Pole-mounted circuit breaker life prediction system based on multi-sensor fusion

Abstract

The present application belongs to the technical field of power equipment state monitoring and life assessment, and discloses a pole-mounted circuit breaker life prediction system based on multi-sensor fusion, which comprises collecting multi-source heterogeneous data during the opening and closing operation of the circuit breaker; calculating the inherent delay compensation value of each collection channel, performing time domain alignment of the multi-source heterogeneous data; and identifying the persistent failure of the collection channel, discriminating the interference outliers, combining the time sequence causal constraint to distinguish the real failure from the abnormal sampling points, completing data correction; constructing a dynamic three-level scale containing a reference value, a deterioration threshold and a failure threshold, extracting deterioration characteristics, and completing the independent state classification of each component; generating a reference deterioration trajectory, combining the equipment comprehensive deterioration degree fine-tuning trajectory, and accordingly carrying out life analysis to obtain the deterioration threshold reaching life and failure life. The present application solves the problems of heterogeneous data time sequence misplacement, field interference and poor static threshold adaptation, outputs layered life results, and supports the state maintenance of distribution network circuit breakers.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and life assessment technology, specifically a pole-mounted circuit breaker life prediction system based on multi-sensor fusion. Background Technology

[0002] As a widely used switching device in power distribution networks, pole-mounted circuit breakers are directly related to the reliability of power supply. At present, multi-sensor condition monitoring and life assessment technology for pole-mounted circuit breakers has made some progress, but there are still many technical shortcomings that need to be addressed.

[0003] First, in existing multi-sensor monitoring solutions, the hardware of various heterogeneous sensors such as vibration, displacement, and current has inherent transmission delays. At the same time, different sensors have different sampling frequency configurations. Most systems rely on simple hardware synchronization triggering without performing quantitative compensation and unified time axis mapping for the inherent delays of each acquisition channel. This can easily cause timing misalignment between the acquired data of different physical quantities, resulting in a shift in the time of the characteristic peaks corresponding to each physical parameter during the opening and closing process. This leads to distortion of the deterioration features extracted subsequently, directly reducing the accuracy of the component deterioration status assessment results.

[0004] Secondly, pole-mounted circuit breakers are deployed in complex outdoor environments for extended periods. Factors such as electromagnetic interference, environmental disturbances, and aging or damage to sensor hardware can cause the collected data to contain a variety of abnormal data, including continuous hardware failures of sensor channels, random electromagnetic interference outliers, and actual sudden failures of the equipment itself. Most existing data processing methods only use a single threshold truncation method to remove out-of-limit data, lacking a multi-dimensional, hierarchical anomaly discrimination mechanism that considers physical boundary constraints, the continuity of sampling point changes, and the inherent temporal causal logic of opening and closing actions. This makes it difficult to effectively distinguish between the different types of abnormal data, and it is easy to mistakenly remove real equipment failures as interference data or identify random interference outliers as equipment degradation failures. Directly involving abnormal data in degradation characteristic calculations will seriously interfere with subsequent state assessments and lifespan projections.

[0005] Furthermore, the thresholds currently used for circuit breaker degradation assessment are mostly static fixed thresholds obtained from accelerated aging tests of laboratory prototypes or factory experience parameters. There are significant differences between laboratory operating conditions and actual on-site operating conditions. The system does not fully utilize historical operating samples of the same model of equipment at different service stages in the target area, and lacks an equivalent full-lifetime degradation sequence formed by splicing samples of multi-gradient commissioning durations. This makes it difficult to accurately identify the physical inflection point where equipment deteriorates from slow to accelerated degradation. At the same time, the traditional threshold calibration process only focuses on the physical laws of equipment degradation itself and does not include the actual on-site maintenance and handling margin in the threshold calibration constraints. This can easily lead to problems such as the calibrated degradation inflection point being too close to failure, leaving insufficient operational space for maintenance personnel to repair and handle the issue, or the inflection point being set too conservatively, resulting in a large number of invalid alarms. Static thresholds cannot keep up with the iterative updates of regional equipment samples and are difficult to adapt to the complex and ever-changing actual operating conditions of on-site equipment, resulting in prominent problems of false alarms and missed alarms. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention aims to provide a pole-mounted circuit breaker life prediction system based on multi-sensor fusion, comprising a cloud monitoring platform, a multi-sensor sensing unit, an edge data time-domain alignment unit, an edge data correction unit, a data processing unit, and a life analysis module. The multi-sensor sensing unit is used to collect multi-source heterogeneous data corresponding to each opening and closing operation of the pole-mounted circuit breaker; The edge data temporal alignment unit is used to detect the actual occurrence time of the corresponding characteristic peak in multi-source heterogeneous data. It obtains the inherent delay compensation value based on the time offset of each acquisition channel relative to the reference benchmark, and uses the inherent delay compensation value to complete the temporal alignment of multi-source heterogeneous data. The edge data correction unit is used to first identify channel-level persistent faults based on the cumulative number of out-of-limit sampling points, then identify interference wild values ​​by the difference between the data before and after the sampling point, and finally, based on the time window defined by the global time zero point, combined with multi-channel time-series causal constraints, distinguish between real faults and abnormal sampling points, and perform data anomaly correction. The data processing unit is used to construct a dynamic three-level scale, extract degradation features from multi-source heterogeneous data, classify the degradation feature indicators of each component independently based on the dynamic three-level scale, and obtain the real-time classification status of each component. The life analysis module is used to construct the baseline degradation trajectory of each component based on a dynamic three-level scale. Based on the comprehensive degradation value of each component of several pole-mounted circuit breakers in the target area, the baseline degradation trajectory of each component is fine-tuned. Based on the fine-tuned baseline degradation trajectory, life analysis is performed to obtain the degradation threshold of the pole-mounted circuit breaker to reach its lifespan or failure life.

[0007] Furthermore, multi-sensor sensing units, edge data temporal alignment units, and edge data correction units are pre-deployed on several pole-mounted circuit breakers within the target area, and data processing units and lifespan analysis modules are built within the cloud monitoring platform; The multi-sensor sensing unit, edge data time domain alignment unit, and edge data correction unit on each pole-mounted circuit breaker are connected to the cloud monitoring platform via IoT nodes. The rising edge of the opening and closing coil driving voltage is used as the global time zero point of the multi-sensor sensing unit. Whenever the opening and closing coil circuit has a rising edge of driving voltage, the rising edge of driving voltage is marked as the global time zero point of this opening and closing operation. When the multi-sensor sensing unit detects the global time zero point, all acquisition channels are started simultaneously through the same clock pulse to acquire multi-source heterogeneous data.

[0008] Furthermore, the edge data temporal alignment unit is used to detect the actual occurrence time of the corresponding feature peak in the multi-source heterogeneous data. Based on the time offset of each acquisition channel relative to the reference benchmark, an inherent delay compensation value is obtained. The process of using the inherent delay compensation value to complete the temporal alignment of the multi-source heterogeneous data includes: The moment when the driving voltage rises in the tripping coil circuit is marked as the moment when the contacts just open; the moment when the driving voltage rises in the closing coil circuit is marked as the moment when the contacts just close; and the moments when the contacts just open and close are marked as reference points. The characteristic peaks corresponding to the moment when the contacts just separate and the moment when the contacts just close are set as the reference references; Real-time feature peak detection is performed on the multi-source heterogeneous data collected by the multi-sensor sensing unit to obtain the occurrence time of the feature peak corresponding to the reference benchmark. Based on the time offset between the reference benchmark and the occurrence time of the corresponding feature peak, the inherent delay compensation value of each acquisition channel in the multi-sensor sensing unit is obtained. Based on the inherent delay compensation value, the time series of the data in the corresponding acquisition channel is shifted and corrected as a whole on the time axis. The time series of the data in each acquisition channel after shift correction is mapped to the same standard time axis.

