Method for monitoring spring failure of GIS circuit breaker based on MEMS sensor

Abstract

The application relates to the technical field of power equipment monitoring, in particular to a GIS circuit breaker spring fault monitoring method based on a MEMS sensor, which solves the technical problem of low accuracy of circuit breaker fault monitoring in the prior art. The method comprises the following steps: performing double-channel separation processing on vibration signals in the operation process of the circuit breaker to obtain a total energy index and a normalized energy spectrum. A set of measured feature points is extracted from the normalized energy spectrum. Based on a preset transmission algorithm, the transmission cost between the set of measured feature points and a set of reference benchmark feature points is calculated, and the transmission cost is decomposed into a stroke lag index and a structure looseness index. According to the total energy index, the stroke lag index and the structure looseness index, the fault state of the circuit breaker is determined.

Description

A MEMS sensor-based method for monitoring spring faults in GIS circuit breakers Technical Field

[0001] This application relates to the field of power equipment monitoring technology, specifically to a method for monitoring spring faults in GIS circuit breakers based on MEMS sensors. Background Technology

[0002] The spring-operated mechanism of a gas-insulated switchgear (GIS) circuit breaker is a core component of power grid control. At the end of the circuit breaker's opening and closing action, the moving contact system carries enormous kinetic energy, which must be absorbed and braked by a hydraulic buffer to prevent rigid impact from damaging the insulation structure. Currently, the industry commonly uses micro-electro-mechanical systems (MEMS) accelerometers to collect vibration signals from the surface of the mechanism housing, and analyzes the vibration waveforms to monitor the mechanical condition.

[0003] However, the characteristics of different types of faults in vibration signals are easily overlapping, making it difficult to accurately distinguish fault types. Furthermore, factors such as changes in ambient temperature can interfere with monitoring results, and newly commissioned or single-phase operating equipment lacks effective reference benchmarks. These factors result in low accuracy in circuit breaker fault monitoring, ultimately leading to inefficient operation and maintenance, and missed or incorrect fault diagnoses. Therefore, improving the accuracy of circuit breaker fault monitoring has become an urgent technical problem to be solved. Summary of the Invention

[0004] To address the low accuracy of fault monitoring in existing circuit breaker technologies, this application aims to provide a method for monitoring spring faults in GIS circuit breakers based on MEMS sensors. The specific technical solution adopted is as follows:

[0005] The vibration signal during circuit breaker operation is processed by dual-channel separation to obtain the total energy index and the normalized energy spectrum. The total energy index is used to characterize the absolute energy of the vibration signal, and the normalized energy spectrum is used to characterize the relative probability distribution of the vibration energy in the time and frequency domain.

[0006] The measured feature point set is extracted from the normalized energy spectrum. The measured feature point set includes multiple feature points, each of which has time coordinates, frequency coordinates and initial energy amplitude.

[0007] Based on the preset transmission algorithm, the transmission cost between the measured feature point set and the reference benchmark feature point set is calculated, and the transmission cost is decomposed into a travel lag index and a structural looseness index. Among them, the travel lag index is used to characterize the time lag of the circuit breaker operation, and the structural looseness index is used to characterize the looseness of the mechanical structure.

[0008] The fault status of the circuit breaker is determined based on the total energy index, travel lag index, and structural looseness index.

[0009] In one possible implementation, the vibration signal during circuit breaker operation is processed through dual-channel separation to obtain a total energy index and a normalized energy spectrum. This includes: determining the start time of mechanical action based on the current signal of the circuit breaker's opening and closing coils; collecting vibration signals within a preset time period from the start time as the vibration signals during circuit breaker operation; performing time-frequency transformation on the vibration signals to obtain the original time-frequency energy matrix; pruning the endpoints of the original time-frequency energy matrix to obtain the effective time-frequency energy matrix; calculating the sum of the values ​​of all elements in the effective time-frequency energy matrix as the total energy index; and dividing each element in the effective time-frequency energy matrix by the total energy index to obtain the normalized energy spectrum.

[0010] In one possible implementation, determining the start time of mechanical action based on the current signal of the circuit breaker's opening and closing coils includes: monitoring the current waveform of the opening and closing coils; determining the start time based on the moment when the current waveform meets preset conditions; the preset conditions include: the current amplitude has exceeded the peak value and decayed to below a first predetermined proportion of the peak value, and the absolute value of the current change rate is less than a preset threshold.

[0011] In one possible implementation, before calculating the transmission cost between the measured feature point set and the reference reference feature point set based on a preset transmission algorithm, the method further includes: when the circuit breaker performs a three-phase linkage operation, calculating the correlation between the normalized energy spectra of the other two phase circuit breakers in the same group; if the correlation is greater than or equal to a preset consistency threshold, then using the average of the normalized energy spectra of the other two phases as the reference reference spectra; if the correlation is less than the preset consistency threshold, then using the historical normalized energy spectra of the circuit breaker as the reference reference spectra; and extracting the reference feature point set from the reference reference spectra to obtain the reference reference feature point set.

[0012] In one possible implementation, the measured feature point set is extracted from the normalized energy spectrum, including: determining multiple target feature points in the normalized energy spectrum; each target feature point has a time coordinate, a frequency coordinate, and an initial energy amplitude; calculating the neighborhood energy dispersion of each target feature point; the neighborhood energy dispersion is used to characterize the uniformity of energy distribution in a local region centered on the target feature point; performing intra-set normalization on the initial energy amplitudes of multiple target feature points to obtain the energy weight of each target feature point; and constructing the measured feature point set based on the time coordinate, frequency coordinate, energy weight, and neighborhood energy dispersion.

[0013] In one possible implementation, determining multiple target feature points in the normalized energy map includes: performing a local maximum search in the normalized energy map to select candidate points whose energy amplitude is greater than other elements in the local neighborhood and greater than a predetermined proportion of the global maximum value; sorting the candidate points in descending order of energy amplitude, and selecting the top K points as multiple target feature points, where K is a preset positive integer.

[0014] In one possible implementation, the neighborhood energy dispersion of each target feature point is calculated by: extracting a neighborhood matrix of a preset size from the normalized energy map centered on the target feature point; normalizing the values ​​in the neighborhood matrix to form a local probability distribution; and using the Shannon entropy value of the local probability distribution as the neighborhood energy dispersion of the target feature point.

