Surge impact analysis-based voltage sampling unit fault detection method and system

Abstract

The application discloses a voltage sampling unit fault detection method and system based on surge impact analysis, relates to the technical field of power detection, and comprises the following steps: obtaining an impact simulation data set by performing multivariate surge impact simulation on a voltage sampling unit; identifying a fault detection strategy according to a voltage sampling unit fault common sense base and the impact simulation data set, taking the identification result as a discrete hidden variable to construct a fault detection strategy matrix set; extracting real-time sampling representation data, performing strategy matching in the fault detection strategy matrix set, determining a matched fault detection strategy, detecting and analyzing the voltage sampling unit, and determining a fault detection result. The application solves the technical problem that the function drift and abnormal characteristics of the voltage sampling unit under the action of a surge impact cannot be accurately identified in the prior art, fault identification is not timely, and the precision is not high, and achieves the technical effect of improving the accuracy and real-time performance of voltage sampling unit fault identification under a surge impact.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power detection, in particular to a voltage sampling unit fault detection method and system based on surge impact analysis. BACKGROUND

[0002] As a key signal acquisition component in a smart electric energy meter, a voltage sampling unit is widely used for accurate sampling and transmission of power grid voltage signals, and directly affects the accuracy and stability of electric energy metering. In actual operation, the voltage sampling unit often faces surge impact interference caused by lightning strikes, power system operations and other factors, which may cause problems such as resistance voltage division ratio change, AD sampling distortion, and abnormal response of the sampling channel. For such impact disturbances, the current common method is to rely on static voltage error checking or limit value judgment, which is difficult to capture the dynamic change characteristics of the voltage sampling signal under the action of the surge impact, and also cannot effectively identify the functional drift or potential failure trend, resulting in obvious deficiencies in the response ability and accurate identification ability of the detection system to impact faults. SUMMARY

[0003] The application provides a voltage sampling unit fault detection method and system based on surge impact analysis, which is used to solve the technical problems that the existing technology cannot accurately identify the functional drift and abnormal characteristics of the voltage sampling unit under the action of the surge impact, resulting in untimely and low-precision fault identification.

[0004] In view of the above problems, the application provides a voltage sampling unit fault detection method and system based on surge impact analysis.

[0005] In a first aspect, the application provides a voltage sampling unit fault detection method based on surge impact analysis, which comprises:

[0006] A surge generator is used to simulate multi-element surge impact on the voltage sampling unit to obtain a set of impact simulation data. A voltage sampling unit fault common sense library is obtained, and a fault detection strategy is identified in combination with the set of impact simulation data. The identification result is taken as a discrete hidden variable to construct a fault detection strategy matrix set. Real-time sampling feature data of the voltage sampling unit is extracted, and the real-time sampling feature data is used as an index to perform strategy matching in the fault detection strategy matrix set to determine a matched fault detection strategy. The voltage sampling unit is detected and analyzed based on the matched fault detection strategy to determine a fault detection result.

[0007] In a second aspect, the application provides a voltage sampling unit fault detection system based on surge impact analysis, which comprises:

[0008] The surge impact simulation module is configured to simulate multiple surges on the voltage sampling unit by using a surge generator to obtain a set of impact simulation data; the strategy identification module is configured to obtain a fault common knowledge base of the voltage sampling unit, identify a fault detection strategy in combination with the set of impact simulation data, and take the identification result as discrete hidden variables to construct a set of fault detection strategy matrices; the strategy matching module is configured to extract real-time sampling feature data of the voltage sampling unit, use the real-time sampling feature data as an index to perform strategy matching in the set of fault detection strategy matrices, and determine a matched fault detection strategy; and the detection analysis module is configured to perform detection analysis on the voltage sampling unit based on the matched fault detection strategy to determine a fault detection result.

[0009] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0010] The present application simulates multiple surges on the voltage sampling unit by using a surge generator to obtain a set of impact simulation data, obtains a fault common knowledge base of the voltage sampling unit, identifies a fault detection strategy in combination with the set of impact simulation data, takes the identification result as discrete hidden variables to construct a set of fault detection strategy matrices, extracts real-time sampling feature data of the voltage sampling unit, uses the real-time sampling feature data as an index to perform strategy matching in the set of fault detection strategy matrices, determines a matched fault detection strategy, and performs detection analysis on the voltage sampling unit based on the matched fault detection strategy to determine a fault detection result. The present application solves the technical problem that the function drift and abnormal characteristics of the voltage sampling unit under the action of surges cannot be accurately identified in the prior art, resulting in untimely and low-precision fault identification. By constructing a set of impact simulation data and generating a detection strategy in combination with a fault common knowledge base, strategy matching and fault analysis are realized with real-time feature data, and the technical effects of improving the accuracy and real-time performance of voltage sampling unit fault identification under surges are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present 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 present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0012] Figure 1 A voltage sampling unit fault detection method flowchart based on surge impact analysis is provided for the embodiments of the present application.

[0013] Figure 2 A voltage sampling unit fault detection system structure diagram based on surge impact analysis is provided for the embodiments of the present application.

[0014] Surge impact simulation module 11, strategy identification module 12, strategy matching module 13, detection analysis module 14. DETAILED DESCRIPTION

[0015] The application provides a voltage sampling unit fault detection method and system based on surge impact analysis, which aims to solve the technical problem that the function drift and abnormal characteristics of the voltage sampling unit under the action of the surge impact cannot be accurately identified in the prior art, resulting in that the fault identification is not timely and the accuracy is not high. The detection strategy is generated by constructing the impact simulation data set and combining the fault common knowledge base, the strategy matching and fault analysis with the real-time characterization data are realized, and the technical effects of improving the accuracy and real-time performance of the voltage sampling unit fault identification under the surge impact are achieved.

