High-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method and system

Abstract

This invention provides a method and system for dynamic modeling and processing of high-voltage vacuum circuit breaker operation and maintenance data, relating to the field of data processing technology. The method involves event-level identification and feature extraction of each opening and closing operation data, constructing an operation event feature vector sequence, and classifying it into mechanism action mutation features and mechanical cumulative deterioration features based on the correspondence between features and physical states. Mutation retention weight sequences and cumulative deterioration weight sequences are generated respectively, and the feature vector sequence is recursively updated to form a dynamic feature set. Based on this, a dynamic reference model with directional expansion and unidirectional offset characteristics is constructed, and operation offset trajectory data is generated through the positional relationship between new operation events and the model, thereby achieving continuous modeling and expression of the circuit breaker's operating state change process.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for dynamic modeling and processing of high-voltage vacuum circuit breaker operation and maintenance data. Background Technology

[0002] Current operation and maintenance data processing for high-voltage vacuum circuit breakers largely relies on static data acquisition and analysis methods based on SCADA systems and online monitoring devices. This involves periodically recording key parameters such as opening and closing times, contact wear, and current and voltage waveforms, and using threshold judgments or empirical models for condition assessment. Generally, centralized database storage and offline analysis are used for equipment health diagnosis, while some systems combine simple trend analysis or expert rule bases for fault early warning.

[0003] In a scenario where a 10kV high-voltage vacuum circuit breaker in a substation is undergoing long-term operational monitoring, this type of method typically summarizes the opening and closing time data daily or weekly and performs averaging. If, during a certain operation, the opening time suddenly increases from the normal 30ms to 55ms due to mechanical jamming, but is diluted to 32ms in the periodic averaging, the preset alarm threshold of 35ms cannot be triggered, resulting in the potential mechanical fault not being identified in time, which may lead to the risk of failure to operate in subsequent frequent operations. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for dynamic modeling and processing of high-voltage vacuum circuit breaker operation and maintenance data, aiming to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for dynamic modeling and processing of high-voltage vacuum circuit breaker operation and maintenance data, the method comprising: Acquire the opening and closing time, current waveform, and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker, and time-mark the opening and closing time, current waveform, and contact travel data according to a single operation event to generate the original operation event data unit; Event-level feature extraction is performed on the original operation event data unit to extract the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, and generate an operation event feature vector sequence. Based on the correspondence between each feature in the operation event feature vector sequence and the physical state of the high-voltage vacuum circuit breaker, the operation event feature vector sequence is divided into feature attributes to generate a mechanism action mutation feature set and a mechanical cumulative deterioration feature set. The preset mechanism action mutation feature set includes the opening and closing time feature value and the coil current waveform peak value, while the mechanical cumulative deterioration feature set includes the contact stroke change feature. Based on the mechanism action mutation feature set, the feature offset between adjacent operation events in the operation event feature vector sequence is calculated, and a mutation preservation weight sequence that changes with operation events is generated based on the feature offset; based on the mechanical cumulative deterioration feature set, the feature change amount generated by the operation event feature vector sequence with the increase of operation number is cumulatively calculated, and the reverse fall amplitude of the feature change amount is limited, generating a cumulative deterioration weight sequence that changes with operation events. Based on the mutation preservation weight sequence and the cumulative degradation weight sequence, the operational event feature vector sequence is dynamically updated by successive recursion to generate a dynamic feature set containing mutation path information and cumulative evolution information. The interval boundary is calculated based on the dynamic feature set, and the boundary is extended by direction correlation based on the mutation-preserving weight sequence. The boundary is offset by unidirectional offset based on the cumulative deterioration weight sequence, generating a dynamic reference model with asymmetric and irreversible characteristics. Based on the relative positional relationship between the feature vector of the new operation event and the dynamic reference model, operation offset trajectory data that evolves with the operation event is calculated and generated.

[0006] Secondly, a dynamic modeling and processing system for high-voltage vacuum circuit breaker operation and maintenance data, the system comprising: The data acquisition module is used to acquire the opening and closing time, current waveform and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker, and to time-mark the opening and closing time, current waveform and contact travel data according to a single operation event to generate the original operation event data unit. The feature extraction module is used to perform event-level feature extraction on the original operation event data units, extracting the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, and generating an operation event feature vector sequence. The feature segmentation module is used to segment the feature vector sequence of operation events according to the correspondence between each feature in the operation event feature vector sequence and the physical state of the high-voltage vacuum circuit breaker, and generate a mechanism action sudden change feature set and a mechanical cumulative deterioration feature set. The preset mechanism action sudden change feature set includes the opening and closing time feature value and the coil current waveform peak value, and the mechanical cumulative deterioration feature set includes the contact stroke change feature. The weight generation module is used to calculate the feature offset between adjacent operation events in the operation event feature vector sequence based on the mechanism action mutation feature set, and generate a mutation-preserving weight sequence that changes with operation events based on the feature offset; based on the mechanical cumulative deterioration feature set, it accumulates and calculates the feature change in the operation event feature vector sequence that increases with the number of operations, limits the reverse fall of the feature change, and generates a cumulative deterioration weight sequence that changes with operation events. The dynamic modeling module is used to perform a dynamic weighted update of the operational event feature vector sequence by recursively updating it based on the mutation retention weight sequence and the cumulative degradation weight sequence, thereby generating a dynamic feature set containing mutation path information and cumulative evolution information. The model building module is used to calculate the interval boundary based on the dynamic feature set, perform directional correlation expansion processing on the boundary based on the mutation-preserving weight sequence, and perform unidirectional offset processing on the boundary based on the cumulative deterioration weight sequence to generate a dynamic reference model with asymmetric and irreversible characteristics. The trajectory generation module is used to calculate and generate operation offset trajectory data that evolves with the operation event based on the relative positional relationship between the operation event feature vector corresponding to the new operation event and the dynamic reference model.

[0007] The above-described solution of the present invention has at least the following beneficial effects: First, by time-identifying and organizing the data of each opening and closing operation of the high-voltage vacuum circuit breaker according to the single operation event, the operation and maintenance data is transformed from a periodic summary to a sequence of operations. This avoids the dilution of single abnormal data by periodic averaging and ensures that abnormal operations can be fully preserved and used in subsequent analysis.

[0008] Furthermore, by extracting features from the operational event data and dividing them into mechanism action mutation features and mechanical cumulative deterioration features, features of different physical attributes are used to participate in the modeling process. The mutation features are used to characterize the abnormal changes in a single operation, while the cumulative features are used to characterize the long-term wear process. This allows the operation and maintenance data analysis results to reflect both instantaneous anomalies and long-term changes.

[0009] Furthermore, by constructing mutation preservation weight sequences and cumulative degradation weight sequences respectively, the mutation features maintain their influence path in subsequent operations, and the cumulative features continuously enhance their influence as the number of operations increases, thus forming a dynamic feature expression that includes both short-term abnormal effects and long-term degradation trends.

[0010] Furthermore, by recursively updating the feature vector sequence of operation events, a dynamic feature set is constructed, enabling the device operating status to continuously evolve and be expressed as the operation process progresses, thus avoiding the information lag problem caused by relying solely on static thresholds or offline statistical analysis.

[0011] Furthermore, by constructing a dynamic reference model with directional correlation extension characteristics and unidirectional offset characteristics, the offset caused by mutation and the offset caused by degradation can be distinguished in the interval structure, thereby realizing a structured expression of different types of state changes.

