics between parameters and only uses a single parameter triggering mechanism, which is easy to cause missed detection, such as early fault false negative or late device false positive.

[0068] The step S40 in the method provided by the embodiments of the present application includes:

[0069] The compensation correction coefficient is taken as the first abnormality recognition accuracy.

[0070] The proportions of the aging leakage parameter and the aging looseness parameter in coupling superposition calculation are calculated, and the mean value is calculated to obtain the aging proportion coefficient as the second abnormality recognition accuracy.

[0071] According to the first abnormality recognition accuracy and the second abnormality recognition accuracy, the abnormality recognition accuracy is calculated and obtained.

[0072] In the pre-trained switch cabinet anomaly identifier, an anomaly identification branch with a proportion of the anomaly identification accuracy is called, and the leakage parameter and the loosening parameter are input respectively to identify and output multiple branch anomaly probabilities, and the average is calculated to obtain an anomaly probability as a monitoring result.

[0073] The pre-training step of the switch cabinet anomaly identifier includes:

[0074] According to the monitoring log of the switch cabinet, a sample leakage parameter set and a sample loosening parameter set are collected, and the proportion of the switch cabinet failure anomaly under different sample leakage parameter and sample loosening parameter combinations is collected and labeled as a sample branch anomaly probability to obtain a sample branch anomaly probability set.

[0075] The sample leakage parameter set, the sample loosening parameter set, and the sample branch anomaly probability set are randomly divided multiple times to obtain multiple anomaly identification training data.

[0076] A plurality of anomaly identification branches are constructed using a BP neural network, and the multiple anomaly identification training data are used to supervise and train and verify the plurality of anomaly identification branches, and the anomaly identification branches are configured in the cloud after convergence to obtain a switch cabinet anomaly identifier.

[0077] In the embodiments of the present application, the compensation correction coefficient is taken as the first anomaly identification accuracy. For example, when the compensation correction coefficient is 10%, the first anomaly identification accuracy = compensation correction coefficient = 10%.

[0078] The proportions of the aging leakage parameter and the aging loosening parameter in the coupling superposition calculation are calculated, and the average is calculated to obtain an aging proportion coefficient as the second anomaly identification accuracy. For example, in the coupling superposition calculation, the aging leakage parameter is 0.5 KPa, and the corrected real-time leakage parameter is 0.5 KPa, so the proportion of the aging leakage parameter = aging leakage parameter ÷ (aging leakage parameter + corrected real-time leakage parameter) 0.5 ÷ (0.5 + 0.5) = 50%; Similarly, in the coupling superposition calculation, the aging loosening parameter is 7 mm, and the corrected real-time loosening parameter is 3 mm, so the proportion of the aging loosening parameter = aging loosening parameter ÷ (aging loosening parameter + corrected real-time loosening parameter) = 7 ÷ (7 + 3) = 70%. The average of the proportions of the aging leakage parameter and the aging loosening parameter is taken as the aging proportion coefficient, for example, the aging leakage parameter proportion is 50%, the aging loosening parameter proportion is 70%, and the aging proportion coefficient = (aging leakage parameter + aging loosening parameter) ÷ 2 = (50% + 70%) ÷ 2 = 60%. The aging proportion coefficient is taken as the second anomaly identification accuracy, and when the aging proportion coefficient is 60%, the second anomaly identification accuracy = aging proportion coefficient = 60%.

[0079] The mean of the first abnormality recognition accuracy and the second abnormality recognition accuracy is taken as the abnormality recognition accuracy. For example, if the first abnormality recognition accuracy is 10% and the second abnormality recognition accuracy is 60%, the abnormality recognition accuracy = (first abnormality recognition accuracy + second abnormality recognition accuracy) ÷ 2 = (10% + 60%) ÷ 2 = 35%.

[0080] Based on ensemble learning and multiple lightweight BP neural network branches, a switch cabinet abnormality recognizer is constructed. Ensemble learning is a method of constructing and combining multiple learners to complete a learning task. BP neural network (Back Propagation) is a kind of multi-layer feedforward neural network trained according to the error back propagation algorithm, and is one of the most widely used neural networks. In the embodiments of the present application, multiple lightweight BP neural network branches are constructed based on ensemble learning, and multiple branches are trained by randomly dividing multiple data. The training data of each branch is different, and the performance is different. The branches are integrated and analyzed, which can effectively improve the accuracy. Moreover, the training data of each branch is small, which can improve the training efficiency and convergence speed. According to the abnormality recognition accuracy, accurate and reliable prediction is realized.