[0016] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the application.

[0017] It should be noted that any variation of the terms "comprise" and "have" is intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0018] Embodiment one, as shown in the application provides a voltage sampling unit fault detection method based on surge impact analysis, which comprises: Figure 1

[0019] Step S100: using a surge generator to perform multi-element surge impact simulation on the voltage sampling unit to obtain an impact simulation data set.

[0020] In the embodiments of the application, by constructing a multi-element surge scene, setting a combination of surge parameters including surge waveform type, impact amplitude and impact time, and connecting the voltage sampling unit to the output end of the surge generator for impact test. The voltage signal under the surge response is collected by using an oscilloscope, a data acquisition card and a spectrum analyzer, and the abnormal signs and function drift characteristics are extracted respectively, and finally the simulation signal, abnormal data and drift data are combined and mapped to form a complete impact simulation data set.

[0021] Further, in the method provided by the application, the voltage sampling unit is simulated by using the surge generator to obtain the impact simulation data set, and the method further comprises:​

[0022] Based on the multi-scene analysis, a multi-surge generator output parameter set is generated, wherein each multi-surge generator output parameter comprises a surge waveform type, an impulse amplitude, and an impulse time; a voltage sampling unit is connected to a surge generator output end, the surge waveform type, the impulse amplitude, and the impulse time of the surge generator are adjusted according to the multi-surge generator output parameter set, and an oscilloscope, a data acquisition card, and a spectrum analyzer are configured to collect voltage sampling signals to obtain an analog voltage sampling signal set; abnormal symptom capture data is obtained by performing abnormal symptom capture on the analog voltage sampling signal set; a function drift data set is determined by performing function drift authentication analysis on the analog voltage sampling signal set; and the abnormal symptom capture data set, the function drift data set, and the analog voltage sampling signal set are respectively mapped and combined to obtain an impulse simulation data set.

[0023] In the embodiments of the present application, first, a parameter set for surge impulse simulation is generated based on multi-scene analysis. Specifically, by archiving historical voltage sampling unit failure data and taking surge impulse failure types, surge waveform types, impulse amplitudes, and impulse times as index fields, a historical multi-failure scene data retrieval method is used to extract a parameter set related to a typical failure event from a database. Subsequently, a decision tree classification algorithm is used to divide the historical surge impulse failure types into patterns to obtain M divided historical surge impulse failure type sets, and these sets are combined to map and classify original historical surge waveform type sets, historical impulse amplitude sets, and historical impulse time sets to generate corresponding M divided sets. On this basis, a three-dimensional parameter space construction method is used to take surge waveform types, impulse amplitudes, and impulse times as three axes, to perform surge impulse scene calibration analysis in M three-dimensional spaces respectively to obtain M surge impulse scene calibration parameters, and to summarize the parameters into a final multi-surge generator output parameter set.

[0024] Subsequently, a test operation is performed, a voltage sampling unit is connected to a surge generator output end using a standard electrical connection method, and the surge waveform type, the impulse amplitude, and the impulse time of the surge generator are adjusted according to the parameters set in the aforementioned multi-surge generator output parameter set. An oscilloscope, a data acquisition card, and a spectrum analyzer are configured synchronously to build a complete high-bandwidth signal acquisition system for real-time acquisition of voltage response signals under each parameter set to obtain an analog voltage sampling signal set formed under the action of a specific impulse.

[0025] After obtaining the set of analog voltage sampling signals, an abnormal symptom capture process is performed, and a feature rule matching method is used to analyze each set of analog voltage sampling signals in multiple dimensions. Specifically, it is determined whether the sampling accuracy decreases, which is achieved by comparing the amplitudes of the analog voltage sampling signals within a 5 ms window before and after the surge injection. If the error exceeds ±1%, it is determined that the sampling accuracy is abnormal. It is determined whether there is signal distortion, which is achieved by using sliding window Fourier transform and wavelet analysis to detect whether there are sudden burrs, jump points or wideband noise. If the amplitude mutation rate is greater than 5 V / μs or the noise frequency band energy rises by more than 10 dB, it is determined that the signal is distorted. It is determined whether there is chip port damage, which is achieved by determining whether there is a situation such as a certain phase voltage signal being zero for a long time, a response data interruption or frequent jumps. The data samples that meet the above determination conditions are classified into the abnormal symptom capture data set, and the abnormal type, associated parameters and trigger conditions are clearly recorded.

[0026] Subsequently, the set of analog voltage sampling signals is analyzed for functional drift authentication. A deviation identification method based on standard comparison is used to compare each analog voltage sampling signal with the preset standard voltage sampling signal point by point. During the comparison process, the key feature indicators of the signal are extracted, including the amplitude mean, the rising edge time and the main frequency component. The difference between the indicators and the corresponding indicators of the standard signal is calculated, and a unified preset tolerance interval is set. If the amplitude mean deviates by more than 3%, the rising edge time is extended by more than 10 μs, or the main frequency component deviates by more than 5 kHz, it is determined that the current signal has functional drift. All signal samples that exceed any of the above drift thresholds are recorded as functional drift data and included in the functional drift data set.