[0009] Furthermore, the edge data correction unit first identifies channel-level persistent faults based on the cumulative number of out-of-limit sampling points, then identifies interfering outliers by the difference between data before and after the sampling point, and finally, based on the time window defined by the global time zero point, combined with multi-channel temporal causal constraints, distinguishes between real faults and abnormal sampling points. The process of performing data anomaly correction includes: Set the physical boundary thresholds for each acquisition channel, the maximum allowable jump amplitude between adjacent sampling points, and the temporal causal constraints between each acquisition channel, and then perform the following steps: Step a1: Compare the data of each acquisition channel with the corresponding physical boundary threshold. If a certain acquisition channel has sampling points that exceed the corresponding physical boundary threshold, extract the cumulative number of sampling points that exceed the corresponding physical boundary threshold for that acquisition channel. If the cumulative number is greater than the preset sampling point number threshold, generate a channel-level continuous fault alarm for that acquisition channel. If the cumulative number is less than or equal to the preset sampling point number threshold, proceed to step a2. Step a2: For the aforementioned sampling points that exceed the corresponding physical boundary threshold, calculate the data difference between the sampling point and the previous and next adjacent sampling points respectively. If any data difference is greater than the maximum allowable jump amplitude of the adjacent sampling points of the acquisition channel, the data corresponding to the sampling point is determined to be an interference field value. If each data difference is less than or equal to the maximum allowable jump amplitude of the adjacent sampling points of the acquisition channel, then step a3 is executed for the sampling point. Step a3: Take the global time zero point of this opening and closing operation as the start timestamp of the preset time window. Within the time window, obtain the sequence and time interval between other sampling points of other acquisition channels and this sampling point. Compare the obtained sequence and time interval with the temporal causal constraints between the acquisition channel to which this sampling point belongs and other acquisition channels. If the temporal causal constraints are satisfied, mark the preset time window to which this sampling point belongs as a real fault state and generate a sudden fault alarm. Otherwise, mark this sampling point as an abnormal state. Step a4: Correct data anomalies for sampling points of channel-level persistent fault alarms, interference outliers, and abnormal states.

[0010] Furthermore, the data processing unit extracts degradation features from multi-source heterogeneous data, performs independent state classification of degradation feature indicators for each component based on a dynamic three-level scale, and obtains the comprehensive degradation value and real-time classification status of each component. The process includes: A dynamic three-level scale is constructed, which includes the baseline value, degradation threshold and failure threshold corresponding to the degradation characteristic index of each component of several pole-mounted circuit breakers in the target area. Based on the degradation mechanism of pole-mounted circuit breakers, data retrieval is used to obtain degradation characteristic indicators of each component. Based on the degradation characteristic indicators of each component, feature extraction is performed on the data collected by all acquisition channels within the time window to obtain the characteristic values ​​corresponding to the degradation characteristic indicators of each component. The characteristic values ​​corresponding to the degradation characteristic indicators of each component are compared with the dynamic three-level scale to obtain the comprehensive degradation degree value of each component. Based on the comprehensive degradation degree value of each component, the real-time classification status of each component is obtained. Real-time component classification status includes normal degradation status, degradation status awaiting warning, and failure status.

[0011] Furthermore, the process of constructing a dynamic three-level scale includes: Obtain all historical time windows in which several pole-mounted circuit breakers in the target area are in a non-real fault state. Obtain the real-time graded status of each component within all historical time windows. Mark the historical time window in which the real-time graded status of each component is in a normal deterioration state as the reference time window. Select the average value of the deterioration characteristic index of each component within all reference time windows as the reference value of the deterioration characteristic index of each component. Obtain all historical time windows of several pole-mounted circuit breakers in the target area under actual fault conditions, as well as all historical time windows of several other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area under actual fault conditions. Mark the characteristic values ​​corresponding to the degradation characteristic indicators of each component in the target area and the several other pole-mounted circuit breakers of the same model as failure sample values ​​within all historical time windows. Compare the failure sample values ​​corresponding to the degradation characteristic indicators of the same component with the benchmark values ​​of the degradation characteristic indicators of the same component, and select the failure sample value closest to the benchmark value as the failure threshold.

[0012] Furthermore, the process of constructing a dynamic three-level scale also includes: Obtain several other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area, remove the pole-mounted circuit breakers that have generated sudden fault alarms in the target area and the pole-mounted circuit breakers that have generated sudden fault alarms in other pole-mounted circuit breakers, and mark the remaining pole-mounted circuit breakers and other pole-mounted circuit breakers in the target area as sample pole-mounted circuit breakers. Obtain all historical time windows of the sample pole-mounted circuit breaker in non-real fault state, as well as the actual cumulative commissioning time and cumulative number of opening and closing operations of the sample pole-mounted circuit breaker within each historical time window. For the characteristic values ​​corresponding to the degradation characteristic indicators of each component within each historical time window, bind the actual cumulative commissioning time and cumulative number of opening and closing operations of the sample pole-mounted circuit breaker within the corresponding historical time window to generate the gradient sample of the sample pole-mounted circuit breaker. The actual cumulative commissioning time includes: 1-3 years, 3-6 years and 6-10 years. For all sample pole-mounted circuit breakers of the same model, the gradient samples are sorted according to the actual cumulative commissioning time and the cumulative number of opening and closing operations in each gradient sample to obtain a mixed time sequence sample sequence. Based on the cumulative number of opening and closing operations, a sliding statistical window is set, and the mixed time series sample sequence is traversed sequentially to obtain the average rate of change of the degradation characteristic index of each component within each sliding statistical window. A rate sequence is constructed based on the average rate of change of the degradation characteristic index of each component within each sliding statistical window. Based on the average rate of change within each sliding statistical window corresponding to the gradient samples with an actual cumulative operation time of 1-3 years, a steady-state baseline rate for the degradation characteristic index of each component is set. Based on the steady-state baseline rate, the rate sequence is traversed. When there are no less than 3 consecutive adjacent sliding statistical windows, and the average rate of change of the degradation characteristic index of the component within the adjacent sliding statistical windows is greater than the preset multiple of the corresponding steady-state baseline rate and is in the range of continuous rate increase, the first sliding statistical window of the 3 adjacent sliding statistical windows is determined to be the initial degradation inflection point. The conditions for the rate of continuous increase are as follows: taking the three adjacent sliding statistical windows as the core and extending two sliding statistical windows forward and backward to form a seven-window continuous interval, the average rate of change of the degradation characteristic index of all components in the seven-window continuous interval is higher than the corresponding steady-state reference rate.

[0013] Furthermore, the process of constructing a dynamic three-level scale also includes: In the mixed time-series sample sequence corresponding to all sample pole-mounted circuit breakers of the same model, query the remaining cumulative number of opening and closing operations between the initial deterioration inflection point of the component's deterioration characteristic index and the corresponding position of the failure threshold of the component's deterioration characteristic index, and mark the remaining cumulative number of opening and closing operations as the manageable operation margin. To preset the minimum maintenance and handling capacity threshold, perform the following steps: Step b1: If the manageable operational margin is less than the minimum maintenance and handling margin threshold, then obtain the offset step size based on the failure threshold and baseline value of the component's degradation characteristic index. Based on the offset step size, iterate the initial degradation inflection point towards the baseline value round by round. After each round of offset, determine whether the initial degradation inflection point after the current offset is in the range of continuously increasing rate. If it is, continue to the next round of offset iteration until the manageable operational margin is greater than or equal to the minimum maintenance and handling margin threshold. Mark the initial degradation inflection point after the current offset as the final degradation inflection point. If the initial degradation inflection point after the current offset is no longer in the range of continuously increasing rate, immediately cancel the current round of offset operation, revert to the position of the initial degradation inflection point in the previous round that was still in the range of continuously increasing rate, and mark the reverted initial degradation inflection point as the final degradation inflection point, and terminate the iterative offset. Step b2: If the manageable operational margin is greater than a preset multiple of the minimum maintenance and handling margin threshold, the initial degradation inflection point is iteratively shifted towards the failure direction in rounds based on the offset step size. After each round of offset, it is determined whether the initial degradation inflection point after the current offset is in the range of continuously increasing rate. If it is, the next round of offset iteration continues until the manageable operational margin is less than or equal to a preset multiple of the minimum maintenance and handling margin threshold. The initial degradation inflection point after the current offset is marked as the final degradation inflection point. If the initial degradation inflection point after the current offset is no longer in the range of continuously increasing rate, the current round of offset operation is immediately canceled, and the position of the initial degradation inflection point in the previous round, which was still in the range of continuously increasing rate, is returned. The initial degradation inflection point after the return is marked as the final degradation inflection point, and the iterative offset is terminated. Step b3: If the manageable operational margin is greater than or equal to the minimum maintenance and operation margin threshold and less than or equal to a preset multiple of the minimum maintenance and operation margin threshold, then the initial degradation inflection point is directly marked as the final degradation inflection point. The characteristic value of the degradation characteristic index of the component corresponding to the final degradation inflection point is taken as the degradation threshold of the degradation characteristic index of that component. A dynamic three-level scale is constructed based on the baseline values, degradation thresholds, and failure thresholds corresponding to the degradation characteristic indicators of each component of each type of pole-mounted circuit breaker.

[0014] Furthermore, based on a dynamic three-level scale, a baseline degradation trajectory for each component is constructed. Based on the comprehensive degradation value corresponding to each component of several pole-mounted circuit breakers within the target area, the process of fine-tuning the baseline degradation trajectory for each component includes: Starting from the baseline value of the degradation characteristic index of each component in the dynamic three-level scale, with the failure threshold as the end point and the degradation threshold as the two-segment dividing inflection point, the baseline degradation trajectory of each component is generated based on gradient samples. The horizontal axis of the baseline degradation trajectory is the cumulative number of opening and closing operations, and the vertical axis is the comprehensive degradation value of each component. The comprehensive degradation value of each component of several pole-mounted circuit breakers in the target area is compared with the baseline degradation trajectory of each component to obtain the relative deviation rate of each component. A relative deviation rate threshold is set. If the relative deviation rate of a component is less than or equal to the relative deviation rate threshold, the baseline degradation trajectory of the component remains unchanged. If the relative deviation rate of a component is greater than the relative deviation rate threshold, the degradation rate of the baseline degradation trajectory is adjusted according to the relative deviation rate to obtain the adjusted baseline degradation trajectory.