[0015] In one possible implementation, based on a pre-defined transmission algorithm, the transmission cost between the measured feature point set and the reference feature point set is calculated, and the transmission cost is decomposed into a travel lag index and a structural slack index. This includes: constructing a transmission cost matrix, where each element represents the unit matching cost required to match a feature point in the measured feature point set to a feature point in the reference feature point set; based on the transmission cost matrix, solving for the optimal transmission scheme that minimizes the total transmission cost between the measured feature point set and the reference feature point set; and according to the optimal transmission scheme, decomposing the total transmission cost into components related to time deviation and components related to waveform discreteness, weighting the components related to time deviation to obtain the travel lag index, and weighting the components related to waveform discreteness to obtain the structural slack index.

[0016] In one possible implementation, the matching cost in the transmission cost matrix includes a time deviation cost and a waveform discrepancy cost. Constructing the transmission cost matrix includes: calculating the time coordinate difference between two feature points and nonlinearly amplifying the time difference to obtain the time deviation cost; calculating the neighborhood energy discrepancy difference between two feature points and nonlinearly amplifying the discrepancy difference to obtain the waveform discrepancy cost; and constructing the transmission cost matrix based on the time deviation cost and the waveform discrepancy cost.

[0017] In one possible implementation, the fault state of the circuit breaker is determined based on the total energy index, the travel hysteresis index, and the structural looseness index, including: comparing the total energy index with a reference total energy value; if the total energy index exceeds a preset range based on the reference total energy value, an energy abnormality fault is determined; comparing the travel hysteresis index with a hysteresis threshold; if the travel hysteresis index is greater than the hysteresis threshold, an abnormal damping characteristic of the buffer is determined; and comparing the structural looseness index with a looseness threshold; if the structural looseness index is greater than the looseness threshold, a mechanical structural looseness fault is determined.

[0018] This application offers the following advantages: It obtains the absolute energy and relative distribution information of vibration signals through dual-channel separation processing, ensuring the integrity of fault characteristic information; it effectively reduces the dimensionality of high-dimensional data by extracting feature point sets containing multi-dimensional attributes; it decomposes the transmission cost into two orthogonal fault indicators using a preset transmission algorithm, achieving accurate differentiation of different fault types; and finally, it comprehensively judges the fault state based on three independent indicators, significantly improving the accuracy of fault diagnosis. This method solves the problems of fault feature aliasing and weak diagnostic targeting in existing technologies, effectively improving the accuracy of circuit breaker fault monitoring, providing maintenance personnel with clear criteria for fault type judgment, and improving maintenance efficiency and equipment reliability. Attached Figure Description

[0019] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 is a flowchart illustrating a method for monitoring spring faults in a GIS circuit breaker based on a MEMS sensor, according to an embodiment of this application.

[0021] Figure 2 is a schematic diagram of the system architecture of a GIS circuit breaker spring fault monitoring system based on MEMS sensors provided in one embodiment of this application. Detailed Implementation

[0022] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a GIS circuit breaker spring fault monitoring method based on MEMS sensors proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0024] In all division and logarithmic operations covered in this application, a smoothing mechanism is employed to prevent computer program crashes or invalid values ​​from being generated due to a zero denominator or zero input. Specifically, a positive correction factor ε (e.g., 0.001) is superimposed on the denominator term of the division operation or the argument term of the logarithmic function; thereby ensuring the robustness and feasibility of the algorithm under extreme conditions.

[0025] Unless otherwise specified, the normalization function Norm() mentioned in this application uses maximum and minimum value normalization. The maximum and minimum values ​​are preset empirical extreme values ​​derived from a large amount of historical experimental data. If the calculation result exceeds the [0,1] interval, a truncation function is used to limit it to the [0,1] range (i.e., if the result is less than 0, it is taken as 0; if it is greater than 1, it is taken as 1) to eliminate the influence of outliers on the evaluation index.

[0026] The following description, in conjunction with the accompanying drawings, details the specific scheme of the GIS circuit breaker spring fault monitoring method based on MEMS sensors provided in this application.

[0027] Please refer to Figure 1, which shows a flowchart of a method for monitoring spring faults in a GIS circuit breaker based on a MEMS sensor according to an embodiment of this application. As shown in Figure 1, the method includes the following steps:

[0028] Step 101: Perform dual-channel separation processing on the vibration signal during the circuit breaker operation to obtain the total energy index and normalized energy spectrum.

[0029] Among them, the total energy index is used to characterize the absolute energy magnitude of the vibration signal, and the normalized energy spectrum is used to characterize the relative probability distribution of vibration energy in the time and frequency domain.

[0030] Optionally, during the circuit breaker's opening and closing operations, vibration signals are collected using a MEMS accelerometer mounted on the surface of the main beam of the mechanism box. The MEMS accelerometer is a commonly used miniature sensor for vibration monitoring, characterized by its small size, fast response, and high accuracy, enabling real-time capture of vibration changes during circuit breaker operation. The sampling rate of the MEMS accelerometer can be set according to the circuit breaker's operating characteristics; for example, a sampling rate of 10kHz can be set to ensure complete capture of vibration details throughout the entire operation. The collected vibration signals are discrete-time series and require preliminary filtering (such as low-pass filtering) to remove environmental noise interference. The filter cutoff frequency can be set according to the main frequency range of the circuit breaker's vibration signals, for example, 500Hz, to retain effective signal components.

[0031] It should be noted that the purpose of the aforementioned dual-channel separation processing is to obtain the absolute energy information and relative distribution information of the vibration signal separately, avoiding monitoring defects caused by single-dimensional information. Among them, the total energy index is a direct representation of the absolute energy magnitude of the vibration signal, which can reflect the impact intensity during mechanical action and is of great significance for identifying destructive faults such as rigid impacts; the normalized energy spectrum focuses on the relative probability distribution of vibration energy in the time and frequency domain, eliminating the influence of absolute amplitude, which facilitates subsequent precise waveform structure analysis.

[0032] Step 102: Extract the set of measured feature points from the normalized energy spectrum.

[0033] The measured feature point set includes multiple feature points, each with time coordinates, frequency coordinates, and initial energy amplitude.

[0034] It should be noted that the normalized energy spectrum is high-dimensional matrix data. Direct calculation is not only inefficient but also susceptible to noise interference. Therefore, it is necessary to extract key measured feature points for dimensionality reduction. Each feature point contains three core attributes: time coordinate, frequency coordinate, and initial energy amplitude. The time coordinate represents the moment when the vibration occurred at the feature point, which is related to the time process of the circuit breaker's operation. The frequency coordinate represents the vibration frequency at the feature point, reflecting the frequency characteristics of the mechanical action. The initial energy amplitude represents the energy intensity at the feature point, reflecting the severity of the vibration at that moment and frequency.

