Vacuum circuit breaker fault detection method and system based on deep fusion perception

Abstract

The present application relates to the technical field of power equipment state monitoring and fault diagnosis, in particular to a vacuum circuit breaker fault detection method and system based on deep fusion perception, which responds to a preset operation instruction, synchronously acquires an electrical control loop transient current signal, a main loop transient voltage drop signal and an envelope signal of a high-frequency electromagnetic field of a near-field space of an arc extinguishing chamber, and constructs a multi-dimensional state matrix; based on an electromechanical coupling energy admittance operator containing an energy conversion coefficient and an admittance matrix, phase space reconstruction is carried out, transient energy flow rate is calculated, and a transient energy trajectory in a three-dimensional energy phase space is generated; the transient energy trajectory is compared with a health state energy manifold, an orthogonal projection residual vector is calculated, and a specific fault type is determined, a quantitative diagnosis report is generated or the health state energy manifold is updated, and the present application realizes interpretable identification of vacuum circuit breaker opening and closing transient faults and improves anti-interference capability.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and fault diagnosis technology, specifically to a method and system for detecting faults in vacuum circuit breakers based on deep fusion sensing. Background Technology

[0002] In the existing monitoring of vacuum circuit breaker operation status, vacuum circuit breakers typically include electrical control circuits, main circuits and arc-extinguishing chambers. Maintenance personnel generally analyze the equipment status through single detection methods such as opening and closing current, stroke, mechanical characteristics or partial discharge, and combine regular power outage maintenance or disassembly and testing methods to determine whether the vacuum level, contact wear and operating mechanism are abnormal.

[0003] Among the existing detection methods, some schemes determine the status of the operating mechanism by collecting coil current or mechanism action signals in the electrical control circuit, some schemes identify abnormal contact of contacts and changes in arc extinguishing performance by detecting voltage and current changes or related electromagnetic signals in the main circuit, and some schemes use empirical thresholds, waveform feature comparison or single-channel data classification to achieve fault early warning.

[0004] However, single-channel detection methods are difficult to reflect the coupled energy transfer relationship between the electrical control circuit, main circuit and arc-extinguishing chamber during the opening and closing of vacuum circuit breakers. In complex substation environments, they are easily affected by external electromagnetic interference, load fluctuations and individual equipment differences, resulting in inaccurate identification of hidden faults such as vacuum level decline, contact wear and mechanical jamming of operating mechanisms. This leads to problems such as high false alarm rate, poor interpretability and insufficient reliability of online diagnosis. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting faults in vacuum circuit breakers based on deep fusion sensing, and to solve the following technical problems:

[0006] It avoids the shortcomings of traditional detection methods that rely on empirical thresholds or lack clear physical constraints, and can effectively identify hidden faults such as vacuum drop, contact wear and mechanical jamming of operating mechanism without power interruption or disassembly of arc-extinguishing chamber. By decoupling the natural operating drift of equipment from the actual abnormal deviation, it achieves more stable and interpretable identification of transient faults and reduces false alarm rate.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A fault detection method for vacuum circuit breakers based on deep fusion sensing, wherein the vacuum circuit breaker includes an electrical control circuit, a main circuit, and an arc-extinguishing chamber, the method comprising the following steps:

[0009] In response to preset operation commands, the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber are acquired synchronously, and a multi-dimensional state matrix is ​​constructed based on the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal.

[0010] The phase space of the multidimensional state matrix is ​​reconstructed based on a preset electromechanical coupling energy admittance operator that includes energy conversion coefficients and admittance matrix. The transient energy flow rate of the multidimensional state matrix within a preset time window is calculated. The transient energy flow rate includes the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal, respectively.

[0011] Using the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal as the first coordinate axis, the second coordinate axis, and the third coordinate axis, a three-dimensional energy phase space is constructed, and a transient energy trajectory is generated in the three-dimensional energy phase space according to the transient energy flow rate.

[0012] The transient energy trajectory is compared with a preset healthy state energy manifold, and the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold is calculated.

[0013] When the magnitude of the orthogonal projection residual vector is higher than or equal to the preset alarm threshold, the specific fault type of the vacuum circuit breaker is determined and a quantitative diagnostic report is generated based on the distribution characteristics of the orthogonal projection residual vector in different dimensions.

[0014] Provided that the magnitude of the orthogonal projection residual vector is lower than the preset alarm threshold, and the current multidimensional state matrix sampling is valid and there is no external transient electromagnetic disturbance, the transient energy trajectory is incorporated into the preset sample set to update the healthy state energy manifold.

[0015] Furthermore, simultaneously acquiring the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber includes:

[0016] The transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal are acquired through a preset high-frequency edge computing node; wherein, the synchronous sampling rate of the high-frequency edge computing node is higher than or equal to a preset sampling rate threshold.

[0017] The transient voltage drop signal is obtained by coupling through a preset capacitive voltage divider.

[0018] Further, based on a preset electromechanical coupling energy admittance operator including energy conversion coefficients and admittance matrix, the phase space of the multidimensional state matrix is ​​reconstructed, the transient energy flux of the multidimensional state matrix within a preset time window is calculated, and a transient energy trajectory is generated in the three-dimensional energy phase space according to the transient energy flux, including:

[0019] Extract multiple time slices of the multidimensional state matrix within the preset time window;

[0020] The instantaneous power gradient corresponding to each time slice is calculated using the electromechanical coupling energy admittance operator.

[0021] The instantaneous power gradient corresponding to each time slice is taken as the transient energy flow rate;

[0022] The transient energy flow rate is mapped onto the three-dimensional energy phase space to connect and generate the transient energy trajectory.

[0023] Further, the transient energy trajectory is compared with a preset healthy state energy manifold, and the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold is calculated, including:

[0024] Extract multiple trajectory feature points on the transient energy trajectory;

[0025] The tangent plane of the energy manifold of the health state at the trajectory feature point is calculated as the reference plane;

[0026] Calculate the orthogonal projection vector of each trajectory feature point from the reference plane;

[0027] All the orthogonal projection vectors are combined to generate the orthogonal projection residual vector.

[0028] Furthermore, based on the distribution characteristics of the orthogonal projection residual vector in different dimensions, the specific fault type of the vacuum circuit breaker is determined, including:

[0029] Extract the first projection component of the orthogonal projection residual vector on the plane formed by the second coordinate axis and the third coordinate axis;

[0030] Calculate the envelope attenuation rate of the high-frequency electromagnetic field envelope signal;

[0031] If the first projection component is greater than a preset planar projection threshold and the envelope attenuation rate is less than a preset attenuation rate threshold, the specific fault type is determined to be an arc extinguishing capability degradation fault; if the first projection component is less than or equal to the preset planar projection threshold, or the envelope attenuation rate is greater than or equal to the preset attenuation rate threshold, it is determined to be another specific fault type.

[0032] Furthermore, determining the specific fault type of the vacuum circuit breaker based on the distribution characteristics of the orthogonal projection residual vector in different dimensions also includes:

[0033] Extract the second projection component of the orthogonal projection residual vector on the first coordinate axis;

[0034] The duration of the transient energy trajectory in the three-dimensional energy phase space is calculated as the energy integration time;

[0035] If the second projection component is greater than a preset axial projection threshold and the energy integration time is greater than a preset time threshold, the specific fault type is determined to be a mechanical fault of the operating mechanism; if the second projection component is less than or equal to the preset axial projection threshold, or the energy integration time is less than or equal to the preset time threshold, it is determined to be another specific fault type.

[0036] Furthermore, a quantitative diagnostic report is generated, including:

[0037] The orthogonal projection residual vector is respectively multiplied by the preset three-dimensional vacuum degree weight coefficient vector and the three-dimensional electrical wear weight coefficient vector that match the three-dimensional energy phase space dimension.

[0038] The results of the inner product operation are used as the equivalent vacuum degree reduction index and the contact electrical wear depth evaluation value, respectively.

[0039] The equivalent vacuum degree reduction index and the contact electrical wear depth assessment value are packaged into the quantitative diagnostic report and output.

[0040] A fault detection system for vacuum circuit breakers based on deep fusion sensing, wherein the vacuum circuit breaker includes an electrical control circuit, a main circuit, and an arc-extinguishing chamber, and the system includes the following modules:

[0041] The matrix construction module is used to respond to preset operation commands, synchronously acquire the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc extinguishing chamber, and construct a multi-dimensional state matrix based on the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal.

[0042] The trajectory generation module is used to reconstruct the phase space of the multidimensional state matrix based on a preset electromechanical coupling energy admittance operator containing energy conversion coefficients and admittance matrix, and to calculate the transient energy flow rate of the multidimensional state matrix within a preset time window. The transient energy flow rate includes the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal, respectively.

[0043] The trajectory generation module is further configured to construct a three-dimensional energy phase space using the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal as the first coordinate axis, the second coordinate axis, and the third coordinate axis, respectively, and generate a transient energy trajectory in the three-dimensional energy phase space according to the transient energy flow rate.