[0081] According to the monitoring log of the switch cabinet, a sample leakage parameter set and a sample loosening parameter set are collected, and the proportion of the switch cabinet failure anomaly under different combinations of sample leakage parameters and sample loosening parameters is collected. The data is labeled as sample branch abnormal probability, and a sample branch abnormal probability set is obtained.

[0082] The sample leakage parameter set, the sample loosening parameter set and the sample branch abnormal probability set are randomly divided multiple times to obtain multiple abnormality recognition training data. For example, 20 times of random division are performed to obtain 20 abnormality recognition training data. Each time, 70% of the data is taken as the training data set, and 30% is taken as the verification data set, so as to ensure that the overlap degree of the training sets of different branches is less than 30%, thereby enhancing the diversity.

[0083] A switch cabinet abnormality recognizer based on ensemble learning and BP neural network is constructed. For example, a switch cabinet abnormality recognizer with 20 branches is constructed. Each recognizer branch adopts a three-layer structure, including an input layer, a hidden layer and an output layer. The number of nodes in the input layer is 2, the leakage parameter and the loosening parameter are input, the number of nodes in the hidden layer is 8, the ReLU activation function is used, the number of nodes in the output layer is 1, the Sigmoid activation function is used, the branch abnormal probability is output, the optimizer uses AdamW, and the loss function uses mean square error.

[0084] The input sample leakage parameter set and the sample loosening parameter set are supervised trained using a separate training strategy, that is, each branch is separately trained on a dedicated training set. The branch weight parameters are adjusted for training until each branch of the switch cabinet anomaly recognizer finally converges, that is, the input leakage parameter set and the loosening parameter set, and the output branch anomaly probability accuracy reaches more than 90%, that is, the training is completed.

[0085] The trained switch cabinet anomaly recognizer is configured in the cloud and identified through Internet of Things transmission. The leakage parameter set and the loosening parameter set are input, and multiple branch anomaly probabilities are output.

[0086] In the pre-trained switch cabinet anomaly recognizer, the branches with the proportion of anomaly recognition accuracy are called, the leakage parameters and the loosening parameters are input respectively, multiple branch anomaly probabilities are identified and output, the average is calculated to obtain the anomaly probability as the monitoring result. For example, the trained switch cabinet anomaly recognizer has 20 branches, and when the anomaly recognition accuracy is 65%, the number of branches called = the total number of branches of the switch cabinet anomaly recognizer x the anomaly recognition accuracy = 20 x 65% = 13 (if the calculation result is not an integer, round up). 13 branches are randomly called to predict the branch anomaly probability, and 13 branch anomaly probabilities are output. The average of the 13 branch anomaly probabilities is taken as the final anomaly probability and output, that is, the switch cabinet anomaly monitoring result is obtained.

[0087] In the embodiments of the present application, based on the compensation correction coefficient and the aging proportion coefficient, the anomaly recognition accuracy is adaptively configured to realize dynamic matching of device state and identification strategy. Based on ensemble learning, multiple lightweight neural network branches are constructed, each branch has different training data and performance, and each branch has less training data, which can improve training efficiency and convergence speed. According to the anomaly recognition accuracy, accurate and reliable prediction is realized. Through the fusion calculation of multiple branch probabilities, the abnormal risk is quantified, and the interpretable anomaly probability index is output to guide the hierarchical operation and maintenance and improve the accuracy of switch cabinet monitoring.

[0088] Embodiment two, as Figure 2 shown, based on the same inventive concept of the switch cabinet monitoring method provided in embodiment one, the present application embodiment further provides a switch cabinet monitoring system, comprising:

[0089] The signal monitoring module 100 is used to monitor the vibration signal in the switch cabinet through the vibration sensor when the switch cabinet operates and performs switching action, and obtain a vibration signal sequence.

[0090] The aging analysis and calculation module 200 is used to collect the historical operation data sequence of the switch cabinet, perform arc-extinguishing medium aging leakage analysis and component aging loosening analysis, and obtain aging leakage parameters and aging loosening parameters.

[0091] The real-time analysis calculation module 300 is configured to perform real-time leakage analysis and real-time looseness analysis on the arc-extinguishing medium and the components according to the vibration signal sequence, obtain real-time leakage parameters and real-time looseness parameters, configure compensation correction coefficients according to the aging leakage parameters and the aging looseness parameters, and perform compensation correction and coupling superposition calculation processing on the real-time leakage parameters and the real-time looseness parameters, thereby obtaining leakage parameters and looseness parameters.