[0035] In one possible implementation, this application can determine feature points through methods such as local maximum search and energy threshold screening. The selected feature points are then used as the set of measured feature points to ensure that the extracted feature points can accurately characterize the key features of the normalized energy spectrum, providing a reliable data foundation for subsequent transmission cost calculation. Optionally, the extraction of feature points should follow the principles of highlighting key information and suppressing noise interference, selecting core points that can reflect fault characteristics through specific screening rules (such as local maximum search and energy threshold screening).

[0036] Step 103: Based on the preset transmission algorithm, calculate the transmission cost between the measured feature point set and the reference benchmark feature point set, and decompose the transmission cost into a travel lag index and a structural looseness index.

[0037] Among them, the travel lag index is used to characterize the time lag of the circuit breaker's operation, and the structural looseness index is used to characterize the looseness of the mechanical structure.

[0038] The aforementioned reference feature point set is standard data used for comparison with the measured feature point set. It originates from the vibration characteristics of the circuit breaker under normal operating conditions and can reflect the health status of the equipment. The preset transmission algorithm is used to construct the matching relationship between the measured feature point set and the reference feature point set, and calculate the transmission cost between the two. This cost comprehensively reflects the differences between the two feature point sets in dimensions such as time, frequency, and energy.

[0039] Optionally, the preset transmission algorithm is an algorithm based on optimal transmission theory. This type of algorithm can effectively handle the matching problem between two sets of distributions, ensuring the accuracy and rationality of the transmission cost calculation. For example, the preset transmission algorithm is the Sinkhorn algorithm, which iteratively scales rows and columns and solves for the optimal transmission flow matrix.

[0040] It should be noted that the travel lag index and the structural looseness index are the results of orthogonal decomposition of transmission cost, and correspond to different types of fault characteristics respectively: the travel lag index mainly characterizes the time lag of the circuit breaker action, which is related to the damping characteristics of the hydraulic buffer. When the buffer oil leaks, deteriorates, or changes in viscosity, the circuit breaker action time will shift, and the index will change accordingly; the structural looseness index mainly characterizes the looseness of the mechanical structure, which is related to the linkage pin clearance, bolt preload, etc. When the mechanical structure is loose, the vibration waveform will exhibit divergent characteristics, and the index will increase significantly.

[0041] Step 104: Determine the fault status of the circuit breaker based on the total energy index, travel lag index, and structural looseness index.

[0042] The three indicators mentioned above—total energy index, stroke hysteresis index, and structural looseness index—reflect the operating status of the circuit breaker from different dimensions. They complement each other, do not interfere with each other, and together constitute the basis for fault diagnosis. The total energy index is used to judge abnormal energy faults, such as rigid impact caused by complete buffer failure or insufficient work due to insufficient spring energy storage; the stroke hysteresis index is used to judge abnormal buffer damping characteristics, such as damping failure caused by oil deterioration or leakage; and the structural looseness index is used to judge mechanical structural looseness faults, such as excessive clearance caused by pin wear or loose bolts.

[0043] As one approach, the judgment process requires setting corresponding judgment criteria for each indicator. These criteria can be determined based on the equipment's factory technical parameters, historical operating data, or industry standards. For example, by statistically analyzing the value range of each indicator under normal operating conditions, a reasonable threshold or threshold range can be set. When the measured indicator exceeds the corresponding range, it is judged as a fault of the corresponding type.

[0044] Based on the above technical solution, this application obtains the absolute energy and relative distribution information of vibration signals through dual-channel separation processing, ensuring the integrity of fault feature information; by extracting feature point sets containing multi-dimensional attributes, it achieves effective dimensionality reduction of high-dimensional data; by using a preset transmission algorithm to decompose the transmission cost into two orthogonal fault indicators, it achieves accurate differentiation of different fault types; finally, it comprehensively judges the fault status based on three independent indicators, significantly improving the accuracy of fault diagnosis. This method solves the problems of fault feature aliasing and weak diagnostic targeting in existing technologies, effectively improving the accuracy of circuit breaker fault monitoring, providing maintenance personnel with clear criteria for fault type judgment, and improving maintenance efficiency and equipment reliability.

[0045] In one possible implementation, the process of performing dual-channel separation processing on the vibration signal during circuit breaker operation in step 101 above to obtain the total energy index and normalized energy spectrum specifically includes:

[0046] Step 201: Determine the start time of mechanical action based on the current signal of the circuit breaker's opening and closing coils.

[0047] The opening and closing actions of the circuit breaker are controlled by the opening and closing coil, and the changes in the coil current signal can accurately reflect the starting state of the mechanical action. Since the response time of the relay and the action time of the electromagnet in the circuit breaker control circuit are discrete, directly using the moment when the control command is issued as the synchronization reference will lead to a large phase error. Therefore, the coil current characteristics are used to lock the absolute zero point of the mechanical action.

[0048] Optionally, this application can monitor the current signal of the opening and closing coils in real time using a Hall current sensor. Hall current sensors are characterized by fast response speed and high measurement accuracy, and can accurately capture changes in the current waveform. The monitored current signal needs to be filtered to remove circuit noise interference and ensure the accuracy of feature recognition.

[0049] In one possible implementation, this step can be specifically implemented as follows: monitoring the current waveform of the opening and closing coils; determining the start time based on the moment when the current waveform meets preset conditions; the preset conditions include: the current amplitude has exceeded the peak value and decayed to below a first predetermined proportion of the peak value, and the absolute value of the current change rate is less than a preset threshold.

[0050] It should be noted that the changing pattern of the current waveform is closely related to the operation of the circuit breaker's electromagnet: when the electromagnet is engaged, the coil current rapidly rises to its peak value; when the electromagnet is released, the coil current gradually decays. Therefore, the peak and decay phases of the current waveform are key to determining the start time of the mechanical action. Based on this, preset conditions ensure the accuracy of the start time from two dimensions: current amplitude and current change rate.

