A method for monitoring the blown state of a projection fuse

Abstract

This invention discloses a method for monitoring the fusing status of jet-type fuses, relating to the field of condition monitoring and relay protection technology for power system distribution network equipment. This invention employs time-series fusion and dynamic feature extraction to effectively identify slowly changing anomalies caused by fuse aging, such as temperature rise residuals and changes in energy integral, thereby providing early warning during the pre-fusing stage and avoiding sudden failures. An adaptive threshold adjustment mechanism, combined with environmental and load factors, enhances the robustness of the monitoring system, reducing false alarms caused by external interference such as temperature fluctuations or load peaks, ensuring stable operation even under harsh conditions. Furthermore, local processing at the edge terminal and low-power design reduce communication dependence and energy consumption, supporting long-term deployment in remote areas and improving power grid operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of condition monitoring and relay protection technology for power system distribution network equipment, and in particular to a method for monitoring the melting status of a jet-type fuse. Background Technology

[0002] Ejector fuses are commonly used overcurrent and short-circuit protection devices on 10kV distribution line branches and on one side of distribution transformers, and can also provide isolation with a clear break. With the digitalization of distribution networks, the demand for online status sensing and rapid alarm of fuses is gradually increasing.

[0003] Under long-term outdoor operation, fuses may gradually deteriorate and enter the pre-arc (pre-melting) stage due to temperature cycling, load fluctuations and environmental aging. Its characteristics include a slow increase in component resistance and power consumption, local temperature rise and intermittent micro-arcs. If not identified in time, it can easily evolve into complete melting, expand the power outage area and increase the complexity of operation and maintenance.

[0004] Existing online monitoring methods often employ multi-source fusion, such as voltage / current and attitude / angle, or incorporate electric field / image information and transmit it wirelessly, which can effectively identify events such as drop / closing / fuse failure. However, during the gradual deterioration / pre-fuse stage, the threshold / rule-based criteria are prone to false alarms / missed alarms due to the influence of on-site load fluctuations, ambient light / obstruction, and sensor noise. Furthermore, algorithms based on complex models face limitations in edge computing power and real-time performance. Therefore, it is necessary to propose a pre-fuse failure monitoring method that is more sensitive to subtle pre-arc signs and more robust to operating disturbances on resource-constrained terminals. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] This invention provides a method for monitoring the melting status of jet-type fuses, which solves the problem that existing jet-type fuse monitoring methods are difficult to capture the pre-melting gradual characteristics caused by aging due to fixed thresholds and environmental interference, thus increasing the risk of missed detection.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for monitoring the fusing status of a jet-type fuse, comprising: Step S1: Collect multi-source operating data of the fuse, wherein the multi-source operating data includes at least electrical quantity data and physical status data; Step S2: Perform time alignment, noise reduction, and sliding window analysis on the multi-source running data to extract a comprehensive feature vector; Step S3: Generate a dynamic health score for the fuse based on the comprehensive feature vector; Step S4: Compare the dynamic health score with the adaptive threshold to obtain a status monitoring signal and output it to the monitoring platform. The status monitoring signal includes at least a normal status and a pre-circuit breaker warning.

[0008] As a preferred embodiment of the method for monitoring the fuse status of a jet-type fuse according to the present invention, the electrical quantity data includes at least one or more of the line current and the voltage across the fuse terminals; The physical state data includes at least one or more of the following: fuse housing temperature and mechanical angle.