[0044] The residual calculation module is used to compare the transient energy trajectory with the preset healthy state energy manifold and calculate the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold;

[0045] The diagnosis and update module is used to determine the specific fault type of the vacuum circuit breaker and generate a quantitative diagnosis report based on the distribution characteristics of the orthogonal projection residual vector in different dimensions when the amplitude of the orthogonal projection residual vector is higher than or equal to a preset alarm threshold; and to update the transient energy trajectory to a new healthy state energy manifold when the amplitude of the orthogonal projection residual vector is lower than the preset alarm threshold, and the current multidimensional state matrix sampling is valid and there is no external transient electromagnetic disturbance.

[0046] The beneficial effects of this invention are:

[0047] 1. This invention simultaneously acquires the transient current of the control loop, the transient voltage drop of the main loop, and the envelope signal of the high-frequency electromagnetic field of the arc-extinguishing chamber to construct a multi-dimensional state matrix, and uses the electromechanical coupling energy admittance operator to reconstruct the phase space to generate a transient energy trajectory. This method overcomes the defect that a single signal is difficult to reflect the electromechanical coupling relationship, and transforms the discrete waveform into a trajectory description that reflects the closed loop of energy transfer, which significantly enhances the physical interpretability of online monitoring.

[0048] 2. This invention extracts transient energy trajectory feature points and calculates their orthogonal projection residual vector on the local tangent plane of the energy manifold in the healthy state. This method decouples the natural operational drift of the equipment under load fluctuations or environmental changes from the actual physical anomalies, avoiding the problem of false alarms caused by the susceptibility of traditional experience thresholds to individual differences and external interference, and improving the anti-interference capability and reliability of online diagnosis in complex environments.

[0049] 3. This invention classifies faults based on the distribution characteristics of orthogonal projection residual vectors in different dimensions; it accurately identifies arc-extinguishing capability degradation faults by combining planar projection components and envelope attenuation rate, and effectively identifies mechanical faults of operating mechanisms by combining axial projection components and energy integration time; this mechanism realizes targeted and accurate analysis of various hidden faults, effectively avoiding misjudgments such as misattributing mechanical jamming to vacuum level decline.

[0050] 4. This invention performs an inner product operation between the residual vector and the preset weight coefficient vector to generate a quantitative diagnostic report that includes the equivalent vacuum degree reduction index and the contact electrical wear assessment value, and updates the health state energy manifold when the threshold is not exceeded. This mechanism provides a quantitative degradation assessment index without power interruption, realizes long-term adaptive health tracking of circuit breakers, and provides reliable decision-making closed-loop support for accurate scheduling of maintenance resources.

[0051] 5. This invention acquires signals through high-frequency edge computing nodes with high synchronization, uses a capacitor voltage divider to couple and obtain transient voltage drops, and calculates instantaneous power gradients by time slices during phase space reconstruction. This design ensures absolute temporal coherence and high transient resolution of multi-source data, prevents false fault features introduced by sampling phase differences, and provides reliable data support for truly reflecting the causal sequence of internal electromechanical physical events. Attached Figure Description

[0052] The invention will now be further described with reference to the accompanying drawings.

[0053] Figure 1 A flowchart illustrating the vacuum circuit breaker fault detection method based on deep fusion perception provided in this application embodiment;

[0054] Figure 2 This is a schematic diagram of a vacuum circuit breaker fault detection system based on deep fusion perception provided in an embodiment of this application. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1 A fault detection method for vacuum circuit breakers based on deep fusion perception, wherein the vacuum circuit breaker includes an electrical control circuit, a main circuit, and an arc-extinguishing chamber, the method includes the following steps: in response to a preset operation command, synchronously acquiring the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber, and constructing a multi-dimensional state matrix based on the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal;

[0057] The phase space of the multidimensional state matrix is ​​reconstructed based on a preset electromechanical coupling energy admittance operator that includes energy conversion coefficients and admittance matrix. The transient energy flow rate of the multidimensional state matrix within a preset time window is calculated. The transient energy flow rate includes the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal, respectively.

[0058] Using the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal as the first coordinate axis, the second coordinate axis, and the third coordinate axis, a three-dimensional energy phase space is constructed, and a transient energy trajectory is generated in the three-dimensional energy phase space according to the transient energy flow rate.

[0059] The transient energy trajectory is compared with a preset healthy state energy manifold, and the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold is calculated.

[0060] When the magnitude of the orthogonal projection residual vector is higher than or equal to the preset alarm threshold, the specific fault type of the vacuum circuit breaker is determined and a quantitative diagnostic report is generated based on the distribution characteristics of the orthogonal projection residual vector in different dimensions.

[0061] Provided that the magnitude of the orthogonal projection residual vector is lower than the preset alarm threshold, and the current multidimensional state matrix sampling is valid and there is no external transient electromagnetic disturbance, the transient energy trajectory is incorporated into the preset sample set to update the healthy state energy manifold.

[0062] This embodiment provides a fault detection mechanism for vacuum circuit breakers based on deep fusion perception. Specifically, this embodiment takes the outgoing line bay vacuum circuit breaker in a 110 kV urban substation as the application scenario. This circuit breaker undertakes the daily opening and closing tasks of the main power supply line of the industrial park. The equipment is under long-term operation under load, complex electromagnetic background, and limited maintenance window conditions.

[0063] In order to identify hidden faults such as vacuum drop, contact wear and mechanical jamming of operating mechanism without power interruption or disassembly of arc-extinguishing chamber, this embodiment does not understand the detection object as a number of isolated waveforms, but regards a complete opening or closing process as a transient energy transfer closed loop between electrical control circuit, main circuit and arc-extinguishing chamber.

[0064] Specifically, when the station control layer issues a tripping or closing command, the system synchronously collects three types of physical quantities within a preset time window. The first type is the transient current signal in the electrical control circuit, which reflects the establishment of electromagnetic driving force, armature movement and mechanical load changes after the coil is energized.

[0065] The second category is the transient voltage drop signal in the main circuit, which reflects the contact state of the contacts, the voltage redistribution during the separation process, and the local impedance change of the conductive path; the third category is the high-frequency electromagnetic field envelope signal in the near-field space of the arc-extinguishing chamber, which reflects the dielectric recovery between contacts, residual arc activity, and partial discharge characteristics.

[0066] The system arranges three types of synchronous time-series data into a multi-dimensional state matrix according to a unified time base. This matrix is ​​not just a data stack, but binds the physical responses of three different observation surfaces at the same operating moment together, so that subsequent analysis has a causal correspondence.

[0067] The start point of the preset time window is set at the moment when the system receives the opening or closing operation command, and the end point is set at the moment when the system detects that the transient signals of the electrical control circuit and the main circuit decay to the steady-state baseline and continue to maintain the preset duration, so as to ensure complete coverage of the entire process of electromechanical response and energy dissipation of a single operation.

[0068] During the phase space reconstruction stage, the system uses a pre-set electromechanical coupling energy admittance operator to process the above-mentioned multidimensional state matrix. The energy conversion coefficient here is used to characterize the typical conversion ratio of coil electrical energy to mechanical kinetic energy, mechanical kinetic energy to contact collision and arc dissipation. The admittance matrix is ​​used to characterize the coupling direction and response hysteresis between various observations. For example, if the coil current suddenly increases at the same time while the main circuit voltage drop has not changed synchronously, it usually indicates that the mechanical transmission chain is starting while the contact position has not changed.

[0069] If the main circuit voltage drop and the near-field high-frequency electromagnetic field both exhibit tailing, it is closer to the physical state where the medium recovery ability is weakened after the contact is separated. Through this operator, the system does not directly output the waveform classification result, but obtains the transient energy flow rate characteristics corresponding to each moment.

[0070] For ease of explanation, an operating time window can be schematically divided into three consecutive segments: the start-up segment, the contact transition segment, and the stabilization segment; if the response of the first coordinate axis direction dominates in the start-up segment, it indicates that the electromagnetic drive is overcoming mechanical inertia.

[0071] If the second and third coordinate axes of the contact transition segment increase synchronously, it indicates that the contact state and the arc extinguishing state change together; if the three-dimensional response in the stable segment converges rapidly, it indicates that the operation conforms to the energy closure law of the healthy device; the system maps the transient energy flow rate of each time slice to the three-dimensional energy phase space in sequence, and connects adjacent time slices to form a transient energy trajectory;

[0072] The trajectory does not reflect the magnitude of a single point, but rather whether the energy flow path evolves along the existing physical topology of the health device throughout the entire operation.

[0073] During the comparison phase, the system calls the health state energy manifold formed by historical health samples; this manifold can be understood as the trajectory cluster that the equipment should fall into during normal operation under a certain model, a certain installation posture, and a certain typical load condition; the system projects the current trajectory onto the vicinity of this manifold and calculates the deviation direction and the degree of deviation.