[0051] The current amplitude condition is: the current amplitude has exceeded the peak value and decayed to below a first predetermined proportion of the peak value. Peak current. This is a hallmark characteristic of electromagnet engagement; after exceeding the peak value, the current begins to decay, indicating that the electromagnet is about to release, and the mechanical linkage mechanism begins to move. The setting of the first predetermined proportion needs to be determined based on the circuit breaker's coil characteristics and operating mechanism. For example, the first predetermined proportion can be set to 5%, meaning that when the current decays to below 5% of the peak value, the electromagnet is considered to have completed its release, and the mechanical action officially begins. This proportion setting needs to balance sensitivity and reliability; too large a proportion may lead to premature judgment of the start time, while too small a proportion may lead to premature judgment. It should be noted that this first predetermined proportion can be determined based on statistical analysis of the current decay curves of typical circuit breaker opening and closing coils, reliably indicating the completion time of electromagnet release.

[0052] The current change rate condition is that the absolute value of the current change rate is less than a preset threshold. The current change rate reflects the rate of current decay. When the current decays to a certain level, the change rate will tend to level off, indicating that the coil current has stabilized at a low level and the electromagnet's operating state no longer changes. The preset threshold needs to be determined based on the coil's electrical parameters. For example, the preset threshold can be set to 0.5 A / ms (Amperes per millisecond), that is, when... When this condition is met, the threshold effectively eliminates fluctuations during the current decay process, ensuring the stability of the initial timing determination.

[0053] The moment is recorded as follows: The current waveform satisfies both of the above preset conditions simultaneously. Considering the buffer duration, this application will Before The moment that is Marked as the start time of the mechanical action. The time specified is the buffer duration. By using a dual-condition judgment, misjudgments that might occur under a single condition are effectively avoided, ensuring accurate locking of the start time and providing a reliable benchmark for subsequent synchronous acquisition of vibration signals. It should be noted that the buffer duration... The setting needs to be greater than the typical duration of the endpoint effect of time-frequency transformation (such as HHT). Based on experience, it is generally taken as 10-20ms. In this embodiment, it is set to 10ms.

[0054] Step 202: Collect vibration signals within a preset time period starting from the start time, as the vibration signals during the operation of the circuit breaker.

[0055] With locked Based on this, the system initiates the data acquisition task using the MEMS accelerometer mounted on the surface of the main beam of the mechanism box. To eliminate the endpoint fly-out effect that may be generated by the subsequent signal processing algorithm (Hilbert-Huang transform), the system sets the acquisition duration to include the actual analysis window and the front and rear buffers.

[0056] Specifically, the effective analysis time defined by the system is: (In this embodiment, it is set to 100ms to cover the entire circuit breaker operation process). The actual signal length intercepted by the system is... ,in The buffer duration is set to 10ms in this embodiment. The system collects and saves data from time [time]. arrive The discrete-time series of accelerations, denoted as the original acceleration series. . This is the preset duration mentioned above; the effective analysis time is... It should cover the entire process of the circuit breaker from the start of its operation to its complete stop, and can be determined based on statistics of the typical operating time of the equipment model.

[0057] Step 203: Perform time-frequency transformation on the vibration signal to obtain the original time-frequency energy matrix.

[0058] As an example, the time-frequency transformation described above is the Hilbert-Huang transform (HHT). In other words, the system transforms the original acceleration sequence... Perform the Hilbert-Huang transform (HHT) to obtain a high-resolution time-frequency distribution.

[0059] Specifically, this application uses the Empirical Mode Decomposition (EMD) algorithm to... The function is decomposed into several intrinsic mode functions (IMF) components. The principal IMF components (usually the first three) with the highest energy contribution rates are selected, and a Hilbert transform is performed on each selected IMF component to calculate its instantaneous frequency. and instantaneous amplitude .

[0060] Construct a two-dimensional time-frequency grid container. The horizontal axis represents time (2ms resolution), and the vertical axis represents frequency (50Hz resolution). Traverse the entire sequence and calculate the square of the instantaneous amplitude at each moment. Accumulate the energy into the corresponding time-frequency grid to generate the original time-frequency energy matrix. Original time-frequency energy matrix The timeline covers the entire length including the buffer. .

[0061] Step 204: Trim the endpoints of the original time-frequency energy matrix to obtain the effective time-frequency energy matrix.

[0062] Optionally, to eliminate data divergence (endpoint effect) at the beginning and end of the sequence caused by EMD decomposition, the system performs a pruning operation, including: directly removing the matrix corresponding to the beginning of the time axis. Time period and after For all columns within a time period, only the middle corresponding column is retained. Data for a given time period. The cropped matrix is ​​denoted as the effective time-frequency matrix. .

[0063] Step 205: Calculate the sum of the values ​​of all elements in the effective time-frequency energy matrix as the total energy index.

[0064] As an example, the total energy index Satisfy the following formula:

[0065]

[0066] in, The value of the element in the i-th row and j-th column of the effective time-frequency energy matrix represents the vibration energy at the corresponding time and frequency point in the matrix. N is the number of rows in the effective time-frequency energy matrix, and M is the number of columns in the effective time-frequency energy matrix.

[0067] Step 206: Divide each element in the effective time-frequency energy matrix by the total energy index to obtain the normalized energy spectrum.

[0068] Specifically, in order to eliminate the common amplitude scaling error introduced by changes in oil viscosity due to changes in ambient temperature and sensor sensitivity drift, this application addresses... The total energy percentage is calculated by dividing the value of each element in the matrix by... The vibration energy density spectrum was obtained. As an example, vibrational energy density maps Satisfy the following formula:

[0069]

[0070] Based on this formula, The sum of all elements in the formula is always equal to 1. This spectrum only represents the relative probability distribution of energy in the time-frequency domain (i.e., waveform shape and phase), and is independent of its absolute amplitude, thus eliminating temperature drift interference. These are parameter tuning coefficients, and their values ​​should be extremely small positive numbers (e.g., 0.01) to avoid denominators of 0.

[0071] Based on the above technical solution, this embodiment accurately locks the start time of mechanical action using coil current signals, ensuring the synchronization of vibration signal acquisition; by reasonably setting the acquisition duration and buffer, the integrity of the vibration signal is guaranteed; time-frequency transformation is used to convert the time-domain signal into a time-frequency domain matrix, achieving synchronous representation of time and frequency dimension features; endpoint clipping eliminates the endpoint effect of time-frequency transformation, improving data reliability; and the calculation of the total energy index and normalized energy spectrum ensures the standardization and operability of the dual-channel separation processing. This application further refines the steps of dual-channel separation processing, making the feature extraction of vibration signals more accurate and laying a solid data foundation for subsequent fault diagnosis.

[0072] In one possible implementation, the process of extracting the measured feature point set from the normalized energy spectrum in step 102 above specifically includes:

[0073] Step 301: Determine multiple target feature points in the normalized energy spectrum.