[0027] Finally, the abnormal symptom capture data set, the functional drift data set and the set of analog voltage sampling signals are respectively mapped and combined. In this process, structured indexing alignment and multi-field vector splicing methods are used to achieve sample integration. Specifically, each surge injection test is taken as a basic unit, a unified test number is set as the primary key index, and the structured indexing alignment method is used to pair the metadata of each set of analog voltage sampling signals, such as sampling time stamp, voltage response curve, surge waveform type, impact amplitude and impact time, with the corresponding abnormal symptom capture labels (such as sampling accuracy decrease, signal distortion, chip port damage) and functional drift authentication labels (such as mean deviation rate, response delay, spectral drift) at the field level. After completing the index mapping, the multi-field vector splicing method is used to combine the three types of data sets into a unified sample format according to the fixed field structure. Each record contains complete surge parameters, analog voltage sampling signals, abnormal type labels and drift indicator results, and finally forms a high-dimensional, fully labeled and structured data sample set, which constructs a complete impulse simulation data set.

[0028] Further, the method provided by the application embodiment further comprises: generating a plurality of surge generator output parameter sets based on the plurality of scene analysis, wherein each of the plurality of surge generator output parameter sets comprises a surge waveform type, an impact amplitude and an impact time, and further comprises:

[0029] performing data retrieval on the historical plurality of fault scene data set with the surge impact fault type, the surge waveform type, the impact amplitude and the impact time as indexes to obtain a historical surge impact fault type set, a historical surge waveform type set, a historical impact amplitude set and a historical impact time set; performing decision tree classification on the historical surge impact fault type to generate M divided historical surge impact fault type sets, wherein M is a positive integer; performing mapping division on the historical surge waveform type set, the historical impact amplitude set and the historical impact time set based on the M divided historical surge impact fault type sets to generate M divided historical surge waveform type sets, M divided historical impact amplitude sets and M divided historical impact time sets; constructing M three-dimensional spaces, and performing surge impact scene calibration analysis on the M divided historical surge waveform type sets, the M divided historical impact amplitude sets and the M divided historical impact time sets in the M three-dimensional spaces respectively to obtain M surge impact scene calibration parameters, and the M surge impact scene calibration parameters are summarized as the plurality of surge generator output parameter sets.

[0030] In the application embodiment, the historical plurality of fault scene data set is first subjected to data retrieval with the surge impact fault type, the surge waveform type, the impact amplitude and the impact time as indexes. The data set refers to the structured historical data collected and stored in the process of voltage sampling unit acceptance surge impact test or actual operation, and specifically includes analog voltage sampling signals, surge injection parameters and manually or automatically labeled fault results. A multi-field joint retrieval method is used to extract all records meeting the conditions from the database with the "fault type", "waveform type", "impact amplitude" and "impact time" as index fields to form the original historical surge impact fault type set, the historical surge waveform type set, the historical impact amplitude set and the historical impact time set.

[0031] Next, the historical surge impact fault type set is classified by decision tree. First, a decision tree model is built and trained using historical fault samples. The training data includes input features and output labels. The input features are signal features extracted from historical simulated voltage sampling signals, such as sampling mean, voltage change rate, signal rise time, and frequency domain main frequency. The output labels are manually annotated or automatically annotated fault type labels, such as "sampling accuracy degradation", "signal distortion", "response drift", "channel loss", etc. The CART algorithm is used to model the training samples, and the information gain or Gini index is used for feature splitting to automatically build a classification path. After the model is trained, historical data is input into the model for classification and judgment, and the corresponding fault type label is output. Finally, the samples are divided into M sub-classes based on the model output, forming M divided historical surge impact fault type sets, where M is a positive integer, usually corresponding to the number of typical types of surge impact faults.

[0032] After obtaining the M divided historical surge impact fault type sets, the original historical surge waveform type set, historical impact amplitude set, and historical impact time set are parameter-mapped and divided. This process uses a label index mapping method, taking the classification label of each fault record as the primary key, and grouping the corresponding surge waveform type, impact amplitude, and impact time into the fault type class to which it belongs. Each fault type class has a corresponding parameter set, forming M divided historical surge waveform type sets, M divided historical impact amplitude sets, and M divided historical impact time sets.

[0033] Finally, when building M three-dimensional spaces, the M divided historical surge waveform type sets, M divided historical impact amplitude sets, and M divided historical impact time sets are used as the parameter dimensions of the three-dimensional coordinate axes. The parameter combination corresponding to each fault type is mapped into an independent three-dimensional space, forming M three-dimensional spaces, each corresponding to a specific surge impact fault mode. In each three-dimensional space, a scene differentiation loss function is constructed based on the mapping vector, and an initial mapping vector is randomly selected for initial calibration analysis to calculate the initial scene differentiation loss. Then, an iterative mapping vector is selected from the corresponding mapping vector set, and the scene differentiation loss function is applied to it for repeated calculation to obtain the iterative scene differentiation loss. By comparing the initial scene differentiation loss and the iterative scene differentiation loss, the calibration vector is optimized and iteratively updated, and finally one or more optimal calibration mapping vectors are obtained in each three-dimensional space. The combination parameters of surge waveform type, impact amplitude, and impact time are extracted from these calibration mapping vectors to form M surge impact scene calibration parameters. Structured aggregation of M surge impact scene calibration parameters forms the final multi-element surge generator output parameter set.

[0034] Further, the method provided by the application further comprises:

[0035] The M divided historical surge waveform type sets, the M divided historical impact amplitude sets and the M divided historical impact time sets are respectively mapped to the M three-dimensional spaces to obtain M mapping vector sets; M initial mapping vectors are respectively randomly selected from the M mapping vector sets, a scene differentiation loss function is constructed, and the M initial mapping vectors are analyzed to obtain M initial scene differentiation loss amounts; M iteration mapping vectors are again respectively randomly selected from the M mapping vector sets, and the M iteration mapping vectors are analyzed by using the scene differentiation loss function to obtain M iteration scene differentiation loss amounts; the M initial mapping vectors and the M iteration mapping vectors are selected and iteratively updated by comparing the M iteration scene differentiation loss amounts and the M initial scene differentiation loss amounts, to obtain M calibrated mapping vectors; elements of the M calibrated mapping vectors are extracted to obtain the M surge impact scene calibration parameters.