[0074] If the deviation reaches the alarm level, it indicates that a repeatable physical anomaly has occurred in the energy conversion closed loop of this operation. At this time, based on the distribution characteristics of the residual in three-dimensional space, it is further determined whether the fault is closer to arc extinguishing capability degradation, contact electrical wear, or mechanical fault of the operating mechanism, and a quantitative diagnostic report is output.

[0075] The preset alarm threshold is set based on the following: statistical analysis of the amplitude of the orthogonal projection residual vector generated by multiple consecutive operations of a specific model vacuum circuit breaker in a healthy state, extracting the maximum distribution envelope of its amplitude, and adding a preset fault tolerance margin coefficient to calculate the preset alarm threshold; or confidence interval estimation based on the residual variance of known healthy samples within the energy manifold of the healthy state.

[0076] If the deviation is still within the allowable range, it is considered a health drift of the device under natural aging, load fluctuations or seasonal environmental changes. This trajectory can be incorporated into the health state energy manifold for subsequent adaptive updates.

[0077] As a boundary treatment, this embodiment specifically handles several abnormal situations; if the operation command has been issued but any of the three types of signals has missing samples, misaligned timestamps, or obvious saturation, then the data will not enter the healthy state energy manifold update process, but will only be retained as an abnormal sampling event and the maintenance personnel will be prompted to check the sensor node.

[0078] If there is an electromagnetic disturbance caused by an external power grid fault that is not dominated by this circuit breaker within the current time window, the system can eliminate it according to the instruction time and the causal sequence of the signal to avoid mistaking non-causal noise as an internal fault of the equipment; if the current residual is close to the alarm threshold but has not formed a stable deviation, the system can mark the equipment as recommended for retesting instead of directly alarming the fault, in order to prevent false alarms caused by a single occasional operation.

[0079] In the aforementioned urban substation scenario, the circuit breaker undertakes multiple industrial load switching tasks during the high-temperature season; after a certain tripping command is issued, the system synchronously collects the coil current, contact voltage drop, and high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber within an operating time window of approximately tens of milliseconds.

[0080] After phase space mapping, it was found that the current trajectory deviates from the existing healthy energy manifold, and the deviation is mainly concentrated in the region related to the main loop voltage drop and near-field high-frequency electromagnetic field. At the same time, there is a significant tailing at the end of the operation.

[0081] Based on this, the system determines that the anomaly is closer to the degradation of arc extinguishing capability than to simple mechanical lag, and generates a diagnostic result that includes the trend of decreasing equivalent vacuum. If only the trajectory in the first coordinate axis direction is elongated in the operation on another day, while the second and third coordinate axis directions are not synchronously abnormal, the system will prioritize classifying it into the warning queue of the mechanical lag direction.

[0082] Furthermore, to avoid the update to a new health state energy manifold being interpreted as a direct replacement of the existing health baseline by a single trajectory, the update in this embodiment is preferably an incremental update;

[0083] Only when the residual amplitude is lower than the preset alarm threshold, and simultaneously the following conditions are met: the sampling link is valid, the operation type is consistent, there is no external fault disturbance in the current operation, and the trajectory is not truncated or missing, can the current transient energy trajectory be added as a new healthy sample and incorporated into the original healthy sample set. The healthy state energy manifold is then regenerated or fine-tuned by the updated sample set.

[0084] Specifically, the system preserves the main structure of the energy manifold in the existing healthy state, and does not overwrite it entirely due to the result of a single operation, thereby avoiding errors in the long-term stable baseline caused by occasional data drift;

[0085] Furthermore, the allowable range is not only constrained by the preset alarm threshold, but can also be verified by combining the consistency of multiple consecutive operations; if the residual of a single operation is lower than the alarm threshold, but its deviation direction is significantly inconsistent with the long-term main change direction of the existing healthy samples, then the trajectory is only temporarily stored as a candidate healthy sample, and will participate in manifold update when the same trend reappears in subsequent similar operations.

[0086] If it does not recur, it will be automatically removed. Through the combination of incremental merging, temporary review and abnormal removal, the health status energy manifold can reflect the natural drift of the equipment over time without losing its sensitivity to new faults.

[0087] The terms "manifold" and "energy manifold" used in the text refer specifically to the aforementioned energy manifold in the healthy state. The terms "operation tail segment," "stable segment," and "convergence segment before the stable phase" used in the text refer to the same physical phase within a preset time window that is close to the end of the operation and the trajectory tends to converge, unless otherwise specified. If the following implementation method uses abbreviations for the purpose of simplifying the text, it is only a compression of the description and does not introduce new detection objects, data types, or reference objects.

[0088] The purpose of this step is to transform the traditional online monitoring, which relies on empirical thresholds or lacks clear physical constraints and is data-driven classification, into a physical measurement process constrained by the laws of energy conservation and electromechanical coupling. This will enable interpretable identification of transient faults in the opening and closing of vacuum circuit breakers and improve their anti-interference capabilities in complex substation environments.

[0089] In a preferred embodiment of the present invention, synchronously acquiring the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber includes: acquiring the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal through a preset high-frequency edge computing node; wherein the synchronous sampling rate of the high-frequency edge computing node is higher than or equal to a preset sampling rate threshold; and the transient voltage drop signal is acquired by coupling a preset capacitive voltage divider.

[0090] This embodiment provides a hardware acquisition mechanism for the synchronous acquisition of three types of transient signals. Specifically, in the aforementioned urban substation scenario, if only the low-speed recording resolution of conventional protection devices is relied upon, only the general trend of coil current change can often be observed. The fluctuation of local impedance voltage drop of the contacts and the high-frequency field change outside the arc extinguishing chamber will be averaged by the sampling interval, resulting in distortion of the subsequent phase space reconstruction.

[0091] Therefore, in this embodiment, a high-frequency edge computing node is deployed near the circuit breaker operating mechanism box and used as a local measurement and preprocessing unit;

[0092] Specifically, the edge node includes at least a synchronous clock module, multiple high-frequency analog-to-digital sampling channels, a front-end conditioning circuit, and a local buffer unit; the coil current signal can be obtained through a through-hole current sampler or a low-inductance shunt to ensure that the steep transition at the beginning of the operation is not affected by the sensor response delay.

[0093] The transient voltage drop signal of the main circuit is obtained by capacitive voltage divider coupling. The reason is that the main circuit of the vacuum circuit breaker is at a high voltage potential. Direct lead wire is not conducive to insulation safety and is also easy to introduce additional circuits that affect the object under test. By capacitive voltage divider coupling, only the high-frequency voltage drop component related to the rapid change of contact state is extracted, which satisfies electrical isolation and focuses the measurement object on the opening and closing transient.

[0094] The envelope signal of the near-field high-frequency electromagnetic field of the arc-extinguishing chamber can be obtained by a near-field probe or a directional small antenna near the shielding shell, which can be used to sense the high-frequency electromagnetic activity leaking out of the arc-extinguishing chamber.

[0095] The synchronous sampling rate is set to be higher than or equal to the preset sampling rate threshold because multiple key physical events during the opening and closing process often occur within milliseconds or even less time. For example, the establishment of the coil current peak, the linkage of the mechanism crossing the characteristic position, and the electric field reconstruction accompanying the contact separation are all occurring in short intervals.

[0096] If the sampling of the three channels is not synchronized, even if the waveforms of each single channel are complete, a pseudo phase difference will be introduced when constructing the multi-dimensional state matrix, which will lead to the fault-free state being misjudged as a mechanical jamming fault, or the real arc-extinguishing tail characteristics being misjudged as sampling delay. Therefore, this embodiment prioritizes ensuring time coherence, rather than just ensuring the accuracy of a single channel.

[0097] For ease of understanding, the three sampling channels of the same tripping operation can be denoted as channel A, channel B, and channel C, respectively. If the marker point at the same moment corresponds to the same time number in the three data channels, then a row in the state matrix can represent the three-dimensional state at the same physical instant.

[0098] If channel B lags behind by several sampling points, the voltage drop in the main circuit of that row will correspond to the lagging physical moment, thus disrupting the causal interpretation of the subsequent energy flow rate. Therefore, edge nodes can perform time alignment and cache correction before writing to the matrix to ensure that the three samples use a unified clock boundary.

[0099] As an anomaly handling mechanism, if the transient voltage drop signal amplitude of the capacitor voltage divider is significantly abnormal due to insulation dampness or loose wiring, the system can detect that it has been deviating from the installation range in the equipment register for a long time through channel self-test. At this time, the judgment based on the main circuit voltage drop is suspended, and only the high-frequency electromagnetic field envelope signal of the coil current and the near field space of the arc extinguishing chamber is retained as an auxiliary observation quantity, and the equipment is placed in the sampling link pending inspection state.

[0100] If the sensitivity of the near-field probe decreases due to changes in the position of the metal shield, the system can trigger recalibration based on the long-term baseline drift trend in the healthy sample, instead of directly interpreting the decrease in amplitude as an improvement in device health; if the synchronization clock loses lock, the entire data of this operation is judged as invalid to avoid incorrectly updating the energy manifold of the health status.