[0092] The monitoring result analysis module 400 is configured to configure switch cabinet abnormality recognition accuracy according to the compensation correction coefficients and the aging proportion coefficients of the aging leakage parameters and the aging looseness parameters in the coupling superposition calculation, and perform switch cabinet abnormality probability recognition according to the leakage parameters and the looseness parameters, thereby obtaining an abnormality probability as a monitoring result.

[0093] In one embodiment, the signal monitoring module 100 is further configured to:

[0094] In one embodiment, the signal monitoring module 100 is further configured to:

[0095] In one embodiment, the signal monitoring module 100 is further configured to:

[0096] In one embodiment, the aging analysis calculation module 200 is further configured to:

[0097] In one embodiment, the aging analysis calculation module 200 is further configured to:

[0098] In one embodiment, the aging analysis calculation module 200 is further configured to:

[0099] In one embodiment, the aging analysis calculation module 200 is further configured to:

[0100] In one embodiment, the real-time analysis calculation module 300 is further configured to:

[0101] In one embodiment, the real-time analysis calculation module 300 is further configured to:

[0102] inputting the vibration signal sequence into the real-time analyzer, outputting an obtained real-time leakage parameter and a real-time looseness parameter;

[0103] inputting the aging looseness parameter into a sensing error table, mapping and configuring an obtained sensing error coefficient, and calculating an obtained compensation correction coefficient, wherein the sensing error table is constructed by using a mapping relationship between a sample looseness parameter and a sample sensing error coefficient;

[0104] amplifying and compensating and correcting the real-time leakage parameter and the real-time looseness parameter according to the compensation correction coefficient, to obtain a corrected real-time leakage parameter and a corrected real-time looseness parameter;

[0105] performing coupling superposition calculation according to the corrected real-time leakage parameter, the corrected real-time looseness parameter, and the aging leakage parameter and the aging looseness parameter, to obtain a leakage parameter and a looseness parameter.

[0106] In one embodiment, the monitoring result analysis module 400 is further configured to:

[0107] use the compensation correction coefficient as a first abnormality recognition accuracy;

[0108] calculate a proportion of the aging leakage parameter and the aging looseness parameter in the coupling superposition calculation, and calculate a mean value to obtain an aging proportion coefficient, which is used as a second abnormality recognition accuracy;

[0109] calculate an abnormality recognition accuracy according to the first abnormality recognition accuracy and the second abnormality recognition accuracy;

[0110] in a pre-trained switch cabinet abnormality recognizer, call an abnormality recognition branch in a proportion of the abnormality recognition accuracy, input the leakage parameter and the looseness parameter respectively, identify and output a plurality of branch abnormality probabilities, calculate a mean value to obtain an abnormality probability, which is used as a monitoring result;

[0111] The pre-training step of the switch cabinet abnormality recognizer includes:

[0112] According to the monitoring log of the switch cabinet, a sample leakage parameter set and a sample looseness parameter set are collected, and the proportion of the switch cabinet failure abnormality under different sample leakage parameter and sample looseness parameter combinations is collected and labeled as a sample branch abnormality probability to obtain a sample branch abnormality probability set;

[0113] The sample leakage parameter set, the sample looseness parameter set, and the sample branch abnormality probability set are randomly divided multiple times to obtain multiple abnormality recognition training data;

[0114] A plurality of abnormality recognition branches are constructed by using a BP neural network, and the multiple abnormality recognition training data are used to supervise and train and verify the plurality of abnormality recognition branches, and the plurality of abnormality recognition branches are configured in the cloud after convergence to obtain a switch cabinet abnormality recognizer.

[0115] In some embodiments, the switch cabinet monitoring method further comprises: Figure 3 In some embodiments, the switch cabinet monitoring method further comprises:

[0116] The memory 510 is configured to store the first computer program 511.

[0117] The processor 520 is configured to read and execute the first computer program 511, thereby implementing the switch cabinet monitoring method of the first embodiment.