[0036] In the embodiments of the application, the M divided historical surge waveform type sets, the M divided historical impact amplitude sets and the M divided historical impact time sets are respectively mapped to the M three-dimensional spaces, which means that, for each divided historical surge impact failure type set, the three parameters of "surge waveform type", "impact amplitude" and "impact time" are taken as three-dimensional coordinate axes, a parameter vector is constructed in the form of a triple, and M independent three-dimensional spaces are respectively established in the parameter space with the failure type as a division boundary. In each three-dimensional space, all historical parameter combinations from the failure type are represented as a vector point set, denoted as a mapping vector set.

[0037] Next, M initial mapping vectors are respectively selected from the M mapping vector sets by using a random balanced sampling method, that is, one mapping vector is randomly selected as an initial representative of the space in each three-dimensional space to obtain the M initial mapping vectors. Then, a scene differentiation loss function is constructed, and the M initial mapping vectors are analyzed. The feature difference between each initial mapping vector and all other vectors in the mapping vector set in which the initial mapping vector is located is taken as input, and the loss amount of the initial mapping vector as a scene representative is output to obtain M initial scene differentiation loss amounts.

[0038] Then, M iteration mapping vectors are respectively selected from the M mapping vector sets by using the same random balanced sampling method, and M iteration scene differentiation loss amounts corresponding to the M iteration mapping vectors are calculated by using the foregoing scene differentiation loss function.

[0039] After comparing the M iteration scene distinction loss amounts and the M initial scene distinction loss amounts, a difference degree determination and direction updating mechanism is adopted. When the iteration loss amount is not better than the initial loss amount but the difference degree is lower than a preset difference degree threshold, the direction from the initial mapping vector to the iteration mapping vector is taken as the parameter updating direction, and a random disturbance updating operation is performed based on the current iteration result in the mapping vector set. After a preset number of updating times is met, the M calibration mapping vectors are finally obtained. For example, when M is equal to 5, the scene distinction loss amount comparison is as shown in Table 1:

[0040] Table 1: M iteration scene distinction loss amount data table

[0041] Mapping vector Initial scene differentiation loss amount Iterative scene differentiation loss amount Difference degree Mapping vector 1 1.351 0.513 -0.838 Mapping vector 2 0.507 1.31 0.803 Mapping vector 3 0.773 0.537 -0.236 Mapping vector 4 0.649 1.411 0.762 Mapping vector 5 1.4 0.711 -0.689

[0042] Finally, the three-dimensional parameter elements of each of the M calibration mapping vectors, i.e., the corresponding surge waveform type, impact amplitude and impact time, are extracted to form M surge impact scene calibration parameters.

[0043] Further, the method provided by the application embodiment further comprises:

[0044] It is judged whether the M iteration scene distinction loss amounts are greater than or equal to the M initial scene distinction loss amounts. If yes, the difference degrees of the M iteration scene distinction loss amounts and the M initial scene distinction loss amounts are calculated. When the calculation result is less than a preset difference degree threshold, the direction from the M initial mapping vectors to the M iteration mapping vectors is taken as the updating direction, and the M iteration mapping vectors are randomly updated in the M mapping vector set until a preset number of updating times is met, and the M calibration mapping vectors are obtained.

[0045] In the application embodiment, the M iteration scene distinction loss amounts and the corresponding M initial scene distinction loss amounts are compared one by one in value to judge whether the performance optimization of each group of parameters is realized. If an iteration scene distinction loss amount is greater than or equal to the corresponding initial scene distinction loss amount, it is considered that the current parameter combination fails to improve the distinguishability of the surge impact scene.

[0046] In the case where the above conditions are met, the difference degree between the iteration scene distinction loss amount and the initial scene distinction loss amount is calculated. The difference degree is the numerical difference value of the two, reflecting the absolute degree of performance change. If the difference degree is less than a preset difference degree threshold, it is explained that the performance change of the parameter group is limited, which meets the acceptable updating condition.

[0047] At this time, the initial mapping vector is determined to point to the direction of the iterative mapping vector as the update direction of the current optimization, and the current iterative mapping vector is updated by perturbation in the corresponding mapping vector set in combination with the direction. The perturbation update is usually performed by a random perturbation method based on a normal distribution to achieve fine adjustment by adding a small change in the given direction.

[0048] The iterative update operation is independently performed in each group and continues until each group reaches the preset update number limit. After the update of all M groups of parameters is finally completed, an optimized output result is obtained from each group, respectively, that is, M calibration mapping vectors are obtained.

[0049] Further, the method provided by the application embodiment, the scene differentiation loss function is:

[0050] ;

[0051] Wherein, is the initial scene differentiation loss amount, is the number of mapping vectors in the mapping vector set corresponding to the initial mapping vector for scene differentiation loss analysis, is the distance between the initial mapping vector for scene differentiation loss analysis and the jth mapping vector in the corresponding mapping vector set, is the number of initial mapping vectors in the comparison set composed of the initial mapping vector after excluding M initial mapping vectors for scene differentiation loss analysis, is the distance between the jth mapping vector and the ith initial mapping vector in the comparison set.