[0074] Each target feature point has a time coordinate, a frequency coordinate, and an initial energy amplitude.

[0075] Optionally, this step can be implemented as follows: performing a local maximum search in the normalized energy spectrum to filter candidate points whose energy amplitude is greater than other elements in the local neighborhood and greater than a predetermined proportion of the global maximum; sorting the candidate points in descending order of energy amplitude, and selecting the top... Each point serves as a feature point for multiple targets. It is a preset positive integer.

[0076] Specifically, in the measured normalized energy spectrum Perform a local maximum search. Define the search window as... Within the grid area, candidate points with amplitudes greater than other elements within the window and greater than 10% of the maximum value of the entire spectrum are selected.

[0077] Candidate points are sorted in descending order of energy amplitude, and the first few are selected. These points constitute the measured feature point set. In this embodiment, All are set to 20. Each feature point It contains three attributes: time coordinates Frequency coordinates and initial energy amplitude .

[0078] Step 302: Calculate the neighborhood energy dispersion of each target feature point.

[0079] Among them, the neighborhood energy dispersion is used to characterize the uniformity of energy distribution in a local region centered on the target feature point.

[0080] Optionally, this step can be implemented as follows: taking the target feature point as the center, extracting a neighborhood matrix of a preset size from the normalized energy spectrum; normalizing the values ​​in the neighborhood matrix to form a local probability distribution; and using the Shannon entropy value of the local probability distribution as the neighborhood energy dispersion of the target feature point.

[0081] Specifically, to characterize the degree of chaos in the waveform (corresponding to mechanical loosening characteristics), the entropy value within the neighborhood of each feature point is calculated. For each point in the set, its coordinates are used to calculate the entropy value within the neighborhood of each feature point. Centered on, extract from the original atlas The micro-neighborhood matrix is ​​then normalized to its nine values, which are then converted into local probability distributions. ( The neighborhood energy dispersion of that point is calculated using the Shannon entropy formula. It satisfies the following formula:

[0082]

[0083] When the mechanical structure is tight, the impact energy is concentrated, and the entropy value is low; when there are gaps or looseness leading to secondary collisions or waveform divergence, the entropy value increases significantly. It represents the logarithm to the base 2.

[0084] Step 303: Perform set-based normalization on the initial energy amplitudes of multiple target feature points to obtain the energy weight of each target feature point.

[0085] Specifically, for the measured feature point set Calculate the sum of the energy amplitudes at all points. Subsequently, the initial energy amplitudes of multiple target feature points are normalized within a set to obtain the energy weight of each target feature point. It satisfies the following formula:

[0086]

[0087] After this processing, the total weight of the measured point set is... This step transforms the physical energy distribution into a mathematically strictly conserved probability distribution.

[0088] Step 304: Construct a set of measured feature points based on time coordinates, frequency coordinates, energy weights, and neighborhood energy dispersion.

[0089] Measured feature point set Each feature point in the dataset contains four attributes: time coordinate (t), frequency coordinate (f), energy weight (w), and neighborhood energy dispersion (f). These four attribute feature points not only retain the original time, frequency, and energy information, but also add discrete information that characterizes the degree of waveform looseness. This allows for a more comprehensive reflection of the characteristics of the vibration signal and provides key data support for the subsequent decomposition of transmission cost into travel lag and structural looseness indices.

[0090] This embodiment improves the extraction process of the measured feature point set by supplementing the calculation of neighborhood energy dispersion and the intra-set normalization process. The introduction of neighborhood energy dispersion enables the feature point set to capture the waveform divergence characteristics caused by mechanical structure loosening, providing a direct basis for the calculation of structural looseness index; the intra-set normalization process ensures that the feature point set meets the computational constraints of the subsequent transmission algorithm, improving the stability and reliability of the algorithm. The above-mentioned four-attribute measured feature point set can more comprehensively and accurately characterize the fault characteristics of vibration signals, laying the foundation for the accurate decomposition of fault indexes and further improving the pertinence and accuracy of fault diagnosis.

[0091] In one possible implementation, before calculating the transmission cost between the measured feature point set and the reference reference feature point set based on a preset transmission algorithm in step 103 above, this application also needs to determine the reference reference feature point set, specifically including:

[0092] Step 401: When the circuit breaker performs a three-phase linkage operation, calculate the correlation between the normalized energy spectra of the other two phase circuit breakers in the same group.

[0093] Optionally, this application uses the two-dimensional Pearson correlation coefficient between the normalized energy spectra of the other two phases of the circuit breakers in the same group. This indicates its relevance.

[0094] Specifically, the current action command type is detected. If the circuit breaker is performing a three-phase linkage operation (i.e., all three phases operate simultaneously under the same command), then the vibration energy density spectra of the other two phase circuit breakers in the same group are retrieved (denoted as follows). and ).

[0095] To prevent misjudgments due to faults in the reference phase itself, a reference reliability check is performed. Calculation and Two-dimensional Pearson correlation coefficient between The calculation formula satisfies:

[0096]

[0097] in, and These are the values ​​for the corresponding grids in the two-phase spectrum. and This represents the mean of the spectrum.

[0098] Step 402: If the correlation is greater than or equal to the preset consistency threshold, the mean of the normalized energy spectra of the other two phases is used as the reference baseline spectra.

[0099] Step 403: If the correlation is less than the preset consistency threshold, the historical normalized energy spectrum of the circuit breaker is used as the reference baseline spectrum.

[0100] Specifically, setting a consistency threshold (In this embodiment) It requires that the two phase spectra be highly consistent (similarity > 95%) to ensure the reliability of the benchmark to the greatest extent. In specific implementation, it can be set according to actual needs.

[0101] like This indicates that the motion characteristics of the other two phases are highly consistent, and they are likely under normal operating conditions. At this point, the arithmetic mean of the two phase spectra is calculated to obtain the average spectra, which is then used as the reference baseline spectra for this diagnosis. Based on this, this application utilizes a strategy of using devices in the same room as references to automatically offset the effects of common environmental temperature drift.

[0102] like This indicates a significant difference between devices in the same group, making the baseline unreliable; or that single-phase / asynchronous operation is currently being performed. In this case, an automatic degradation strategy is executed, accessing the non-volatile memory and retrieving the historical normalized energy spectrum recorded during the device's factory test or most recent health check as the reference baseline spectrum. .

[0103] Step 404: Extract the reference feature point set from the reference reference map to obtain the reference reference feature point set.