[0012] Finally, by mapping new operation events to a dynamic reference model and generating operation offset trajectory data, changes in the device's operating state can be described in the form of continuous trajectories. This allows for timely identification of potential risks caused by a single abnormal operation and performance change trends caused by long-term wear during actual operation. Attached Figure Description

[0013] Figure 1 This is a flowchart of the dynamic modeling and processing method for high-voltage vacuum circuit breaker operation and maintenance data provided in an embodiment of the present invention. Detailed Implementation

[0014] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0015] like Figure 1 As shown, embodiments of the present invention propose a dynamic modeling and processing method for high-voltage vacuum circuit breaker operation and maintenance data, the method comprising: Acquire the opening and closing time, current waveform, and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker, and time-mark the opening and closing time, current waveform, and contact travel data according to a single operation event to generate the original operation event data unit; Event-level feature extraction is performed on the original operation event data unit to extract the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, and generate an operation event feature vector sequence. Based on the correspondence between each feature in the operation event feature vector sequence and the physical state of the high-voltage vacuum circuit breaker, the operation event feature vector sequence is divided into feature attributes to generate a mechanism action mutation feature set and a mechanical cumulative deterioration feature set. The preset mechanism action mutation feature set includes the opening and closing time feature value and the coil current waveform peak value, while the mechanical cumulative deterioration feature set includes the contact stroke change feature. Based on the mechanism action mutation feature set, the feature offset between adjacent operation events in the operation event feature vector sequence is calculated, and a mutation preservation weight sequence that changes with operation events is generated based on the feature offset; based on the mechanical cumulative deterioration feature set, the feature change amount generated by the operation event feature vector sequence with the increase of operation number is cumulatively calculated, and the reverse fall amplitude of the feature change amount is limited, generating a cumulative deterioration weight sequence that changes with operation events. Based on the mutation preservation weight sequence and the cumulative degradation weight sequence, the operational event feature vector sequence is dynamically updated by successive recursion to generate a dynamic feature set containing mutation path information and cumulative evolution information. The interval boundary is calculated based on the dynamic feature set, and the boundary is extended by direction correlation based on the mutation-preserving weight sequence. The boundary is offset by unidirectional offset based on the cumulative deterioration weight sequence, generating a dynamic reference model with asymmetric and irreversible characteristics. Based on the relative positional relationship between the feature vector of the new operation event and the dynamic reference model, operation offset trajectory data that evolves with the operation event is calculated and generated.

[0016] In this embodiment of the invention, by identifying the opening and closing time, current waveform, and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker at the single operation event level, the operation and maintenance data is transformed from a periodic statistical form into data units organized according to the operation process. This ensures that each operation is treated as an independent object in subsequent analysis, and avoids the uniform processing of multiple operation data during the statistical process, which would mask individual differences.

[0017] By extracting event-level features from the raw operational event data units, the characteristic values ​​of opening and closing time, peak values ​​of coil current waveform, and contact stroke variation are uniformly expressed as a sequence of operational event feature vectors. This allows different types of operation and maintenance data to form a consistent data representation, providing a foundation for subsequent multi-dimensional correlation analysis. Simultaneously, each feature is correlated with the physical state of the circuit breaker, categorizing them into abrupt changes in mechanical action and cumulative mechanical degradation features. This ensures that different types of features can participate in modeling processing according to their physical change patterns.

[0018] By calculating the feature offset between adjacent operation events for the abrupt change characteristics of the mechanism's actions and generating a change preservation weight sequence that changes with the operation events, the feature offset corresponding to a single abnormal operation is continuously retained in multiple subsequent operation processes, thus forming a change impact path. By calculating the cumulative change of mechanical degradation characteristics with the increase of the number of operations and limiting the reverse change amplitude, a cumulative degradation weight sequence is generated, so that cumulative changes such as contact wear show a continuous accumulation and non-completely reversible change trend in the modeling process, thereby reflecting the irreversible characteristics of the mechanical degradation process.

[0019] In the dynamic modeling process, the feature vector sequence of operation events is updated successively according to the mutation retention weight sequence and the cumulative degradation weight sequence, so that different types of features form differentiated evolution paths in the time dimension. Among them, mutation features maintain their continuous influence in multiple operations, and the weight ratio of cumulative features gradually increases with the number of operations. Thus, a dynamic feature set that contains both mutation path information and cumulative evolution information is constructed, so that operation and maintenance data can reflect the continuous evolution relationship of equipment status changes with the operation process.

[0020] During the model construction process, the interval boundaries are calculated based on the dynamic feature set, and the boundaries corresponding to the mutation features are subjected to directional correlation expansion processing, while the boundaries corresponding to the cumulative features are subjected to unidirectional offset processing. This results in the interval structure having directional asymmetry characteristics and irreversible characteristics of changing along the deterioration direction, thereby constructing a dynamic reference model that can simultaneously reflect instantaneous anomalies and long-term deterioration, enabling the equipment operating status to form a multi-dimensional interval expression with constraints in the spatial distribution.

[0021] During the operation status characterization process, the operation event feature vector corresponding to the new operation event is mapped to the dynamic reference model to obtain the offset trajectory data that changes with the operation process, so that the equipment status change can be expressed through continuous trajectory, thereby realizing the continuous description of the operation status change process.

[0022] In a preferred embodiment of the present invention, the opening and closing time, current waveform, and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker are acquired, and the opening and closing time, current waveform, and contact travel data are time-identified according to a single operation event to generate an original operation event data unit, including: The trip coil current signal, closing coil current signal, and contact displacement signal are synchronously collected by an online monitoring device. The start and end times of the circuit breaker tripping are determined based on the initial rise point and the stable segment of the current signal. The tripping time is calculated, for example, the tripping time is determined by the time interval from the start point to the current falling back to the steady state point. The start and end positions of the contact movement are extracted based on the change range of the contact displacement signal to form the corresponding stroke data; Each opening or closing action is treated as an independent operation event, and numbered and identified according to the timestamp to form an operation event identification sequence. The opening and closing times, current waveform data sequences, and contact travel data in the same operation event are uniformly encapsulated to generate the original operation event data unit.

[0023] In a preferred embodiment of the present invention, event-level feature extraction is performed on the original operation event data unit to extract the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, generating an operation event feature vector sequence, including: The opening and closing times in each original operation event data unit are directly extracted as time feature values. Peak detection is performed on the current waveform, and the maximum current value is extracted as the peak feature of the waveform; The displacement change is calculated from the contact stroke data, and the stroke change characteristics are obtained by using the difference between the starting and ending positions. The opening and closing time characteristics, current waveform peak characteristics, and contact stroke change characteristics are combined in a fixed order to form a single operation event feature vector. The feature vectors of each operation event are arranged sequentially according to the time order of the operation events to generate a sequence of operation event feature vectors.

[0024] In a preferred embodiment of the present invention, based on the relative positional relationship between the operation event feature vector corresponding to the new operation event and the dynamic reference model, operation offset trajectory data evolving with the operation event is calculated and generated, including: Obtain the feature vector of the operation event corresponding to the new operation event, and extract the values ​​of each feature component; Read the interval boundaries of the corresponding feature components from the dynamic reference model, including the upper and lower bounds of each feature component; Each feature component in the feature vector of the operation event is compared with the corresponding interval boundary one by one to determine the offset direction and offset distance of each feature component relative to the interval. The offset distance of each feature component is normalized to obtain the offset value of each feature component. According to the preset weight combination rules, the offset values ​​of each feature component are weighted and combined to generate a single operation offset vector. According to the time sequence of operation events, multiple consecutive single operation offset vectors are arranged to generate an operation offset trajectory sequence, which is used to characterize the state change path during the operation process.

[0025] In a preferred embodiment of the present invention, based on the mechanism action mutation feature set, the feature offset between adjacent operation events in the operation event feature vector sequence is calculated, and a mutation-preserving weight sequence that changes with the operation events is generated based on the feature offset, including: Based on the feature set of sudden changes in mechanism action, the corresponding difference is calculated for the characteristic values ​​of opening and closing time and the peak value of coil current waveform in two consecutive operation events to generate feature offset. The system is segmented and judged based on the correspondence between the feature offset and the preset stable interval of the action, and a weight level judgment result is generated. Based on the weight level determination results, operation events that exceed the preset action stability range are assigned corresponding maintenance weight levels, and mutation marker sequences are generated. Based on the mutation marker sequence, the retention weight level remains unchanged in the subsequent preset number of operation events, and the retention weight level is updated according to the preset decreasing rule each time a new operation event is added, thus generating a mutation retention weight sequence.

[0026] In this embodiment of the invention, the feature offset between adjacent operation events is calculated based on the mechanism action mutation feature set, and a mutation retention weight sequence is generated. This allows the opening and closing time feature values ​​and coil current waveform peak features to be identified as changes in a single operation behavior when an offset occurs. By segmenting the feature offset and the stable operation interval, different offset degrees correspond to different weight levels, thereby distinguishing between normal fluctuations and abnormal changes. After marking operation events that exceed the stable operation interval and assigning retention weights, the abnormal operation continues to participate in the modeling process in multiple subsequent operation events. Its impact is gradually weakened through a decreasing rule, ensuring that the mutation feature persists in the short term without continuously amplifying in the long term, thus achieving time control over the impact range of a single abnormal operation.