[0101] In the renovation of the same outgoing line bay, the edge node is installed inside the side plate of the mechanism box, the coil current sampler is arranged in the closing and opening coil branch, the capacitor voltage divider is installed inside the main circuit isolation cover, and the high-frequency near-field probe is arranged in a position adjacent to the arc-extinguishing chamber shell but without damaging the original insulation distance.

[0102] During a nighttime load switching operation, the three channels start sampling simultaneously with a unified clock. The edge nodes first complete the timestamp encapsulation locally, and then send the state matrix that constitutes the complete time series into the subsequent diagnostic process.

[0103] The purpose of this step is to provide a raw measurement basis with the same time reference and sufficient transient resolution for subsequent energy phase space reconstruction, thereby enabling reliable external observation of the electromechanical coupling process inside the circuit breaker.

[0104] In a preferred embodiment of the present invention, the phase space of the multidimensional state matrix is ​​reconstructed based on a preset electromechanical coupling energy admittance operator containing energy conversion coefficients and admittance matrix, the transient energy flow rate of the multidimensional state matrix within the preset time window is calculated, and a transient energy trajectory is generated in the three-dimensional energy phase space according to the transient energy flow rate, including: extracting multiple time slices of the multidimensional state matrix within the preset time window;

[0105] The instantaneous power gradient corresponding to each time slice is calculated using the electromechanical coupling energy admittance operator; the instantaneous power gradient corresponding to each time slice is used as the transient energy flow rate; the transient energy flow rate is mapped to the three-dimensional energy phase space to connect and generate the transient energy trajectory.

[0106] This embodiment provides a transient energy trajectory generation mechanism; specifically, in the aforementioned scenario, obtaining only synchronous three-channel signals is still insufficient to stably distinguish between vacuum degradation and mechanical jamming.

[0107] The reason is that different models of circuit breakers have different coil rated currents, contact structures and arc-extinguishing chamber shapes. If the original time-domain waveforms are directly compared, it is easy to misjudge the inherent physical differences between the equipment as fault characteristics. Therefore, this embodiment further uses an electromechanical coupling energy admittance operator to uniformly map different physical quantities into trajectory representations in the sense of energy flow rate.

[0108] Specifically, the system extracts multiple consecutive time slices within a preset time window; each time slice can cover an operation segment of a preset time length, used to describe the local coupling relationship between electromagnetic drive establishment, mechanical response propagation and medium recovery within that segment.

[0109] The energy conversion coefficient is used to convert the measurements of each channel to a similar energy description scale, while the admittance matrix is ​​used to express the mutual coupling direction between the three channels; for example, the rapid rise of the coil current usually precedes the mechanical motion, so its contribution to the energy flow rate in the first coordinate axis direction is stronger during the start-up phase.

[0110] Significant changes in the main circuit voltage drop often occur when the contacts are close to separation or closure, thus having a higher weight in the transition phase; the envelope tail of the near-field high-frequency electromagnetic field characterizes the arc extinguishing and dielectric recovery process more, thus revealing the internal state of the arc extinguishing chamber better at the end of the operation.

[0111] For ease of explanation, a time window can be divided into four time slices, referred to as slice one to slice four respectively. If slice one corresponds to the establishment of coil excitation, the instantaneous power gradient is most prominent on the first coordinate axis. If slice two corresponds to the transmission mechanism driving the contact to move, the first coordinate axis direction and the second coordinate axis direction rise together.

[0112] If the electric field is reconstructed after the contact of segment three is separated, the direction of the second coordinate axis and the direction of the third coordinate axis will be linked; if segment four is restored to stability, the three-dimensional power gradient will gradually fall back; the system will place the four segments in the three-dimensional energy phase space in sequence and connect them to form a directional transient energy trajectory.

[0113] The trajectory of a health device is usually continuous, convergent, and has stable turning points; when a fault exists, the trajectory may be elongated in a certain dimension, expanded in or out of a certain plane, or stuck at the end.

[0114] Unlike simple statistical characteristics, the instantaneous power gradient here does not emphasize the complex calculation process itself, but rather its physical meaning, namely the speed and direction of energy from the input of the electrical control circuit, through the mechanical chain, and then to the main circuit and arc-extinguishing chamber in a short time segment;

[0115] In this way, even if the absolute amplitudes of different devices differ, as long as the energy flow path follows the same mechanism, they can still be compared in a unified space.

[0116] As a boundary treatment, if the time window is divided into time spans greater than the first preset threshold, multiple key events may be mixed into the same time slice, resulting in unclear trajectory transitions; if the time window is divided into time spans less than the second preset threshold, noise may be amplified in individual slices; if the time window is divided into time spans less than or equal to the first preset threshold and greater than or equal to the second preset threshold, the current slice granularity is determined to be appropriate and subsequent instantaneous power gradient calculations are performed based on the current slice.

[0117] Therefore, the system can select an appropriate slice size based on the equipment model, rated operating time, and historical sample stability.

[0118] When the duration of an operation is abnormally shortened, such as when there is a refusal to move or incomplete completion, the system can truncate subsequent slices in advance and mark the missing parts as empty segments, without forcibly filling them in, in order to avoid generating physically non-existent trajectory tail segments; if an individual slice is missing a valid value for a certain channel, the slice will not enter the healthy state energy manifold update, but can be retained as a fault clue.

[0119] During a single closing test operation at the city's substation, the system divided the entire operation process into four time slices: excitation, drive, contact establishment, and stabilization.

[0120] The trajectory formed during the healthy period initially rises rapidly along the first coordinate axis, then changes simultaneously along both the first and second coordinate axes, exhibiting only a brief, small response along the third coordinate axis before quickly returning to the stable region. Several months later, when the same device was closed again, the system found that the trajectory between the third and fourth slices was significantly elongated along both the second and third coordinate axes, indicating that the contact transition and arc extinguishing recovery no longer converged as rapidly as before. This change provided a clear physical entry point for subsequent fault classification.

[0121] Furthermore, in this embodiment, the instantaneous power gradient is preferably obtained according to the incremental rule within the time slice, rather than an abstract computational amount that lacks physical directionality and is detached from the business scenario;

[0122] Specifically, this can be understood as follows: First, the three synchronous measurement values ​​in each time slice are scaled up. Then, the energy contribution direction of each channel in the slice is determined based on the energy conversion coefficient. The three contributions are coupled and combined in the same slice using the admittance matrix to obtain the local rate of change of the time slice in the first coordinate axis direction, the second coordinate axis direction, and the third coordinate axis direction.

[0123] The specific calculation process is as follows: the three transient data after being scaled are constructed into an input column vector. First, the vector is multiplied by a diagonal transformation matrix composed of preset energy conversion coefficients to unify it to the energy flow rate dimension. Then, it is multiplied by a preset electromechanical admittance matrix to achieve cross-coupling mapping between the three-dimensional channels. Finally, the output is a column vector containing local rate of change of three-dimensional components.

[0124] The local rate of change corresponds to the degree of enhancement or weakening of the control loop drive, the acceleration or deceleration of the main loop contact state change, and the convergence or tailing of the high-frequency activity of the arc-extinguishing chamber, and can therefore be directly used as the transient energy flow rate.

[0125] Furthermore, to avoid different dimensions causing a certain channel to dominate the phase space coordinates for a long time, this embodiment preferably performs amplitude normalization processing on the three types of channels according to the equipment ledger or health sample baseline before mapping, so that the three-dimensional flow rate distribution of the same model of equipment during the healthy period falls into a comparable range.

[0126] The normalization process here does not change the original measurement facts, but rather converts the high-frequency electromagnetic field envelope signals of the coil current, main circuit voltage drop, and near-field space of the arc-extinguishing chamber into a unified flow rate scale, thereby ensuring that the trajectory shape mainly reflects the energy transfer structure, rather than simply reflecting the absolute magnitude difference of a certain channel; thus, the trajectory's turning point, elongation, and stagnation can all correspond to specific physical stage changes.

[0127] Furthermore, in this embodiment, the instantaneous power gradient is the slice-level result calculated according to a single time slice, and the transient energy flux is the name used after the slice-level result is mapped to the three-dimensional energy phase space. The two correspond one-to-one in numerical terms, without introducing a second set of calculations.

[0128] Correspondingly, the descriptions of local change rate, three-dimensional flow rate distribution, etc., appearing in the text are all descriptions of the same slice-level quantity from the perspective of computation or spatial distribution, rather than new parameters or new features. Through this unified approach, the instantaneous power gradient, local change rate, and transient energy flow rate can be avoided from being misunderstood as multiple independent computational objects.

[0129] The purpose of this step is to unify transient measurements from different sources and with different dimensions into an energy flow path description, thereby achieving comparability across models and operational phases, and providing a stable trajectory object for subsequent residual analysis.

