Magnetic control type pole-mounted circuit breaker mechanical fault diagnosis system based on coil current characteristics

Abstract

This invention relates to the field of circuit breaker fault diagnosis technology, specifically disclosing a mechanical fault diagnosis system for magnetically controlled pole-mounted circuit breakers based on coil current characteristics. The system includes a feature extraction and analysis module, used to acquire coil current characteristic parameters of the magnetically controlled pole-mounted circuit breaker, and to perform environmental decoupling based on the correlation between the coil current characteristic parameters, real-time voltage, and ambient temperature to extract core features. In actual power grid operation, this invention first decouples the coil current characteristics from the environment by correlating real-time voltage and ambient temperature, eliminating data pollution caused by external operating condition fluctuations and obtaining core features that purely reflect the mechanical physical state.

Description

Technical Field

[0001] This invention relates to the field of circuit breaker fault diagnosis technology, and in particular to a mechanical fault diagnosis system for magnetically controlled pole-mounted circuit breakers based on coil current characteristics. Background Technology

[0002] Existing mechanical fault diagnosis for magnetically controlled pole-mounted circuit breakers typically relies on collecting single coil current data and directly inputting the raw data into a fixed diagnostic model, or simply comparing it with a preset static threshold, in order to determine whether the equipment has a fault risk and arrange a maintenance plan accordingly.

[0003] However, in actual outdoor operation, the coil current characteristics of circuit breakers will experience normal physical deviations due to grid voltage fluctuations and changes in ambient temperature. Furthermore, the natural aging of the equipment over time will cause the operating baseline to continuously change. This means that if a single, fixed judgment standard is used, and the sensitivity of the diagnostic system is increased to proactively prevent minor mechanical problems, then parameter fluctuations caused by environmental temperature and pressure interference and normal equipment aging will frequently break through the diagnostic threshold, triggering numerous false alarms and wasting maintenance resources.

[0004] However, if the judgment criteria are relaxed in order to reduce the false alarm rate, it is very easy to miss early-stage real mechanical wear or jamming faults, leading to serious power outages. The lack of means to isolate environmental interference and the lack of a dynamic secondary screening mechanism for suspected faults mean that the existing diagnostic methods cannot ensure both the timeliness and accuracy of early warnings under actual complex operating conditions. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a mechanical fault diagnosis system for magnetically controlled pole-mounted circuit breakers based on coil current characteristics, in order to improve the efficiency and safety of power grid operation and maintenance.

[0006] To achieve the above objectives, a first aspect of the present invention provides a mechanical fault diagnosis system for a magnetically controlled pole-mounted circuit breaker based on coil current characteristics, comprising:

[0007] The feature extraction and analysis module is used to obtain the coil current characteristic parameters of the magnetically controlled pole-mounted circuit breaker, and to perform environmental decoupling based on the correlation between the coil current characteristic parameters, real-time voltage and ambient temperature to extract the core features.

[0008] The fault diagnosis and trend prediction module is used to predict the probability of fault occurrence based on the core features and the dynamic aging coefficient of the equipment; in response to the probability of fault occurrence being greater than or equal to a preset probability threshold, a verification process is triggered; in the verification process, the weight parameters of the core features corresponding to the fault type in the prediction model are increased, and the fault judgment is re-executed in combination with the dynamic threshold generated based on the historical fault assessment index, so as to output a comprehensive equipment status conclusion.

[0009] The predictive maintenance decision engine module is used to generate a predictive maintenance decision scheme that includes maintenance timing and resource allocation scheme based on the overall status conclusion of the equipment and preset constraints.

[0010] To achieve the above objectives, a second aspect of the present invention proposes a method for diagnosing mechanical faults in a magnetically controlled pole-mounted circuit breaker based on coil current characteristics. The method includes:

[0011] The coil current characteristic parameters of the magnetically controlled pole-mounted circuit breaker are obtained, and environmental decoupling is performed based on the correlation between the coil current characteristic parameters, real-time voltage, and ambient temperature to extract core features. Based on the core features and the equipment dynamic aging coefficient, the probability of fault occurrence is predicted. In response to the probability of fault occurrence being greater than or equal to a preset probability threshold, a verification process is triggered. In the verification process, the weight parameters of the core features corresponding to the fault type in the prediction model are increased, and the fault judgment is re-executed in combination with a dynamic threshold generated based on historical fault assessment indices to output a comprehensive equipment status conclusion. Based on the comprehensive equipment status conclusion and preset constraints, a predictive maintenance decision scheme including maintenance timing and resource allocation schemes is generated.

[0012] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein when the computer program is executed by the processor, it implements the above-described method for mechanical fault diagnosis of a magnetically controlled pole-mounted circuit breaker based on coil current characteristics.

[0013] In actual power grid operation, the technical solution of this invention first decouples the coil current characteristics from the environment by associating real-time voltage with ambient temperature, eliminating data pollution caused by external operating condition fluctuations and obtaining core characteristics that purely reflect the mechanical and physical state.

[0014] Subsequently, after predicting initial risks based on the equipment's dynamic aging coefficient, the solution innovatively initiated a closed-loop verification process for suspected fault stages. In actual operation, when the system detects early, subtle fault signs, it automatically amplifies the weight parameters of the features corresponding to the suspected fault and retrieves dynamic thresholds matching historical operating states for rigorous secondary cross-verification. This multi-level adaptive diagnostic mechanism, which involves early environmental decoupling and noise reduction, mid-term dynamic aging correction, and late-stage weighted re-verification, successfully resolves the technical conflict between high sensitivity in capturing early hidden dangers and the high false alarm rate easily caused by complex environments in practical applications, significantly improving the accuracy and reliability of the final output equipment status conclusions.