[0027] In a preferred embodiment of the present invention, the method for setting a preset stable range of motion includes: The opening and closing time characteristics and coil current waveform peak characteristics of the high-voltage vacuum circuit breaker under normal operating conditions are collected to form a benchmark dataset; Perform statistical processing on each feature component in the benchmark dataset to calculate the mean and dispersion parameter; Using the average value as the center, upper and lower boundary intervals are constructed based on the dispersion parameter to obtain the stable range of each feature component; The stable ranges corresponding to each feature component are combined to form a preset stable range for the action.

[0028] In a preferred embodiment of the present invention, the method for setting the preset quantity includes: Obtain typical operating time intervals and mechanism recovery time parameters of high-voltage vacuum circuit breakers; Based on the correspondence between the institution's recovery time and the operation time interval, the corresponding range of operation times is calculated. The range of operation counts is adjusted by combining the number of operations with the duration of abnormal impact in historical abnormal operation data; The modified number of operations is used as a preset number to control the duration of the mutation retention weight.

[0029] In a preferred embodiment of the present invention, the method for setting the preset decreasing rule includes: Set the corresponding decreasing step size based on the initial level of the mutation preservation weight; Each time a new operation event is added, the weight level is updated in decreasing increments; When the weight level decreases to the preset termination value, the mutation hold state of this operation event is terminated; The corresponding decreasing period is set according to different initial weight levels, so that the mutation characteristics of different levels have different durations.

[0030] In a preferred embodiment of the present invention, segmentation is performed based on the correspondence between feature offsets and preset action stability intervals to generate weight level determination results, including: Obtain the feature offset corresponding to each operation event, and extract the offset values ​​of the opening and closing time feature values ​​and the peak value feature of the coil current waveform respectively; Obtain the upper and lower boundaries of the corresponding feature components in the preset action stability interval, and calculate the distance value of the feature offset relative to the boundary. The degree of deviation is determined based on the distance value, and the degree of deviation is divided into multiple level intervals; Each level interval is assigned a preset weight level, and its weight level is determined based on the interval where the feature offset is located. The weight level of each operation event is recorded in chronological order of operation time, and a weight level determination result is generated.

[0031] The rule for dividing the grade intervals is based on the degree of deviation of the feature offset from the stable interval. By statistically analyzing historical operating data, the offset range is divided into multiple intervals, so that different degrees of offset correspond to different weight levels, thereby reflecting the differences in the degree of anomaly.

[0032] The preset weight levels are set based on the degree of deviation of the feature offset from the stable range. By statistically analyzing historical abnormal operation data, corresponding weights are set for each level range, so that the degree of offset and the weight level are correlated.

[0033] In a preferred embodiment of the present invention, based on the weight level determination result, operation events exceeding the preset action stability range are assigned a corresponding maintenance weight level, and a mutation marker sequence is generated, including: Obtain the weight level determination results for each operation event; Determine whether the feature offset of the operation event exceeds the preset action stability range; When the feature offset of an operation event exceeds the preset stable range of the action, the operation event is marked as a mutation event, and its weight level determination result is used as its retention weight level. The labeling results of each mutation event are arranged according to the time sequence of the operation events to generate a mutation label sequence, and the corresponding preservation weight level is recorded synchronously.

[0034] In a preferred embodiment of the present invention, based on the mechanical cumulative degradation feature set, the feature changes in the operation event feature vector sequence generated with the increase of the number of operations are cumulatively calculated, and the reverse decline amplitude of the feature changes is limited, to generate a cumulative degradation weight sequence that changes with the operation events, including: Based on the mechanical cumulative deterioration feature set, the difference of the contact stroke change characteristics in multiple consecutive operation events is calculated to generate a single operation change sequence. Based on the sequence of changes in a single operation, directional consistency analysis is performed to generate the sequence of main change directions. Based on the main change direction sequence, the changes of single operations with the same direction are accumulated to generate a cumulative change sequence. Based on the correspondence between the cumulative change sequence and the preset fallback threshold, the magnitude of the change in the opposite direction is limited to generate a constrained change sequence. The sequence of constraint changes is segmented according to the changes in the number of operations to generate a deterioration stage interval sequence. Assign a cumulative degradation weight level to the corresponding operation event based on the degradation stage interval sequence, and maintain the weight level without decreasing during subsequent operation event updates to generate a cumulative degradation weight sequence.

[0035] In this embodiment of the invention, the changes in features generated by the number of operations in the feature vector sequence of operation events are cumulatively calculated based on the mechanical cumulative degradation feature set, so that the contact stroke change characteristics form a gradual cumulative trend during continuous operation. By performing directional consistency analysis on the changes in a single operation, the cumulative process maintains a uniform direction of change, thereby reflecting the continuity of mechanical wear. After limiting the amplitude of changes in opposite directions, the cumulative process avoids significant regression due to local fluctuations, thus maintaining the continuity of the overall trend. By dividing the cumulative changes into stages and assigning cumulative degradation weight levels, different degradation stages have different degrees of influence, and the weight levels are maintained without decreasing during subsequent updates, so that the cumulative degradation process exhibits irreversible change characteristics during modeling, thereby achieving a continuous expression of the long-term mechanical degradation process.

[0036] In a preferred embodiment of the present invention, directional consistency analysis is performed based on the sequence of changes in a single operation to generate a main change direction sequence, including: Obtain the sequence of changes in a single operation across multiple consecutive operation events; The sign of each change in the sequence of changes in a single operation is determined, and positive changes are defined as positive directions, while negative changes are defined as negative directions. Set the length of the direction analysis window, and count the number of positive and negative changes within each window; When the number of changes in the positive direction meets the preset ratio condition, the main change direction of the window is defined as the positive direction; when the number of changes in the negative direction meets the preset ratio condition, the main change direction of the window is defined as the negative direction. Arrange the main change directions corresponding to each window according to the time sequence of the operation events to generate a sequence of main change directions.

[0037] The preset setting allows the main change direction to reflect the main change trend in continuous operation by setting the window length and direction ratio thresholds.

[0038] In a preferred embodiment of the present invention, based on the main change direction sequence, the changes in single operations with the same direction are accumulated to generate a cumulative change sequence, including: Determine the main change direction based on the main change direction sequence; The cumulative change is obtained by successively accumulating the single operation change quantities that are in the same direction as the main change in the single operation change quantity sequence; Single operation changes that are inconsistent with the main change direction are not included in the current accumulation process and are processed separately in subsequent processing steps; When each operation event is updated, the change amount of a single operation that meets the direction consistency condition is added to the cumulative result, forming a cumulative change amount sequence that updates over time.

[0039] In a preferred embodiment of the present invention, based on the correspondence between the cumulative change sequence and a preset fallback threshold, amplitude limiting processing is applied to changes in opposite directions to generate a constrained change sequence, including: Obtain the cumulative change sequence and the corresponding single-operation change sequence; Determine whether the direction of change of a single operation is consistent with the direction of the main change; When the direction of the change in a single operation is opposite to the direction of the main change, calculate the magnitude of the change in that single operation. The magnitude of the change in a single operation is compared with a preset fallback threshold. When the magnitude of the change exceeds the preset fallback threshold, it is limited to a value within the preset fallback threshold range. The constrained change is added to the cumulative change sequence to generate the constrained change sequence.

[0040] The preset fallback threshold is set based on the natural fluctuation range during equipment operation and the irreversible nature of mechanical degradation. By statistically analyzing historical stable data, the normal fluctuation range is determined, and the fallback limit ratio is set in combination with the degradation trend so that reverse changes do not change the overall degradation direction.

[0041] In a preferred embodiment of the present invention, the degradation stage interval sequence is generated by segmenting the constraint change sequence according to the change of the constraint change amount sequence with the number of operations, including: The sequence of constraint changes is accumulated to obtain a sequence of cumulative changes as a function of the number of operations; Multiple stages are divided into intervals based on the range of cumulative changes; Determine the stage interval where the cumulative change corresponding to each operation event falls, and determine the corresponding deterioration stage; Arrange the degradation stages corresponding to each operation event according to the time sequence of the operation events to generate a degradation stage interval sequence.

[0042] The basis for pre-setting the degradation stage intervals lies in the stage characteristics of the equipment performance degradation process. By combining the equipment design parameters and historical operating data, the cumulative change is divided into multiple intervals, so that each interval corresponds to a different degree of degradation.