[0130] In a preferred embodiment of the present invention, comparing the transient energy trajectory with a preset healthy state energy manifold and calculating the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold includes: extracting multiple trajectory feature points on the transient energy trajectory;

[0131] The tangent plane of the health state energy manifold at the trajectory feature point is calculated as the reference plane; the orthogonal projection vector of each trajectory feature point deviating from the reference plane is calculated; all the orthogonal projection vectors are combined to generate the orthogonal projection residual vector.

[0132] This embodiment provides a mechanism for energy manifold comparison and orthogonal projection residual generation in a healthy state. Specifically, if ordinary distance or simple threshold judgment is still used on the basis of the aforementioned energy trajectory already generated, it is easily affected by individual equipment installation errors, changes in ambient temperature, and normal aging drift.

[0133] Because the healthy state is not a fixed point, but rather a trajectory surface with a preset tolerance range and directionality, this embodiment further introduces the concept of the energy manifold of the healthy state and its local tangent plane to distinguish between natural drift along the healthy trend and abnormal escape from the healthy mechanism.

[0134] Specifically, the system extracts multiple trajectory feature points from the current transient energy trajectory; these feature points can be selected at physically meaningful locations such as trajectory turning points, near energy peaks, and before and after stable convergence.

[0135] Since the features that can effectively characterize the degradation of the internal state of the equipment during the opening and closing process are mainly concentrated in a few key mechanism nodes rather than globally uniformly distributed data points, this embodiment selectively extracts trajectory feature points with physical characterization significance for analysis; the system finds the local region corresponding to each feature point on the energy manifold of the healthy state and obtains the tangent plane of the region as the health reference plane at that point;

[0136] The tangent plane here can be understood as the natural direction of change that the health trajectory is allowed to exist in this local stage; for example, as the temperature drops, the rate at which the coil current builds up may change slightly, but as long as the change still follows the common evolutionary direction of health devices, it belongs to the manifold variation.

[0137] If the deviation direction of a certain feature point is significantly inconsistent with the healthy tangent plane, it is more likely to correspond to a new fault mechanism. Based on this, the system calculates the orthogonal projection vector of each trajectory feature point relative to the local reference plane. By using the orthogonal projection direction, the shortest abnormal deviation path of the current trajectory from the healthy reference plane can be extracted, thereby effectively removing the conventional data drift caused by fluctuations in operating conditions and more accurately characterizing the true fault characteristics of the equipment.

[0138] For ease of understanding, we can schematically select three feature points from a current trajectory: the front segment, the middle segment, and the tail segment. If the deviation of the front segment point relative to the tangent plane is less than the preset deviation threshold, it indicates that the electromagnetic starting part is still normal. If the deviation of the middle segment point reaches the preset deviation threshold, but the direction is basically consistent with the tangent plane, it indicates that it may only be a change in load conditions.

[0139] The tail point is clearly perpendicular to the tangent plane and extends outward, indicating that energy retention, which is not common in healthy samples, occurred in the tail of the operation. After combining the orthogonal projection vectors of these three points in chronological order, a set of residual vectors with directional and magnitude meanings is obtained. This vector not only indicates how much the deviation is, but also indicates which physical dimension the deviation is in.

[0140] As an anomaly handling mechanism, if there are too few samples in the corresponding region of the healthy state energy manifold and the local tangent plane is not stable enough, the system can expand the neighborhood range or retreat to a coarser-grained healthy cluster for comparison. However, such results are not directly used to update the healthy state energy manifold.

[0141] If the current trajectory is too short, for example, if the device fails to move, is in a half-state, or the command is interrupted, the complete set of feature points cannot be extracted. The system can generate local residuals for the completed part and mark it as an incomplete action event. If the deviation directions of multiple feature points conflict with each other, it indicates that the anomaly may be a compound fault or there is sampling contamination. The system can output a retest suggestion first and not immediately issue a single fault conclusion.

[0142] After several months of continuous operation at the substation, the system has accumulated healthy trajectories under multiple seasonal operating conditions for the same type of circuit breaker and formed a stable healthy energy manifold. During a planned trip, the first part of the current trajectory basically overlapped with the historical normal trajectory, the middle part showed an outward expansion exceeding the first preset tolerance, and the tail part deviated from the local tangent plane in the direction of the third coordinate axis exceeding the second preset tolerance.

[0143] Based on this, the system generates a residual vector dominated by the orthogonal projection of the tail section, and uses it as the basis for subsequent judgment of the relevant degradation of the arc-extinguishing chamber; if another device has a large-scale detachment in the front section, but the data in the middle and tail sections is insufficient, the system marks it as a mechanism abnormality and prioritizes its investigation.

[0144] Furthermore, in this embodiment, the trajectory feature points are preferably selected according to fixed business rules, rather than arbitrarily sampled; preferably, they include at least the first peak point of the start-up phase, the turning point of the contact transition phase, and the convergence inflection point before the stabilization phase; when there is obvious tailing in an operation, tailing points can be added as supplementary feature points; the purpose of this point selection is to make each feature point correspond to a clear physical phase, which facilitates the establishment of a one-to-one correspondence between the residual vector and the fault mechanism in the future.

[0145] Furthermore, finding the local region corresponding to each feature point on the energy manifold of the health state is preferably done by selecting several health feature points that are temporally adjacent and spatially close from health samples with the same equipment, operation type, installation posture, and similar load conditions to form a neighborhood, and then determining the tangent plane from this neighborhood; specifically, the dimensionality reduction and calculation of multiple health feature points in the neighborhood can be performed by principal component analysis or a plane fitting algorithm based on least squares, thereby obtaining the tangent plane that can represent the local evolution direction;

[0146] After this processing, the tangent plane represents the local direction in which the health trajectory is allowed to change under this physical stage, rather than the global average reference plane obtained from global sample statistics; thus, even if the equipment is affected by seasonal temperature or load changes and causes a small drift along the manifold direction, the system can still maintain a low false alarm; only when the current feature point deviates significantly from the local allowable direction will a large residual be formed in the orthogonal direction.

[0147] The purpose of this step is to decouple the natural operational drift that is allowed in health devices from actual abnormal deviations, thereby achieving more stable fault identification and reducing false alarms.

[0148] In a preferred embodiment of the present invention, determining the specific fault type of the vacuum circuit breaker based on the distribution characteristics of the orthogonal projection residual vector in different dimensions includes: extracting the first projection component of the orthogonal projection residual vector on the plane formed by the second coordinate axis and the third coordinate axis; and calculating the envelope attenuation rate of the high-frequency electromagnetic field envelope signal.

[0149] If the first projection component is greater than a preset planar projection threshold and the envelope attenuation rate is less than a preset attenuation rate threshold, the specific fault type is determined to be an arc extinguishing capability degradation fault; if the first projection component is less than or equal to the preset planar projection threshold, or the envelope attenuation rate is greater than or equal to the preset attenuation rate threshold, it is determined to be another specific fault type.

[0150] This embodiment provides a fault classification mechanism for arc extinguishing capability degradation. Specifically, based on the aforementioned residual vector, if the judgment is made solely based on the global amplitude of the residual vector, although the abnormal state of the operation process can be identified, it is difficult to accurately locate the root cause of the fault as either the degradation of the internal medium recovery capability of the arc extinguishing chamber or the sluggish mechanical response of the operating mechanism.

[0151] Especially in high-load switching scenarios, the main circuit voltage redistribution and the near-field electromagnetic activity of the arc-extinguishing chamber often occur simultaneously. Therefore, it is necessary to further subdivide the residual distribution characteristics in the plane formed by the second and third coordinate axes.

[0152] Specifically, the second coordinate axis corresponds to the energy flow rate characteristics related to the transient voltage drop of the main circuit, and the third coordinate axis corresponds to the energy flow rate characteristics related to the envelope of the near-field high-frequency electromagnetic field of the arc-extinguishing chamber. If the orthogonal projection residual accounts for a high proportion on the plane formed by these two dimensions, it indicates that the anomaly is mainly concentrated in the changes in the contact state of the contacts and the electromagnetic recovery process of the arc-extinguishing chamber, rather than in the coil driving stage.

[0153] Furthermore, the system calculates the envelope attenuation rate of the high-frequency electromagnetic field envelope signal to determine whether the high-frequency activity after arc extinguishing can quickly subside. The specific calculation method of the envelope attenuation rate is as follows: extract the tail segment envelope data of the high-frequency electromagnetic field envelope signal after contact separation, perform curve fitting on the tail segment envelope data based on the preset exponential decay model, and extract the reciprocal of the decay time constant in the fitting equation as the envelope attenuation rate.

[0154] Physically, in a healthy arc-extinguishing chamber, after the contacts separate, the medium recovers within a preset recovery period, and the near-field high-frequency electromagnetic activity decays at a preset rate. If the vacuum level decreases, the arc-extinguishing capability degrades, or stronger residual discharge conditions form around the contacts, the envelope decay will slow down, resulting in a continuous trailing effect.