[0104] It should be noted that the process of extracting the reference feature point set from the reference reference spectrum to obtain the reference reference feature point set can refer to the process of extracting the measured feature point set from the normalized energy spectrum described above, and will not be elaborated upon in this application; the obtained reference reference feature point set is denoted as: .

[0105] This embodiment solves the problems of poor environmental adaptability and missing reference spectra caused by fixed references in existing technologies by dynamically selecting reference spectra through correlation verification during three-phase linkage operation. When the status of equipment in the same group is consistent, the mean spectrum is used as the reference to automatically cancel common interference; when the status of equipment is inconsistent or single-phase operation occurs, historical fingerprint data is used as the reference to ensure monitoring continuity. This dynamic reference generation strategy significantly improves the reliability and adaptability of the reference reference, provides a stable comparison basis for subsequent transmission cost calculation and fault index decomposition, and further improves the accuracy of fault diagnosis and engineering applicability.

[0106] In one possible implementation, step 103 above, based on a preset transmission algorithm, calculates the transmission cost between the measured feature point set and the reference benchmark feature point set, and decomposes the transmission cost into a travel lag index and a structural slack index. Specifically, this process includes:

[0107] Step 501: Construct the transmission cost matrix.

[0108] In this matrix, each element represents the unit matching cost required to match a feature point in the measured feature point set to a feature point in the reference feature point set.

[0109] Optionally, the matching cost in the transmission cost matrix includes time deviation cost and waveform discrepancy cost. In this case, the step can be implemented as follows: calculate the time coordinate difference between two feature points and amplify the time difference nonlinearly to obtain the time deviation cost; calculate the neighborhood energy discrepancy difference between two feature points and amplify the discrepancy difference nonlinearly to obtain the waveform discrepancy cost; construct the transmission cost matrix based on the time deviation cost and the waveform discrepancy cost.

[0110] Specifically, construct a dimension as Transmission cost matrix , of which Line number Column elements The representative will be the first 1 measured point Moved to the benchmark points The unit cost. This cost consists of two parts:

[0111] Time Deviation Cost Used to penalize displacement on the time axis. Calculates the time coordinate difference between two points. To be highly sensitive to hysteresis exceeding mechanical tolerances, this embodiment uses a squared form to construct the cost:

[0112]

[0113] in, This represents the maximum tripping time deviation allowed by the circuit breaker technical protocol (2ms in this embodiment). The design of the square term makes the cost of small time jitters extremely low, while once a significant lag occurs, the cost will increase dramatically in a parabolic manner.

[0114] Waveform Discrete Cost Used to construct a loose potential barrier. Calculates the absolute value of the difference in neighborhood energy dispersion between two points. To prevent the algorithm from mismatching high-entropy loose points with low-entropy normal points, this embodiment uses an exponential form to construct the cost:

[0115]

[0116] in, This is the entropy weighting coefficient (taken as 1.5). exp represents the entropy normalization factor (taken as 0.5). The design of the exponential term creates a computational barrier, making matching across different waveform structures extremely expensive. This forces the algorithm to prioritize matching points with similar waveform structures. exp represents the exponential function.

[0117] The total cost is obtained by adding the time deviation cost and the waveform discretization cost: .in, The reference frequency value is 50Hz, for example.

[0118] Step 502: Based on the transmission cost matrix, find the optimal transmission scheme that minimizes the total transmission cost between the measured feature point set and the reference benchmark feature point set.

[0119] Specifically, in obtaining the cost matrix Then, a Discrete Optimal Transport (DOP) model is established, and a solution is obtained. 3D transport flow matrix Elements in the matrix Indicates from the measured point Assigned to reference point The probabilistic quality. The goal of the solution is to minimize the total transmission cost. :

[0120]

[0121] This optimization problem must satisfy the following linear constraints (mass conservation):

[0122] Outflow constraint: for any measured point The sum of flows to all reference points equals the normalized weight of that point. (Right now ).

[0123] Inflow constraint: for any reference point The sum of the flows from all measured points equals the normalized weight of that point. (Right now ).

[0124] Non-negativity constraint: All traffic .

[0125] Since the aforementioned steps have ensured Therefore, the above linear programming problem must have a feasible solution. This application can use the simplex method or the network simplex method to quickly solve for the optimal matrix. .

[0126] Step 503: Based on the optimal transmission scheme, decompose the total transmission cost into components related to time deviation and components related to waveform discrepancy.

[0127] Specifically, to obtain the optimal transport flow matrix Subsequently, this matrix actually describes the optimal geometric mapping relationship between the measured spectrum and the reference spectrum. Based on this mapping, the total cost is decomposed into two independent components: the component related to time deviation is caused by the difference in time coordinates between feature points, reflecting the degree of time lag in circuit breaker operation; the component related to waveform discrepancy is caused by the difference in neighborhood energy dispersion between feature points, reflecting the degree of looseness of the mechanical structure.

[0128] Step 504: Weighted calculation of the components related to time deviation to obtain the travel lag index, and weighted calculation of the components related to waveform discreteness to obtain the structural looseness index.

[0129] For travel lag indicators This metric represents the weighted average time deviation of the entire system under the optimal matching path. As an example, the travel lag metric... It satisfies the following formula:

[0130]

[0131] The travel lag index is only sensitive to the overall time shift. When the buffer damping increases or decreases, the entire waveform shifts along the time axis. The general increase led to The value increases; however, a simple loose waveform will not significantly change this value.

[0132] Regarding the structural looseness index This index represents the weighted average waveform dispersion of the entire system under the optimal matching path. As an example, the structural looseness index... It satisfies the following formula:

[0133]

[0134] The structural looseness index is only sensitive to waveform divergence. When mechanical looseness causes a general increase in waveform entropy, Enlargement, leading to The entropy of the waveform increases; however, a simple time shift does not change the entropy value of the waveform, and therefore does not affect the index.

[0135] This embodiment constructs a transmission cost matrix that comprehensively reflects the differences in characteristic points and uses linear programming to solve for the optimal transmission scheme, ensuring the accuracy and rationality of transmission cost calculation. By decomposing the total transmission cost into two orthogonal fault indicators, it achieves precise decoupling of time lag and structural loosening fault characteristics. This method solves the diagnostic confusion problem caused by the aliasing of fault features in existing technologies, enabling each fault indicator to specifically reflect a fault type, providing a clear quantitative basis for subsequent fault status judgment, and further improving the accuracy and reliability of fault diagnosis.