[0015] Furthermore, based on the fine-tuned baseline degradation trajectory, the process of obtaining the degradation threshold of the pole-mounted circuit breaker to reach its lifespan or failure life includes: If all components of the pole-mounted circuit breaker are in a normal deterioration state, then based on the current cumulative number of opening and closing operations of the pole-mounted circuit breaker and the baseline deterioration trajectory of each component, the deterioration threshold of each component is obtained to reach the remaining number of opening and closing operations. From the deterioration thresholds of each component to reach the remaining number of opening and closing operations, the minimum deterioration threshold to reach the remaining number of opening and closing operations is selected as the deterioration threshold of the pole-mounted circuit breaker to reach its lifespan. If a component of a pole-mounted circuit breaker is in a state of pending warning deterioration, the remaining number of opening and closing operations for each component in the pending warning deterioration state is obtained based on the current cumulative number of opening and closing operations of the pole-mounted circuit breaker and the baseline deterioration trajectory of each component in the pending warning deterioration state. From the remaining number of opening and closing operations of each component in the pending warning deterioration state, the minimum number of remaining opening and closing operations is selected as the failure life of the pole-mounted circuit breaker.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. Using the rising edge of the opening and closing coil driving voltage as the global time zero point, and combined with the reference benchmark when the contacts are just opening and closing, the inherent delay compensation value is obtained by the time offset of the characteristic peaks of each acquisition channel. The timing shift is completed and uniformly mapped to the same standard time axis, eliminating the timing misalignment problem caused by the inherent delay of different sensor hardware and the inconsistency of sampling frequency. This provides a consistent original data foundation for subsequent feature extraction and degradation analysis, avoiding feature distortion and misjudgment of degradation assessment caused by timing deviation.

[0017] 2. The collected data is verified and distinguished step by step from three levels: physical boundary, sampling continuity, and multi-channel temporal causal relationship. This can not only identify channel continuous faults caused by sensor hardware abnormalities, but also identify random outliers caused by on-site electromagnetic interference. At the same time, based on the inherent physical timing logic of the opening and closing operation, it can distinguish between real sudden equipment failures and invalid abnormal sampling points.

[0018] 3. A dynamic three-level benchmark adapted to in-operation equipment in the field is constructed. It does not rely on accelerated aging samples from the laboratory, but instead utilizes historical operating data from the target area and in-operation circuit breakers of the same model to calibrate baseline values, degradation thresholds, and failure thresholds. An equivalent full-lifecycle evolution sequence is formed by stitching together gradient samples from multiple operational durations, identifying physical inflection points where the degradation rate continuously increases. The degradation threshold is adaptively adjusted based on maintenance and repair margin constraints, balancing the physical accuracy of degradation inflection points with on-site maintenance operability. Furthermore, the benchmark can iteratively update with regional equipment samples, unlike traditional fixed static thresholds. It can adapt to the degradation differences caused by actual on-site operating conditions, reducing the risk of false alarms and missed alarms caused by a single fixed threshold, and providing an objective and dynamic evaluation basis for component degradation classification. Attached Figure Description

[0019] Figure 1This is a flowchart of a pole-mounted circuit breaker life prediction system based on multi-sensor fusion, according to an embodiment of this application. Figure 2 This is a flowchart illustrating how the edge data correction unit performs data anomaly correction in an embodiment of this application. Figure 3 A flowchart illustrating the process of identifying the final degradation inflection point during the construction of a dynamic three-level scale in this application embodiment. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] like Figure 1 As shown, the pole-mounted circuit breaker life prediction system based on multi-sensor fusion includes a cloud monitoring platform, a multi-sensor sensing unit, an edge data temporal domain alignment unit, an edge data correction unit, a data processing unit, and a life analysis module. The multi-sensor sensing unit is used to collect multi-source heterogeneous data corresponding to each opening and closing operation of the pole-mounted circuit breaker; The edge data temporal alignment unit is used to detect the actual occurrence time of the corresponding characteristic peak in multi-source heterogeneous data. It obtains the inherent delay compensation value based on the time offset of each acquisition channel relative to the reference benchmark, and uses the inherent delay compensation value to complete the temporal alignment of multi-source heterogeneous data. The edge data correction unit is used to first identify channel-level persistent faults based on the cumulative number of out-of-limit sampling points, then identify interference wild values ​​by the difference between the data before and after the sampling point, and finally, based on the time window defined by the global time zero point, combined with multi-channel time-series causal constraints, distinguish between real faults and abnormal sampling points, and perform data anomaly correction. The data processing unit is used to construct a dynamic three-level scale, extract degradation features from multi-source heterogeneous data, classify the degradation feature indicators of each component independently based on the dynamic three-level scale, and obtain the real-time classification status of each component. The life analysis module is used to construct the baseline degradation trajectory of each component based on a dynamic three-level scale. Based on the comprehensive degradation value of each component of several pole-mounted circuit breakers in the target area, the baseline degradation trajectory of each component is fine-tuned. Based on the fine-tuned baseline degradation trajectory, life analysis is performed to obtain the degradation threshold of the pole-mounted circuit breaker to reach its lifespan or failure life.

[0022] For the five core failure components of pole-mounted circuit breakers (vacuum interrupter, operating mechanism, insulating support, conductive circuit, and control circuit), a layout logic combining point and surface approaches and main and auxiliary sensors is adopted. Each type of sensor corresponds to one or more degradation modes, while redundant verification sensors are set up to eliminate environmental interference. Specific sensor configurations and monitoring targets are as follows: Vibration sensors: 3 in total, installed on the side of the operating mechanism housing, the upper flange of the insulating support, and the stationary end cover of the arc-extinguishing chamber, respectively. They are used to collect the time-domain waveform of impact vibration during the opening and closing process, corresponding to monitoring mechanical deterioration such as mechanism jamming, spring fatigue, connecting rod loosening, and contact bounce inside the arc-extinguishing chamber.

[0023] Contact temperature sensors: 4 in total, embedded in the upper and lower terminals, inside the moving conductive rod, and on the surface of the spring-loaded motor inside the mechanism housing. Used to monitor electrical contact and thermal degradation such as increased contact resistance in the conductive circuit and motor overload heating.

[0024] Current transformers: Two in total, connected in series in the opening and closing coil circuit and the main circuit phase line, respectively. They are used to collect the coil current waveform and the main circuit load current, and to monitor electrical and mechanical coupling deterioration such as inter-turn short circuits, core jamming, and contact erosion.

[0025] Displacement sensor: 1 unit, installed at the end of the moving contact transmission link, using the non-contact laser displacement principle. It is used to collect mechanical characteristic parameters such as the opening and closing stroke curve, the speed of just opening / closing, and overtravel, directly reflecting the transmission efficiency of the mechanism and the amount of contact wear.

[0026] Partial discharge sensor: 1 unit, installed on the middle surface of the insulating support, using the ultra-high frequency coupling principle. Used to monitor insulation degradation such as vacuum level drop in the vacuum interrupter and contamination discharge on the insulating surface.

[0027] Environmental sensor: 1 unit, integrating temperature, humidity, salt spray, and light intensity data acquisition, installed on the shaded side of the equipment. Used to collect environmental stress parameters, providing environmental input for subsequent degradation rate correction.

[0028] Layout logic: All high-voltage side sensors adopt passive or low-power design and communicate with edge units through optical fiber to avoid electromagnetic interference; mechanical sensors and electrical sensors are paired one-to-one to form a verification pair. For example, vibration sensors and displacement sensors jointly verify the mechanical state of the mechanism, and temperature sensors and current sensors jointly verify the degradation of conductive circuits.

[0029] It should be further explained that, in the specific implementation process, multi-sensor sensing units, edge data time domain alignment units, and edge data correction units are pre-deployed on several pole-mounted circuit breakers in the target area, and data processing units and life analysis modules are built in the cloud monitoring platform; The multi-sensor sensing unit, edge data time domain alignment unit, and edge data correction unit on each pole-mounted circuit breaker are connected to the cloud monitoring platform via IoT nodes. The rising edge of the opening and closing coil drive voltage is used as the global time zero point of the multi-sensor sensing unit. The physical basis for selecting this signal as the reference is that the coil energization is the electrical starting point of the opening and closing operation. The signal is stable and has no mechanical delay. It can be directly acquired through the electrical circuit without the need for additional sensors. Whenever the opening and closing coil circuit has a rising edge of the drive voltage, the rising edge of the drive voltage is marked as the global time zero point of this opening and closing operation. When the multi-sensor sensing unit detects the global time zero point, all acquisition channels (vibration, displacement, main circuit current) are started simultaneously through the same clock pulse to acquire multi-source heterogeneous data. From the hardware level, the acquisition start time of each channel is consistent, and the coarse alignment error is controlled within ±10μs, which can offset most of the hardware trigger delay.