[0155] For ease of explanation, the residual after a certain operation can be understood in three parts: the first coordinate axis has a small offset, while the second and third coordinate axes have significant offsets. If it is observed that the high-frequency electromagnetic field envelope signal in the near-field space of the arc-extinguishing chamber does not quickly return to the background level at the end of the operation, but continues for several consecutive time slices, then it can be considered that the medium recovery speed after contact separation is lower than that during the healthy period.

[0156] At this point, the system classifies it as a direction of arc extinguishing capability degradation; conversely, if the in-plane offset in the second and third coordinate axis directions is not prominent, or if there is a plane offset but the high-frequency electromagnetic field envelope signal of the near-field space of the arc extinguishing chamber still decays relatively quickly, it indicates that the anomaly has failed to form dual evidence consistent with vacuum degradation, and is not directly judged as arc extinguishing capability degradation.

[0157] As an anomaly handling mechanism, if the near-field high-frequency channel is subjected to external transient interference caused by the operation of nearby equipment, the envelope may be raised for a short time. However, this raising is usually not synchronized with the voltage drop change of the main circuit of this circuit breaker, nor does it follow the natural attenuation law of the end of the operation. Before making a judgment, the system can first check the time sequence correspondence between the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber and the current operation command and the main circuit event. If the causal consistency is not satisfied, the attenuation rate is not used as a valid criterion.

[0158] If the planar projection component is close to the threshold boundary, the system can mark the result as suspicious for the arc-extinguishing chamber state, and it is recommended to review it in conjunction with the trend of subsequent operations; if the second coordinate axis direction and the third coordinate axis direction are abnormal at the same time but one of the channels is missing, no definitive conclusion will be output to avoid misusing incomplete evidence for vacuum state determination.

[0159] During the peak summer season at the city's substation, the target circuit breaker undertook frequent switching tasks. After a certain trip, the system found that the current residual was mainly distributed in the plane formed by the main circuit voltage drop and the near-field high-frequency electromagnetic field. At the same time, the high-frequency electromagnetic field envelope attenuated significantly slower at the end of the operation.

[0160] Based on the recent log of a high number of contact interruptions for this equipment, the system prioritizes classifying this anomaly as a degradation of arc-extinguishing capability and prompts maintenance personnel to focus on checking the vacuum status of the arc-extinguishing chamber and the contact erosion. If another circuit breaker exhibits only limited planar projection under similar operating conditions, while the high-frequency electromagnetic field envelope signal in the near-field space of the arc-extinguishing chamber converges normally, the system will not classify it as an arc-extinguishing fault, but will instead proceed to the analysis process for other fault types.

[0161] The purpose of this step is to combine the voltage reconstruction and near-field electromagnetic recovery processes, which are most relevant to the internal state of the arc-extinguishing chamber, for analysis, so as to achieve targeted identification of vacuum level decline or arc-extinguishing capability degradation.

[0162] In a preferred embodiment of the present invention, determining the specific fault type of the vacuum circuit breaker based on the distribution characteristics of the orthogonal projection residual vector in different dimensions further includes: extracting the second projection component of the orthogonal projection residual vector on the first coordinate axis; and calculating the duration of the transient energy trajectory in the three-dimensional energy phase space as the energy integration time.

[0163] If the second projection component is greater than a preset axial projection threshold and the energy integration time is greater than a preset time threshold, the specific fault type is determined to be a mechanical fault of the operating mechanism; if the second projection component is less than or equal to the preset axial projection threshold, or the energy integration time is less than or equal to the preset time threshold, it is determined to be another specific fault type.

[0164] This embodiment provides a fault classification mechanism for mechanical failures of operating mechanisms. Specifically, in the previous embodiment, it was already possible to identify abnormalities in the direction of arc extinguishing capability degradation. However, in actual operation and maintenance, a large number of faults still originate from wear of the operating mechanism, poor lubrication, spring fatigue, or slight jamming of the connecting rod.

[0165] Such faults do not necessarily lead to obvious high-frequency tailing in the arc-extinguishing chamber, but they will leave energy output characteristics with extended action time in the electrical control circuit; therefore, this embodiment further extracts mechanical abnormality features from the residual components and trajectory duration in the first coordinate axis direction.

[0166] Specifically, the first coordinate axis corresponds to the energy flow rate characteristics related to the transient current signal, which directly reflects the energy input change of the operating coil when driving the mechanism; in the health device, after the coil current is established, the mechanism completes the action within the predetermined resistance range, the stretching degree of the energy trajectory on the first coordinate axis is limited, and the duration of the entire trajectory in the phase space is relatively stable.

[0167] If the mechanism experiences jamming, increased friction, or a decrease in the performance of the return elastic element, the coil drive holding time will exceed the preset normal operating time, which manifests as an increase in the residual in the first coordinate axis direction, and at the same time, the energy integration time of the trajectory will be prolonged. Here, the energy integration time reflects the effective duration of energy closure in one operation, which essentially corresponds to the dynamic time required for the mechanical chain to complete displacement and stabilization.

[0168] To facilitate understanding, two identical closing operations can be compared schematically. During the healthy period, the trajectory rises rapidly in the first coordinate axis direction and then quickly enters the dominant area in the second and third coordinate axis directions, and ends quickly. However, when there is mechanical jamming, the trajectory lingers longer in the first coordinate axis direction, indicating that the equipment repeatedly consumes energy during the start-up and transmission phases, but the change in contact state is not synchronously and fully advanced.

[0169] If the duration of the overall trajectory in phase space is observed to be longer, it indicates that the offset of the first coordinate axis is not a random current fluctuation, but more likely an energy retention caused by the increased resistance of the mechanical transmission chain; the system will then classify the anomaly as a mechanical failure of the operating mechanism.

[0170] As an anomaly handling mechanism, if the residual in the first coordinate axis direction is significant, but the trajectory duration is not prolonged, it may be caused by fluctuations in coil power supply conditions, transient deviations in operating power supply, or zero-point drift of the sensor, and should not be directly judged as a mechanical fault; if the duration is prolonged, but the residual in the first coordinate axis direction is not prominent, it may be due to delay in the arc-extinguishing tail section, lag in mechanism positioning detection, or other non-mechanical factors, and should be further judged by combining the characteristics of the second and third coordinate axis directions.

[0171] If the equipment is in an extremely low temperature environment, the increased viscosity of the lubricating grease may cause seasonal slowdowns in operation. In this case, the system can call up healthy samples from the same temperature range for comparison to avoid mistaking common changes caused by the environment for individual failures.

[0172] After the aforementioned outgoing line intervals underwent the autumn and winter season switching, a vacuum circuit breaker showed residual growth in the first coordinate axis direction during several trips, and the energy integration time was significantly increased compared to the healthy summer sample; at the same time, the main circuit voltage drop and the plane related to the near-field high-frequency electromagnetic field did not show any significant anomalies consistent with arc extinguishing degradation.

[0173] Based on this, the system prioritizes identifying the equipment as having a mechanical fault in the operating mechanism and points to checking the brake spring, shaft pin lubrication, and transmission link clearance in the maintenance recommendations; if subsequent maintenance confirms an increase in mechanism damping, the judgment chain will be consistent with the actual physical cause.

[0174] Furthermore, in this embodiment, the unified energy integration time refers to the duration from the effective starting point of the transient energy trajectory entering the three-dimensional energy phase space to its return to the stable convergence region;

[0175] The descriptions of trajectory duration, overall duration, effective duration, trajectory end time, etc. used in the preceding or following text for ease of description all correspond to the same judgment quantity. They are merely descriptions of the same time quantity in different contexts and do not represent new duration parameters.

[0176] This unified standard is used to ensure that the axial projection threshold and the preset time threshold always act on the same time measure, avoiding inconsistencies in the determination of mechanical faults due to changes in terminology;

[0177] The purpose of this step is to use the energy input characteristics of the electric drive side to reflect the internal blockage of the mechanical chain, thereby enabling the differentiation and diagnosis of mechanical faults and avoiding misattributing mechanical problems to arc-extinguishing chamber problems.

[0178] In a preferred embodiment of the present invention, generating a quantitative diagnostic report includes: performing inner product operations on the orthogonal projection residual vector with a preset three-dimensional vacuum degree weight coefficient vector and a three-dimensional electrical wear weight coefficient vector that match the three-dimensional energy phase space dimension; using the results of the inner product operations as the equivalent vacuum degree decrease index and the contact electrical wear depth assessment value, respectively; and encapsulating the equivalent vacuum degree decrease index and the contact electrical wear depth assessment value into the quantitative diagnostic report and outputting it.

[0179] This embodiment provides a quantitative diagnostic report generation mechanism; specifically, based on the aforementioned fault classification, if the system only provides textual conclusions such as the existence of arc extinguishing capability degradation or mechanical fault, although it can be used for alarms, it is not conducive to maintenance sequencing, spare parts preparation and life assessment.