[0009] As a preferred embodiment of the method for monitoring the fuse status of a jet-type fuse according to the present invention, the window length and step size of the sliding window analysis are adaptively set according to the operating load and environmental conditions, and the comprehensive feature vector includes at least one of the following: mean square value of current / voltage, slope and fluctuation, energy integral, equivalent resistance change rate, temperature rise residual, angle change rate and attitude stability index. The calculation method for the comprehensive feature vector includes: Step C1: In the sliding analysis, set the window and step size according to load and environmental fluctuations: , , in, Indicates the first Window length, Indicates the first The step size of each slide Minimum window length, The scaling factor. This is the step size scaling factor, dimensionless. The loading factor is dimensionless. For the first The RMS current within the window, expressed in amperes (A). Rated current, in amperes (A). This is an environmental index, dimensionless. This is the floor operator; For window index, subscript Indicates the rated value; overline This represents the RMS or mean value within the corresponding window. Refers to RMS; Step C2: Extract heat accumulation correlation features within each adaptive window: , in, For the first The energy integral of the window, in units of , For the discrete current sequence, the first One sample, in units of A. The sampling period is expressed in seconds. For the first Window starting sample number, dimensionless, subscript For sample index; Step C3: Model the temperature with a first-order smoothed electro-thermal baseline and generate residuals: , in, For the first Window temperature rise residual, in °C. For the shell temperature at the The mean value of the window, in °C. The baseline temperature estimate for the previous window, in °C. The smoothing coefficient is dimensionless. For the first Average ambient temperature at the window, in °C. This is the electro-thermal gain coefficient, in units of... , The square of the current in the first place The mean of the window, in units of Subscript Indicates temperature-related quantities, subscript Indicates environmental quantity; Step C4: Dimensionless transformation is performed on the robust baseline, and task weights are applied to obtain the comprehensive feature vector: , in, For the first The comprehensive feature vector of the window, dimension Dimensionless For the first The original feature vector of the window, The baseline median array for each component, and Same dimension, Let be the scale vector of each component, and be ... Same dimension, For the first The weights of each feature component are dimensionless. For feature dimension, For feature component index, For diagonal matrix operators, It is the reciprocal of each element; Step C5: The slope and volatility are obtained using the least squares straight line slope and the detrended standard deviation within the window. The equivalent resistance change rate, angle change rate, and attitude stability are formed into sub-vectors by taking the difference between adjacent windows and the variance family index. Finally, the electrical, thermal, and mechanical subsets are trimmed and concatenated as needed. .

[0010] As a preferred embodiment of the method for monitoring the fusing status of a jet-type fuse according to the present invention, the dynamic health score is based on a baseline model established by historical operating data and trend analysis is performed. The trend analysis includes modeling and updating of long-term slow changes and periodic disturbances.

[0011] As a preferred embodiment of the method for monitoring the fusing status of a jet-type fuse according to the present invention, the adaptive threshold is dynamically adjusted according to the ambient temperature, ambient humidity and current load level, and hysteresis and rate of change limits are set. The step of adjusting the adaptive threshold according to environment and load includes: Step D1: Use robust statistics of health scores as a threshold baseline: , in, Indicates the first The threshold baseline of the window is dimensionless. Indicates the first The previously selected set of historical health scores is dimensionless. The scaling factor is dimensionless. For the sample median operator, For absolute median difference operator; Step D2: Overlay environmental and load effects onto the baseline to form an initial adjustment threshold: , in, For the first The threshold for adjusting the window once, dimensionless. These are the influence coefficients for temperature, humidity, and load, respectively; they are dimensionless. For the first The normalized quantity of the ambient temperature at the window, dimensionless. For the first Normalized measure of ambient humidity at the window, dimensionless. The loading factor is dimensionless. For the first Window current RMS, in amperes (A). Rated current, unit is A, subscript Indicates environmental quantity, subscript Indicates the rated value; Step D3: Set the hysteresis band for different entry criteria for uplink and downlink: , in, This represents the uplink trigger threshold, which is dimensionless. This represents the downlink release threshold, which is dimensionless. Hysteresis half-amplitude, dimensionless, subscript They represent upward and downward movement, respectively. Step D4: Limit the rate of change of the threshold in adjacent windows to obtain the final threshold: , in, For the first The final threshold of the window, dimensionless. The final threshold of the previous window, dimensionless. The maximum rate of change in the upward direction. The maximum rate of change in the downward direction. To be Cut off to interval The operator.

[0012] As a preferred embodiment of the method for monitoring the fusing status of a jet-type fuse according to the present invention, the status monitoring signal includes one or more of five states: normal, pre-fusing, fusing, drop, and abnormal; wherein, the pre-fusing state is generated when the dynamic health score exceeds the limit and the duration meets the criterion.

[0013] In a preferred embodiment of the method for monitoring the fusing state of a jet-type fuse according to the present invention, the mechanical angle is acquired by an inertial measurement unit, which is used to detect changes in the mechanical position of the fuse and trigger candidate event markers when a rapid angle change is detected.