[0043] In a preferred embodiment of the present invention, a cumulative degradation weight sequence is generated by assigning a cumulative degradation weight level to the corresponding operation event according to the degradation stage interval sequence, and maintaining the weight level without decreasing during subsequent operation event updates. Preset corresponding cumulative degradation weight levels for each degradation stage; Assign the corresponding cumulative degradation weight level to the operation event based on the degradation stage to which it belongs; During subsequent operation event updates, determine the relationship between the current operation event's degradation stage and the corresponding stage of the previous weight level; When the current degradation stage is higher than the previous stage, update the cumulative degradation weight level to the cumulative degradation weight level of the corresponding stage. When the current degradation stage is lower than the previous stage, the original cumulative degradation weight level remains unchanged. Arrange the cumulative degradation weight levels corresponding to each operation event according to the time sequence of the operation events to generate a cumulative degradation weight sequence.

[0044] The preset cumulative degradation weight level is based on the difference in the degree of impact of different degradation stages on the equipment's operating status. By analyzing the degree of performance change corresponding to each stage, different weights are assigned to different stages so that the weight level can reflect the continuous enhancement characteristics of the degradation process.

[0045] In a preferred embodiment of the present invention, the operational event feature vector sequence is dynamically and recursively updated based on the mutation preservation weight sequence and the cumulative degradation weight sequence to generate a dynamic feature set, including: Based on the sequence of feature vectors of operational events and the corresponding mutation preservation weight sequence and cumulative degradation weight sequence, the feature components in each feature vector of operational events are weighted to generate a component-weighted feature vector sequence. Construct an initial dynamic feature set based on the component-weighted feature vector corresponding to the first operation event in the component-weighted feature vector sequence; Following the chronological order of operations, the component-weighted feature vectors corresponding to each subsequent operation event are recursively updated sequentially: During each recursive update, the current weighted feature vector and the previous recursive result are differentially fused according to the feature component type to generate the current dynamic feature set; During the recursive update process, the weights of the corresponding feature components in multiple consecutive operation events are maintained according to the mutation preservation weight sequence, so that the mutation features form a continuous influence path during the recursive update process. During the recursive update process, the weight ratio of the corresponding feature components is gradually increased according to the operation order based on the cumulative deterioration weight sequence, so that the cumulative change forms a monotonically enhanced evolution result during the recursive update process. Based on the evolution results of persistent influence paths and monotonically enhanced evolution, a dynamic feature set containing mutation path information and cumulative evolution information is obtained.

[0046] In this embodiment of the invention, the operational event feature vector sequence is dynamically and recursively updated based on the mutation preservation weight sequence and the cumulative degradation weight sequence, enabling each operational event feature to form a continuous update relationship in the time dimension. By weighting each feature component, different types of features participate in the modeling process according to their corresponding weights. During the recursive update process, the current component weighted feature vector and the previous recursive result are differentially fused according to the feature component type, so that mutation features form a short-term persistent path through coverage, and cumulative features form a long-term enhanced path through superposition. By maintaining the mutation feature weight and gradually increasing the cumulative feature weight during the recursive process, different features form differentiated evolution trajectories in the dynamic feature set, thereby enabling the dynamic feature set to simultaneously reflect the combined effects of instantaneous abnormal changes and long-term degradation processes.

[0047] In a preferred embodiment of the present invention, based on the operational event feature vector sequence and the corresponding mutation preservation weight sequence and cumulative degradation weight sequence, the feature components in each operational event feature vector are weighted to generate a component-weighted feature vector sequence, including: Obtain the sequence of feature vectors of operational events, as well as the corresponding mutation preservation weight sequence and cumulative degradation weight sequence; For each feature component in the feature vector of each operation event, match the corresponding mutation preservation weight value and cumulative degradation weight value respectively; According to the preset weighted calculation rules, weighted calculation is performed on each feature component to obtain the weighted feature component values; The weighted feature components are combined in their original order to form a component-weighted feature vector. Arrange the weighted feature vectors of each component according to the time sequence of the operation events to generate a sequence of weighted feature vectors.

[0048] The preset weighted calculation rules are based on the differences in the degree of influence of different feature components on the operating status. The influence weight of each feature component is determined by statistical analysis of historical data, and the feature components are comprehensively weighted by combining the mutation retention weight and the cumulative degradation weight, so that the weighted result can reflect the importance of different features.

[0049] In a preferred embodiment of the present invention, during the recursive update process, the weights of corresponding feature components in multiple consecutive operation events are maintained according to the mutation preservation weight sequence, so that the mutation features form a continuous influence path during the recursive update process, including: During the recursive update process, obtain the mutation preservation weight value corresponding to the current operation event; Determine whether the mutation preservation weight value is greater than the preset mutation preservation weight threshold; When the mutation retention weight value meets the condition, the corresponding mutation feature component value in the current operation event is recorded as the retention value; In the subsequent recursive update process of multiple operation events, the preserved value is directly used to replace the update result of the corresponding feature component; After each recursive update, the mutation preservation weight value is updated. When the mutation preservation weight value drops to the termination condition, the corresponding feature component is restored to participate in the normal update. Through the above processing, the mutation feature forms a persistent impact path in continuous operational events.

[0050] The preset mutation retention weight threshold and retention duration range are based on the duration of the mutation characteristics' impact on subsequent operations. This is determined by statistical analysis of the impact range of historical abnormal events, so that the mutation impact is maintained within a limited range.

[0051] In a preferred embodiment of the present invention, during the recursive update process, according to the cumulative degradation weight sequence, the weight ratio of the corresponding feature component is gradually increased in the order of operation, so that the cumulative change forms a monotonically enhancing evolution result during the recursive update process, including: In each operation event, obtain the corresponding cumulative degradation weight value; The weight percentage of the corresponding feature component is determined based on the cumulative degradation weight value; During the recursive update process, the feature component values ​​of the current operation event are added to the previous recursive result according to their weight proportions. In subsequent operation events, as the cumulative degradation weight value increases, the weight ratio of the corresponding feature component is gradually increased; During the iterative update process, the weight percentage is kept constant so that the cumulative change continues to strengthen. Through the above processing, the cumulative features are transformed into a monotonically enhanced evolutionary result in the operation sequence.

[0052] In a preferred embodiment of the present invention, based on the evolution results of the persistent influence path and monotonically enhanced evolution, a dynamic feature set containing mutation path information and cumulative evolution information is obtained, including: After completing the recursive update, obtain the update feature component values ​​corresponding to each operation event; Identify feature components that remain unchanged across multiple consecutive operational events and label them as mutation path features; Identify feature components that exhibit increasing changes with operational events and label them as cumulative evolutionary features; The mutation path features and cumulative evolution features are combined according to a unified structure to form a dynamic feature vector corresponding to a single operation event; The dynamic feature vectors are arranged in chronological order of operation to generate a dynamic feature set, which is used to describe the evolution of the device state.

[0053] In a preferred embodiment of the present invention, interval boundaries are calculated based on a dynamic feature set, and the boundaries are subjected to directional correlation expansion processing based on a mutation-preserving weight sequence, and unidirectional offset processing is performed on the boundaries based on a cumulative degradation weight sequence, thereby generating a dynamic reference model with asymmetric and irreversible characteristics, including: Construct a value distribution sequence based on the operation time order of each feature component in the dynamic feature set, and extract the upper and lower bounds of each feature component based on the value distribution sequence to generate the initial boundary; Based on the mutation-preserving weight sequence, the boundaries of the corresponding feature components in the initial boundary are expanded in a direction-related manner, so that the expansion magnitude of the boundary corresponding to the mutation direction is greater than that of the boundary in the opposite direction, thereby generating an asymmetric interval structure. Based on the cumulative degradation weight sequence, the boundaries of the corresponding feature components in the asymmetric interval structure are unidirectionally offset, so that the boundaries move along the degradation direction and are restricted from retreating in the opposite direction, thereby generating an offset interval structure with irreversible characteristics. By combining the characteristic components based on the asymmetric interval structure and the offset interval structure, a multidimensional interval model is generated to characterize the operating state of the high-voltage vacuum circuit breaker.