[0180] Therefore, in this embodiment, the orthogonal projection residual vector is further mapped into a quantitative index with engineering readability, including at least the equivalent vacuum degree reduction index and the contact electrical wear depth assessment value.

[0181] Specifically, the vacuum degree weighting coefficient vector is used to characterize the strength of the correlation between each component of the residual and the change in the vacuum state of the arc-extinguishing chamber; generally, the residual component that better reflects the main circuit voltage drop reconstruction and near-field high-frequency electromagnetic tail has a higher weight in this vector.

[0182] The electrical wear weighting coefficient vector places more emphasis on the components related to contact state, post-arc recovery characteristics, and operational repeatability shift, because contact erosion and surface morphology changes will alter contact transition and arc root distribution.

[0183] The vacuum degree weight coefficient vector and the electrical wear weight coefficient vector are obtained as follows: using a historical fault sample library with known vacuum degree decline and known contact electrical wear depth, the orthogonal projection residual vector corresponding to each fault sample is extracted. Through multivariate statistical methods such as partial least squares regression or principal component analysis, the correlation mapping weight between each dimension component in the residual vector and the physical degradation amount is determined, thereby pre-generating the vacuum degree weight coefficient vector and the electrical wear weight coefficient vector respectively.

[0184] After the system fuses the residual vector with the corresponding weight vectors of the two sets, two quantitative results can be obtained.

[0185] The results here are not directly equivalent to physical vacuum pressure or actual ablation depth, but rather are assessment indicators that have a stable correlation with these physical degradation amounts, used to reflect the degree of deterioration of the equipment relative to the healthy baseline.

[0186] For ease of explanation, it can be illustrated that a certain residual vector can be understood as consisting of three sets of deviations: the front segment, the middle segment, and the tail segment. If the component related to high-frequency tailing dominates in the tail segment, it will contribute more to the vacuum condition assessment. If the component related to contact transition instability is high in the middle segment, it will contribute more significantly to the electrical wear assessment.

[0187] The system will write both types of evaluation results into a unified diagnostic report, which can include information such as the operation time, operation type, equipment number, dominant direction of residual, recommended maintenance level, and whether retesting is required. For dispatchers and maintenance personnel, this report is more practical and instructive than a simple fault label.

[0188] As an anomaly handling mechanism, if the data is only sufficient to support the fault direction determination but not to support stable quantification, such as a missing key channel, insufficient healthy baseline samples, or residuals near the threshold boundary, the report can only output the trend level and not the fine evaluation value. If the results of multiple operations on the same device fluctuate greatly within a short period, the system can use the trend envelope of multiple valid results to form a comprehensive report to avoid a jump in the quantitative indicators caused by a single occasional working condition.

[0189] If the equipment has just undergone maintenance and the arc-extinguishing chamber or contact assembly has been replaced, the original quantization baseline should be re-initialized to prevent the mixing of old and new device states.

[0190] In the monthly status assessment of the substation, the system generated a summary report on the multiple opening and closing records of the target circuit breaker over the past week. The report showed that the equivalent vacuum degree decline index continued to rise, while the contact electrical wear depth assessment value was at a medium level, indicating that the current main risk is more inclined to the deterioration of the arc-extinguishing chamber medium condition than simply contact wear.

[0191] Based on this, the maintenance department will prioritize the special inspection of the arc-extinguishing chamber and appropriately postpone the disassembly and inspection plan of the mechanical mechanism; if the report of another piece of equipment shows that the electrical wear assessment value has increased while the vacuum index is basically stable, then the operation and maintenance resources can be allocated more effectively.

[0192] The purpose of this step is to transform abstract residual deviations into quantitative health indicators that can be used for maintenance decisions, thereby achieving closed-loop support from alarm identification to condition assessment.

[0193] Please see Figure 2 A vacuum circuit breaker fault detection system based on deep fusion perception, wherein the vacuum circuit breaker includes an electrical control circuit, a main circuit, and an arc-extinguishing chamber, the system includes the following modules: a matrix construction module, used to respond to preset operation commands, synchronously acquire transient current signals of the electrical control circuit, transient voltage drop signals of the main circuit, and high-frequency electromagnetic field envelope signals of the near-field space of the arc-extinguishing chamber, and construct a multi-dimensional state matrix based on the transient current signals, the transient voltage drop signals, and the high-frequency electromagnetic field envelope signals;

[0194] The trajectory generation module is used to reconstruct the phase space of the multidimensional state matrix based on a preset electromechanical coupling energy admittance operator containing energy conversion coefficients and admittance matrix, and to calculate the transient energy flow rate of the multidimensional state matrix within a preset time window. The transient energy flow rate includes the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal, respectively.

[0195] The trajectory generation module is further configured to construct a three-dimensional energy phase space using the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal as the first coordinate axis, the second coordinate axis, and the third coordinate axis, respectively, and generate a transient energy trajectory in the three-dimensional energy phase space according to the transient energy flow rate.

[0196] The residual calculation module is used to compare the transient energy trajectory with the preset healthy state energy manifold and calculate the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold;

[0197] The diagnosis and update module is used to determine the specific fault type of the vacuum circuit breaker and generate a quantitative diagnosis report based on the distribution characteristics of the orthogonal projection residual vector in different dimensions when the amplitude of the orthogonal projection residual vector is higher than or equal to a preset alarm threshold; and to update the transient energy trajectory to a new healthy state energy manifold when the amplitude of the orthogonal projection residual vector is lower than the preset alarm threshold, and the current multidimensional state matrix sampling is valid and there is no external transient electromagnetic disturbance.

[0198] This embodiment provides a vacuum circuit breaker fault detection system based on deep fusion perception; specifically, the system can be deployed in the aforementioned urban substation to perform on-site online monitoring of multiple vacuum circuit breakers of the same or different models; the system can be installed in a single bay to form an independent diagnostic unit, or it can be connected to the monitoring host through the station's communication network to form a centralized management platform.

[0199] Specifically, the matrix construction module is connected to the field sensing chain and is responsible for collecting coil current, main circuit transient voltage drop and arc-extinguishing chamber near-field high-frequency electromagnetic field envelope according to a unified clock after receiving station control layer operation instructions or mechanism auxiliary contact triggers, and generating a multi-dimensional state matrix.

[0200] To ensure real-time performance, this module is preferably installed in edge-side hardware to perform data time-stamping correction, channel validity checks, and short-term buffering. After receiving the state matrix, the trajectory generation module calls the preset energy conversion coefficients and admittance matrix to convert the original signal into transient energy flow rates for each time slice and form a trajectory object in the three-dimensional energy phase space. This module can be implemented by a field-programmable logic device, a digital signal processor, or an industrial controller with real-time computing capabilities.

[0201] The residual calculation module is used to retrieve the energy manifold of the health status of the device and perform local tangent plane construction and orthogonal projection calculation at the trajectory feature point level to generate residual vectors. The role of this module in the system is to transform the difference between the current operation and the health operation into a deviation that can be interpreted in a directional way, rather than a simple waveform inconsistency.

[0202] The diagnosis and update module is responsible for taking subsequent actions based on the residual amplitude and distribution characteristics. When the residual exceeds the alarm threshold, the module outputs the arc extinguishing capability degradation, mechanical failure or other specific fault types according to the aforementioned classification rules, and generates a quantitative diagnosis report. When the residual is below the threshold, the module incorporates the current trajectory into the health status energy manifold update to track the natural evolution of the equipment during long-term service.

[0203] For ease of understanding, the system can be schematically understood as a four-level processing chain; the first level performs data acquisition and time-series alignment to generate a multi-dimensional state matrix; the second level maps the multi-dimensional time-series features to the spatial dimension to generate a directional energy trajectory; the third level performs a local comparison between the energy trajectory and the energy manifold of the healthy state and calculates the deviation vector; the fourth level outputs a specific fault type determination result based on the deviation vector features, or performs a manifold update operation if the threshold is not exceeded.

[0204] Each stage in this link corresponds to a specific measurement or physical interpretation task, so the whole system can maintain the properties of the measuring device rather than an abstract data analysis platform.

[0205] As an exception handling mechanism, if the matrix construction module detects a sensor malfunction or time synchronization failure, subsequent modules will not perform health updates and will only record invalid samples. If the trajectory generation module lacks adaptation parameters due to equipment model switching, it can load the energy conversion coefficient template of the corresponding model from the equipment ledger. If loading fails, it will switch to the sampling-only mode without judgment.

[0206] If the residual calculation module finds that the current number of healthy samples is insufficient, for example, the newly commissioned equipment has not yet completed baseline accumulation, the diagnosis and update module can first establish a temporary manifold with the initial factory test baseline and a small number of early healthy samples, and then switch to the formal baseline after several consecutive normal operations.

[0207] If multiple intervals operate simultaneously at the same time, the system can isolate the data according to the device number and local trigger source to avoid cross-attribution.