[0052] In the application embodiment, represents the scene differentiation loss amount in the current initial scene, which is used to evaluate whether the initial parameter combination has good differentiation. represents the initial mapping parameter combination, is the number of mapping vectors in the mapping vector set corresponding to the initial mapping vector for scene differentiation loss analysis. is the distance between the initial mapping vector for scene differentiation loss analysis and the jth mapping vector in the corresponding mapping vector set, which is calculated by the Euclidean distance. is the number of initial mapping vectors in the comparison set composed of the initial mapping vector after excluding M initial mapping vectors for scene differentiation loss analysis, which represents the number of samples used to construct the comparison set after excluding itself from the current scene category, that is, the number of remaining samples after excluding the current sample in the category to which the initial sample belongs, which is used to form the non-self comparison group of the same category. is the distance between the jth mapping vector and the jth initial mapping vector in the comparison set, which is calculated by the Euclidean distance.

[0053] Further, the method provided by the application embodiment further comprises:

[0054] The analog voltage sampling signals in the analog voltage sampling signal set are compared with the standard voltage sampling signal, and data exceeding a preset tolerance interval in a comparison result is taken as functional drift data, to obtain the functional drift data set.

[0055] In the application embodiment, first, a standard voltage sampling signal is constructed, and the standard signal is generated by collecting a plurality of groups of voltage sampling response signals under normal working conditions and using a template average method. The template average method is a modeling method based on statistical average, and a representative standard curve is obtained by taking an arithmetic average of amplitudes at the same time point after a plurality of sampling curves are aligned on a time axis through interpolation, so as to describe the typical response behavior of the voltage sampling unit under an abnormal state.

[0056] Subsequently, each sampling curve in the analog voltage sampling signal set is compared with the standard voltage sampling signal one by one. The comparison process uses a time normalization comparison method, and first, the analog sampling signal and the standard signal are unified to the same time resolution and sampling point number using linear interpolation, and then the key feature error is calculated in a point-by-point manner.

[0057] Then, the error information after comparison is screened through a single feature threshold determination method. Specifically, a plurality of preset tolerance intervals are set, including upper tolerance limits for amplitude error, response delay, steady-state offset, and waveform distortion. When the analog signal exceeds the corresponding tolerance threshold in any index, it is considered that the signal has functional abnormalities deviating from the standard state.

[0058] Finally, all analog sampling signals exceeding the preset tolerance interval in any comparison feature are uniformly marked as functional drift data, and are collected into a functional drift data set.

[0059] Step S200: Acquire a voltage sampling unit fault common sense library, combine the impact simulation data set to identify a fault detection strategy, and take the identification result as a discrete hidden variable to construct a fault detection strategy matrix set.

[0060] In the application embodiment, first, a voltage sampling unit fault common sense library is acquired, and the common sense library is established by a technical expert by summarizing a large amount of historical test data, and covers typical features of different types of faults (such as precision drift, circuit break, temperature drift, surge impact, etc.) in voltage response signals, for example, waveform amplitude change, continuous response anomaly, signal frequency offset, ripple enhancement, abnormal noise, or response time delay, etc. These features are used as the basis for characterizing faults, and are used as a mode reference for subsequent strategy identification.

[0061] Next, based on the existing impact simulation data set, the abnormal symptom capture data set is extracted as the initial fault characterization basis. The pre-established voltage sampling unit fault common sense library is used to identify the multiple fault directionality of the abnormal symptom capture data set, and determine whether there is a clear single fault direction. If the identification result shows that there is multiple fault direction, the voltage sampling unit fault common sense library is called to perform directionality data missing analysis to identify the missing key fault discrimination features in the current data. According to the analysis result, the corresponding supplementary features are extracted from the simulated voltage sampling signal set to generate a complete impact simulation data set. The detection index of the completed data is extracted, and the obtained result is coded as a discrete latent variable, and finally a structured fault detection strategy matrix set is constructed.

[0062] Further, the method provided by the application embodiment further comprises:

[0063] extracting the abnormal symptom capture data set in the impact simulation data set; using the voltage sampling unit fault common sense library to identify the multiple fault directionality of the abnormal symptom capture data set, if the identification result shows that there is multiple fault direction, if the identification result shows that there is multiple fault direction, then using the voltage sampling unit fault common sense library to perform directionality data missing analysis on the impact simulation data set, and according to the analysis result, extracting data from the simulated voltage sampling signal set to obtain a complete impact simulation data set; extracting the detection index of the complete impact simulation data set, and coding the extracted result as a latent variable to construct a fault detection strategy matrix set.

[0064] In the application embodiment, first, the abnormal symptom capture processing is performed, which analyzes each group of simulated voltage sampling signals in the impact simulation data set based on the instantaneous amplitude mutation detection algorithm. Specifically, the instantaneous voltage change rate in each window is calculated in units of 0.5ms sliding window, and the average of the normal response curve calculated by the historical fault-free sample template average method is taken as the reference. If the voltage change rate in a window exceeds ±20% compared with the reference value, it is determined that the data corresponding to the window has abnormal voltage response behavior, which is marked as an abnormal segment and extracted, and finally the abnormal symptom capture data set is obtained.

[0065] Next, the multiple fault direction identification method is used to analyze the abnormal symptom capture data set. The method calls the voltage sampling unit fault common sense library, which is obtained by classifying and counting historical impulse test data, and includes various typical voltage abnormal response modes and their corresponding fault mechanisms, such as response distortion, peak feedback, and delayed response. The pattern matching method is used to compare the abnormal data with the templates in the common sense library, and to determine whether there is a situation where an abnormal performance corresponds to multiple fault explanations. If multiple possible fault mechanisms are identified, the corresponding multiple direction identification labels are output, and the multiple fault direction identification result is obtained.

[0066] When there is multiple fault direction, the missing data analysis of the directionality is performed. This step continues to call the voltage sampling unit fault common sense library, and uses the feature coverage analysis method, that is, to analyze whether the abnormal data contains the key feature fragments required by each candidate fault mechanism, such as the front steep region, the hysteresis region, or the stable segment. If it is found that the key data segment corresponding to a candidate mechanism is missing in the abnormal symptom, the data is extracted from the original simulated voltage sampling signal set according to the time stamp and channel index. Finally, through this step, the complete impulse simulation data set can be obtained.