[0054] In this embodiment of the invention, interval boundaries are calculated based on a dynamic feature set, and the boundaries are constrained by combining the mutation preservation weight sequence and the cumulative degradation weight sequence, so that each feature component forms an interval structure with directional differences in spatial distribution. By performing directional correlation extension processing on the boundaries corresponding to mutation features, the boundary range of the mutation direction is extended more than that of the opposite direction boundary, thereby reflecting the influence range of a single abnormal operation on the distribution of the operating state; by performing unidirectional offset processing on the boundaries corresponding to cumulative features, the interval boundaries gradually move along the degradation direction and restrict reverse retreat, thereby maintaining the continuity of long-term changes. By combining the asymmetric interval structure and the offset interval structure, different features are formed into a multi-dimensional interval expression with constraints in a unified model, thereby realizing a structured description of the spatial distribution characteristics of the operating state.

[0055] In a preferred embodiment of the present invention, a value distribution sequence is constructed based on the operation time order of each feature component in the dynamic feature set, and the upper and lower bounds of each feature component are extracted based on the value distribution sequence to generate an initial boundary, including: Obtain the feature component values ​​corresponding to each operation event in the dynamic feature set; According to the time sequence of the operation events, the values ​​of the same feature component are extracted to form a value distribution sequence; Set the length of the boundary statistics window, and select the corresponding number of operation events within the window to participate in the statistics; Perform statistical processing on the value distribution sequence, extract the maximum value in the sequence as the upper bound, and extract the minimum value in the sequence as the lower bound; Perform the above processing on each feature component to obtain the corresponding upper and lower bounds; The upper and lower bounds of each feature component are combined to generate the initial boundary.

[0056] The setting of the preset boundary statistical window length is based on the time distribution of operation events and data stability. By analyzing historical operation data, a suitable window range is determined so that the boundary can reflect the state distribution of the current stage.

[0057] In a preferred embodiment of the present invention, based on the mutation-preserving weight sequence, the boundaries of corresponding feature components in the initial boundary are subjected to direction-dependent expansion processing, such that the expansion amplitude of the boundary corresponding to the mutation direction is greater than that of the boundary in the opposite direction, thereby generating an asymmetric interval structure, including: Obtain mutation direction information for the corresponding operation events in the mutation preservation weight sequence; The direction of expansion of the corresponding feature component is determined based on the direction of mutation; The boundary expansion magnitude is obtained by multiplying the mutation preservation weight value with the preset boundary expansion coefficient. Perform an expansion process on the boundary corresponding to the mutation direction to adjust it according to the boundary expansion magnitude; The boundary expansion range in the opposite direction is adjusted by keeping the original value or by calculating the boundary expansion range using a lower expansion factor; The above processing is performed on each feature component to generate an asymmetric interval structure with directional differences.

[0058] The preset boundary expansion coefficient is set based on the correspondence between the degree of deviation of mutation characteristics and its impact on the operating status. The expansion ratio is determined by analyzing historical mutation events so that the boundary expansion can reflect the scope of abnormal impact.

[0059] In a preferred embodiment of the present invention, based on the cumulative degradation weight sequence, the boundaries of corresponding feature components in the asymmetric interval structure are subjected to unidirectional offset processing, causing the boundaries to move along the degradation direction and restricting the boundaries from retreating in the opposite direction, thereby generating an offset interval structure with irreversible characteristics, including: Obtain the degradation direction information of the corresponding operation event in the cumulative degradation weight sequence; The boundary offset direction of each feature component is determined based on the degradation direction; The actual offset is calculated by proportionally dividing the cumulative degradation weight value by the preset boundary offset step size. Offset processing is performed on the corresponding boundaries in the asymmetric interval structure, so that the boundaries are moved along the deterioration direction; Record historical boundary values ​​and compare the current boundary with the historical boundary during each update. When the current boundary is less than the historical boundary, keep the historical boundary unchanged to limit reverse rollback. The above processing is performed on each feature component to generate an offset interval structure with unidirectional variation characteristics.

[0060] The preset boundary offset step size is based on the rate of change of the cumulative degradation process. The offset magnitude is determined by analyzing the long-term operating data of the equipment, so that the boundary change is consistent with the degree of degradation.

[0061] In a preferred embodiment of the present invention, during each recursive update, the current component-weighted feature vector and the previous recursive result are differentially fused according to the feature component type to generate the current dynamic feature set, including: Based on the correspondence between each feature component in the current component weighted feature vector and the corresponding feature component in the previous recursive result, a feature component correspondence relationship is generated. Based on the correspondence of feature components, the feature components corresponding to the abrupt change features of the mechanism action are classified as abrupt change processing components, and the feature components corresponding to the cumulative deterioration features of the machinery are classified as cumulative processing components, thus generating feature component classification results. Based on the feature component classification results, priority coverage processing is performed on the mutation processing component, so that the mutation processing component in the current component weighted feature vector directly replaces the corresponding feature component in the previous recursive result during fusion, and the replacement result of the feature component remains unchanged in the subsequent preset number of recursive updates, generating mutation path preservation results; Based on the feature component classification results, the cumulative processing components are subjected to consistent direction superposition processing, so that the cumulative processing components in the current component weighted feature vector are continuously superimposed with the corresponding feature components in the previous recursive result in the order of operation time, and the reverse change amplitude is restricted during the superposition process to generate cumulative evolution results. Based on the mutation path preservation results and cumulative evolution results, continuous transition processing is performed on the feature components other than the mutation processing components and cumulative processing components, so that the current component weighted feature vector and the previous recursive result form a continuous change relationship, generating a transition fusion result. Based on the results of mutation path preservation, cumulative evolution, and transition fusion, the feature components are combined to generate the current dynamic feature set.

[0062] In this embodiment of the invention, during each recursive update process, the current weighted feature vector and the previous recursive result are differentially fused according to the feature component type, so that different features evolve in different ways during the update process. By performing priority coverage processing on the mutation processing component, the mutation feature directly acts on the current state expression during the fusion process and maintains this effect in subsequent updates, thereby forming a continuous mutation influence path; by performing consistent direction superposition processing on the cumulative processing component, this type of feature forms a gradually enhanced cumulative result in multiple operations, and the consistency of change direction is maintained by limiting the magnitude of reverse change; by performing continuous transition processing on the remaining feature components, the overall feature change remains continuous in time. By combining the above results, the dynamic feature set forms an expression structure with path continuity and change consistency during the recursive process.

[0063] In a preferred embodiment of the present invention, a feature component correspondence relationship is generated by matching each feature component in the current component-weighted feature vector with the corresponding feature component in the previous recursive result, including: Each feature component in the current component weighted feature vector is numbered and labeled according to a preset order; Each feature component in the previous recursive result is numbered and labeled in the same order; Establish a one-to-one correspondence based on the number, and match each feature component in the current component weighted feature vector with the corresponding feature component in the previous recursive result; Record the correspondence between each feature component to generate a feature component correspondence table.

[0064] The preset order is based on the structural definition of the operation event feature vector. By unifying the feature arrangement order, the consistent correspondence between feature components at different time steps is ensured.

[0065] In a preferred embodiment of the present invention, based on the feature component classification result, a priority overlay process is performed on the mutation processing component, so that the mutation processing component in the current component weighted feature vector directly replaces the corresponding feature component in the previous recursive result during fusion, and the feature component replacement result remains unchanged in the subsequent preset number of recursive updates, generating a mutation path preservation result, including: Identify mutation treatment components based on the feature component classification results; Obtain the mutation preservation weight value corresponding to the mutation treatment component; When the mutation retention weight value meets the retention condition, the component value in the current component weighted feature vector is directly written into the current dynamic feature set, replacing the corresponding component value in the previous recursive result; During subsequent recursive updates, a hold operation is performed on this component, meaning it does not participate in the update within a preset number of operation events; After each recursive update, the mutation preservation weight value is updated. When the mutation preservation weight value drops to the termination condition, the component is restored to participate in the normal update. The above processing generates results that preserve mutation paths.

[0066] In a preferred embodiment of the present invention, based on the feature component classification result, a consistent direction superposition process is performed on the accumulated processing components, so that the accumulated processing components in the current component weighted feature vector are continuously superimposed with the corresponding feature components in the previous recursive result in the order of operation time, and the reverse change amplitude is limited during the superposition process to generate an accumulated evolution result, including: Identify the cumulative processing component based on the feature component classification results; Obtain the cumulative degradation weight value corresponding to the cumulative processed component; The superposition ratio is calculated based on the correspondence between the cumulative degradation weight value and the preset superposition ratio; The cumulative processed component in the current component weighted feature vector is added to the corresponding component in the previous recursive result according to the superposition ratio to obtain the updated value; Determine if there are changes in opposite directions during the update process. If there are changes in opposite directions, limit their magnitude according to a preset fallback threshold. Repeat the above process in chronological order to generate cumulative evolution results.