[0014] As a preferred embodiment of the method for monitoring the fuse status of a jet-type fuse according to the present invention, the method includes: identifying abnormal fluctuations in the multi-source operating data, wherein the identification of abnormal fluctuations includes robust statistical discrimination of short-term impacts and external interferences, and the results are used to correct the dynamic health score.

[0015] As a preferred embodiment of the method for monitoring the fuse status of a jet-type fuse according to the present invention, the method is executed on an edge terminal installed near the fuse, the edge terminal having local storage and wireless communication functions, the wireless communication being one or more of cellular, NB-IoT or LoRa.

[0016] As a preferred embodiment of the method for monitoring the fuse status of a jet-type fuse according to the present invention, the edge terminal is powered by a power harvesting module and a battery, and enters low-power sleep mode during non-collection periods. When the state is determined to be pre-fuse or fuse-out, the edge terminal immediately sends the status monitoring signal to the monitoring platform in a reporting priority manner.

[0017] The beneficial effects of this invention are as follows: This invention significantly improves the accuracy and reliability of circuit breaker status monitoring through multi-source data fusion and adaptive processing mechanisms.

[0018] This invention employs time-series fusion and dynamic feature extraction to effectively identify slowly changing anomalies caused by fuse aging, such as temperature rise residuals and changes in energy integral, thereby providing early warnings during the pre-fuse stage and preventing sudden failures. An adaptive threshold adjustment mechanism, combined with environmental and load factors, enhances the robustness of the monitoring system, reduces false alarms caused by external interference (such as temperature fluctuations or load peaks), and ensures stable operation even under harsh conditions. Furthermore, local processing at the edge terminal and low-power design reduce communication dependence and energy consumption, supporting long-term deployment in remote areas and improving power grid operation and maintenance efficiency.

[0019] This invention achieves continuous monitoring of the health status of fuses through intelligent algorithm optimization, extends the service life of equipment, and provides more accurate fault protection capabilities for power distribution systems. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation on the scope of this application.

[0021] Figure 1 This is a flowchart illustrating the method for monitoring the melting status of the jet-type fuse in the embodiment. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0023] All terms used in this application (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0024] For example, the terms “first” and “second” used in this application are only used to distinguish and describe similar objects, to differentiate the first object from another object, and are not used to describe a specific order or sequence, nor should they be interpreted as indicating or implying relative importance.

[0025] This application proposes a method for monitoring the fusing condition of a jet-type fuse, combined with... Figure 1 As shown, the method includes: Step S1: Collect multi-source operating data of the fuse. The multi-source operating data includes at least electrical quantity data and physical status data. In this embodiment, the engineering meaning of multi-source operating data is as follows: electrical quantity data refers to the synchronous sequence of line current and fuse voltage obtained from existing measurement channels; physical state data refers to the temperature and mechanical angle sequences obtained from the shell-attached temperature sensor and inertial measurement unit. The default sampling period is one-tenth of a second, which can be adjusted within the range of 1 / 10th of a second to one-twentieth of a second, determined based on the on-site communication and storage capabilities, as well as the verification results of short-term impact capture. Temperature sampling is one second by default, and angle sampling is one-twentieth of a second by default. If a certain type of data is temporarily unavailable, only other acquired data of the same type will be used for subsequent processing, and will be automatically merged after recovery; if the missing measurement gap does not exceed a few seconds, it will be filled by interpolation according to the time neighborhood; if it exceeds the upper limit, the window will be marked as low confidence to suppress false alarms.

[0026] Step S2 involves performing time alignment, denoising, and sliding window analysis on the multi-source operational data to extract a comprehensive feature vector. Specifically, time alignment uses timestamps as a reference to perform unified time base resampling on different channel sequences, and employs one-sided interpolation at start and end boundaries to avoid phase distortion. Denoising prioritizes a combination of median and low-pass noise types that are insensitive to pulse interference, with a default length of no less than two sampling intervals and no more than one grid fundamental frequency cycle. The window length and step size are provided by the subsequent adaptive process. If clock drift or jitter exceeding equipment specifications is detected during operation, forced resynchronization is performed using the most recent successful alignment time as the anchor point, and the weights are reduced within this window to avoid abrupt changes in the health score.