[0015] Ultimately, the system automatically generates maintenance decisions based on these decisions and on-site constraints. This directly guides maintenance personnel to conduct accurate predictive condition-based maintenance, effectively avoiding blind outages and over-maintenance, and significantly improving the efficiency and safety of power grid operation and maintenance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the implementation architecture of the magnetically controlled pole-mounted circuit breaker mechanical fault diagnosis system based on coil current characteristics provided by the present invention.

[0017] Figure 2 This is a comparison chart of the original coil current waveform and the data before and after Butterworth filtering and outlier removal processing in the mechanical fault diagnosis system for magnetically controlled pole-mounted circuit breakers based on coil current characteristics provided by this invention.

[0018] Figure 3 This invention provides a three-dimensional surface diagram showing the decoupled relationship between coil current characteristics and temperature and pressure environment in a magnetically controlled pole-mounted circuit breaker mechanical fault diagnosis system based on coil current characteristics, using multiple linear regression.

[0019] Figure 4 This invention provides an adaptive dynamic threshold following and fault over-limit trend diagram based on the equipment dynamic aging coefficient in the mechanical fault diagnosis system of magnetically controlled pole-mounted circuit breaker based on coil current characteristics.

[0020] Figure 5 This is a schematic diagram of the transient current waveform and Joule integral energy region of the coil under extreme short-circuit breaking conditions in the mechanical fault diagnosis system of a magnetically controlled pole-mounted circuit breaker based on coil current characteristics provided by the present invention.

[0021] Figure 6 This invention provides a nonlinear reconstructed step response diagram of the dynamic aging coefficient weight allocation of the equipment before and after extreme operating conditions in the mechanical fault diagnosis system of a magnetically controlled pole-mounted circuit breaker based on coil current characteristics.

[0022] Figure 7This is a multi-axis curve diagram of environmental phase change isolation and cross-cycle aging compensation baseline adaptive migration under extreme cold wave climate in the mechanical fault diagnosis system of magnetically controlled pole-mounted circuit breaker based on coil current characteristics provided by the present invention.

[0023] Figure 8 This is a flowchart illustrating the mechanical fault diagnosis method for magnetically controlled pole-mounted circuit breakers based on coil current characteristics provided by the present invention.

[0024] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0026] The mechanical fault diagnosis system, method, and electronic device of the magnetically controlled pole-mounted circuit breaker based on coil current characteristics according to embodiments of the present invention are described below with reference to the accompanying drawings.

[0027] Example 1:

[0028] Magnetic pole-mounted circuit breakers play a crucial role in controlling and protecting lines in power distribution networks. The mechanical performance of their internal operating mechanisms directly determines whether the circuit breaker can reliably perform opening and closing actions. This embodiment provides a mechanical fault diagnosis system for magnetic pole-mounted circuit breakers based on coil current characteristics. This system aims to achieve early diagnosis and predictive maintenance of faults in the mechanical operating mechanism of the circuit breaker by collecting and deeply analyzing the current characteristics of the circuit breaker's operating coil, combined with multi-dimensional operating environment data.

[0029] like Figure 1 As shown, the system in this embodiment includes: a multi-source data acquisition and preprocessing module, a feature extraction and analysis module, a fault diagnosis and trend prediction module, and a predictive maintenance decision engine module. The modules interact and transmit commands through an internal data bus or network communication protocol, collectively forming a closed-loop fault diagnosis and decision execution system.

[0030] It is important to note that this system employs an architecture that includes both edge devices and a cloud platform. Edge devices are deployed in intelligent distribution terminals or local control boxes near the magnetically controlled pole-mounted circuit breakers, while the cloud platform is deployed in the power system's dispatch center or remote data center. A secure bidirectional data transmission channel is established between the edge devices and the cloud platform via wireless or wired communication networks. The edge devices are equipped with lightweight predictive models for real-time data acquisition, preprocessing, and preliminary local fault probability calculations. The cloud platform deploys a complete trend prediction model and a predictive maintenance decision engine module. In response to the edge device calculating a fault probability greater than or equal to a preset probability threshold, the data is uploaded to the cloud platform in real time, where it performs high-precision prediction and verification. This edge-cloud collaborative architecture ensures real-time data processing at the field, reduces bandwidth consumption under normal conditions, and leverages the powerful computing capabilities of the cloud to provide high-precision diagnostics and decision support at critical moments.

[0031] For example, for the front-end data access of the system, the multi-source data acquisition and preprocessing module is used to: acquire the original coil current, ambient temperature and real-time voltage data; sequentially perform low-pass filtering, cyclic redundancy check code verification and sliding window outlier filtering, and then perform data normalization processing to obtain preprocessed data, which is then used by the feature extraction and analysis module to obtain coil current characteristic parameters.

[0032] In actual power system operating environments, the signals acquired by sensors are often accompanied by a large amount of electromagnetic interference, high-frequency harmonics, and random bit errors during communication transmission. Therefore, the multi-source data acquisition and preprocessing module first acquires the original coil current through a high-frequency Hall current sensor arranged in the circuit breaker control circuit, acquires the real-time voltage data of the operating power supply through a voltage transformer, and acquires the ambient temperature inside or outside the circuit breaker enclosure through a patch-type or ambient temperature probe.

[0033] After acquiring the aforementioned raw multi-source data, low-pass filtering is performed sequentially. This low-pass filtering can employ a Butterworth low-pass filter, with a reasonable cutoff frequency set to filter out high-frequency white noise coupled into the signal loop from the switching power supply or external high-frequency electromagnetic fields. Next, cyclic redundancy check (CRC) is performed to verify whether there are packet losses or bit flip errors in the data packet during analog-to-digital conversion and initial transmission. If the check fails, the data frame is discarded and a retransmission is requested, or interpolation is used to repair it. Then, sliding window outlier filtering is performed. A fixed-length data window slides across the time series, calculating the statistical characteristics of the data within the window. When the data at a new sampling point deviates from the main distribution within the window by more than a set multiple, it is identified as a transient spike and removed, thus smoothing the data sequence.