[0067] The preset superposition ratio is based on dividing the degradation process into stages and mapping them to corresponding ratios, so that the superposition process reflects the changes in the degree of degradation.

[0068] In a preferred embodiment of the present invention, based on the mutation path preservation result and the cumulative evolution result, continuous transition processing is performed on the feature components other than the mutation processing component and the cumulative processing component, so that the current component weighted feature vector and the previous recursive result form a continuous change relationship, generating a transition fusion result, including: Identify feature components that were not classified into mutation treatment components and cumulative treatment components; Get the weighted feature vector of the current component and the value of the corresponding component in the previous recursive result; Based on the preset transition coefficient, the current value and the previous recursive result are weighted and calculated to obtain the transition update value; Use the transitional update value as the corresponding component value in the current dynamic feature set; The above process is repeated in subsequent recursive updates to maintain the continuous change relationship of this type of feature component in the time series and generate transition fusion results.

[0069] The preset transition coefficient is set based on the smoothness requirement of feature component changes. It is determined by analyzing the fluctuation of historical data to ensure that the update results remain continuous between adjacent operation events.

[0070] In a preferred embodiment of the present invention, based on the mutation path preservation result, cumulative evolution result, and transition fusion result, each feature component is combined to generate the current dynamic feature set, including: The feature components corresponding to the mutation path preservation result, the feature components corresponding to the cumulative evolution result, and the feature components corresponding to the transition fusion result are arranged in a preset order; Each feature component is uniformly encapsulated to form a dynamic feature vector corresponding to a single operation event; Output the current dynamic feature vector as the recursive update result; According to the time sequence of the operation events, each dynamic feature vector is stored sequentially to form a dynamic feature set for subsequent modeling processing.

[0071] In a preferred embodiment of the present invention, the feature components are combined according to the asymmetric interval structure and the offset interval structure to generate a multi-dimensional interval model for characterizing the operating state of the high-voltage vacuum circuit breaker, including: Based on the asymmetric interval structure corresponding to each feature component, the directional extension boundary of each feature component is determined, and each feature component is classified and labeled according to the extension direction to generate a set of directional boundaries. Based on the offset interval structure corresponding to each feature component, the unidirectional offset boundary of each feature component is determined, and the feature components are arranged in layers according to the offset direction to generate an offset boundary set. Based on the set of directional boundaries and the set of offset boundaries, the boundaries of each feature component are constrained and combined so that the directional extension boundary and the unidirectional offset boundary of each feature component form a corresponding relationship, thereby generating a feature interval constraint structure. Based on the feature interval constraint structure, each feature component is hierarchically sorted according to its weight ratio in the feature vector of the operation event, so that the interval boundary of the high-influence feature component first defines the overall interval range, thus generating a hierarchical interval structure. Based on the hierarchical interval structure, the intervals of each feature component are nested and combined so that the intervals of lower-level feature components are within the range of higher-level feature components, thus generating a multi-dimensional interval model.

[0072] In this embodiment of the invention, the feature components are combined according to the asymmetric interval structure and the offset interval structure, so that different features form a hierarchical multidimensional structure in the interval representation. By classifying the directional expansion boundaries of each feature component, the boundaries of the mutation direction and the opposite direction are structurally distinguished; by arranging the unidirectional offset boundaries in layers, the degree of change of different features in the degradation direction is differentiated. By constraining and combining the directional boundary set and the offset boundary set, each feature component simultaneously possesses directional expansion characteristics and unidirectional offset characteristics; by hierarchically sorting each feature component according to its weight ratio, high-weight features preferentially limit the overall interval range; by nesting and combining the intervals of each level, lower-level features are constrained by higher-level features, thereby forming a unified multidimensional interval model to characterize the distribution relationship of the operating state in different feature dimensions.

[0073] In a preferred embodiment of the present invention, based on the asymmetric interval structure corresponding to each feature component, the directional extension boundary of each feature component is determined, and each feature component is classified and identified according to the extension direction to generate a set of directional boundaries, including: Obtain the upper and lower bounds of the asymmetric interval structure corresponding to each feature component; Determine the direction of abrupt change for each feature component in the current operation event; The corresponding directional expansion boundary is determined based on the mutation direction. That is, when the feature component undergoes a positive mutation, the upper bound of the expansion is selected as the directional expansion boundary, and when it undergoes a reverse mutation, the lower bound of the expansion is selected as the directional expansion boundary. Each feature component is classified and labeled according to its expansion direction to form directional labeling information; The directional extension boundaries of each feature component and its corresponding identifier are combined to generate a set of directional boundaries.

[0074] In a preferred embodiment of the present invention, based on the offset interval structure corresponding to each feature component, the unidirectional offset boundary of each feature component is determined, and the feature components are arranged in layers according to the offset direction to generate an offset boundary set, including: Obtain the boundary values ​​in the offset interval structure corresponding to each feature component; The offset direction of each feature component is determined based on the cumulative degradation weight sequence; Select the boundary after moving along the offset direction as the unidirectional offset boundary; The level to which each feature component belongs is determined based on the offset magnitude of each feature component; Arrange the feature components in hierarchical order and record the corresponding offset boundary values; The offset boundaries and their hierarchical information of each feature component are combined to generate an offset boundary set.

[0075] In a preferred embodiment of the present invention, the boundaries of each feature component are constrained and combined according to the set of directional boundaries and the set of offset boundaries, so that the directional extension boundary and the unidirectional offset boundary of each feature component form a corresponding relationship, thereby generating a feature interval constraint structure, including: Index and identify each feature component in the directional boundary set; Search for feature components with the same index in the offset boundary set; Match the directional extension boundary of the corresponding feature component with the unidirectional offset boundary; Based on the preset constraint combination rules, the matched boundaries are integrated to form a unified boundary expression; Perform the above processing on all feature components to generate a feature interval constraint structure that includes directional constraints and offset constraints.

[0076] The constraint combination rules are set based on the constraint relationship between different boundaries and the operating state. The combination method is determined by analyzing the boundary action mode so that the boundary retains both the directional expansion characteristics and the offset constraint characteristics.

[0077] In a preferred embodiment of the present invention, based on the feature interval constraint structure, each feature component is hierarchically sorted according to its weight ratio in the operation event feature vector, so that the interval boundaries of high-influence feature components preferentially define the overall interval range, generating a hierarchical interval structure, including: Obtain the weight ratio of each feature component in the feature vector of the operation event; Sort the feature components according to their weight percentage; The feature component with the highest weight is divided into the first layer; The remaining feature components are then sequentially divided into subsequent layers; Record the interval boundaries corresponding to each level to generate a hierarchical interval structure.

[0078] In a preferred embodiment of the present invention, based on a hierarchical interval structure, the intervals of each feature component are nested and combined so that the intervals of lower-level feature components are located within the intervals of higher-level feature components, thereby generating a multi-dimensional interval model, including: The interval of the first layer feature components is used as the base interval; The interval of the second-layer feature components is restricted to the range of the basic interval; Embed the intervals of subsequent feature components into the intervals of the previous layer in hierarchical order; The intervals at each level are combined to form a multi-dimensional interval structure; Output a multidimensional interval model to describe the spatial constraint relationships between the feature components.

[0079] Embodiments of the present invention also provide a dynamic modeling and processing system for high-voltage vacuum circuit breaker operation and maintenance data, the system comprising: The data acquisition module is used to acquire the opening and closing time, current waveform and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker, and to time-mark the opening and closing time, current waveform and contact travel data according to a single operation event to generate the original operation event data unit. The feature extraction module is used to perform event-level feature extraction on the original operation event data units, extracting the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, and generating an operation event feature vector sequence. The feature segmentation module is used to segment the feature vector sequence of operation events according to the correspondence between each feature in the operation event feature vector sequence and the physical state of the high-voltage vacuum circuit breaker, and generate a mechanism action sudden change feature set and a mechanical cumulative deterioration feature set. The preset mechanism action sudden change feature set includes the opening and closing time feature value and the coil current waveform peak value, and the mechanical cumulative deterioration feature set includes the contact stroke change feature. The weight generation module is used to calculate the feature offset between adjacent operation events in the operation event feature vector sequence based on the mechanism action mutation feature set, and generate a mutation-preserving weight sequence that changes with operation events based on the feature offset; based on the mechanical cumulative deterioration feature set, it accumulates and calculates the feature change in the operation event feature vector sequence that increases with the number of operations, limits the reverse fall of the feature change, and generates a cumulative deterioration weight sequence that changes with operation events. The dynamic modeling module is used to perform a dynamic weighted update of the operational event feature vector sequence by recursively updating it based on the mutation retention weight sequence and the cumulative degradation weight sequence, thereby generating a dynamic feature set containing mutation path information and cumulative evolution information. The model building module is used to calculate the interval boundary based on the dynamic feature set, perform directional correlation expansion processing on the boundary based on the mutation-preserving weight sequence, and perform unidirectional offset processing on the boundary based on the cumulative deterioration weight sequence to generate a dynamic reference model with asymmetric and irreversible characteristics. The trajectory generation module is used to calculate and generate operation offset trajectory data that evolves with the operation event based on the relative positional relationship between the operation event feature vector corresponding to the new operation event and the dynamic reference model.