[0030] It should be further explained that, in the specific implementation process, the edge data temporal alignment unit is used to detect the actual occurrence time of the corresponding characteristic peak in the multi-source heterogeneous data. Based on the time offset of each acquisition channel relative to the reference benchmark, an inherent delay compensation value is obtained. The process of using the inherent delay compensation value to complete the temporal alignment of the multi-source heterogeneous data includes: The moment when the driving voltage rises in the tripping coil circuit is marked as the moment when the contacts just open; the moment when the driving voltage rises in the closing coil circuit is marked as the moment when the contacts just close; and the moments when the contacts just open and close are marked as reference points. The characteristic peaks corresponding to the moment when the contacts just separate and the moment when the contacts just close are set as the reference references; When the moment the contact just opens is set as the reference, the characteristic peaks corresponding to this moment include: the inflection point of the displacement curve from static to dynamic (corresponding to the displacement channel), the drop point of the main circuit current (corresponding to the main circuit current channel), and the peak value of the opening impact output by the vibration sensor (corresponding to the vibration channel). When the moment the contacts just close is set as the reference, the characteristic peaks corresponding to these peaks include: the inflection point of the sudden drop in velocity of the displacement curve (corresponding to the displacement channel), the point at which the main circuit current rises from zero (corresponding to the main circuit current channel), and the peak value of the closing impact output by the vibration sensor (corresponding to the vibration channel). Real-time characteristic peak detection is performed on the multi-source heterogeneous data collected by the multi-sensor sensing unit to obtain the occurrence time of the characteristic peak corresponding to the reference. Based on the time offset between the reference and the occurrence time of the corresponding characteristic peak, the inherent delay compensation value of each acquisition channel in the multi-sensor sensing unit is obtained. For example, at the moment when the contact just separates, the characteristic peak corresponding to the vibration channel is the peak value of the opening impact output by the vibration sensor. Based on the time offset between the moment when the contact just separates and the occurrence time of the opening impact peak output by the vibration sensor, the inherent delay compensation value of the vibration channel is obtained. Based on the inherent delay compensation value, the time series of the data in the corresponding acquisition channel is shifted and corrected on the overall time axis. Since the sampling frequencies of different physical quantities are quite different (vibration 20kHz, coil current 10kHz, displacement 1kHz), the number of data points and time intervals of each channel are inconsistent after alignment, making it impossible to perform point-by-point corresponding calculation. Therefore, the time series of the data in each acquisition channel after shift correction is mapped to the same standard time axis. The standard time axis uses the time interval of the highest sampling channel in the system (e.g., the 50μs interval of the vibration channel) as the reference scale. For low sampling rate channels, linear interpolation is used to fill in the data points: based on the values ​​and time intervals of two adjacent real sampling points, the approximate value at each standard scale point is calculated. Finally, all channels form time series of the same length and time, achieving point-by-point time domain alignment.

[0031] It should be further explained that, in the specific implementation process, the edge data correction unit first identifies channel-level persistent faults based on the cumulative number of out-of-limit sampling points, then identifies interfering outliers by the difference between the data before and after the sampling point, and finally, based on the time window defined by the global time zero point, combined with multi-channel temporal causal constraints, distinguishes between real faults and abnormal sampling points. The process of performing data anomaly correction includes: Set the physical boundary thresholds for each acquisition channel, the maximum allowable jump amplitude between adjacent sampling points, and the temporal causal constraints between each acquisition channel, and then perform the following steps, such as... Figure 2 As shown: Step a1: Compare the data of each acquisition channel with the corresponding physical boundary threshold. If a certain acquisition channel has sampling points that exceed the corresponding physical boundary threshold, extract the cumulative number of sampling points that exceed the corresponding physical boundary threshold for that acquisition channel. If the cumulative number is greater than the preset sampling point number threshold (100), generate a channel-level continuous fault alarm for that acquisition channel. If the cumulative number is less than or equal to the preset sampling point number threshold, proceed to step a2. Taking the ZW32 pole-mounted circuit breaker as an example, the physical boundary threshold is set as follows: Opening and closing coil current: 0~10A, moving contact travel displacement: 0~12mm (rated travel); Vibration acceleration: -500g to +500g; Terminal temperature: -40℃~150℃.

[0032] The determination in step a2 is based on the continuity of physical quantity change: any real physical quantity change is limited by the mechanical / electrical inertia of the equipment, and there is no possibility of instantaneous jump without transition. Therefore, the threshold of the maximum rate of change of adjacent sampling points is used to distinguish single-point outliers from real features.

[0033] Step a2: For the aforementioned sampling points that exceed the corresponding physical boundary threshold, calculate the data difference between the sampling point and the previous and next adjacent sampling points respectively. If any data difference is greater than the maximum allowable jump amplitude of the adjacent sampling points of the acquisition channel, the data corresponding to the sampling point is determined to be an interference field value. If each data difference is less than or equal to the maximum allowable jump amplitude of the adjacent sampling points of the acquisition channel, then step a3 is executed for the sampling point. The maximum allowable jump amplitude between adjacent sampling points is derived from existing physical mechanisms, for example: The current in the opening and closing coils is limited by the coil inductance, and the rate of current rise di / dt does not exceed 50A / ms. The maximum speed of the moving contact shall not exceed 2 m / s, and the adjacent jump between corresponding displacement sampling points shall not exceed 0.1 mm / 50 μs; Although the vibration signal is a high-frequency oscillation, the actual oscillation is a continuous alternation of positive and negative values. If the absolute value of a single point is much larger than that of the adjacent points and there is no continuous oscillation, it is an interference field value.

[0034] Step a3: Using the global time zero point of this opening and closing operation as the start timestamp of the preset time window (200ms), within the time window, obtain the sequence and time interval between other sampling points of other acquisition channels and this sampling point. Compare the obtained sequence and time interval with the temporal causal constraints between the acquisition channel to which this sampling point belongs and other acquisition channels. If the temporal causal constraints are met, that is, in a normal opening and closing operation, the changes of various physical parameters must follow a fixed sequence and a fixed time window, and time reversal, exceeding the window, and disordered abrupt changes are not allowed. If the occurrence order of the current channel data abrupt change completely conforms to the fixed sequence and the relative time delay before and after is within the fixed time window, then the preset time window to which this sampling point belongs is marked as a real fault state. The original data is retained within the preset time window without correction and directly participates in subsequent feature and degradation calculations to generate a sudden fault alarm. Otherwise, the sampling point is marked as an abnormal state. For example, the normal closing sequence is "coil current rises → contacts begin to move → initial closing impact vibration → main circuit conduction," and the time interval between each sampling point should be within a reasonable range (e.g., the time from coil energization to contact movement should be 10~30ms). If the sampling sequence of a channel is reversed or the interval exceeds the reasonable range, the data for that channel is considered abnormal. Step a4: Perform data anomaly correction on sampling points for channel-level persistent fault alarms, interference outliers, and abnormal states. Specifically: Persistent channel failure: The edge device generates a local alarm and uploads it to the cloud. At the same time, it suspends the data of that channel from participating in the component degradation calculation and replaces it with other redundant sensors of the same component until the sensor is repaired.

[0035] Interference outliers: Replace them with linear interpolation of two adjacent normal sampling points to preserve the integrity of the data sequence and not affect subsequent overall feature extraction; Sampling points for abnormal states: If the abnormal segment is located at the beginning or end of the opening and closing phases and does not cover core feature points such as the point of opening / closing and characteristic peaks, then the average template of the historical normal operation of the same channel will be used to fill it. If the abnormal segment covers core feature points such as the point of opening / closing or the characteristic peak, and the true features cannot be restored by interpolation, then all process data of this operation will be discarded, the log record will be retained, and the data will be collected again in the next opening / closing operation.

[0036] It should be further explained that, in the specific implementation process, the data processing unit extracts degradation features from multi-source heterogeneous data, performs independent state classification of degradation feature indicators for each component based on a dynamic three-level scale, and obtains the comprehensive degradation value and real-time classification status of each component. The process includes: A dynamic three-level scale is constructed, which includes the baseline value, degradation threshold and failure threshold corresponding to the degradation characteristic index of each component of several pole-mounted circuit breakers in the target area. Based on the degradation mechanism of pole-mounted circuit breakers, data retrieval is used to obtain degradation characteristic indicators of each component. The global time zero point is used as the start timestamp of the preset time window. Based on the degradation characteristic indicators of each component, feature extraction is performed on the data collected by all acquisition channels within the time window to obtain the feature values ​​corresponding to the degradation characteristic indicators of each component. The feature values ​​corresponding to the degradation characteristic indicators of each component are compared with the dynamic three-level scale to obtain the comprehensive degradation value of each component. Based on the comprehensive degradation value of each component, the real-time graded status of each component is obtained. Real-time component classification status includes normal degradation status, degradation status awaiting warning, and failure status.