[0034] like Figure 2This diagram illustrates the comparison of data from the multi-source data acquisition and preprocessing module before and after Butterworth low-pass filtering and sliding window outlier removal of the raw coil current in a harsh industrial electromagnetic environment. The horizontal axis represents sampling time in milliseconds, and the vertical axis represents current values ​​in amperes.

[0035] The gray curve in the figure represents the original coil current waveform acquired by the sensor. It can be observed that within the sampling period of 0 to 100 milliseconds, the original waveform is not only superimposed with dense high-frequency white noise spikes, but also has transient abnormal spike pulses with values ​​as high as 8.5 amperes and 7.2 amperes at positions such as 15 milliseconds and 45 milliseconds, respectively. These interferences are caused by communication random bit errors or external high-frequency electromagnetic field coupling.

[0036] The blue curve in the figure represents the waveform after preprocessing. By comparing the waveform transformations of the two curves, it can be found that the blue curve retains the physical peak value of 5 amps around 30 milliseconds when the circuit breaker coil operates, as well as the subsequent holding current characteristic of stabilizing at 2 amps, while smoothing out background high-frequency noise and eliminating spike pulses.

[0037] This data graph objectively reflects the actual effect of preprocessing methods in removing electromagnetic interference and bit error contamination, indicating that the processed data can provide a reliable data source guarantee for extracting core features that purely reflect the mechanical physical state.

[0038] It is important to note that, in order to eliminate the impact of gradient explosion or weight imbalance caused by data of different dimensions and orders of magnitude on subsequent machine learning model training and numerical calculation, the multi-source data acquisition and preprocessing module then performs data normalization to obtain preprocessed data.

[0039] The data normalization process is performed using the following calculations:

[0040] ;

[0041] in, For the normalized data, This refers to the original individual data after filtering outliers using a sliding window. It is the minimum value of the corresponding class of data in the historical records. This represents the maximum value of the corresponding data type in the historical records. This is a preset minimum positive constant to prevent zero, such as a value of .

[0042] The above formula can be used to map all parameters with different physical dimensions and value ranges, such as current, voltage, and temperature, to a dimensionless standard range of 0 to 1.

[0043] After data preprocessing is completed, the feature extraction and analysis module receives the preprocessed data. Specifically, the feature extraction and analysis module is used to obtain the coil current characteristic parameters of the magnetically controlled pole-mounted circuit breaker, and to perform environmental decoupling based on the correlation between the coil current characteristic parameters, real-time voltage, and ambient temperature to extract the core features.

[0044] In actual circuit breaker operation, the current waveform generated by the coil during operation, such as starting current, peak current, steady-state current, and operating time, is not only affected by mechanical physical conditions such as friction and jamming, but also greatly influenced by real-time voltage and ambient temperature at the moment of operation. For example, increased ambient temperature leads to increased DC resistance of the coil's copper conductors, thereby reducing the steady-state current amplitude; a drop in real-time control voltage causes a decrease in the rate of change of the driving magnetic flux, resulting in a slower current waveform rise. If these normal fluctuations caused by environmental and electrical conditions are not properly isolated, they can easily be misdiagnosed as faults caused by increased mechanical friction.

[0045] Therefore, in the feature extraction and analysis module, the correlation is a multivariate linear relationship, and the specific process of environmental decoupling is calculated using the following two formulas:

[0046] ;

[0047] in, These are estimated values ​​for the coil current characteristic parameters. For real-time voltage, This is the real-time ambient temperature output by the multi-source data acquisition and preprocessing module. The first corresponding coefficient associated with voltage. This is the second corresponding coefficient related to temperature. This is the third corresponding coefficient, which serves as a constant compensation term.

[0048] The aforementioned coefficients can be pre-calibrated using the basic data accumulated from repeated operation tests of circuit breakers under different temperature and pressure conditions in their factory healthy state, through a multiple linear regression algorithm.

[0049] After obtaining the estimated value, the next step is to perform a subtraction operation to obtain the core features:

[0050] ;

[0051] in, These are the core features that have been initially extracted. These are the actual coil current characteristic parameters obtained.

[0052] This subtraction process, in a physical sense, removes the reasonable current bias caused by the change in resistance due to temperature and the change in driving force due to voltage. The remaining numerical difference purely represents the distortion of the current waveform caused by purely mechanical damage such as changes in the coefficient of mechanical friction, deformation of the linkage mechanism, or fatigue of the contact spring.

[0053] like Figure 3 This diagram illustrates the three-dimensional spatial relationship of coil current characteristic parameters decoupled from the environment using a multiple linear regression algorithm in the feature extraction and analysis module. In the figure, the X-axis represents ambient temperature in degrees Celsius, ranging from -20 to +60; the Y-axis represents real-time voltage in volts, ranging from 180 to 240; and the Z-axis represents the coil current characteristic parameter in amperes.

[0054] The smooth, colored 3D surface in the figure represents the theoretical reference surface, reflecting the reasonable bias law of current characteristics fluctuating with external temperature and pressure. The black scattered points in the figure represent actual sampling points. The red vertical solid line segment connecting the actual sampling points and the theoretical reference surface represents the core decoupling feature, and its length quantifies the difference between the actual acquired parameters and the estimated values.

[0055] This three-dimensional geometric analysis operation shows that the system can separate data offsets caused by temperature changes or voltage fluctuations. The extracted vertical difference reflects the waveform changes of the circuit breaker caused by factors such as increased mechanical friction or spring fatigue, which helps to reduce the false alarm rate caused by environmental fluctuations.

[0056] Optionally, the process of extracting core features by the feature extraction and analysis module further includes: calculating the corrected instantaneous power, current change rate, and temperature change rate containing total harmonic distortion, and using them as candidate features; calculating the Pearson correlation coefficient between each candidate feature and the preset fault assessment index; and selecting candidate features whose absolute value of the Pearson correlation coefficient is greater than or equal to the set correlation coefficient threshold, and using them as core features.