[0027] Step S3: Generate a dynamic health score for the fuse based on the comprehensive feature vector; Step S4: Compare the dynamic health score with the adaptive threshold to obtain the status monitoring signal and output it to the monitoring platform. The status monitoring signal includes at least the normal status and the pre-circuit breaker warning. In one embodiment, electrical quantity data includes at least one or more of line current and voltage across the fuse; Physical condition data include at least one or more of the following: fuse housing temperature and mechanical angle. In one embodiment, the window length and step size of the sliding window analysis are adaptively set based on the operating load and environmental conditions. The comprehensive feature vector includes at least one of the following: mean square value of current / voltage, slope and fluctuation, energy integral, equivalent resistance change rate, temperature rise residual, angle change rate, and attitude stability index. For example, the attitude stability index is a combined statistical measure of the dispersion and abrupt change frequency of the angle sequence within the window, used to distinguish stable closure, jitter, and drop trends; the energy integral is used to characterize the intensity of heat accumulation, facilitating the identification of continuous overload scenarios; the equivalent resistance change rate measures the slow degradation of the conductive circuit by the relative change between adjacent windows. By default, the above statistics are normalized to the same dimension space and tuned once with empirical weights, and then fine-tuned quarterly or by cumulative operating hours during the self-calibration phase of equipment operation.

[0028] The methods for calculating the comprehensive feature vector include: Step C1: In the sliding analysis, set the window and step size according to load and environmental fluctuations: , , in, Indicates the first Window length, Indicates the first The step size of each slide Minimum window length, The scaling factor. This is the step size scaling factor, dimensionless. The loading factor is dimensionless. For the first The RMS current within the window, expressed in amperes (A). Rated current, in amperes (A). An environmental index, dimensionless. This is the floor operator; For window index, subscript Indicates the rated value; overline This represents the RMS or mean within the corresponding window, as explained by the variable meaning here. Refers to RMS; Step C2: Extract heat accumulation correlation features within each adaptive window: , in, For the first The energy integral of the window, in units of , For the discrete current sequence, the first One sample, in units of A. The sampling period is expressed in seconds. For the first Window starting sample number, dimensionless, subscript For sample index; Step C3: Model the temperature with a first-order smoothed electro-thermal baseline and generate residuals: , in, For the first Window temperature rise residual, in °C. For the shell temperature at the The mean value of the window, in °C. The baseline temperature estimate for the previous window, in °C. The smoothing coefficient is dimensionless. For the first Average ambient temperature at the window, in °C. This is the electro-thermal gain coefficient, in units of... , The square of the current in the first place The mean of the window, in units of Subscript Indicates temperature-related quantities, subscript The equivalent resistance is represented as an electrical sub-characteristic, calculated as the ratio of the voltage RMS to the current RMS, and added to the original feature set. Optionally, to avoid instability caused by an excessively small denominator under low-load conditions, the equivalent resistance is only calculated when the effective current value is several times higher than the equipment's measurement resolution. When this condition is not met, the component is held at the most recent effective value and marked as frozen, while a lower weight is assigned to this component during the normalization stage of the comprehensive feature vector. If short-term saturation or packet loss occurs in voltage sampling, the statistical value of the most recent effective segment is used to fill the gap across the window, and the confidence of the window is reduced with an additional penalty in the subsequent health scoring stage.

[0029] Step C4: Dimensionless transformation is performed on the robust baseline, and task weights are applied to obtain the comprehensive feature vector: , in, For the first The comprehensive feature vector of the window, dimension Dimensionless For the first The original feature vector of the window (concatenated with steps C2, C3 and other statistics). The baseline median array for each component, and Same dimension, Let be the scale vector of each component, and be ... Same dimension, For the first The weights of each feature component are dimensionless. The feature dimension (positive integer). For feature component index, For diagonal matrix operators, It is the reciprocal of each element; Step C5: The slope and volatility are obtained using the least squares straight line slope and the detrended standard deviation within the window. The equivalent resistance change rate, angle change rate, and attitude stability are formed into sub-vectors by taking the difference between adjacent windows and the variance family index. Finally, the electrical, thermal, and mechanical subsets are trimmed and concatenated as needed. Similarly, the environmental index is obtained by mapping the ambient temperature and relative humidity at the installation point to a range of 0-1 based on historical quantile intervals, and applying saturation limits under extreme weather conditions to ensure numerical stability. The initial assignment of task weights can be set to a decreasing distribution based on the feature contribution ranking of historical data, with each subset retaining at least a few dominant components by default to avoid overfitting. When a subset is found to have long-term missing data or abnormal fluctuations, the weight of that subset in the comprehensive feature vector is reduced proportionally, and then gradually increases smoothly after data recovery to ensure scoring continuity.