[0037] The degradation characteristics of each component include: Operating mechanism characteristics: opening and closing time, opening and closing speed, number of vibration peaks, vibration energy distribution range, peak coil current and energizing time; Characteristics of conductive circuits: contact temperature rise rate, circuit resistance conversion value, and cumulative load current amplitude; Characteristics of the insulation system: number of partial discharges, discharge amplitude, and discharge phase distribution range; Environmental stress characteristics: daily average temperature difference, cumulative humidity duration, and salt spray concentration level.

[0038] The process of obtaining the comprehensive degradation value corresponding to the degradation characteristic index of each component and obtaining the real-time graded status of each component includes: Read the three-level thresholds corresponding to the degradation characteristic indicators of each component in the dynamic three-level scale: baseline value B, degradation threshold W, and failure threshold F. Based on the physical change characteristics of the characteristic quantities, divide the degradation characteristic indicators of each component into positive degradation characteristics and negative degradation characteristics. Use a fixed segmented mapping rule to calculate the degradation degree d in the 0-1 interval, where 0 corresponds to a brand new, undegraded state and 1 corresponds to a completely failed state.

[0039] Calculation process for positive degradation characteristics (the larger the value, the more severe the degradation): Positive characteristics include: vibration peak value, coil current peak value, temperature rise rate, circuit resistance, discharge quantity, etc.

[0040] If the eigenvalue corresponding to the positive degradation feature is less than or equal to the baseline value B, it is determined that there is no degradation and the degradation degree d=0. If the baseline value B < the characteristic value corresponding to the positive degradation feature ≤ the degradation threshold W, it is in the slow degradation range, and a low-speed range linear mapping is adopted: d = 0.5 × (measured value − B) / (W − B); If the degradation threshold W < the characteristic value corresponding to the positive degradation feature ≤ the failure threshold F, it is in the accelerated degradation range, and the high-speed range linear superposition mapping is adopted: d = 0.5 + 0.5 × (measured value − W) / (F − W); If the feature value corresponding to the positive degradation feature is greater than the failure threshold F, the function of that dimension is determined to be faulty, and the degradation degree d=1.

[0041] Negative degradation feature calculation process (the smaller the value, the more severe the degradation): Negative characteristics include: opening and closing speed, insulation equivalent parameters, and effective travel margin.

[0042] If the eigenvalue corresponding to the negative degradation feature is greater than or equal to the baseline value B, the degradation degree d = 0. If the degradation threshold W ≤ the feature value corresponding to the positive degradation feature < the baseline value B, the slow degradation interval is: d = 0.5 × (B − measured value) / (B − W); If the failure threshold F ≤ the characteristic value corresponding to the positive degradation feature < degradation threshold W, the accelerated degradation interval is: d = 0.5 + 0.5 × (W − measured value) / (W − F); If the eigenvalue corresponding to the positive degradation feature is less than the failure threshold F, the degradation degree d = 1.

[0043] Single component comprehensive degradation fusion calculation process: Retrieve the weight coefficients corresponding to the pre-set degradation characteristic indicators of each component. These weights are fixedly configured based on the component degradation mechanism, and the sum of all characteristic weights of each component is 1. Perform weighted fusion calculation: Overall degradation: In the formula, w i Let be the weight coefficient corresponding to the i-th degradation feature index, and di be the degradation degree corresponding to the i-th degradation feature index. Execution failure veto judgment: If the degree of deterioration corresponding to any deterioration characteristic index of the component is di=1, the overall deterioration degree of the component is directly and forcibly assigned to D=1; Based on the final calculated overall degradation level of 0 to 1, the automated status classification is determined and corresponding handling measures are implemented: When 0≤D<0.5, the component is judged to be in a normal deterioration state. The corresponding component is in a slow steady-state deterioration range with no accelerated deterioration trend, and the routine inspection strategy is maintained. When 0.5≤D<1, the component is determined to be in a state of deterioration awaiting warning. The key ledger on the cloud platform is uploaded, indicating that the component has entered the self-driven accelerated deterioration stage and its performance continues to degrade irreversibly. When D=1, the component is determined to be in a failure state, and a component failure alarm is immediately generated, marking the loss of core functions and requiring repair or replacement.

[0044] It should be further noted that, in the specific implementation process, the preset parameters and preset thresholds in this embodiment are set by those skilled in the art based on the actual situation or obtained by simulation of a large amount of data. These preset values ​​can be adjusted by those skilled in the art according to actual needs.

[0045] It should be further explained that, in the specific implementation process, the process of constructing a dynamic three-level benchmark includes: Obtain all historical time windows in which several pole-mounted circuit breakers in the target area are in a non-real fault state. Obtain the real-time graded status of each component within all historical time windows. Mark the historical time window in which the real-time graded status of each component is in a normal deterioration state as the reference time window. Select the average value of the deterioration characteristic index of each component within all reference time windows (obtained by summing and averaging the characteristic values ​​of the deterioration characteristic index of each component within all reference time windows) as the reference value of the deterioration characteristic index of each component. Obtain all historical time windows of several pole-mounted circuit breakers in the target area under actual fault conditions, as well as all historical time windows of several other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area under actual fault conditions. Mark the characteristic values ​​corresponding to the degradation characteristic indicators of each component within all historical time windows of several pole-mounted circuit breakers in the target area and several other pole-mounted circuit breakers of the same model as failure sample values. Compare the failure sample values ​​corresponding to the degradation characteristic indicators of the same component with the benchmark values ​​of the degradation characteristic indicators of the same component, and select the failure sample value closest to the benchmark value as the failure threshold. Wherein: for positive degradation characteristics where the larger the value, the higher the degree of degradation, the failure threshold value closest to the benchmark is the minimum value among all failure sample values; for negative degradation characteristics where the smaller the value, the higher the degree of degradation, the failure threshold value closest to the benchmark is the maximum value among all failure sample values.

[0046] It should be further explained that, in the specific implementation process, the construction of the dynamic three-level benchmark also includes: Obtain several other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area, remove the pole-mounted circuit breakers that have generated sudden fault alarms in the target area and the pole-mounted circuit breakers that have generated sudden fault alarms in other pole-mounted circuit breakers, and mark the remaining pole-mounted circuit breakers and other pole-mounted circuit breakers in the target area as sample pole-mounted circuit breakers. Obtain all historical time windows of the sample pole-mounted circuit breaker in a non-real fault state, as well as the actual cumulative commissioning time and cumulative number of opening and closing operations of the sample pole-mounted circuit breaker within each historical time window. For the characteristic values ​​corresponding to the degradation characteristic indicators of each component within each historical time window, bind the actual cumulative commissioning time and cumulative number of opening and closing operations of the sample pole-mounted circuit breaker within the corresponding historical time window to generate the gradient sample of the sample pole-mounted circuit breaker (characteristic values ​​corresponding to the degradation characteristic indicators of each component and the bound actual cumulative commissioning time and cumulative number of opening and closing operations). The actual cumulative commissioning time includes: 1-3 years, 3-6 years, and 6-10 years. For all sample pole-mounted circuit breakers of the same model, the gradient samples are sorted according to the actual cumulative commissioning time and cumulative number of opening and closing operations in each gradient sample to obtain a mixed time-series sample sequence. That is, the mixed time-series sample sequence is spliced ​​together by sorting all gradient samples of 1-3 years + 3-6 years + 6-10 years according to their equivalent life. The equivalent life of each gradient sample is equal to the cumulative commissioning time / the cumulative number of opening and closing operations. The equivalent life is the shortest at the leftmost end of the mixed time-series sequence, and the equivalent life gradually increases from left to right. The equivalent life is the longest at the rightmost end of the mixed time-series sample sequence. A sliding statistical window is set based on the cumulative number of opening and closing operations (the sliding statistical window contains N consecutive valid opening and closing operations, with N preset to 15). The mixed time-series sample sequence is traversed sequentially to obtain the average rate of change of the degradation characteristic indicators of each component within each sliding statistical window. Among them, the rate of change of positive degradation characteristics (the larger the value, the more severe the degradation, such as coil energizing time, contact temperature rise rate, and vibration energy) is the ratio of the increment of the characteristic value within the window to the increment of the cumulative number of operations corresponding to the window; the rate of change of negative degradation characteristics (the smaller the value, the more severe the degradation, such as insulation resistance and opening and closing speed) is the ratio of the decrease of the characteristic value within the window to the increment of the cumulative number of operations corresponding to the window. A rate sequence is constructed based on the average rate of change of the degradation characteristic indicators of each component within each sliding statistical window. Based on the average rate of change within each sliding statistical window corresponding to the gradient samples with an actual cumulative operating time of 1-3 years, a steady-state baseline rate for the degradation characteristic indicators of each component is set (obtained by averaging and summing the average rates of change within each sliding statistical window corresponding to the gradient samples with an actual cumulative operating time of 1-3 years). Based on the steady-state baseline rate, the rate sequence is traversed to find the position where the rate continues to rise. When there are no less than 3 consecutive adjacent sliding statistical windows, and the average rate of change of the degradation characteristic indicators of the components within the adjacent sliding statistical windows is greater than the preset multiple of the corresponding steady-state baseline rate of 2.0, and is in the range of continuous rate increase, then the first sliding statistical window of the 3 adjacent sliding statistical windows is determined to be the initial degradation inflection point. The conditions for the rate of continuous increase are as follows: taking the three adjacent sliding statistical windows as the core and extending two sliding statistical windows forward and backward to form a seven-window continuous interval, the average rate of change of the degradation characteristic index of all components in the seven-window continuous interval is higher than the corresponding steady-state reference rate.