[0057] For example, relying solely on the time and amplitude characteristics of a waveform may not be sufficient to fully reflect the complex mechanical degradation process. Therefore, the system calculates the corrected instantaneous power, including total harmonic distortion (THD). At the moment of actuation of the magnetic control mechanism, the nonlinear magnetic circuit causes harmonics in the current, and changes in harmonic components often correspond to abnormal changes in the air gap of the iron core. The corrected instantaneous power, including THD, is calculated using the following formula:

[0058] ;

[0059] in, To correct instantaneous power, For real-time voltage, This refers to the actual coil current obtained. The power factor, which reflects the proportion of active components, The total harmonic distortion of the current is calculated by the fast Fourier transform.

[0060] Simultaneously calculate the rate of change of current and the rate of change of temperature:

[0061] ;

[0062] ;

[0063] in, The rate of change of current, The current at the current moment, The current at the previous sampling time. The sampling time interval, For the rate of temperature change, The temperature at the current moment. The temperature is the temperature at the previous sampling time. The decoupled basic features, along with the corrected instantaneous power, current rate of change, and temperature rate of change, are collectively summarized to construct a candidate feature set.

[0064] To further reduce the computational dimensionality of the prediction model and eliminate redundant noise, the Pearson correlation coefficient between each candidate feature and the preset fault assessment index was calculated. The preset fault assessment index is a specific quantitative score characterizing the severity of mechanical deterioration of the equipment, recorded in historical ledgers, manually labeled by experts, or obtained from offline power outage detection. The Pearson correlation coefficient was calculated using the following formula:

[0065] ;

[0066] in, The Pearson correlation coefficient is used. Let be the value of a candidate feature for the i-th sample in the sequence. This is the average value of the option's feature in the sequence. The preset fault assessment index corresponds to the i-th sample in the sequence. This is the average value of the preset fault assessment index in the sequence.

[0067] The system selects candidate features whose absolute value of the Pearson correlation coefficient is greater than or equal to a set correlation coefficient threshold (e.g., a correlation coefficient threshold of 0.6), discards redundant features with low correlation, and finally uses them as the core features input into the subsequent model.

[0068] Specifically, the fault diagnosis and trend prediction module receives the core features extracted above. Based on the core features and the equipment's dynamic aging coefficient, the fault diagnosis and trend prediction module predicts the probability of a fault occurring. In response to a fault occurrence probability greater than or equal to a preset probability threshold, a verification process is triggered. During the verification process, the weight parameters of the core features corresponding to the fault type in the prediction model are increased, and the fault judgment is re-executed based on a dynamic threshold generated from historical fault assessment indices, to output a comprehensive conclusion on the equipment's status.

[0069] However, relying solely on current sampling characteristics for prediction often overlooks the long-term historical wear and tear cumulative effects experienced by the equipment. Therefore, in the fault diagnosis and trend prediction module, the dynamic aging coefficient of the equipment is calculated using the following formula:

[0070] ;

[0071] in, This refers to the dynamic aging coefficient of the equipment. The correlation coefficient based on the age of manufacture exhibits a monotonically increasing property as the equipment is put into use. As the first preset weighting factor, This is the operating condition correlation coefficient calculated based on the equipment's load rate and the number of short-circuit current impacts it experiences. As the second preset weighting factor, The correlation coefficient for the number of repairs changes as the frequency of recorded inspections and component replacements increases. As the third preset weighting factor, The health correlation coefficient issued by the equipment inspection system. This is the fourth preset weighting factor.

[0072] Based on the obtained dynamic aging coefficient of the equipment, a correction coefficient is generated, and then the preliminary calculated probability of failure is adjusted using the correction coefficient.

[0073] The fault diagnosis and trend prediction module internally deploys forward inference algorithms such as long short-term memory neural network models or support vector machine models. The core features selected above are input into this prediction model. After the model outputs an initial probability of fault occurrence, the system uses the equipment's dynamic aging coefficient to generate a correction coefficient to adjust the initially calculated probability of fault occurrence. This adjustment process uses the following formula:

[0074] ;

[0075] in, This is the final probability of failure after adjustment. The initial probability of failure is calculated by the model. This is a proportional constant used to smooth the adjustment amplitude. The above-derived dynamic aging coefficient of the equipment is given.

[0076] By incorporating aging lifespan correction, the calculated failure probability for the same amplitude waveform distortion is significantly higher for equipment that has been in service for ten years than for new equipment that has been in operation for only six months, which is more in line with the physical wear and tear patterns in industrial settings.

[0077] Subsequently, the system monitors the adjusted probability of fault occurrence. Under normal, stable operation, this probability remains low. If the probability of fault occurrence is greater than or equal to a preset probability threshold (this threshold can be set according to the reliability requirements of the power grid, for example, 0.6 for important lines and 0.75 for general lines), the system determines that the equipment has an extremely high potential for anomalies. To avoid false alarms caused by occasional errors in a single measurement, the system will not immediately issue an alarm but will instead trigger a very rigorous verification process.

[0078] In the verification process, the system first increases the weight parameters of the core features corresponding to the fault type in the prediction model. For example, if the initial prediction tends to identify it as a tripping jam fault, this fault is usually highly correlated with two core features: the total tripping time and the peak time of the coil current. The system will temporarily increase the bias weights of the input nodes containing these two features within the model to enhance the prediction model's sensitivity to the detailed features of this type of fault.

[0079] Meanwhile, to provide a more objective judgment benchmark in cross-validation, the system re-executes fault judgment by combining a dynamic threshold generated based on historical fault assessment indices. Specifically, the process of re-executing fault judgment by combining a dynamic threshold generated based on historical fault assessment indices includes: statistically analyzing historical fault assessment indices within a continuously set time window, calculating the mean and standard deviation of the fault assessment indices; adding the product of a preset adjustment coefficient and the standard deviation of the fault assessment indices to the mean of the fault assessment indices to calculate the dynamic threshold; when the value corresponding to the currently extracted core feature is greater than the dynamic threshold, it is judged as a confirmed fault, and the confirmed result is used as the overall equipment status conclusion.