[0208] In the actual deployment of this substation, each circuit breaker mechanism box is equipped with an edge sampling node, and the station control layer server deploys residual calculation and reporting management programs; during a planned load switch, the target bay circuit breaker receives a tripping command, and the matrix construction module immediately starts three-way synchronous sampling;

[0209] The trajectory generation module generates and uploads the energy trajectory locally within milliseconds; the residual calculation module compares it with the energy manifold of the device's health status; the diagnosis and update module identifies that the anomalies are mainly concentrated in the planes of the second and third coordinate axes, and at the same time, the high-frequency electromagnetic field envelope signal of the near-field space of the arc extinguishing chamber converges more slowly. Therefore, it pushes an alarm on the arc extinguishing capability degradation to the operation and maintenance platform and generates a diagnostic report with the device number, timestamp and quantitative indicators.

[0210] If the residual of another device does not exceed the limit in the same batch of operations, its trajectory is automatically merged into the corresponding health state energy manifold for subsequent baseline iterations;

[0211] The purpose of this system is to integrate field high-frequency measurement, phase space reconstruction, residual analysis, and state update into an engineering-deployable closed-loop platform, thereby enabling online fault detection and continuous health tracking of vacuum circuit breakers during the opening and closing transient process.

[0212] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for detecting faults in vacuum circuit breakers based on deep fusion sensing, characterized in that, The vacuum circuit breaker includes an electrical control circuit, a main circuit, and an arc-extinguishing chamber. The method includes the following steps: In response to preset operation commands, the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber are acquired synchronously, and a multi-dimensional state matrix is ​​constructed based on the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal. The phase space of the multidimensional state matrix is ​​reconstructed based on a preset electromechanical coupling energy admittance operator that includes energy conversion coefficients and admittance matrix. The transient energy flow rate of the multidimensional state matrix within a preset time window is calculated. The transient energy flow rate includes the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal, respectively. Using the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal as the first coordinate axis, the second coordinate axis, and the third coordinate axis, a three-dimensional energy phase space is constructed, and a transient energy trajectory is generated in the three-dimensional energy phase space according to the transient energy flow rate. The transient energy trajectory is compared with a preset healthy state energy manifold, and the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold is calculated. If the magnitude of the orthogonal projection residual vector is higher than or equal to the preset alarm threshold, the fault type of the vacuum circuit breaker is determined and a quantitative diagnostic report is generated based on the distribution characteristics of the orthogonal projection residual vector in different dimensions. Under the premise that the magnitude of the orthogonal projection residual vector is lower than the preset alarm threshold, and the current multidimensional state matrix sampling is valid and there is no external transient electromagnetic disturbance, the transient energy trajectory is incorporated into the preset sample set to update the healthy state energy manifold. The electromechanical coupling energy admittance operator, based on a preset energy conversion coefficient and admittance matrix, reconstructs the phase space of the multidimensional state matrix, calculates the transient energy flux of the multidimensional state matrix within a preset time window, and generates a transient energy trajectory in the three-dimensional energy phase space based on the transient energy flux, including: Extract multiple time slices of the multidimensional state matrix within the preset time window; The instantaneous power gradient corresponding to each time slice is calculated using the electromechanical coupling energy admittance operator. The instantaneous power gradient corresponding to each time slice is taken as the transient energy flow rate; The transient energy flow rate is mapped onto the three-dimensional energy phase space to connect and generate the transient energy trajectory.

2. The vacuum circuit breaker fault detection method based on deep fusion sensing according to claim 1, characterized in that, The synchronous acquisition of the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc-extinguishing chamber includes: The transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal are acquired through a preset high-frequency edge computing node. Wherein, the synchronization sampling rate of the high-frequency edge computing node is higher than or equal to a preset sampling rate threshold; The transient voltage drop signal is obtained by coupling through a preset capacitive voltage divider.

3. The vacuum circuit breaker fault detection method based on deep fusion sensing according to claim 1, characterized in that, The step of comparing the transient energy trajectory with a preset healthy state energy manifold and calculating the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold includes: Extract multiple trajectory feature points on the transient energy trajectory; The tangent plane of the energy manifold of the health state at the trajectory feature point is calculated as the reference plane; Calculate the orthogonal projection vector of each trajectory feature point from the reference plane; All the orthogonal projection vectors are combined to generate the orthogonal projection residual vector.

4. The vacuum circuit breaker fault detection method based on deep fusion sensing according to claim 1, characterized in that, The step of determining the fault type of the vacuum circuit breaker based on the distribution characteristics of the orthogonal projection residual vector in different dimensions includes: Extract the first projection component of the orthogonal projection residual vector on the plane formed by the second coordinate axis and the third coordinate axis; Calculate the envelope attenuation rate of the high-frequency electromagnetic field envelope signal; If the first projection component is greater than a preset planar projection threshold and the envelope attenuation rate is less than a preset attenuation rate threshold, the fault type is determined to be an arc extinguishing capability degradation fault; if the first projection component is less than or equal to the preset planar projection threshold, or the envelope attenuation rate is greater than or equal to the preset attenuation rate threshold, it is determined to be another fault type.

5. The vacuum circuit breaker fault detection method based on deep fusion sensing according to claim 4, characterized in that, The step of determining the fault type of the vacuum circuit breaker based on the distribution characteristics of the orthogonal projection residual vector in different dimensions further includes: Extract the second projection component of the orthogonal projection residual vector on the first coordinate axis; The duration of the transient energy trajectory in the three-dimensional energy phase space is calculated as the energy integration time; If the second projection component is greater than a preset axial projection threshold and the energy integration time is greater than a preset time threshold, the fault type is determined to be a mechanical fault of the operating mechanism; if the second projection component is less than or equal to the preset axial projection threshold, or the energy integration time is less than or equal to the preset time threshold, it is determined to be another fault type.

6. The vacuum circuit breaker fault detection method based on deep fusion sensing according to claim 1, characterized in that, The generation of the quantitative diagnostic report includes: The orthogonal projection residual vector is respectively multiplied by the preset three-dimensional vacuum degree weight coefficient vector and the three-dimensional electrical wear weight coefficient vector that match the three-dimensional energy phase space dimension. The results of the inner product operation are used as the equivalent vacuum degree reduction index and the contact electrical wear depth evaluation value, respectively. The equivalent vacuum degree reduction index and the contact electrical wear depth assessment value are packaged into the quantitative diagnostic report and output.

7. A vacuum circuit breaker fault detection system based on deep fusion sensing, characterized in that, The vacuum circuit breaker includes an electrical control circuit, a main circuit, and an arc-extinguishing chamber. The system includes the following modules: The matrix construction module is used to respond to preset operation commands, synchronously acquire the transient current signal of the electrical control circuit, the transient voltage drop signal of the main circuit, and the high-frequency electromagnetic field envelope signal of the near-field space of the arc extinguishing chamber, and construct a multi-dimensional state matrix based on the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal. The trajectory generation module is used to reconstruct the phase space of the multidimensional state matrix based on a preset electromechanical coupling energy admittance operator containing energy conversion coefficients and admittance matrix, and to calculate the transient energy flow rate of the multidimensional state matrix within a preset time window. The transient energy flow rate includes the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal, respectively. The trajectory generation module is further configured to construct a three-dimensional energy phase space using the energy flow rate characteristics corresponding to the transient current signal, the transient voltage drop signal, and the high-frequency electromagnetic field envelope signal as the first coordinate axis, the second coordinate axis, and the third coordinate axis, respectively, and generate a transient energy trajectory in the three-dimensional energy phase space according to the transient energy flow rate. The residual calculation module is used to compare the transient energy trajectory with the preset healthy state energy manifold and calculate the orthogonal projection residual vector of the transient energy trajectory deviating from the healthy state energy manifold; The diagnosis and update module is used to determine the fault type of the vacuum circuit breaker and generate a quantitative diagnosis report based on the distribution characteristics of the orthogonal projection residual vector in different dimensions when the amplitude of the orthogonal projection residual vector is higher than or equal to a preset alarm threshold; and to update the transient energy trajectory to a new healthy state energy manifold when the amplitude of the orthogonal projection residual vector is lower than the preset alarm threshold, and the current multidimensional state matrix sampling is valid and there is no external transient electromagnetic disturbance. The electromechanical coupling energy admittance operator, based on a preset energy conversion coefficient and admittance matrix, reconstructs the phase space of the multidimensional state matrix, calculates the transient energy flux of the multidimensional state matrix within a preset time window, and generates a transient energy trajectory in the three-dimensional energy phase space based on the transient energy flux, including: Extract multiple time slices of the multidimensional state matrix within the preset time window; The instantaneous power gradient corresponding to each time slice is calculated using the electromechanical coupling energy admittance operator. The instantaneous power gradient corresponding to each time slice is taken as the transient energy flow rate; The transient energy flow rate is mapped onto the three-dimensional energy phase space to connect and generate the transient energy trajectory.

Images