[0118] In some embodiments, the switch cabinet monitoring method further comprises: Figure 4 In some embodiments, the switch cabinet monitoring method further comprises:

[0119] In some embodiments, the switch cabinet monitoring method further comprises:

[0120] The present application provides a switch cabinet monitoring method, system, intelligent terminal and storage medium. Through fusion analysis of historical operation data and real-time vibration signals, dynamic compensation correction coefficient configuration, and abnormal recognition precision adjustment driven by aging proportion coefficient, the accuracy and early warning capability of switch cabinet fault recognition are significantly improved. Compared with the traditional method, the technical scheme provided by the present application uses the cloud-deployed aging analyzer and real-time analyzer to extract irreversible aging parameters from the historical vibration signals and capture recoverable deformation parameters from the current vibration signals, and combines the dynamic compensation correction of the sensing error table to effectively separate the hardware error and the real fault feature. Based on the aging proportion coefficient, the number of abnormal recognition branches is dynamically adjusted to optimize the calculation efficiency and recognition accuracy of the model, and ensure that the switch cabinet at different aging stages can adapt to the optimal monitoring strategy. The technology provided by the present application significantly overcomes the single data dimension dependence problem and the sensor error accumulation problem, uses the multi-branch BP neural network of ensemble learning to predict the abnormal probability, and outputs the final result by comprehensively considering the abnormal probability of multiple outputs, which realizes the active prediction of progressive failure and the intelligent filtering of interference signals, and improves the accuracy and robustness of switch cabinet abnormal monitoring.

[0121] The present application achieves the technical effect of fusing historical aging trend and real-time monitoring data to realize accurate, effective and timely switch cabinet monitoring.

[0122] It should be noted that the above-mentioned embodiment sequences of the present application are merely for description only, but not for representing the advantages and disadvantages of the embodiments. And the above-mentioned embodiments of the present specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0123] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0124] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that are within the scope of the present application should be included. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A switchgear monitoring method characterized by, The method comprises: During the switching operation of the switch cabinet, the vibration signals in the switch cabinet are monitored through the vibration sensor to obtain a vibration signal sequence; A historical operation data sequence of the switch cabinet is collected to perform arc-extinguishing medium aging leakage analysis and component aging loosening analysis, and aging leakage parameters and aging loosening parameters are obtained; According to the vibration signal sequence, real-time leakage analysis and real-time loosening analysis of the arc-extinguishing medium are performed to obtain real-time leakage parameters and real-time loosening parameters, compensation correction coefficients are configured according to the aging leakage parameters and the aging loosening parameters, and the real-time leakage parameters and the real-time loosening parameters are subjected to compensation correction and coupling superposition calculation processing to obtain leakage parameters and loosening parameters; According to the compensation correction coefficients and the aging proportion coefficients of the aging leakage parameters and the aging loosening parameters in the coupling superposition calculation, switch cabinet abnormality recognition accuracy is configured, and switch cabinet abnormality probability recognition is performed according to the leakage parameters and the loosening parameters to obtain an abnormality probability as a monitoring result, comprising: The compensation correction coefficients are taken as first abnormality recognition accuracy; The proportions of the aging leakage parameters and the aging loosening parameters in the coupling superposition calculation are calculated, and a mean value is calculated to obtain an aging proportion coefficient as second abnormality recognition accuracy; According to the first abnormality recognition accuracy and the second abnormality recognition accuracy, abnormality recognition accuracy is calculated and obtained; In a pre-trained switch cabinet abnormality recognizer, an abnormality recognition branch in a proportion of the abnormality recognition accuracy is called, and the leakage parameters and the loosening parameters are input respectively to recognize and output multiple branch abnormality probabilities, a mean value is calculated to obtain an abnormality probability as a monitoring result; The pre-training steps of the switch cabinet abnormality recognizer comprise: According to the monitoring log of the switch cabinet, sample leakage parameter sets and sample loosening parameter sets are collected, and the proportions of switch cabinet failures under different sample leakage parameter and sample loosening parameter combinations are collected and labeled as sample branch abnormality probabilities to obtain a sample branch abnormality probability set; The sample leakage parameter sets, the sample loosening parameter sets, and the sample branch abnormality probability set are randomly divided multiple times to obtain multiple abnormality recognition training data; A BP neural network is used to construct multiple abnormality recognition branches, the multiple abnormality recognition training data are used respectively to supervise training and verification of the multiple abnormality recognition branches, and the multiple abnormality recognition branches are configured in the cloud after convergence to obtain a switch cabinet abnormality recognizer.

2. The switchgear monitoring method according to claim 1, characterized in that, During the switching operation of the switch cabinet, the vibration signals in the switch cabinet are monitored through the vibration sensor to obtain a vibration signal sequence, comprising: During the switching operation of the switch cabinet, the vibration signals in the switch cabinet are monitored through the vibration sensor; The vibration signals obtained through monitoring are recorded in time sequence to obtain a vibration signal sequence.