[0080] The above calculation of the dynamic threshold is performed using the following formula:

[0081] ;

[0082] in, To calculate the generated dynamic threshold, To continuously set the average historical failure assessment index within a time window (such as the past 30 days), To continuously set the standard deviation of historical fault assessment index within a time window, These are preset adjustment coefficients set according to the reliability requirements of the data distribution.

[0083] Compared to the rigid and unchanging static thresholds set at the factory, this dynamic threshold, which is generated based on the historical statistical characteristics of the device itself, can adapt to the reasonable baseline drift caused by the natural and slow aging of the device over time.

[0084] like Figure 4 This demonstrates the process of dynamic threshold adjustment based on equipment uptime in the fault diagnosis and trend prediction module, as well as the characteristic exceeding the limit when a fault occurs. The horizontal axis represents uptime per month, ranging from 0 to 60; the vertical axis represents the characteristic evaluation value.

[0085] In the figure, the red dashed line represents the upper limit of the adaptive dynamic threshold generated by the system, and the blue solid line represents the core feature extraction value. During the stable operation period from 0 to 50 months, with the normal wear and tear of the equipment's transmission components, the core feature extraction value shows a trend of slowly increasing from 40 to 55. At this time, the red adaptive dynamic threshold upper limit shifts synchronously with this aging baseline and remains above the feature value.

[0086] In the 52nd month, the core feature extraction value represented by the blue solid line suddenly rose to over 70, exceeding the upper limit of the red adaptive dynamic threshold, which was around 64 at the time.

[0087] This waveform transformation and crossover indicates that while adapting to normal aging, the system can capture sudden physical anomalies in the operating mechanism through dynamic thresholds, providing a basis for outputting diagnostic conclusions.

[0088] When, during the verification process, the numerical values ​​of the extracted core features not only achieve high-weighted tendency scores in the model but also exceed the upper limit of the device's specific dynamic threshold at the statistical level, the system effectively eliminates interference from occasional events and normal aging, classifies it as a confirmed fault, and uses the diagnosis as the overall status conclusion of the device. For example, the conclusion might be: the fatigue level of the operating mechanism's energy storage spring has reached level three, accompanied by moderate jamming of the transmission main shaft. If the dynamic threshold is not exceeded, it is determined to be in the suspected fault observation period, no confirmed conclusion is issued, and the subsequent sampling frequency is increased.

[0089] Finally, the predictive maintenance decision engine module in the system backend receives the overall equipment status conclusion. Specifically, the predictive maintenance decision engine module is used to generate a predictive maintenance decision scheme that includes maintenance timing and resource allocation plans based on the overall equipment status conclusion and preset constraints.

[0090] The predictive maintenance decision engine module generates predictive maintenance decision schemes by: extracting fault risk assessment values, equipment importance level weights, maintenance windows, and resource matching degrees as constraints; constructing a multi-objective optimization function, whose input variables include the reduction in fault risk after scheme execution, maintenance cost consumption, and expected power outage time; substituting multiple alternative maintenance schemes into the multi-objective optimization function to calculate a comprehensive score, and outputting the scheme with the highest comprehensive score as the final predictive maintenance decision scheme.

[0091] In the actual scenario of power system on-site operation and maintenance scheduling, diagnosing equipment failure is only the first step; how to scientifically arrange maintenance is even more crucial. The predictive maintenance decision engine module extracts the equipment importance level weight of the line from the dispatch management system. For example, lines supplying power to hospitals or hub substations have extremely high weights. It also extracts the allowable outage maintenance window period of the current power grid load curve, as well as the quantity of spare parts in the current material warehouse and the resource matching degree of the team members as constraint boundary conditions.

[0092] The calculation of the multi-objective optimization function is performed using the following formula:

[0093] ;

[0094] in, The overall score for the alternative maintenance solutions, This represents the reduction in failure risk after the alternative solution is implemented. The amount of manpower and material maintenance costs required to implement this alternative plan. To implement this alternative plan, it is necessary to coordinate with the power grid's application for the expected power outage time. To assign the first optimization weight to reduce risk, To assign a second optimization weight based on cost considerations, A third optimization weight is assigned to consider the impact of power outages.

[0095] The decision engine module uses a heuristic search algorithm to generate multiple alternative maintenance plans, such as immediate power outage for emergency repair and complete replacement, reduced operation with coil replacement scheduled for Sunday morning during off-peak hours, and no replacement but with increased inspection frequency and lubrication replenishment. Data from each of these alternative maintenance plans is then fed into the aforementioned multi-objective optimization function for calculation, ultimately outputting the plan with the highest overall score.

[0096] The plan clearly defines the optimal timing for on-site maintenance and the resource allocation plan for specific spare parts, tools, and equipment required. It is directly distributed to the mobile terminals of front-line maintenance teams, forming an executable predictive maintenance decision-making plan.

[0097] In summary, the technical solution provided in this embodiment addresses the problems in the prior art where simple comparisons easily lead to false warnings under environmental fluctuations and make it impossible to scientifically schedule maintenance. The system first uses multiple linear regression to decouple and denoise the original signal affected by temperature and pressure interference, ensuring the purity of the feature source. Then, it uses a combination of equipment dynamic aging coefficient and dynamic threshold to perform cross-judgment when the secondary verification process is triggered.