[0030] Specifically, a feature extraction chain is constructed here; the adaptive window is driven by load factor and environmental index, which automatically encrypts data fragments under heavy load and high environmental thermal pressure, while reducing the computational and communication burden during the stable phase; the energy integral quantity adopts the I²t approach to reflect heat accumulation, which is closely related to thermal degradation and is suitable for overcurrent scenarios of jet-type fuses; the temperature rise residual is referenced by the electro-thermal first-order smoothing model, which can reflect the deviation of slow aging and short-term anomalies, and has few parameters and simple calculation, making it suitable for real-time operation at the edge; in the normalization and weighting stage, a dimensionless space is constructed based on robust statistics, and the influence of different features in subsequent scoring is adjusted by diagonal weights, which is convenient for online fine-tuning; non-core statistics are used as a supplement to provide sensitive quantities for mechanical triggering and external disturbances; furthermore, the feature extraction chain is executed as a timed task on the edge device, with the default update time consistent with the window step size; when the device is in power-limited or bandwidth-limited mode, the three main features of energy integral quantity, temperature rise residual and attitude stability are retained first, and the update frequency of other components is reduced to control resource consumption. For installation points with frequent abnormal jitter, the window overlap can be increased as needed to enhance robustness, while only compressed key statistics are uploaded on the reporting side to reduce the link load.

[0031] In one embodiment, the dynamic health score establishes a baseline model based on historical operating data and performs trend analysis. The trend analysis includes modeling and updating long-term gradual changes and periodic disturbances. In this embodiment, the baseline model uses historical normal period data as samples, obtains a reference trajectory through seasonal and load-based stratified modeling, and is updated on a rolling basis according to a fixed period or cumulative operating hours. The default baseline window covers at least one week of continuous data, with a recommended range of one week to one month. The update cycle is once a day by default. The modeling of periodic disturbances can be achieved by estimating the amplitude and phase of the fundamental wave and a small number of its lower harmonics. If the site load regularity is weak, it automatically degenerates to an aperiodic assumption and only retains the gradual change term.