3. The switchgear monitoring method according to claim 1, characterized in that, A historical operation data sequence of the switch cabinet is collected to perform arc-extinguishing medium aging leakage analysis and component aging loosening analysis, and aging leakage parameters and aging loosening parameters are obtained, comprising: According to the monitoring log of the switch cabinet, vibration signal sequences of the switch cabinet in a historical time are collected to obtain a historical vibration signal sequence set; According to the operation and maintenance data of the same type switch cabinet, the aging analyzer is trained and configured in the cloud, wherein the aging analyzer includes an arc extinguishing medium aging analysis branch and a component aging analysis branch, the input data of the arc extinguishing medium aging analysis branch and the component aging analysis branch is a sample historical vibration signal sequence set, and the output data is a sample aging leakage parameter and a sample aging loosening parameter respectively; The historical vibration signal sequence set is input into the trained aging analyzer to output the obtained aging leakage parameter and aging loosening parameter.

4. The switchgear monitoring method according to claim 1, characterized by, According to the vibration signal sequence, real-time arc extinguishing medium leakage analysis and real-time component loosening analysis are performed to obtain real-time leakage parameters and real-time loosening parameters, including: According to the operation and maintenance data of the same type switch cabinet, the real-time analyzer is trained and configured in the cloud, wherein the real-time analyzer includes an arc extinguishing medium real-time analysis branch and a component real-time analysis branch, the input data of the arc extinguishing medium real-time analysis branch and the component real-time analysis branch is a sample vibration signal sequence, and the output data is a sample real-time leakage parameter and a sample real-time loosening parameter respectively; The vibration signal sequence is input into the real-time analyzer to output the obtained real-time leakage parameter and real-time loosening parameter.

5. The switchgear monitoring method according to claim 1, characterized by, According to the aging loosening parameter, a compensation correction coefficient is configured, and compensation correction and coupling superposition calculation processing are performed on the real-time leakage parameter and the real-time loosening parameter to obtain leakage parameters and loosening parameters, including: The aging loosening parameter is input into a sensing error table to map and configure a sensing error coefficient to calculate a compensation correction coefficient, wherein the sensing error table is constructed by using a mapping relationship between a sample loosening parameter and a sample sensing error coefficient; According to the compensation correction coefficient, amplification compensation correction calculation is performed on the real-time leakage parameter and the real-time loosening parameter to obtain corrected real-time leakage parameters and corrected real-time loosening parameters; According to the corrected real-time leakage parameters, the corrected real-time loosening parameters, and the aging leakage parameters and the aging loosening parameters, coupling superposition calculation is performed to obtain leakage parameters and loosening parameters.

6. A switchgear monitoring system characterized by, The system is used to implement the switch cabinet monitoring method of any one of claims 1-5, and the system comprises: A signal monitoring module is configured to monitor vibration signals in the switch cabinet through a vibration sensor when the switch cabinet operates and a switch operates to obtain a vibration signal sequence; An aging analysis calculation module is configured to collect historical operation data sequences of the switch cabinet, perform arc extinguishing medium aging leakage analysis and component aging loosening analysis, and obtain aging leakage parameters and aging loosening parameters; A real-time analysis calculation module is configured to perform arc extinguishing medium real-time leakage analysis and component real-time loosening analysis according to the vibration signal sequence to obtain real-time leakage parameters and real-time loosening parameters, configure a compensation correction coefficient according to the aging leakage parameters and the aging loosening parameters, and perform compensation correction and coupling superposition calculation processing on the real-time leakage parameters and the real-time loosening parameters to obtain leakage parameters and loosening parameters. The monitoring result analysis module is configured to calculate the aging proportion coefficients of the aging leakage parameter and the aging loosening parameter in the coupling superposition calculation according to the compensation correction coefficients, configure the switch cabinet abnormality identification accuracy, identify the switch cabinet abnormality probability according to the leakage parameter and the loosening parameter, and obtain the abnormality probability as the monitoring result.

7. A smart terminal, characterized in that Comprise: a memory for storing a first computer program; a processor for reading and executing the first computer program, thereby realizing the switch cabinet monitoring method according to any one of claims 1-5.

8. A non-transitory computer-readable storage medium, comprising: The storage medium stores a second computer program, and the second computer program is executed by the processor to realize the switch cabinet monitoring method according to any one of claims 1-5.

Images