[0098] This series of combined measures not only enables the system to keenly detect even the slightest signs of degradation in the mechanical structure of pole-mounted circuit breakers, but also allows it to adapt to the aging baseline drift of the equipment itself, effectively resolving the technical contradiction between high diagnostic sensitivity and high false alarm rate. Finally, the decision-making scheme output by a multi-objective optimization function, combined with actual field constraints, can directly guide power grid maintenance personnel to take the most cost-effective maintenance actions at the most appropriate time, greatly improving the safety management level of power grid assets and the reliability of power supply.

[0099] Example 2:

[0100] In a conventional power distribution network operating environment, the dynamic aging coefficient of the equipment obtained by linear weighted summation of multiple coefficients in Example 1 can better fit the natural aging process of equipment under long-term stable operation due to the passage of time, normal mechanical action and routine maintenance.

[0101] However, in actual power distribution network operations, pole-mounted circuit breakers inevitably encounter extreme abnormal situations such as metallic short circuits at the end of the line and large-current ground discharges caused by lightning overvoltages. Under these extreme conditions, the protective breaking action performed by the circuit breaker is accompanied by extremely large short-circuit breaking currents. The huge short-circuit current generates extremely high-energy arcing heat effects on the contact surfaces of the circuit breaker's vacuum interrupter, causing rapid melting, vaporization, and splashing losses of the contact materials. At the same time, the huge electrodynamic force generated by the peak short-circuit current acts directly on the connecting rods, main shaft, and buffer springs of the magnetically controlled operating mechanism in the form of mechanical impact, leading to a step-like nonlinear deterioration of the fatigue of the mechanical structure.

[0102] If the diagnostic system still uses the conventional static proportional weighting factor to calculate the dynamic aging coefficient of the equipment, then because the circuit breaker may have just been put into operation (with an extremely low correlation coefficient for the number of years since manufacture) or has never undergone a major overhaul (with an extremely low correlation coefficient for the number of maintenance visits), these low coefficient values ​​will severely dilute the dramatic deterioration of the correlation coefficient of operating conditions caused by extreme short-circuit interruptions during the weighted average mathematical calculation. This will result in the system outputting a still low risk rating for the equipment's dynamic aging coefficient, thus seriously underestimating the true mechanical and electrical degradation of the equipment after a severe blow, leading to serious underreporting of predictive maintenance mechanisms after extreme operating conditions.

[0103] To resolve the technical conflict between this conventional static linear weighted algorithm and the nonlinear step aging of equipment under extreme short-circuit stress, the system is also equipped with a weight reconstruction mechanism in the fault diagnosis and trend prediction module for adjusting the aging calculation rules under extreme operating conditions.

[0104] Specifically, the execution process of the weight reconstruction mechanism includes: real-time acquisition of the coil interruption current of the circuit breaker during the interruption operation, and calculation of the peak value of the single interruption current and the square integral value of the current within the corresponding interruption time.

[0105] In the actual monitoring and protection system of the distribution network, the evaluation of the circuit breaker's breaking capacity requires not only attention to its steady-state load current, but also high-frequency capture of the transient current characteristics at the moment of breaking. The system's multi-source data acquisition and preprocessing module is equipped with a current transformer with a high dynamic range and an independent high-speed analog-to-digital conversion channel, specifically designed for high-frequency sampling during the entire breaking operation period from when the circuit breaker receives the tripping command pulse issued by the relay protection device until the main circuit current completely crosses zero and extinguishes the arc.

[0106] The acquisition process of the coil breaking current corresponds to a discrete data sequence that varies over time. The digital signal processing unit in the system's microprocessor traverses this discrete data sequence in real time, optimizes it through a comparison algorithm, and calculates the peak value of the single breaking current. The peak value of the single breaking current is calculated using the following formula:

[0107] ;

[0108] in, This represents the peak value of the single interruption current. This represents the discrete value of the coil breaking current acquired in real time at the j-th sampling point during the breaking operation. This is the function for finding the maximum value.

[0109] In practical applications, the peak value of the single breaking current directly determines the maximum transient electrodynamic force that the magnetic control mechanism experiences at the moment of breaking. The electrodynamic force is proportional to the square of the peak current. When this peak value reaches a certain level, it can easily cause plastic deformation of the transmission linkage or shear damage to the pin. It is the core physical quantity for assessing the instantaneous mechanical damage depth of the circuit breaker.

[0110] Simultaneously, the system calculates the square integral of the current during the corresponding breaking time. In physics, the square integral of the current is equivalent to the Joule integral, used to characterize the cumulative total thermal stress and energy impact experienced by the contact system during the breaking arc. Since the acquired signal is a discrete digital signal, the system uses the trapezoidal rule or Simpson's rule to approximate the true continuous integral in the discrete time domain. The square integral of the current is calculated using the following formula:

[0111] ;

[0112] in, This represents the integral of the square of the current during the corresponding breaking time. These are the discrete index points of the sampled time series. This is the starting sampling index point corresponding to the moment the electric arc ignites during the interruption operation. This is the termination sampling index point corresponding to the moment when the arc extinguishing current completely crosses zero during the interruption operation. Let j be the discrete value of the coil breaking current at the j-th sampling point. This is the time interval constant between two adjacent samples in the high-speed analog-to-digital conversion channel.

[0113] In practical industrial applications, this square fraction of the current directly reflects the volume of vaporization and burn-off of the metal plating material of the circuit breaker contacts. Excessive burn-off of the contacts will change the overtravel parameters of the magnetic control mechanism, leading to increased closing bounce or abnormal opening speed, which is a major hidden danger that may induce mechanical operation failures.

[0114] like Figure 5 This figure illustrates the waveform changes of the transient current in the coil of a magnetically controlled pole-mounted circuit breaker under short-circuit breaking conditions, as well as the Joule integral energy region used to quantify thermal stress. The horizontal axis represents time in milliseconds, ranging from 0 to 100; the vertical axis represents coil current in amperes, ranging from 0 to 30.

[0115] The solid blue line in the figure represents the transient current waveform. After the circuit breaker receives the tripping command, the waveform changes within an action period of 10 to 50 milliseconds. The current rises and reaches a peak single-break current of 25 amperes in about 20 milliseconds.