[0032] In one embodiment, the adaptive threshold is dynamically adjusted based on ambient temperature, ambient humidity and current load level, and hysteresis and rate of change limits are set to suppress threshold jitter. The steps involved in adaptive threshold adjustment based on environment and load include: Step D1: Use robust statistics of health scores as a threshold baseline: , in, Indicates the first The threshold baseline of the window is dimensionless. Indicates the first The previously selected set of historical health scores is dimensionless. The scaling factor is dimensionless. For the sample median operator, For absolute median difference operator; Step D2: Overlay environmental and load effects onto the baseline to form an initial adjustment threshold: , in, For the first The threshold for adjusting the window once, dimensionless. These are the influence coefficients for temperature, humidity, and load, respectively; they are dimensionless. For the first The normalized quantity of the ambient temperature at the window, dimensionless. For the first Normalized measure of ambient humidity at the window, dimensionless. The loading factor is dimensionless. For the first Window current RMS, in amperes (A). Rated current, unit is A, subscript Indicates environmental quantity, subscript Indicates the rated value; and Map the installation point's experience range to the interval. ; Step D3: Set the hysteresis band for different entry criteria for uplink and downlink: , in, This represents the uplink trigger threshold, which is dimensionless. This represents the downlink release threshold, which is dimensionless. Hysteresis half-amplitude, dimensionless, subscript They represent upward and downward movement, respectively. Step D4: Limit the rate of change of the threshold in adjacent windows to obtain the final threshold: , in, For the first The final threshold of the window, dimensionless. The final threshold of the previous window, dimensionless. This represents the maximum rate of change in the upward direction (the dimensionless increment per window). This represents the maximum rate of change during the downward movement (the dimensionless decrease per window). To be Cut off to interval Operators; The confidence level used to limit false alarms during the healthy period can be determined by a grid search on historical normal samples, ensuring that the pre-circuit breaker trigger rate meets operational goals. The values ​​were obtained from environmental binning regression, and positive coefficients were selected to make the threshold more sensitive to high temperature and humidity. The scale and normalization method were matched. This data is obtained from load stratification statistics and reflects the need for early warning under heavy load conditions. Related to the noise scale, it can be set as a proportion of the steady-state fluctuation of the health score within a short window. Based on the minimum discernible time constant settings for two scenarios—slow drift and rapid deterioration—the uplink time constant is set slightly longer than the downlink time constant to better track anomalies. The upper and lower limits of the interval mapping are derived from the annual statistical quantile of the installation point; finally, the above coefficients are updated and written to the local configuration at the edge through periodic self-calibration; for example, the self-calibration process is triggered during low-load periods at night, and is executed once every 24 hours by default, and the alarm situation of the past day is verified before execution to avoid learning deviation parameters during abnormal periods; the hysteresis half amplitude is recommended to be several times the steady-state fluctuation of the health score, and the commonly used range is 2 to 3 times; the upward value of the rate of change limit can be slightly higher than the downward value, so as to follow more sensitively when the anomaly develops rapidly, and to avoid releasing too quickly during the recovery phase. If environmental and load statistics drift across seasons, the baseline window is automatically extended until the new stable segment is covered before the parameters are fixed.

[0033] Specifically, this threshold scheme uses robust statistics to build a baseline, avoiding the influence of extreme samples. On this basis, environmental and load effects are superimposed, allowing the threshold to adaptively shift upward or downward according to operating conditions, thus conforming to the actual risk level of the device under high temperature, high humidity, and heavy load scenarios. The hysteresis band provides a separation criterion for entry and release, suppressing jitter and frequent switching near the boundary. The rate of change limit constrains the threshold's transition amplitude in adjacent windows, which can both follow slow drift and avoid sharp adjustments caused by short-term disturbances. The parameter selection adopts a hierarchical regression and quantile calibration path based on historical data of the site, which is easy to migrate to different installation points. The normalization interval is taken from long-term statistics to reduce the impact of cross-seasonal differences. In one embodiment, the status monitoring signal includes one or more of five states: normal, pre-circuit break, circuit breaker, drop, and abnormal. The pre-circuit break state is generated when the dynamic health score exceeds a limit and the duration meets a criterion. Similarly, the duration criterion is, in engineering practice, taken as the lower limit of the number of adjacent windows, defaulting to no less than two consecutive windows and no more than a certain number of windows. In cases of consecutive limit exceedances but interruptions shorter than one step, the events are merged into a single event. To reduce boundary jitter, warning cancellation requires the score to be below the release threshold and a grace period of at least one window to be maintained. When a warning repeatedly occurs within a limited time and reaches a set number of times, the event level can be upgraded to a high-priority warning and reported immediately.

[0034] In one embodiment, the mechanical angle is acquired by an inertial measurement unit (IMU), which detects changes in the mechanical position of the fuse and triggers candidate event marking when a rapid angle change is detected. Optionally, the IMU performs a static calibration after installation to determine the gravity reference direction, and during operation, it uses attitude estimation combining low-pass and amplitude limiting to suppress vibration effects. The trigger threshold for rapid angle changes can be adjusted according to the device type and installation method, typically ranging from tens of degrees to several tens of degrees of equivalent transformation, supplemented by a minimum holding time to filter out instantaneous jitter. If the angular velocity or angular acceleration sensing channel is temporarily unavailable, candidate event marking is completed only using the angle channel, and the weight of this component is reduced accordingly during the scoring phase.