[0047] It should be further explained that, in the specific implementation process, the construction of the dynamic three-level benchmark also includes: In the mixed time-series sample sequence corresponding to all sample pole-mounted circuit breakers of the same model, query the remaining cumulative number of opening and closing operations between the initial deterioration inflection point of the component's deterioration characteristic index and the corresponding position of the failure threshold of the component's deterioration characteristic index, and mark the remaining cumulative number of opening and closing operations as the manageable operation margin. A minimum maintenance handling capacity threshold is preset, which is 80 circuit breaker opening and closing operations. The following steps are then executed: Figure 3 As shown: Step b1: If the available operational margin is less than the minimum maintenance operational margin threshold, it means that there are not enough operation times left for maintenance after the inflection point is triggered. Then, obtain the offset step size based on the failure threshold and the baseline value of the component's degradation characteristic index. The offset step size is 2% of (failure threshold - baseline value). Based on the offset step size, the initial degradation inflection point is iteratively offset towards the baseline value round by round. After each round of offset, it is determined whether the initial degradation inflection point after the current offset is in the range of continuously increasing rate. If it is, continue to the next round of offset iteration until the available operational margin is greater than or equal to the minimum maintenance operational margin threshold. Mark the initial degradation inflection point after the current offset as the final degradation inflection point. If the initial degradation inflection point after the current offset is no longer in the range of continuously increasing rate, immediately cancel the current round of offset operation, revert to the position of the initial degradation inflection point in the previous round that was still in the range of continuously increasing rate, and mark the reverted initial degradation inflection point as the final degradation inflection point, and terminate the iterative offset. Step b2: If the manageable operational margin is greater than a preset multiple (2 times) of the minimum maintenance and handling margin threshold, it means that the physical inflection point is too early. Even after the equipment enters accelerated degradation, it still has a very long safe lifespan. Premature warnings will cause a large number of false alarms and occupy maintenance and handling resources. At this time, the degradation threshold is slightly shifted backward towards the failure threshold to compress the redundant warning interval and improve the accuracy of the threshold. Based on the offset step size, the initial degradation inflection point is iteratively shifted towards the failure direction in rounds. After each round of offset, it is determined whether the initial degradation inflection point after the current offset is in the range of continuously increasing rate. If it is, the next round of offset iteration continues until the manageable operational margin is less than or equal to a preset multiple of the minimum maintenance and handling margin threshold. The initial degradation inflection point after the current offset is marked as the final degradation inflection point. If the initial degradation inflection point after the current offset is no longer in the range of continuously increasing rate, the current round of offset operation is immediately canceled, and the initial degradation inflection point position in the previous round, which was still in the range of continuously increasing rate, is returned to the position. The initial degradation inflection point after the return is marked as the final degradation inflection point, and the iterative offset is terminated. Step b3: If the manageable operational margin is greater than or equal to the minimum maintenance and operation margin threshold and less than or equal to a preset multiple of the minimum maintenance and operation margin threshold, then the initial degradation inflection point is directly marked as the final degradation inflection point. The characteristic value of the degradation characteristic index of the component corresponding to the final degradation inflection point is taken as the degradation threshold of the degradation characteristic index of that component. A dynamic three-level scale is constructed based on the baseline values, degradation thresholds, and failure thresholds corresponding to the degradation characteristic indicators of each component of each type of pole-mounted circuit breaker.

[0048] It should be further explained that since it is impossible to track the entire 10-year lifespan of a single device in the field, this implementation method uses multiple sets of field-operated devices with different commissioning times to splice an equivalent full lifespan sequence to replace long-term tracking of a single device, without the need for laboratory accelerated aging prototypes.

[0049] It should be further explained that the initial reference value, degradation threshold and failure threshold of the dynamic three-level scale are set according to power industry standards and equipment factory parameters.

[0050] It should be further noted that, in the specific implementation process, the other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area all meet the following prerequisites: 1) The rated operating voltage is the same; 2) The load levels are similar, and the annual load fluctuation range does not differ by more than ±20%; 3) The environmental conditions are similar: the ambient temperature range, humidity, salt spray level, and outdoor installation conditions exposed to sun and rain belong to the same environmental zone; 4) The frequency of opening and closing operations is in the same range, so as to avoid mixing frequently operated equipment and rarely operated equipment into the same sample.

[0051] It should be further explained that, in the specific implementation process, the process of constructing the baseline degradation trajectory of each component based on the dynamic three-level scale, and fine-tuning the baseline degradation trajectory of each component based on the comprehensive degradation value of each component of several pole-mounted circuit breakers in the target area includes: Starting from the baseline value (deterioration degree = 0) of the degradation characteristic index of each component in the dynamic three-level scale, and ending at the failure threshold (deterioration degree = 1), the degradation threshold is a two-segment dividing inflection point. Based on gradient samples, the baseline degradation trajectory of each component is generated. The 0~deterioration threshold segment corresponds to the slow degradation stage, the degradation rate is approximately constant, and the trajectory shows a gentle linear increase. Deterioration threshold to failure threshold segment: This corresponds to the accelerated deterioration stage, where the deterioration rate increases continuously with the degree of deterioration, and the trajectory shows a downward convex acceleration curve. The location of the trajectory inflection point corresponds strictly to the degradation threshold of the dynamic three-level scale, ensuring that the physical mechanism of threshold calibration and trajectory deduction is consistent.

[0052] The horizontal axis of the baseline degradation trajectory represents the cumulative number of opening and closing operations, and the vertical axis represents the comprehensive degradation value of each component. After the three key nodes are determined, the trajectory is divided into a slow degradation segment and an accelerated degradation segment. Continuous curves are fitted based on gradient sample data to obtain the values ​​of all intermediate points on the trajectory. (1) Slow deterioration stage: Extract all gradient samples from the gradient samples for 1-3 years and 3-6 years of operation to obtain multiple sets of discrete data points (cumulative number of opening and closing operations - corresponding comprehensive deterioration value); Using the overall degradation degree as the dependent variable and the cumulative number of opening and closing operations as the independent variable, a linear fit was performed using the least squares method to obtain the linear equation for the slowly deteriorating section: Where k1 is the steady-state degradation rate, i.e. the degradation increment corresponding to a unit number of operations, D represents the comprehensive degradation value, and N represents the cumulative number of opening and closing operations; Based on this linear equation, the degradation degree corresponding to any cumulative number of operations is calculated, that is, the coordinates of any point on the slow degradation segment.

[0053] (2) Accelerated deterioration stage: Extract all gradient samples that have been in operation for 6-10 years from the gradient samples to obtain multiple sets of discrete data points (cumulative number of opening and closing operations - corresponding comprehensive deterioration value), and the data range covers the complete range from the deterioration threshold to the failure threshold. By fitting the discrete points with a quadratic polynomial, the curve equation of the accelerated degradation segment is obtained. Based on this curve equation, the degradation degree corresponding to any cumulative number of operations within the accelerated segment is calculated, which is the coordinate of any point on the accelerated degradation segment.

[0054] The comprehensive degradation value of each component of several pole-mounted circuit breakers within the target area is compared with the baseline degradation trajectory of each component to obtain the relative deviation rate of each component. In the formula, D represents the relative deviation rate. real D represents the overall degradation value. theory This represents the theoretical comprehensive degradation value on the baseline degradation trajectory. A relative deviation rate threshold (5%) is set. If the relative deviation rate of a component is less than or equal to the threshold, the baseline degradation trajectory remains unchanged. If the relative deviation rate is greater than the threshold, the degradation rate of the baseline degradation trajectory is adjusted based on the relative deviation rate to obtain the adjusted baseline degradation trajectory. The degradation rate is the increment of the comprehensive degradation value corresponding to each opening and closing operation. The adjustment coefficient is determined based on the relative deviation rate. or Specifically: If the overall degradation value > the theoretical overall degradation value (actual degradation is faster): the degradation rate is amplified by an adjustment factor of equal magnitude. The adjustment factor is... This ensures that the rate of degradation of the subsequent trajectory matches the measured trend. If the overall degradation value is less than the theoretical overall degradation value (actual degradation is slower): the degradation rate will be reduced by an equal amount using an adjustment factor. The adjustment factor is... .