[0116] This peak value corresponds to the mechanical stress borne by the internal operating mechanism of the circuit breaker when overcoming the short-circuit electrodynamic force. When the system detects that this peak value exceeds the set current damage threshold, it generates an extreme operating condition signal.

[0117] The graphic in the figure, filled with a red semi-transparent area between 10 milliseconds and 50 milliseconds, represents the Joule integral energy region, which characterizes the square integral value of the current during the time from the arc ignition to complete extinction during the breaking action, and quantifies the thermal stress borne by the contact system.

[0118] By calculating the peak and energy integral regions, the system can quantify the state changes of the mapped circuit breaker after being subjected to short-circuit impacts, providing data input for subsequent adjustment of weighting factors.

[0119] Optionally, after completing the real-time calculation of the aforementioned core physical quantities, the system's data discrimination unit begins to execute the high-risk state interception and analysis logic. In response to the single-break current peak value being greater than or equal to a set current damage threshold, or the increment of the current square integral value being greater than a set energy mutation threshold, an extreme operating condition signal is generated. Here, the set current damage threshold is not an arbitrary empirical value, but is strictly calculated based on the rated short-circuit breaking current parameters marked on the nameplate of this type of magnetically controlled pole-mounted circuit breaker, combined with the mechanical yield limit of materials mechanics. The setting of the current damage threshold is described by the following formula:

[0120] ;

[0121] in, The set current damage threshold, The rated short-circuit breaking current peak value specified at the time of manufacture. This is the mechanical damage tolerance coefficient derived from a large number of destructive tests.

[0122] When the system judges When it is established, it means that the circuit breaker has just experienced a strong electrodynamic impact, and its internal buffer or energy storage structure has most likely developed irreversible micro-cracks or macro-deformation.

[0123] On the other hand, for the assessment of thermal stress, the system focuses on the energy loss jump caused by a single interruption action. A set energy mutation threshold is used to intercept abnormal interruption events where the peak current may not exceed the limit, but due to system faults causing arc reignition or abnormally prolonged interruption time, a huge amount of accumulated heat energy is generated. The increment here refers to the squared current fraction generated by this interruption action alone. The system executes the following logical judgment: ;

[0124] in, This is the set energy mutation threshold. In practical applications, this threshold is set based on the conversion equivalent relationship between the maximum allowable single burn-off volume of the vacuum interrupter contacts and the arc energy.

[0125] The two conditions mentioned above are logically ORed in the system's logic gates. This means that regardless of whether the circuit breaker is subjected to violent tearing by transient electrical forces (peak value exceeding limits) or deep ablation by a prolonged high-temperature arc (energy integral exceeding limits), as long as either condition is met, the system will immediately interrupt the normal steady-state monitoring process and generate an extreme condition signal. Internally, this extreme condition signal is represented by a high-level digital instruction with the highest interrupt priority or a specific system bus flag, used to forcibly trigger the subsequent weight reconfiguration process.

[0126] It is also important to note that in the system's normal logic flow, the dynamic aging coefficient of the equipment is linearly determined by four dimensions: manufacturing age, operating conditions, number of maintenance visits, and health status. However, upon receiving the extreme operating condition signal, the calculation rules for the dynamic aging coefficient are adjusted: the preset weighting factor corresponding to the operating condition correlation coefficient is set as the first extreme value weight, and the preset weighting factors corresponding to the manufacturing age correlation coefficient and the number of maintenance visits correlation coefficient are set as the second extreme value weight. The dynamic aging coefficient is then recalculated based on the adjusted extreme value weights; wherein, the value of the first extreme value weight is greater than the value of the second extreme value weight.

[0127] This adjustment process is the core mechanism for coping with complex and harsh power grid environments in this embodiment. Upon receiving an extreme operating condition signal, the system determines that the aging state of the equipment is no longer dominated by the passage of time (years since manufacture) or the frequency of routine maintenance (number of repairs), but is completely taken over by the just-occurred destructive short-circuit interruption event (operating condition). If the original weight allocation is maintained, a severe mathematical dilution effect will occur. Therefore, the system must perform a forced weight reallocation based on a sudden change in physical stress.

[0128] The weight management subroutine in the system first executes the overwrite instruction. It overwrites the preset weight factors of the operating condition correlation coefficients, which were originally set to a lower order of magnitude, with the first extreme value weight. The first extreme value weight is assigned using the following formula: ;

[0129] in, The first extreme value weight is used to characterize the importance of the operating condition. It is a high-order constant that approximates the value of 1.

[0130] By setting the first extreme value weight to an extremely high value, the system forcibly assigns absolute dominance to the correlation coefficient of operating conditions in subsequent multiplication and addition calculations. Next, the system overwrites the preset weight factors corresponding to the correlation coefficients of manufacturing age and maintenance frequency with the second extreme value weight. The second extreme value weight is assigned using the following formula:

[0131] ;

[0132] in, The second extreme value weight is used to characterize the importance of the manufacturing age and the number of repairs. It is a low-order constant that approximates the value of 0. By setting the weight of the second extreme value to an extremely low value, the system mathematically almost completely shields the objective assessment of the current damage status from the historical fact that the device is a new device or has just undergone maintenance.

[0133] Therefore, after the above weight adjustment, the condition must be met: the weight value of the first extreme value must be greater than the weight value of the second extreme value. Furthermore, in actual engineering configurations, the difference between the two values ​​often spans multiple orders of magnitude to ensure the effectiveness of extreme value reconstruction.

[0134] like Figure 6 This diagram illustrates the reconstruction and step response process of the weight allocation of various components of the dynamic aging coefficient of the equipment before and after the triggering of an extreme short-circuit interruption condition. The horizontal axis of the diagram represents the event trigger time series, with values ​​ranging from 0 to 10; the vertical axis represents the weight coefficient values, with values ​​ranging from 0 to 1.