[0035] In one embodiment, abnormal fluctuations are identified in multi-source operational data. This identification includes robust statistical discrimination of short-term shocks and external interference, the results of which are used to correct the dynamic health score. Further, the abnormal fluctuation identification first calculates robust estimates of steady-state fluctuations within each channel, then identifies short-term shocks using a fixed-multiple threshold. Identified samples participate in scoring within their respective windows using either weight reduction or truncation. A low-quality data state is triggered when widespread anomalies occur across multiple consecutive windows, preventing the misinterpretation of external factors such as communication jitter and strong electromagnetic interference as equipment degradation. Anomaly markers are gradually removed after the data stabilizes, avoiding long-tail effects on subsequent scoring.

[0036] In one embodiment, the method is performed on an edge terminal installed near the fuse, the edge terminal having local storage and wireless communication capabilities, the wireless communication being one or more of cellular, NB-IoT, or LoRa; In one embodiment, the edge terminal uses a power harvesting module and a battery for power supply, and enters low-power sleep mode during non-collection periods to extend battery life. When the edge terminal is determined to be in a pre-fuse or fuse-out state, it immediately sends a status monitoring signal to the monitoring platform in a reporting priority manner. In this embodiment, priority reporting is implemented through an independent queue and a retry strategy. In the event of immediate transmission failure, exponential backoff is used, and the most recent critical events are retained in a local cache queue. When the cache is full, non-critical heartbeat information is discarded in a time-priority manner. To save energy, a low duty cycle communication strategy is adopted during non-event periods. After entering a high-priority warning, the reporting frequency is temporarily increased until the event is resolved or the upper limit duration is reached. If communication is unavailable for an extended period, the status indication is maintained locally and re-reported all at once after the link is restored.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0038] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of this application and form different embodiments. For example, all the embodiments above can be used in any combination. The information disclosed in this background section is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

Claims

1. A method of monitoring the blown state of a projection fuse, characterized by, include: Step S1: Collect multi-source operating data of the fuse, wherein the multi-source operating data includes at least electrical quantity data and physical status data; Step S2: Perform time alignment, noise reduction, and sliding window analysis on the multi-source running data to extract a comprehensive feature vector; Step S3: Generate a dynamic health score for the fuse based on the comprehensive feature vector; Step S4: Compare the dynamic health score with the adaptive threshold to obtain a status monitoring signal and output it to the monitoring platform. The status monitoring signal includes at least a normal status and a pre-circuit breaker warning.

2. The method of claim 1, wherein the method comprises: The electrical quantity data includes at least one or more of the line current and the voltage across the fuse. The physical state data includes at least one or more of the following: fuse housing temperature and mechanical angle.

3. The method for monitoring the fusing status of a jet-type fuse as described in claim 1, characterized in that, The window length and step size of the sliding window analysis are adaptively set according to the operating load and environmental conditions. The comprehensive feature vector includes at least one of the following: mean square value of current / voltage, slope and fluctuation, energy integral, equivalent resistance change rate, temperature rise residual, angle change rate and attitude stability index. The calculation method for the comprehensive feature vector includes: Step C1: In the sliding analysis, set the window and step size according to load and environmental fluctuations: ， ， in, Indicates the first Window length, Indicates the first The step size of each slide Minimum window length, The scaling factor. This is the step size scaling factor, dimensionless. The loading factor is dimensionless. For the first The RMS current within the window, expressed in amperes (A). Rated current, in amperes (A). This is an environmental index, dimensionless. This is the floor operator; For window index, subscript Indicates the rated value; overline This represents the RMS or mean value within the corresponding window. RMS refers to the effective value of the current sequence within the k-th sliding window, i.e., the root mean square value, in amperes (A). Step C2: Extract heat accumulation correlation features within each adaptive window: ， in, For the first The energy integral of the window, in units of , For the discrete current sequence, the first One sample, in units of A. The sampling period is expressed in seconds. For the first Window starting sample number, dimensionless, subscript For sample index; Step C3: Model the temperature with a first-order smoothed electro-thermal baseline and generate residuals: ， in, For the first Window temperature rise residual, in °C. For the shell temperature at the The mean value of the window, in °C. The baseline temperature estimate for the previous window, in °C. The smoothing coefficient is dimensionless. For the first Average ambient temperature at the window, in °C. This is the electro-thermal gain coefficient, in units of... , The square of the current in the first place The mean of the window, in units of Subscript Indicates temperature-related quantities, subscript Indicates environmental quantity; Step C4: Dimensionless transformation is performed on the robust baseline, and task weights are applied to obtain the comprehensive feature vector: ， in, For the first The comprehensive feature vector of the window, dimension Dimensionless For the first The original feature vector of the window, The baseline median array for each component, and Same dimension, Let be the scale vector of each component, and be... Same dimension, For the first The weights of each feature component are dimensionless. For feature dimension, For feature component index, For diagonal matrix operators, It is the reciprocal of each element; Step C5: The slope and volatility are obtained using the least squares straight line slope and the detrended standard deviation within the window. The equivalent resistance change rate, angle change rate, and attitude stability are formed into sub-vectors by taking the difference between adjacent windows and the variance family index. Finally, the electrical, thermal, and mechanical subsets are trimmed and concatenated as needed. .