[0080] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0081] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0082] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0083] The method of the present invention will be specifically described below in conjunction with actual operating scenarios: In a 110kV substation, continuous operation monitoring is conducted on a 10kV outgoing line bay high-voltage vacuum circuit breaker. This circuit breaker mainly undertakes frequent switching tasks for capacitor banks and load lines, with an average of approximately 80-120 operations per day. The system collects the opening coil current, closing coil current, and contact displacement signals in real time through online monitoring devices, and generates independent operation event data units for each opening or closing operation.

[0084] (1) Basic data collection and event construction In the first 1000 consecutive operations, the system obtained the following typical data range: Opening time: 28ms~32ms; Peak current waveform: 4.5A~5.2A; Contact stroke variation: 9.8mm~10.2mm; The system automatically identifies the start and end times of the operation based on the rising edge and falling point of the current, and determines the movement range by combining the contact displacement signal. Each operation is encapsulated as an operation event data unit and formed into an operation event sequence in chronological order.

[0085] (2) Modeling of mutation features During the 1053rd operation, a mechanism jamming occurred: The tripping time suddenly changed from 30ms to 56ms; The peak current increased from 4.8A to 6.3A; The system calculates the characteristic offset between the current operation and the previous operation, and compares it with the preset stable action range (e.g., tripping time ±3ms, current peak ±0.5A). If it is determined to be a high-level anomaly, a mutation retention weight level (e.g., level 4) is assigned.

[0086] The system then generates mutation markers and performs the following processing in subsequent operations: In the subsequent 10 operations, this outlier continued to participate in the dynamic modeling; After each operation, the weight is gradually reduced according to a preset decreasing rule (e.g., 4-3-2-1); Even after the 1054th to 1060th operations returned to normal (the tripping time returned to 31ms), the anomaly continued to be reflected in the model through the "mutation path".

[0087] (3) Cumulative degradation modeling During continuous operation, contact wear gradually accumulates:

[0088] The system performs the following processing on changes in contact stroke: 1. Calculate the change in a single operation; 2. Conduct directional consistency analysis (all changes are positive); 3. Accumulate to form the cumulative change; 4. Limit the amplitude of local reverse fluctuations (e.g., a brief drop of 0.1mm); The degradation stages are then determined based on the cumulative changes: 0~0.3mm: Slight degradation (weight 1); 0.3~0.7mm: Moderate degradation (weight 2); 0.7mm: Severe degradation (weight 3); And in the subsequent process: the weight is only allowed to increase, not decrease.

[0089] (4) Construction of dynamic feature set The system fuses the feature vector and weights for each operation: For mutation characteristics: use the "coverage preservation" method to preserve outliers in subsequent operations; For cumulative characteristics: an "incremental superposition" method is adopted, gradually increasing the impact with the number of operations; For common features: a smooth transition is applied; This ultimately forms a dynamic feature set, whose characteristics are as follows: Mutation pathway: Short-term anomalies persist; Cumulative path: Long-term changes gradually strengthen.

[0090] (5) Construction of dynamic reference model Based on the dynamic feature set, the system constructs an interval model: 1. Mutation characterization: For example, the tripping time deviates abnormally upward: the upper limit expands to 60ms, while the lower limit remains around 30ms, forming an asymmetric interval.

[0091] 2. Cumulative feature processing: The contact travel changes over time: the interval shifts upward as a whole and the historical maximum value does not regress, forming a unidirectional shift interval.

[0092] 3. Multidimensional interval combination: The system sorts the data according to its weight percentage: Opening time (0.5) – First layer; Peak current (0.3) – Second layer; Changes in travel (0.2) – Third layer; And construct a nested interval model: the tripping time interval, current interval, and travel interval are nested in sequence.

[0093] (6) Offset trajectory generation When the 3200th operation occurs: Opening time: 33ms (normal); Stroke variation: 11.3mm (too large); The system calculates its position relative to the model: In terms of circuit breaker operation: close to the center; In terms of travel distance: close to the upper boundary; Generate a single offset vector: (Switch-off offset ≈ 0.1, current offset ≈ 0.2, travel offset ≈ 0.8) As the operation continues: The 3201st time – (0.1, 0.2, 0.82); The 3202nd time — (0.1, 0.2, 0.85); This eventually forms a continuous trajectory.

[0094] Through this embodiment, this method achieves the following: single anomalies will not be masked by averaging, long-term wear and tear will not be offset by short-term fluctuations, and state changes will be upgraded from "point judgment" to "trajectory expression".

[0095] Ultimately, it can be used to predict failure to operate, identify jamming trends, and determine the life stage of contacts.

[0096] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method, characterized in that, The method includes: Acquire the opening and closing time, current waveform, and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker, and time-mark the opening and closing time, current waveform, and contact travel data according to a single operation event to generate the original operation event data unit; Event-level feature extraction is performed on the original operation event data unit to extract the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, and generate an operation event feature vector sequence. Based on the correspondence between each feature in the operation event feature vector sequence and the physical state of the high-voltage vacuum circuit breaker, the operation event feature vector sequence is divided into feature attributes to generate a mechanism action mutation feature set and a mechanical cumulative deterioration feature set. The preset mechanism action mutation feature set includes the opening and closing time feature value and the coil current waveform peak value, while the mechanical cumulative deterioration feature set includes the contact stroke change feature. Based on the mechanism action mutation feature set, the feature offset between adjacent operation events in the operation event feature vector sequence is calculated, and a mutation preservation weight sequence that changes with operation events is generated based on the feature offset; based on the mechanical cumulative deterioration feature set, the feature change amount generated by the operation event feature vector sequence with the increase of operation number is cumulatively calculated, and the reverse fall amplitude of the feature change amount is limited, generating a cumulative deterioration weight sequence that changes with operation events. Based on the mutation preservation weight sequence and the cumulative degradation weight sequence, the operational event feature vector sequence is dynamically updated by successive recursion to generate a dynamic feature set containing mutation path information and cumulative evolution information. The interval boundary is calculated based on the dynamic feature set, and the boundary is extended by direction correlation based on the mutation-preserving weight sequence. The boundary is offset by unidirectional offset based on the cumulative deterioration weight sequence, generating a dynamic reference model with asymmetric and irreversible characteristics. Based on the relative positional relationship between the feature vector of the new operation event and the dynamic reference model, operation offset trajectory data that evolves with the operation event is calculated and generated.

2. The high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method according to claim 1, characterized in that, Based on the mechanism action mutation feature set, the feature offset between adjacent operation events in the operation event feature vector sequence is calculated, and a mutation preservation weight sequence that changes with operation events is generated based on the feature offset, including: Based on the feature set of sudden changes in mechanism action, the corresponding difference is calculated for the characteristic values ​​of opening and closing time and the peak value of coil current waveform in two consecutive operation events to generate feature offset. The system is segmented and judged based on the correspondence between the feature offset and the preset stable interval of the action, and a weight level judgment result is generated. Based on the weight level determination results, operation events that exceed the preset action stability range are assigned corresponding maintenance weight levels, and mutation marker sequences are generated. Based on the mutation marker sequence, the retention weight level remains unchanged in the subsequent preset number of operation events, and the retention weight level is updated according to the preset decreasing rule each time a new operation event is added, thus generating a mutation retention weight sequence.