[0067] After the data completion, the detection index extraction step is performed, and the rule feature extraction method is used to extract the key performance indicators from the complete impulse simulation data set, including the peak voltage (reflecting the response intensity), the first-order difference slope (describing the rising / descending trend), the response duration (characterizing the fault influence interval), and the waveform energy index (used to distinguish the energy accumulation type abnormality). The index extraction result of each sample is coded and quantitatively classified according to a unified format, as a set of hidden variables representing the characteristics of the sample, for describing its potential fault performance. Finally, the hidden variables of all samples are numbered and classified, and a set of fault detection strategy matrices is constructed according to the sample dimension.

[0068] Further, the method provided by the application embodiment further comprises:

[0069] If it is identified that there is no multiple fault direction, the detection index extraction is performed on the impulse simulation data set, and the extraction result is used as a hidden variable to construct a fault detection strategy matrix set.

[0070] In the embodiments of the present application, if the identification result shows that there is no multiple fault direction, it indicates that the abnormal characteristics shown in the analog voltage sampling signal are only related to a single fault type. At this time, the detection index extraction is directly performed on the impact analog data set. In this process, first, each analog voltage sampling signal is analyzed, and a rule-based feature extraction method is used to extract multiple key performance indicators from each signal. These indicators include peak voltage, which is used to reflect the intensity of the signal; response duration, which is the duration from the trigger point to the stable voltage value, used to evaluate the system's ability to respond to faults; first-order differential slope, which calculates the rate of signal change, helping to judge the stability of the signal; and waveform energy indicator, which evaluates the energy distribution and abnormal characteristics in the signal through spectral analysis.

[0071] All extracted detection indicators are then uniformly processed by standardization methods, using maximum-minimum normalization or Z-score standardization to ensure that different indicators are compared on the same scale. These standardized detection indicators are encoded as latent variables, with each latent variable representing the characteristic performance of the signal under a specific fault condition. In this way, the characteristics of each signal are converted into digital form, facilitating subsequent fault pattern recognition and strategy reasoning.

[0072] Finally, all extracted latent variables are organized by sample number and fault characteristics to form a set of fault detection strategy matrices.

[0073] Step S300: Extract real-time sampling representation data of the voltage sampling unit, and perform strategy matching in the fault detection strategy matrix set indexed by the real-time sampling representation data to determine the matched fault detection strategy.

[0074] In the embodiments of the present application, first, the real-time sampling representation data of the voltage sampling unit is extracted through a data acquisition card or a digital oscilloscope and other devices. The voltage signal is collected in real time, and signal processing algorithms such as low-pass filtering and band-pass filtering are used to denoise and preprocess the signal. The real-time sampling representation data of the voltage sampling unit includes multiple key features, such as amplitude (reflecting the intensity of the voltage signal), frequency (reflecting the periodic change of the signal), response time (used to evaluate the time from initial mutation to stability), waveform morphology (reflecting the morphological change of the signal, such as slope, mutation, etc.), and other related parameters.

[0075] Subsequently, a feature vector similarity matching method is used, with the extracted real-time sampling representation data as the index, to perform strategy matching operations in the fault detection strategy matrix set that has been constructed. In the matching process, the Euclidean distance is used to calculate the difference between the real-time data vector and each strategy feature vector in the matrix, and the matching item with the smallest distance is used as the determination standard. Finally, the closest matching fault detection strategy to the current real-time sampling state is determined by this method.

[0076] Step S400: detecting and analyzing the voltage sampling unit based on the matching fault detection strategy to determine the fault detection result.

[0077] In the embodiments of the present application, the real-time voltage signal of the voltage sampling unit is detected and analyzed based on the matching fault detection strategy. First, according to the matched fault detection strategy, the corresponding detection method and rule are selected. For example, if the matched fault mode is voltage drift, the amplitude change detection method is used to judge whether there is voltage drift phenomenon by comparing the change of signal amplitude. If the voltage change exceeds the preset range, it is considered that the signal has a fault.

[0078] Then the key features such as peak voltage, frequency offset, waveform change, etc. are extracted from the real-time sampling signal, and then threshold judgment is performed. For voltage drift fault, a tolerance interval of voltage change is set, such as ±10%, if the real-time signal exceeds the interval, the fault detection is triggered. Similarly, for signal distortion or other fault modes, the corresponding detection method (such as spectrum analysis, waveform distortion detection, etc.) is applied for fault recognition.

[0079] Finally, the fault detection result is output according to the detection result to determine whether the signal has a fault type (such as voltage drift, signal distortion, etc.).

[0080] In the embodiments of the present application, as described above, the embodiments of the present application have at least the following technical effects:

[0081] The present application uses a surge generator to simulate multi-element surge impact on the voltage sampling unit to obtain an impact simulation data set; acquires a voltage sampling unit fault common knowledge base, identifies a fault detection strategy in combination with the impact simulation data set, and takes the identification result as a discrete hidden variable to construct a fault detection strategy matrix set; extracts real-time sampling feature data of the voltage sampling unit, indexes the real-time sampling feature data in the fault detection strategy matrix set to perform strategy matching, and determines a matching fault detection strategy; and detects and analyzes the voltage sampling unit based on the matching fault detection strategy to determine a fault detection result. The present application solves the technical problem that the existing technology cannot accurately identify the function drift and abnormal characteristics of the voltage sampling unit under the action of surge impact, leading to untimely and low-precision fault identification. By constructing an impact simulation data set and generating a detection strategy in combination with a fault common knowledge base, strategy matching and fault analysis with real-time feature data are realized to achieve the technical effects of improving the accuracy and real-time performance of voltage sampling unit fault identification under surge impact.