[0136] In one possible implementation, step 104 above, which involves determining the fault state of the circuit breaker based on the total energy index, travel lag index, and structural looseness index, specifically includes:

[0137] Step 601: Compare the total energy index with the benchmark total energy value. If the total energy index exceeds the preset range based on the benchmark total energy value, it is determined that there is an energy abnormality fault.

[0138] Specifically, obtain the baseline total energy under the current operating conditions. If the reference spectrum is derived from the three-phase mean, then This is the average of the total energy of the other two phases; if it comes from historical fingerprints, then the historical baseline total energy is read directly.

[0139] Calculate the ratio of the measured total energy to the reference total energy. :

[0140]

[0141] Set the upper limit threshold for energy deviation (Take 1.5) and lower threshold (Take 0.5). The upper and lower thresholds for energy deviation are set based on the statistical distribution of the energy ratio of a large number of normal and fault cases. Exceeding this range indicates that the energy anomaly has a clear fault indication significance.

[0142] like This indicates that the energy released by the mechanical action far exceeds the normal level, usually caused by the complete failure of the damper (oil leakage and idle stroke), resulting in the moving contact directly impacting the cylinder bottom. In this case, it is judged as: severe rigid impact / dampening failure, and the highest level alarm is output.

[0143] like This indicates that the energy released by the mechanical action is significantly insufficient, usually caused by insufficient energy storage in the opening / closing springs or severe jamming in the transmission chain. In this case, it is determined as: mechanism jamming / insufficient work, and the highest level alarm is output.

[0144] like Within the normal range ( This indicates that there is no catastrophic energy anomaly in the equipment, and the system continues to execute subsequent precision diagnostic channels.

[0145] Step 602: Compare the travel lag index with the lag threshold. If the travel lag index is greater than the lag threshold, it is determined that there is an abnormality in the damping characteristics of the buffer.

[0146] Specifically, return to the maximum permissible deviation value of the opening and closing time specified in the equipment technical manual. (Typically ±2ms). Due to travel lag indicators Essentially, it is a weighted average cost of normalized flow, with physical dimensions equivalent to single-point cost. Therefore, the system assumes that an ideal feature point experiences an event exactly equal to... By shifting the time and substituting it into the time cost formula, the threshold can be calculated. :

[0147]

[0148] Due to this embodiment Values ​​and The same (both 2ms), therefore This threshold represents the theoretical average cost that should be calculated when the mechanical properties are at the edge of being acceptable.

[0149] like This indicates that the measured spectrum has undergone an overall translation exceeding the tolerance on the time axis, but the energy amplitude and waveform structure may be normal. This indicates that the abnormal macroscopic movement speed is caused by deterioration (viscosity change) or minor leakage of the buffer oil. Maintenance recommendations: Check the hydraulic buffer oil level and properties, and verify the opening and closing speed characteristics.

[0150] Step 603: Compare the structural looseness index with the looseness threshold. If the structural looseness index is greater than the looseness threshold, it is determined that there is a mechanical structural looseness fault.

[0151] Specifically, fingerprint data is acquired when the device is in a factory-installed zero-backlash, tightened state. The average entropy value of the signal-free regions (background noise) in this fingerprint map is calculated as the system noise floor. Assuming the entropy increment of the feature points reaches the noise floor level (i.e., the signal begins to be overwhelmed by noise), substituting this into the waveform cost formula, the threshold is calculated:

[0152]

[0153] This threshold represents the level of thermal noise generated by the inherent assembly gaps of the equipment.

[0154] If the structural looseness index This indicates that the measured spectrum is normal in terms of time axis and energy amplitude, but the local waveform exhibits high-entropy divergence characteristics. This indicates that secondary collisions in the gap are caused by wear of the connecting rod pin, loosening of the cotter pin, or reduction of the preload of the fastening bolts. The output maintenance suggestion is to check the fastening torque of the mechanical connecting rod and inspect the pin connection gap.

[0155] Optional, if and If both thresholds are exceeded simultaneously, it indicates that the equipment simultaneously faces the risk of sluggish movement and structural loosening, typically foreshadowing impending structural disintegration. The system triggers a combined severe fault alarm and locks remote operation access to the circuit breaker via relay nodes until on-site maintenance is completed.

[0156] This embodiment clarifies the correspondence between three indicators and fault types by setting judgment thresholds based on equipment technical parameters and historical data, ensuring the objectivity and accuracy of fault diagnosis. The judgment of the total energy index can quickly identify catastrophic faults such as rigid impacts and mechanical jamming; the judgment of the stroke lag index and structural looseness index can accurately locate common faults such as abnormal buffer damping and loose mechanical structures; and the judgment of complex faults can identify severe multiple faults, providing maintenance personnel with clear fault types and handling suggestions. This method solves the problems of vague fault judgment standards and weak maintenance targeting in existing technologies, effectively improving maintenance efficiency and equipment operational safety.

[0157] Please refer to Figure 2, which shows a system architecture diagram of a GIS circuit breaker spring fault monitoring system based on MEMS sensors according to an embodiment of the present invention. This GIS circuit breaker spring fault monitoring system based on MEMS sensors includes: a map generation unit 201, a feature extraction unit 202, a transmission cost decomposition unit 203, and a fault determination unit 204. The units communicate bidirectionally via a communication link to ensure real-time interaction of collected data and analysis results. The communication link can employ wired or wireless transmission methods to meet the communication needs of different monitoring scenarios.

[0158] The spectrum generation unit 201 is used to perform dual-channel separation processing on the vibration signal during the circuit breaker operation to obtain the total energy index and the normalized energy spectrum. The total energy index is used to characterize the absolute energy magnitude of the vibration signal, and the normalized energy spectrum is used to characterize the relative probability distribution of the vibration energy in the time-frequency domain.

[0159] The feature extraction unit 202 is used to extract a set of measured feature points from the normalized energy spectrum; the set of measured feature points includes multiple feature points, each feature point having a time coordinate, a frequency coordinate, and an initial energy amplitude.

[0160] The transmission cost decomposition unit 203 is used to calculate the transmission cost between the measured feature point set and the reference benchmark feature point set based on a preset transmission algorithm, and decompose the transmission cost into a travel lag index and a structural looseness index; wherein, the travel lag index is used to characterize the time lag of the circuit breaker operation, and the structural looseness index is used to characterize the looseness of the mechanical structure.

[0161] The fault determination unit 204 is used to determine the fault status of the circuit breaker based on the total energy index, travel lag index and structural looseness index.