[0055] It should be further explained that, in the specific implementation process, the process of performing life analysis based on the fine-tuned baseline degradation trajectory to obtain the degradation threshold of the pole-mounted circuit breaker reaching its lifespan or failure life includes: If all components of the pole-mounted circuit breaker are in a normal deterioration state, then based on the current cumulative number of opening and closing operations of the pole-mounted circuit breaker and the baseline deterioration trajectory of each component, the deterioration threshold of each component is obtained to reach the remaining number of opening and closing operations. ,in, D1 represents the remaining number of opening and closing operations when the degradation threshold is reached, D2 represents the current overall degradation degree, and Vn represents the degradation rate at the corresponding position of the current cumulative number of opening and closing operations on the baseline degradation trajectory. From the degradation thresholds of each component reaching the remaining number of opening and closing operations, the minimum degradation threshold reaching the remaining number of opening and closing operations is selected as the degradation threshold reaching the life of the pole-mounted circuit breaker. If a component of a pole-mounted circuit breaker is in a state of pending warning deterioration, the remaining number of opening and closing operations for each component in this state is obtained based on the current cumulative number of opening and closing operations and the baseline deterioration trajectory of each component. Specifically, along the baseline deterioration trajectory, the remaining number of opening and closing operations is calculated point by point from the position corresponding to the current cumulative number of opening and closing operations towards the failure threshold until the deterioration degree reaches D=1. The cumulative number of opening and closing operations corresponding to this trajectory segment is counted as the remaining number of opening and closing operations. The minimum remaining number of opening and closing operations is selected from the remaining number of opening and closing operations of each component in this state as the failure life of the pole-mounted circuit breaker.

[0056] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A pole-mounted circuit breaker life prediction system based on multi-sensor fusion, characterized in that, It includes a cloud monitoring platform, a multi-sensor sensing unit, an edge data temporal domain alignment unit, an edge data correction unit, a data processing unit, and a lifetime analysis module; The multi-sensor sensing unit is used to collect multi-source heterogeneous data corresponding to each opening and closing operation of the pole-mounted circuit breaker; The edge data temporal alignment unit is used to detect the actual occurrence time of the corresponding characteristic peak in multi-source heterogeneous data. It obtains the inherent delay compensation value based on the time offset of each acquisition channel relative to the reference benchmark, and uses the inherent delay compensation value to complete the temporal alignment of multi-source heterogeneous data. The edge data correction unit is used to first identify channel-level persistent faults based on the cumulative number of out-of-limit sampling points, then identify interference wild values ​​by the difference between the data before and after the sampling point, and finally, based on the time window defined by the global time zero point, combined with multi-channel time-series causal constraints, distinguish between real faults and abnormal sampling points, and perform data anomaly correction. The data processing unit is used to construct a dynamic three-level scale, extract degradation features from multi-source heterogeneous data, classify the degradation feature indicators of each component independently based on the dynamic three-level scale, and obtain the real-time classification status of each component. The life analysis module is used to construct the baseline degradation trajectory of each component based on a dynamic three-level scale. Based on the comprehensive degradation value of each component of several pole-mounted circuit breakers in the target area, the baseline degradation trajectory of each component is fine-tuned. Based on the fine-tuned baseline degradation trajectory, life analysis is performed to obtain the degradation threshold of the pole-mounted circuit breaker to reach its lifespan or failure life.

2. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 1, characterized in that, Multi-sensor sensing units, edge data temporal alignment units, and edge data correction units are pre-deployed on several pole-mounted circuit breakers within the target area, and data processing units and life analysis modules are built within the cloud monitoring platform; The multi-sensor sensing unit, edge data time domain alignment unit, and edge data correction unit on each pole-mounted circuit breaker are connected to the cloud monitoring platform via IoT nodes. The rising edge of the opening and closing coil driving voltage is used as the global time zero point of the multi-sensor sensing unit. Whenever the opening and closing coil circuit has a rising edge of driving voltage, the rising edge of driving voltage is marked as the global time zero point of this opening and closing operation. When the multi-sensor sensing unit detects the global time zero point, all acquisition channels are started simultaneously through the same clock pulse to acquire multi-source heterogeneous data.

3. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 2, characterized in that, The edge data temporal alignment unit is used to detect the actual occurrence time of corresponding feature peaks in multi-source heterogeneous data. Based on the time offset of each acquisition channel relative to the reference benchmark, an inherent delay compensation value is obtained. The process of using this inherent delay compensation value to complete the temporal alignment of multi-source heterogeneous data includes: The moment when the driving voltage rises in the tripping coil circuit is marked as the moment when the contacts just open; the moment when the driving voltage rises in the closing coil circuit is marked as the moment when the contacts just close; and the moments when the contacts just open and close are marked as reference points. The characteristic peaks corresponding to the moment when the contacts just separate and the moment when the contacts just close are set as the reference references; Real-time feature peak detection is performed on the multi-source heterogeneous data collected by the multi-sensor sensing unit to obtain the occurrence time of the feature peak corresponding to the reference benchmark. Based on the time offset between the reference benchmark and the occurrence time of the corresponding feature peak, the inherent delay compensation value of each acquisition channel in the multi-sensor sensing unit is obtained. Based on the inherent delay compensation value, the time series of the data in the corresponding acquisition channel is shifted and corrected as a whole on the time axis. The time series of the data in each acquisition channel after shift correction is mapped to the same standard time axis.

4. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 3, characterized in that, The edge data correction unit first identifies channel-level persistent faults based on the cumulative number of out-of-limit sampling points, then identifies interfering outliers by the difference between the data before and after the sampling point, and finally, based on the time window defined by the global time zero point and combined with multi-channel temporal causal constraints, distinguishes between real faults and abnormal sampling points. The process of performing data anomaly correction includes: Set the physical boundary thresholds for each acquisition channel, the maximum allowable jump amplitude between adjacent sampling points, and the temporal causal constraints between each acquisition channel, and then perform the following steps: Step a1: Compare the data of each acquisition channel with the corresponding physical boundary threshold. If a certain acquisition channel has sampling points that exceed the corresponding physical boundary threshold, extract the cumulative number of sampling points that exceed the corresponding physical boundary threshold for that acquisition channel. If the cumulative number is greater than the preset sampling point number threshold, generate a channel-level continuous fault alarm for that acquisition channel. If the cumulative number is less than or equal to the preset sampling point number threshold, proceed to step a2. Step a2: For the aforementioned sampling points that exceed the corresponding physical boundary threshold, calculate the data difference between the sampling point and the previous and next adjacent sampling points respectively. If any data difference is greater than the maximum allowable jump amplitude of the adjacent sampling points of the acquisition channel, the data corresponding to the sampling point is determined to be an interference field value. If each data difference is less than or equal to the maximum allowable jump amplitude of the adjacent sampling points of the acquisition channel, then step a3 is executed for the sampling point. Step a3: Take the global time zero point of this opening and closing operation as the start timestamp of the preset time window. Within the time window, obtain the sequence and time interval between other sampling points of other acquisition channels and this sampling point. Compare the obtained sequence and time interval with the temporal causal constraints between the acquisition channel to which this sampling point belongs and other acquisition channels. If the temporal causal constraints are satisfied, mark the preset time window to which this sampling point belongs as a real fault state and generate a sudden fault alarm. Otherwise, mark this sampling point as an abnormal state. Step a4: Correct data anomalies for sampling points of channel-level persistent fault alarms, interference outliers, and abnormal states.

5. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 4, characterized in that, The data processing unit extracts degradation features from multi-source heterogeneous data, performs independent state classification of degradation feature indicators for each component based on a dynamic three-level scale, and obtains the comprehensive degradation value and real-time classification status of each component. The process includes: A dynamic three-level scale is constructed, which includes the baseline value, degradation threshold and failure threshold corresponding to the degradation characteristic index of each component of several pole-mounted circuit breakers in the target area. Based on the degradation mechanism of pole-mounted circuit breakers, data retrieval is used to obtain degradation characteristic indicators of each component. Based on the degradation characteristic indicators of each component, feature extraction is performed on the data collected by all acquisition channels within the time window to obtain the characteristic values ​​corresponding to the degradation characteristic indicators of each component. The characteristic values ​​corresponding to the degradation characteristic indicators of each component are compared with the dynamic three-level scale to obtain the comprehensive degradation degree value of each component. Based on the comprehensive degradation degree value of each component, the real-time classification status of each component is obtained. Real-time component classification status includes normal degradation status, degradation status awaiting warning, and failure status.

6. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 5, characterized in that, The process of constructing a dynamic three-level scale includes: Obtain all historical time windows in which several pole-mounted circuit breakers in the target area are in a non-real fault state. Obtain the real-time graded status of each component within all historical time windows. Mark the historical time window in which the real-time graded status of each component is in a normal deterioration state as the reference time window. Select the average value of the deterioration characteristic index of each component within all reference time windows as the reference value of the deterioration characteristic index of each component. Obtain all historical time windows of several pole-mounted circuit breakers in the target area under actual fault conditions, as well as all historical time windows of several other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area under actual fault conditions. Mark the characteristic values ​​corresponding to the degradation characteristic indicators of each component in the target area and the several other pole-mounted circuit breakers of the same model as failure sample values ​​within all historical time windows. Compare the failure sample values ​​corresponding to the degradation characteristic indicators of the same component with the benchmark values ​​of the degradation characteristic indicators of the same component, and select the failure sample value closest to the benchmark value as the failure threshold.

7. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 6, characterized in that, The process of constructing a dynamic three-level scale also includes: Obtain several other pole-mounted circuit breakers of the same model as each pole-mounted circuit breaker in the target area, remove the pole-mounted circuit breakers that have generated sudden fault alarms in the target area and the pole-mounted circuit breakers that have generated sudden fault alarms in other pole-mounted circuit breakers, and mark the remaining pole-mounted circuit breakers and other pole-mounted circuit breakers in the target area as sample pole-mounted circuit breakers. Obtain all historical time windows of the sample pole-mounted circuit breaker in non-real fault state, as well as the actual cumulative commissioning time and cumulative number of opening and closing operations of the sample pole-mounted circuit breaker within each historical time window. For the characteristic values ​​corresponding to the degradation characteristic indicators of each component within each historical time window, bind the actual cumulative commissioning time and cumulative number of opening and closing operations of the sample pole-mounted circuit breaker within the corresponding historical time window to generate the gradient sample of the sample pole-mounted circuit breaker. The actual cumulative commissioning time includes: 1-3 years, 3-6 years and 6-10 years. For all sample pole-mounted circuit breakers of the same model, the gradient samples are sorted according to the actual cumulative commissioning time and the cumulative number of opening and closing operations in each gradient sample to obtain a mixed time sequence sample sequence. Based on the cumulative number of opening and closing operations, a sliding statistical window is set, and the mixed time series sample sequence is traversed sequentially to obtain the average rate of change of the degradation characteristic index of each component within each sliding statistical window. A rate sequence is constructed based on the average rate of change of the degradation characteristic index of each component within each sliding statistical window. Based on the average rate of change within each sliding statistical window corresponding to the gradient samples with an actual cumulative operation time of 1-3 years, a steady-state baseline rate for the degradation characteristic index of each component is set. Based on the steady-state baseline rate, the rate sequence is traversed. When there are no less than 3 consecutive adjacent sliding statistical windows, and the average rate of change of the degradation characteristic index of the component within the adjacent sliding statistical windows is greater than the preset multiple of the corresponding steady-state baseline rate and is in the range of continuous rate increase, the first sliding statistical window of the 3 adjacent sliding statistical windows is determined to be the initial degradation inflection point. The conditions for the rate of continuous increase are as follows: taking the three adjacent sliding statistical windows as the core and extending two sliding statistical windows forward and backward to form a seven-window continuous interval, the average rate of change of the degradation characteristic index of all components in the seven-window continuous interval is higher than the corresponding steady-state reference rate.

8. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 7, characterized in that, The process of constructing a dynamic three-level scale also includes: In the mixed time-series sample sequence corresponding to all sample pole-mounted circuit breakers of the same model, query the remaining cumulative number of opening and closing operations between the initial deterioration inflection point of the component's deterioration characteristic index and the corresponding position of the failure threshold of the component's deterioration characteristic index, and mark the remaining cumulative number of opening and closing operations as the manageable operation margin. To preset the minimum maintenance and handling capacity threshold, perform the following steps: Step b1: If the manageable operational margin is less than the minimum maintenance and handling margin threshold, then obtain the offset step size based on the failure threshold and baseline value of the component's degradation characteristic index. Based on the offset step size, iterate the initial degradation inflection point towards the baseline value round by round. After each round of offset, determine whether the initial degradation inflection point after the current offset is in the range of continuously increasing rate. If it is, continue to the next round of offset iteration until the manageable operational margin is greater than or equal to the minimum maintenance and handling margin threshold. Mark the initial degradation inflection point after the current offset as the final degradation inflection point. If the initial degradation inflection point after the current offset is no longer in the range of continuously increasing rate, immediately cancel the current round of offset operation, revert to the position of the initial degradation inflection point in the previous round that was still in the range of continuously increasing rate, and mark the reverted initial degradation inflection point as the final degradation inflection point, and terminate the iterative offset. Step b2: If the manageable operational margin is greater than a preset multiple of the minimum maintenance and handling margin threshold, the initial degradation inflection point is iteratively shifted towards the failure direction in rounds based on the offset step size. After each round of offset, it is determined whether the initial degradation inflection point after the current offset is in the range of continuously increasing rate. If it is, the next round of offset iteration continues until the manageable operational margin is less than or equal to a preset multiple of the minimum maintenance and handling margin threshold. The initial degradation inflection point after the current offset is marked as the final degradation inflection point. If the initial degradation inflection point after the current offset is no longer in the range of continuously increasing rate, the current round of offset operation is immediately canceled, and the position of the initial degradation inflection point in the previous round, which was still in the range of continuously increasing rate, is returned. The initial degradation inflection point after the return is marked as the final degradation inflection point, and the iterative offset is terminated. Step b3: If the manageable operational margin is greater than or equal to the minimum maintenance and operation margin threshold and less than or equal to a preset multiple of the minimum maintenance and operation margin threshold, then the initial degradation inflection point is directly marked as the final degradation inflection point. The characteristic value of the degradation characteristic index of the component corresponding to the final degradation inflection point is taken as the degradation threshold of the degradation characteristic index of that component. A dynamic three-level scale is constructed based on the baseline values, degradation thresholds, and failure thresholds corresponding to the degradation characteristic indicators of each component of each type of pole-mounted circuit breaker.

9. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 8, characterized in that, The process of constructing a baseline degradation trajectory for each component based on a dynamic three-level scale, and fine-tuning the baseline degradation trajectory for each component based on the comprehensive degradation value of each component of several pole-mounted circuit breakers within the target area, includes: Starting from the baseline value of the degradation characteristic index of each component in the dynamic three-level scale, with the failure threshold as the end point and the degradation threshold as the two-segment dividing inflection point, the baseline degradation trajectory of each component is generated based on gradient samples. The horizontal axis of the baseline degradation trajectory is the cumulative number of opening and closing operations, and the vertical axis is the comprehensive degradation value of each component. The comprehensive degradation value of each component of several pole-mounted circuit breakers in the target area is compared with the baseline degradation trajectory of each component to obtain the relative deviation rate of each component. A relative deviation rate threshold is set. If the relative deviation rate of a component is less than or equal to the relative deviation rate threshold, the baseline degradation trajectory of the component remains unchanged. If the relative deviation rate of a component is greater than the relative deviation rate threshold, the degradation rate of the baseline degradation trajectory is adjusted according to the relative deviation rate to obtain the adjusted baseline degradation trajectory.

10. The pole-mounted circuit breaker life prediction system based on multi-sensor fusion according to claim 9, characterized in that, Based on the fine-tuned baseline degradation trajectory, the process of obtaining the degradation threshold of the pole-mounted circuit breaker to reach its lifespan or failure life includes: If all components of the pole-mounted circuit breaker are in a normal deterioration state, then based on the current cumulative number of opening and closing operations of the pole-mounted circuit breaker and the baseline deterioration trajectory of each component, the deterioration threshold of each component is obtained to reach the remaining number of opening and closing operations. From the deterioration thresholds of each component to reach the remaining number of opening and closing operations, the minimum deterioration threshold to reach the remaining number of opening and closing operations is selected as the deterioration threshold of the pole-mounted circuit breaker to reach its lifespan. If a component of a pole-mounted circuit breaker is in a state of pending warning deterioration, the remaining number of opening and closing operations for each component in the pending warning deterioration state is obtained based on the current cumulative number of opening and closing operations of the pole-mounted circuit breaker and the baseline deterioration trajectory of each component in the pending warning deterioration state. From the remaining number of opening and closing operations of each component in the pending warning deterioration state, the minimum number of remaining opening and closing operations is selected as the failure life of the pole-mounted circuit breaker.

Images