[0082] Embodiment two, based on the same inventive concept as the voltage sampling unit fault detection method based on surge impact analysis in the foregoing embodiments, such as Figure 2As shown, the application provides a voltage sampling unit fault detection system based on surge impact analysis, and the system and method embodiments in the application are based on the same inventive concept. The system includes:

[0083] A surge impact simulation module 11 is configured to simulate multi-element surge impact on the voltage sampling unit by using a surge generator to obtain a set of impact simulation data. A strategy identification module 12 is configured to obtain a voltage sampling unit fault common sense library, identify a fault detection strategy in combination with the set of impact simulation data, and construct a fault detection strategy matrix set by taking the identification result as a discrete hidden variable. A strategy matching module 13 is configured to extract real-time sampling feature data of the voltage sampling unit, perform strategy matching in the fault detection strategy matrix set by taking the real-time sampling feature data as an index, and determine a matched fault detection strategy. A detection analysis module 14 is configured to perform detection analysis on the voltage sampling unit based on the matched fault detection strategy, and determine a fault detection result.

[0084] Further, the system is further configured to implement the following functions:

[0085] Based on multi-element scene analysis, a set of multi-element surge generator output parameters is generated, wherein each multi-element surge generator output parameter includes a surge waveform type, an impact amplitude, and an impact time. The voltage sampling unit is connected to the surge generator output end, the surge waveform type, the impact amplitude, and the impact time of the surge generator are adjusted according to the set of multi-element surge generator output parameters, and an oscilloscope, a data acquisition card, and a spectrum analyzer are configured to collect voltage sampling signals to obtain a set of simulated voltage sampling signals. Abnormal symptom capture data is obtained by capturing abnormal symptoms of the set of simulated voltage sampling signals. A set of function drift data is determined by performing function drift authentication analysis on the set of simulated voltage sampling signals. The set of abnormal symptom capture data, the set of function drift data, and the set of simulated voltage sampling signals are respectively mapped and combined to obtain a set of impact simulation data.

[0086] Further, the system is further configured to implement the following functions:

[0087] The data retrieval of the historical multi-element fault scene data set is performed with the surge impact fault type, the surge waveform type, the impact amplitude and the impact time as indexes, to obtain a historical surge impact fault type set, a historical surge waveform type set, a historical impact amplitude set and a historical impact time set; the historical surge impact fault type is classified by a decision tree to generate M divided historical surge impact fault type sets, wherein M is a positive integer; the historical surge waveform type set, the historical impact amplitude set and the historical impact time set are mapped and divided based on the M divided historical surge impact fault type sets to generate M divided historical surge waveform type sets, M divided historical impact amplitude sets and M divided historical impact time sets; M three-dimensional spaces are constructed, and the M divided historical surge waveform type sets, the M divided historical impact amplitude sets and the M divided historical impact time sets are subjected to surge impact scene calibration analysis in the M three-dimensional spaces respectively to obtain M surge impact scene calibration parameters, and the M surge impact scene calibration parameters are summarized as the multi-element surge generator output parameter set.

[0088] Further, the system is also used to implement the following functions:

[0089] The M divided historical surge waveform type sets, the M divided historical impact amplitude sets and the M divided historical impact time sets are respectively mapped to M three-dimensional spaces to obtain M mapping vector sets; M initial mapping vectors are respectively randomly selected from the M mapping vector sets, a scene differentiation loss function is constructed to analyze the M initial mapping vectors, and M initial scene differentiation loss amounts are obtained; M iterative mapping vectors are again respectively randomly selected from the M mapping vector sets, and the M iterative mapping vectors are analyzed by using the scene differentiation loss function to obtain M iterative scene differentiation loss amounts; the M initial mapping vectors and the M iterative mapping vectors are selected and iteratively updated by comparing the M iterative scene differentiation loss amounts and the M initial scene differentiation loss amounts, and M calibrated mapping vectors are obtained; elements of the M calibrated mapping vectors are extracted to obtain the M surge impact scene calibration parameters.

[0090] Further, the system is also used to implement the following functions:

[0091] It is judged whether the M iterative scene differentiation loss amounts are greater than or equal to the M initial scene differentiation loss amounts, if yes, a difference degree of the M iterative scene differentiation loss amounts and the M initial scene differentiation loss amounts is calculated, when a calculation result is less than a preset difference degree threshold, a direction from the M initial mapping vectors to the M iterative mapping vectors is taken as an update direction, the M iterative mapping vectors are randomly updated in the M mapping vector sets until a preset update number is met, and the M calibrated mapping vectors are obtained.

[0092] Further, the system is also used to realize the following functions:

[0093] The scene differentiation loss function is:

[0094] ; wherein, is the initial scene differentiation loss amount, is the number of mapping vectors in the mapping vector set corresponding to the initial mapping vector for scene differentiation loss analysis, is the distance between the initial mapping vector for scene differentiation loss analysis and the jth mapping vector in the corresponding mapping vector set, is the number of initial mapping vectors in the comparison set formed by eliminating the initial mapping vector from the M initial mapping vectors, is the distance between the jth mapping vector and the ith initial mapping vector in the comparison set.

[0095] Further, the system is also used to realize the following functions:

[0096] The analog voltage sampling signals in the analog voltage sampling signal set are compared with the standard voltage sampling signal, and data outside the preset tolerance interval is taken as functional drift data, and the functional drift data set is obtained.