4. The method for monitoring the fusing status of a jet-type fuse as described in claim 1, characterized in that, The dynamic health score establishes a baseline model based on historical operational data and performs trend analysis, which includes modeling and updating long-term gradual changes and periodic disturbances.

5. The method for monitoring the fusing status of a jet-type fuse as described in claim 1, characterized in that, The adaptive threshold is dynamically adjusted based on ambient temperature, ambient humidity and current load level, and hysteresis and rate of change limits are set. The step of adjusting the adaptive threshold according to environment and load includes: Step D1: Use robust statistics of health scores as a threshold baseline: ， in, Indicates the first The threshold baseline of the window is dimensionless. Indicates the first The previously selected set of historical health scores is dimensionless. The scaling factor is dimensionless. For the sample median operator, For absolute median difference operator; Step D2: Overlay environmental and load effects onto the baseline to form an initial adjustment threshold: ， in, For the first The threshold for adjusting the window once, dimensionless. These are the influence coefficients for temperature, humidity, and load, respectively; they are dimensionless. For the first The normalized quantity of the ambient temperature at the window, dimensionless. For the first Normalized measure of ambient humidity at the window, dimensionless. The loading factor is dimensionless. For the first Window current RMS, in amperes (A). Rated current, unit is A, subscript Indicates environmental quantity, subscript Indicates the rated value; Step D3: Set the hysteresis band for different entry criteria for uplink and downlink: ， in, This represents the uplink trigger threshold, which is dimensionless. This represents the downlink release threshold, which is dimensionless. Hysteresis half-amplitude, dimensionless, subscript They represent upward and downward movement, respectively. Step D4: Limit the rate of change of the threshold in adjacent windows to obtain the final threshold: ， in, For the first The final threshold of the window, dimensionless. The final threshold of the previous window, dimensionless. The maximum rate of change in the upward direction. The maximum rate of change in the downward direction. To be Cut off to interval The operator.

6. The method for monitoring the fusing status of a jet-type fuse as described in claim 1, characterized in that, The status monitoring signal includes one or more of the following five states: normal, pre-fuse, fuse, drop, and abnormal; wherein, the pre-fuse state is generated when the dynamic health score exceeds the limit and the duration meets the criterion.

7. The method for monitoring the fusing status of a jet-type fuse as described in claim 2, characterized in that, The mechanical angle is acquired by an inertial measurement unit, which is used to detect changes in the mechanical position of the fuse and trigger candidate event markers when a rapid angle change is detected.

8. The method for monitoring the fusing status of a jet-type fuse as described in claim 4, characterized in that, The multi-source operational data is subjected to abnormal fluctuation identification, which includes robust statistical discrimination of short-term shocks and external interference, and the results are used to correct the dynamic health score.

9. The method for monitoring the fusing status of a jet-type fuse as described in claim 1, characterized in that, The method is performed on an edge terminal installed near the fuse, the edge terminal having local storage and wireless communication capabilities, the wireless communication being one or more of cellular, NB-IoT, or LoRa.

10. The method for monitoring the fusing status of a jet-type fuse as described in claim 9, characterized in that, The edge terminal is powered by a combination of an energy harvesting module and a battery, and enters a low-power sleep mode during non-collection periods. When the state is determined to be pre-fuse or fuse-out, the edge terminal immediately sends the status monitoring signal to the monitoring platform in a reporting priority manner.

Images