3. The high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method according to claim 1, characterized in that, Based on the mechanical cumulative degradation feature set, the changes in features in the operation event feature vector sequence that increase with the number of operations are cumulatively calculated, and the magnitude of the reverse decline of the feature changes is limited to generate a cumulative degradation weight sequence that changes with the operation events, including: Based on the mechanical cumulative deterioration feature set, the difference of the contact stroke change characteristics in multiple consecutive operation events is calculated to generate a single operation change sequence. Based on the sequence of changes in a single operation, directional consistency analysis is performed to generate the sequence of main change directions. Based on the main change direction sequence, the changes of single operations with the same direction are accumulated to generate a cumulative change sequence. Based on the correspondence between the cumulative change sequence and the preset fallback threshold, the magnitude of the change in the opposite direction is limited to generate a constrained change sequence. The sequence of constraint changes is segmented according to the changes in the number of operations to generate a deterioration stage interval sequence. Assign a cumulative degradation weight level to the corresponding operation event based on the degradation stage interval sequence, and maintain the weight level without decreasing during subsequent operation event updates to generate a cumulative degradation weight sequence.

4. The high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method according to claim 1, characterized in that, Based on the mutation-preserving weight sequence and the cumulative degradation weight sequence, the operational event feature vector sequence is dynamically updated through successive recursive iterations to generate a dynamic feature set, including: Based on the sequence of feature vectors of operational events and the corresponding mutation preservation weight sequence and cumulative degradation weight sequence, the feature components in each feature vector of operational events are weighted to generate a component-weighted feature vector sequence. Construct an initial dynamic feature set based on the component-weighted feature vector corresponding to the first operation event in the component-weighted feature vector sequence; Following the chronological order of operations, the component-weighted feature vectors corresponding to each subsequent operation event are recursively updated sequentially: During each recursive update, the current weighted feature vector and the previous recursive result are differentially fused according to the feature component type to generate the current dynamic feature set; During the recursive update process, the weights of the corresponding feature components in multiple consecutive operation events are maintained according to the mutation preservation weight sequence, so that the mutation features form a continuous influence path during the recursive update process. During the recursive update process, the weight ratio of the corresponding feature components is gradually increased according to the operation order based on the cumulative deterioration weight sequence, so that the cumulative change forms a monotonically enhanced evolution result during the recursive update process. Based on the evolution results of persistent influence paths and monotonically enhanced evolution, a dynamic feature set containing mutation path information and cumulative evolution information is obtained.

5. The high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method according to claim 1, characterized in that, The interval boundaries are calculated based on the dynamic feature set, and the boundaries are extended with directional correlation based on the mutation-preserving weight sequence. A unidirectional offset is then applied to the boundaries based on the cumulative degradation weight sequence, generating a dynamic reference model with asymmetric and irreversible properties, including: Construct a value distribution sequence based on the operation time order of each feature component in the dynamic feature set, and extract the upper and lower bounds of each feature component based on the value distribution sequence to generate the initial boundary; Based on the mutation-preserving weight sequence, the boundaries of the corresponding feature components in the initial boundary are expanded in a direction-related manner, so that the expansion magnitude of the boundary corresponding to the mutation direction is greater than that of the boundary in the opposite direction, thereby generating an asymmetric interval structure. Based on the cumulative degradation weight sequence, the boundaries of the corresponding feature components in the asymmetric interval structure are unidirectionally offset, so that the boundaries move along the degradation direction and are restricted from retreating in the opposite direction, thereby generating an offset interval structure with irreversible characteristics. By combining the characteristic components based on the asymmetric interval structure and the offset interval structure, a multidimensional interval model is generated to characterize the operating state of the high-voltage vacuum circuit breaker.

6. The high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method according to claim 4, characterized in that, During each recursive update, the current weighted feature vector and the previous recursive result are differentially fused according to the feature component type to generate the current dynamic feature set, including: Based on the correspondence between each feature component in the current component weighted feature vector and the corresponding feature component in the previous recursive result, a feature component correspondence relationship is generated. Based on the correspondence of feature components, the feature components corresponding to the abrupt change features of the mechanism action are classified as abrupt change processing components, and the feature components corresponding to the cumulative deterioration features of the machinery are classified as cumulative processing components, thus generating feature component classification results. Based on the feature component classification results, priority coverage processing is performed on the mutation processing component, so that the mutation processing component in the current component weighted feature vector directly replaces the corresponding feature component in the previous recursive result during fusion, and the replacement result of the feature component remains unchanged in the subsequent preset number of recursive updates, generating mutation path preservation results; Based on the feature component classification results, the cumulative processing components are subjected to consistent direction superposition processing, so that the cumulative processing components in the current component weighted feature vector are continuously superimposed with the corresponding feature components in the previous recursive result in the order of operation time, and the reverse change amplitude is restricted during the superposition process to generate cumulative evolution results. Based on the mutation path preservation results and cumulative evolution results, continuous transition processing is performed on the feature components other than the mutation processing components and cumulative processing components, so that the current component weighted feature vector and the previous recursive result form a continuous change relationship, generating a transition fusion result. Based on the results of mutation path preservation, cumulative evolution, and transition fusion, the feature components are combined to generate the current dynamic feature set.

7. The high-voltage vacuum circuit breaker operation and maintenance data dynamic modeling processing method according to claim 5, characterized in that, By combining the characteristic components based on the asymmetric interval structure and the offset interval structure, a multi-dimensional interval model for characterizing the operating state of a high-voltage vacuum circuit breaker is generated, including: Based on the asymmetric interval structure corresponding to each feature component, the directional extension boundary of each feature component is determined, and each feature component is classified and labeled according to the extension direction to generate a set of directional boundaries. Based on the offset interval structure corresponding to each feature component, the unidirectional offset boundary of each feature component is determined, and the feature components are arranged in layers according to the offset direction to generate an offset boundary set. Based on the set of directional boundaries and the set of offset boundaries, the boundaries of each feature component are constrained and combined so that the directional extension boundary and the unidirectional offset boundary of each feature component form a corresponding relationship, thereby generating a feature interval constraint structure. Based on the feature interval constraint structure, each feature component is hierarchically sorted according to its weight ratio in the feature vector of the operation event, so that the interval boundary of the high-influence feature component first defines the overall interval range, thus generating a hierarchical interval structure. Based on the hierarchical interval structure, the intervals of each feature component are nested and combined so that the intervals of lower-level feature components are within the range of higher-level feature components, thus generating a multi-dimensional interval model.

8. A dynamic modeling and processing system for operation and maintenance data of high-voltage vacuum circuit breakers, characterized in that, The system, used in any one of claims 1 to 7, comprises: The data acquisition module is used to acquire the opening and closing time, current waveform and contact travel data generated by each opening and closing operation of the high-voltage vacuum circuit breaker, and to time-mark the opening and closing time, current waveform and contact travel data according to a single operation event to generate the original operation event data unit. The feature extraction module is used to perform event-level feature extraction on the original operation event data units, extracting the opening and closing time feature values, coil current waveform peak features, and contact stroke change features, and generating an operation event feature vector sequence. The feature segmentation module is used to segment the feature vector sequence of operation events according to the correspondence between each feature in the operation event feature vector sequence and the physical state of the high-voltage vacuum circuit breaker, and generate a mechanism action sudden change feature set and a mechanical cumulative deterioration feature set. The preset mechanism action sudden change feature set includes the opening and closing time feature value and the coil current waveform peak value, and the mechanical cumulative deterioration feature set includes the contact stroke change feature. The weight generation module is used to calculate the feature offset between adjacent operation events in the operation event feature vector sequence based on the mechanism action mutation feature set, and generate a mutation-preserving weight sequence that changes with operation events based on the feature offset; based on the mechanical cumulative deterioration feature set, it accumulates and calculates the feature change in the operation event feature vector sequence that increases with the number of operations, limits the reverse fall of the feature change, and generates a cumulative deterioration weight sequence that changes with operation events. The dynamic modeling module is used to perform a dynamic weighted update of the operational event feature vector sequence by recursively updating it based on the mutation retention weight sequence and the cumulative degradation weight sequence, thereby generating a dynamic feature set containing mutation path information and cumulative evolution information. The model building module is used to calculate the interval boundary based on the dynamic feature set, perform directional correlation expansion processing on the boundary based on the mutation-preserving weight sequence, and perform unidirectional offset processing on the boundary based on the cumulative deterioration weight sequence to generate a dynamic reference model with asymmetric and irreversible characteristics. The trajectory generation module is used to calculate and generate operation offset trajectory data that evolves with the operation event based on the relative positional relationship between the operation event feature vector corresponding to the new operation event and the dynamic reference model.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images