[0097] Further, the system is also used to realize the following functions:

[0098] Extract the abnormal symptom capture data set in the impact simulation data set; use the voltage sampling unit fault common sense library to perform multiple fault directionality identification on the abnormal symptom capture data set, if it is identified that there is multiple fault direction, if it is identified that there is multiple fault direction, then use the voltage sampling unit fault common sense library to perform directional data missing analysis on the impact simulation data set, and according to the analysis result, data extraction is performed from the analog voltage sampling signal set to obtain a complete impact simulation data set; extract the detection index of the complete impact simulation data set, take the extraction result as a hidden variable, and construct a fault detection strategy matrix set.

[0099] Further, the system is also used to realize the following functions:

[0100] If it is identified that there is no multiple fault direction, then the impact simulation data set is extracted for detection index, and the extraction result is taken as a hidden variable to construct a fault detection strategy matrix set.

[0101] 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. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0102] 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.

[0103] The specification and drawings of the present application are merely exemplary descriptions of the present application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. A method for detecting a fault in a voltage sampling unit based on surge impact analysis, characterized by, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises:

2. The surge-impulse-analysis-based voltage-sampling-unit fault detection method according to claim 1, wherein, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: Perform data retrieval on the historical multi-element fault scene data set with surge impact fault type, surge waveform type, impact amplitude and impact time as indexes to obtain a historical surge impact fault type set, a historical surge waveform type set, a historical impact amplitude set and a historical impact time set; Classify the historical surge impact fault type by decision tree to generate M divided historical surge impact fault type sets, wherein M is a positive integer; Map and divide the historical surge waveform type set, historical impact amplitude set and historical impact time set based on the M divided historical surge impact fault type sets to generate M divided historical surge waveform type sets, M divided historical impact amplitude sets and M divided historical impact time sets; Construct M three-dimensional spaces, and perform surge impact scene calibration analysis on the M divided historical surge waveform type sets, M divided historical impact amplitude sets and M divided historical impact time sets in the M three-dimensional spaces respectively to obtain M surge impact scene calibration parameters, and aggregate the M surge impact scene calibration parameters into the multi-element surge generator output parameter set.

3. The surge-impulse-analysis-based voltage-sampling-unit fault detection method of claim 2, wherein, Construct M three-dimensional spaces, and perform surge impact scene calibration analysis on the M divided historical surge waveform type sets, M divided historical impact amplitude sets and M divided historical impact time sets in the M three-dimensional spaces respectively to obtain M surge impact scene calibration parameters, including: Map the M divided historical surge waveform type sets, M divided historical impact amplitude sets and M divided historical impact time sets to M three-dimensional spaces respectively to obtain M mapping vector sets; Randomly select M initial mapping vectors from the M mapping vector sets respectively, construct a scene differentiation loss function to analyze the M initial mapping vectors, and obtain M initial scene differentiation loss amounts; Randomly select M iteration mapping vectors from the M mapping vector sets again, and analyze the M iteration mapping vectors by using the scene differentiation loss function to obtain M iteration scene differentiation loss amounts; Compare the M iteration scene differentiation loss amounts and the M initial scene differentiation loss amounts, select and iteratively update the M initial mapping vectors and the M iteration mapping vectors, and obtain M calibration mapping vectors; Extract elements of the M calibration mapping vectors to obtain the M surge impact scene calibration parameters.

4. The surge-impulse-analysis-based voltage-sampling-unit fault detection method of claim 3, wherein, Determine whether the M iteration scene differentiation loss amounts are greater than or equal to the M initial scene differentiation loss amounts, if yes, calculate the difference degree of the M iteration scene differentiation loss amounts and the M initial scene differentiation loss amounts, when the calculation result is less than a preset difference degree threshold, take the direction from the M initial mapping vectors to the M iteration mapping vectors as an update direction, randomly update the M iteration mapping vectors in the M mapping vector sets, until a preset update number is met, and obtain the M calibration mapping vectors.

5. The surge-impulse-analysis-based voltage-sampling-unit fault detection method according to claim 3, wherein, The scene differentiation loss function is: ； wherein, is the loss amount of the initial scene, is the number of mapping vectors in the mapping vector set corresponding to the initial mapping vector for the loss analysis of the scene division, is the distance between the initial mapping vector for the loss analysis of the scene division and the jth mapping vector in the corresponding mapping vector set, is the number of initial mapping vectors in the comparison set composed after the initial mapping vector for the loss analysis of the scene division is removed from the M initial mapping vectors, is the distance between the jth mapping vector and the ith initial mapping vector in the comparison set.

6. The surge-impulse-analysis-based voltage-sampling-unit fault detection method according to claim 1, wherein, The analog voltage sampling signals in the analog voltage sampling signal set are compared with a standard voltage sampling signal, and data beyond a preset tolerance interval is taken as function drift data, to obtain the function drift data set.

7. A voltage sampling unit fault detection system based on surge impact analysis, characterized by, The system is used to execute the voltage sampling unit fault detection method based on surge impact analysis as claimed in any one of claims 1-6, and the system comprises: A surge impact simulation module is used to simulate multi-element surge impact on the voltage sampling unit by using a surge generator, to obtain an impact simulation data set; A strategy recognition module is used to acquire a voltage sampling unit fault common sense library, to recognize fault detection strategies in combination with the impact simulation data set, to take the recognition result as a discrete hidden variable, and to construct a fault detection strategy matrix set; A strategy matching module is used to extract real-time sampling representation data of the voltage sampling unit, to take the real-time sampling representation data as an index, to perform strategy matching in the fault detection strategy matrix set, and to determine a matching fault detection strategy; A detection analysis module is used to perform detection analysis on the voltage sampling unit based on the matching fault detection strategy, to determine a fault detection result.

Images