[0162] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0163] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for monitoring spring faults in GIS circuit breakers based on MEMS sensors, characterized in that, The method includes: performing dual-channel separation processing on the vibration signal during circuit breaker operation to obtain a total energy index and a normalized energy spectrum; wherein, the total energy index is used to characterize the absolute energy magnitude of the vibration signal, and the normalized energy spectrum is used to characterize the relative probability distribution of vibration energy in the time-frequency domain; extracting a set of measured feature points from the normalized energy spectrum; the set of measured feature points includes multiple feature points, each feature point having a time coordinate, a frequency coordinate, and an initial energy amplitude; calculating the transmission cost between the set of measured feature points and a reference feature point set based on a preset transmission algorithm, and decomposing the transmission cost into a travel lag index and a structural looseness index; wherein, the travel lag index is used to characterize the time lag degree of circuit breaker action, and the structural looseness index is used to characterize the looseness degree of mechanical structure; determining the fault state of the circuit breaker based on the total energy index, the travel lag index, and the structural looseness index; wherein, based on a preset transmission algorithm, calculating the transmission cost between the set of measured feature points and the reference feature point set, and decomposing the transmission cost into a travel lag index and a structural looseness index, including... The process involves: constructing a transmission cost matrix, where each element represents the unit matching cost required to match a feature point in the measured feature point set to a feature point in the reference feature point set; based on the transmission cost matrix, finding the optimal transmission scheme that minimizes the total transmission cost between the measured feature point set and the reference feature point set; according to the optimal transmission scheme, decomposing the total transmission cost into a component related to time deviation and a component related to waveform discrepancy; weighting the component related to time deviation to obtain the travel lag index; and weighting the component related to waveform discrepancy to obtain the structural slack index; the matching cost in the transmission cost matrix includes time deviation cost and waveform discrepancy cost; constructing the transmission cost matrix includes: calculating the time coordinate difference between two feature points and nonlinearly amplifying the time difference to obtain the time deviation cost; calculating the neighborhood energy dispersion difference between two feature points and nonlinearly amplifying the dispersion difference to obtain the waveform discrepancy cost; and constructing the transmission cost matrix based on the time deviation cost and the waveform discrepancy cost.

2. The method for monitoring spring faults in GIS circuit breakers based on MEMS sensors according to claim 1, characterized in that, The vibration signal during circuit breaker operation is processed using a dual-channel separation method to obtain a total energy index and a normalized energy spectrum. This includes: determining the start time of mechanical action based on the current signal of the circuit breaker's opening and closing coils; collecting vibration signals within a preset time period starting from the start time as the vibration signals during circuit breaker operation; performing time-frequency transformation on the vibration signals to obtain an original time-frequency energy matrix; truncating the endpoints of the original time-frequency energy matrix to obtain an effective time-frequency energy matrix; calculating the sum of the values ​​of all elements in the effective time-frequency energy matrix as the total energy index; and dividing each element in the effective time-frequency energy matrix by the total energy index to obtain the normalized energy spectrum.

3. The method for monitoring GIS circuit breaker spring faults based on MEMS sensors according to claim 2, characterized in that, Determining the start time of mechanical action based on the current signal of the circuit breaker's opening and closing coils includes: monitoring the current waveform of the opening and closing coils; determining the start time based on the moment when the current waveform meets preset conditions; the preset conditions include: the current amplitude has exceeded the peak value and decayed to below a first predetermined proportion of the peak value, and the absolute value of the current change rate is less than a preset threshold.

4. The method for monitoring GIS circuit breaker spring faults based on MEMS sensors according to claim 1, characterized in that, Before calculating the transmission cost between the measured feature point set and the reference benchmark feature point set based on a preset transmission algorithm, the method further includes: when the circuit breaker performs a three-phase linkage operation, calculating the correlation between the normalized energy spectra of the other two phase circuit breakers in the same group; if the correlation is greater than or equal to a preset consistency threshold, then the mean of the normalized energy spectra of the other two phases is used as the reference benchmark spectra; if the correlation is less than the preset consistency threshold, then the historical normalized energy spectra of the circuit breaker is used as the reference benchmark spectra; extracting the benchmark feature point set from the reference benchmark spectra to obtain the reference benchmark feature point set.

5. The method for monitoring GIS circuit breaker spring faults based on MEMS sensors according to claim 1, characterized in that, Extracting a set of measured feature points from the normalized energy spectrum includes: determining multiple target feature points in the normalized energy spectrum; each target feature point has a time coordinate, a frequency coordinate, and an initial energy amplitude; calculating the neighborhood energy dispersion of each target feature point; the neighborhood energy dispersion is used to characterize the uniformity of energy distribution in a local region centered on the target feature point; performing intra-set normalization on the initial energy amplitudes of the multiple target feature points to obtain the energy weight of each target feature point; and constructing the set of measured feature points based on the time coordinate, the frequency coordinate, the energy weight, and the neighborhood energy dispersion.

6. The method for monitoring spring faults in GIS circuit breakers based on MEMS sensors according to claim 5, characterized in that, Determining multiple target feature points in the normalized energy spectrum includes: performing a local maximum search in the normalized energy spectrum to select candidate points whose energy amplitude is greater than other elements in the local neighborhood and greater than a predetermined proportion of the global maximum value; sorting the candidate points in descending order of energy amplitude, and selecting the top K points as the multiple target feature points, where K is a preset positive integer.

7. The method for monitoring GIS circuit breaker spring faults based on MEMS sensors according to claim 5, characterized in that, Calculating the neighborhood energy dispersion of each target feature point includes: extracting a neighborhood matrix of a preset size from the normalized energy spectrum with the target feature point as the center; normalizing the values ​​in the neighborhood matrix to form a local probability distribution; and using the Shannon entropy value of the local probability distribution as the neighborhood energy dispersion of the target feature point.

8. The method for monitoring spring faults in GIS circuit breakers based on MEMS sensors according to claim 1, characterized in that, Determining the fault status of the circuit breaker based on the total energy index, the travel hysteresis index, and the structural looseness index includes: comparing the total energy index with a reference total energy value; if the total energy index exceeds a preset range based on the reference total energy value, an energy abnormality fault is determined; comparing the travel hysteresis index with a hysteresis threshold; if the travel hysteresis index is greater than the hysteresis threshold, an abnormal buffer damping characteristic is determined; and comparing the structural looseness index with a looseness threshold; if the structural looseness index is greater than the looseness threshold, a mechanical structure looseness fault is determined.

Images