[0135] The vertical black dashed line in the figure represents the exact location where the system generates an extreme operating condition. Before this trigger point, the green solid line, representing the weights of the manufacturing year and the number of maintenance visits, and the blue solid line, representing the weights of the correlation coefficient of the operating conditions, both remain at an initial level of 0.25.

[0136] After the triggering moment, the blue solid line shows an upward step response, rising and stabilizing at the first extreme weight position of 0.9; the green solid line shows a downward step response, falling to the second extreme weight position of 0.05, reducing the influence of conventional historical data in the current assessment.

[0137] The orange solid line representing the preset weighting factor for health status remains at a level of 0.25 throughout the calculation. The state transition of the curve at the trigger point reflects the system's response process in adjusting the weights of the mathematical model, indicating that the weight reconstruction mechanism can increase the evaluation weight of current operating condition parameters and reduce the influence of average conventional health data on the assessment of extreme physical stress damage.

[0138] After completing the nonlinear extreme value reconstruction of the weighting factors, the system calls the underlying arithmetic logic unit to recalculate the dynamic aging coefficient of the equipment based on the adjusted extreme value weights. The recalculated dynamic aging coefficient of the equipment is solved using the following formula:

[0139] ;

[0140] in, This refers to the dynamic aging factor of the equipment, recalculated after the system encounters extreme operating conditions. To set a second extreme value weight that approximates 0, The first extreme value weight is set to approximate the value of 1.

[0141] The reconstructed dynamic aging coefficient of the equipment is output and fed back to the fault diagnosis and trend prediction module, serving as a key parameter for correcting the initial probability of fault occurrence in the prediction model. In practical applications, due to... The maximum amplification effect and The extreme compression effect, even the circuit breaker and It exhibits an extremely healthy new product condition, but as long as The value spiked due to the short circuit breaking, and the final result was... It will still exhibit extremely high aging and deterioration values.

[0142] When this aging factor, which indicates severe damage to the equipment, is input into the subsequent correction formula, it significantly increases the final failure probability output by the model, causing it to quickly exceed the preset probability threshold and thus forcibly triggering a rigorous closed-loop verification process. In subsequent cross-validation and dynamic threshold comparison, the system will conduct a more stringent review of the circuit breaker's next opening and closing coil current waveform after extreme operating conditions. Even if the waveform shows only an extremely slight increase in frictional resistance, the system will diagnose it as a fault and immediately generate a predictive maintenance decision requiring emergency replacement or a major power outage repair.

[0143] In summary, considering the common diagnostic blind spots in existing technologies, traditional mechanical fault diagnosis systems mostly rely on static statistical models or fixed multi-parameter weighting systems. These traditional solutions have some reference value under ideal conditions of stable environment and natural equipment aging, but they essentially treat equipment aging as a linear process that progresses at a constant rate. Once a destructive physical event such as phase-to-phase short circuit or lightning overcurrent occurs in the distribution network, traditional systems cannot break free from the constraints of static weights in their mathematical models. This causes severe local physical losses to be masked by massive historical health data, ultimately resulting in erroneous diagnostic reports. Circuit breakers with serious internal defects are allowed to continue operating with these defects, which can easily lead to serious equipment damage or large-scale power outages during the next operation.

[0144] The extreme condition response mechanism based on dynamic weight reconstruction, as detailed in this embodiment, endows the diagnostic system with the ability to proactively break the inertia of conventional mathematical calculations when faced with sudden and enormous electromagnetic and thermal stress shocks. This scheme accurately quantifies irreversible instantaneous mechanical tearing and contact burning damage by frequently capturing the peak value of a single interruption current and deeply calculating the square integral value of the current. Furthermore, by establishing a strict destructive threshold system, once an extreme stress shock is diagnosed, the conventional preset weights reflecting stable historical patterns are decisively abandoned. An extreme first-value weight is used to infinitely amplify the current damage condition, while a very low second-value weight is used to completely shield against false youthful states or maintenance record interference.

[0145] This technology has significant practical implications for actual power system dispatching and asset operation and maintenance management. It ensures that the fatigue and aging status of circuit breakers after undergoing major tests can be accurately and intuitively reflected by the diagnostic system, effectively avoiding the hidden danger of risk being diluted by mathematical averaging, and greatly improving the absolute reliability of power grid fault prediction and the timeliness of emergency repair decisions.

[0146] Example 3:

[0147] In Example 1, the system generates dynamic thresholds based on historical fault assessment indices within continuously set time windows, thereby adapting to reasonable baseline drift caused by the natural and slow aging of equipment over time. However, in actual outdoor operating environments of power distribution networks, equipment not only faces slow natural aging but also often faces extremely drastic changes in meteorological conditions.

[0148] This embodiment focuses on resolving the conflict scenario where the natural offset of characteristic baselines caused by drastic changes in the external environment and climate leads to confusion with the actual mechanical fault determination. In northern my country or high-altitude areas, sudden drops in temperature or extreme cold waves frequently occur in winter. Pole-mounted circuit breakers are exposed to the outdoor environment for extended periods, and the physical media such as mechanical grease and damping oil applied inside their operating mechanisms are highly sensitive to ambient temperature. In extremely cold weather, the physical viscosity coefficient of the grease increases non-linearly and sharply, and may even undergo microscopic physical phase transitions or icing at extremely low temperatures. This phase transition of the physical media leads to a significant increase in the mechanical resistance of the circuit breaker operating mechanism during opening and closing operations.

[0149] Reflected in the coil current signal sampled by the system, the increased mechanical resistance leads to a higher peak value of the coil drive current and a longer action time, resulting in a significant step-like baseline shift in the core features extracted by the system and the corresponding fault assessment index. This shift is a normal, natural shift caused by changes in the physical material properties due to extreme low temperatures, rather than a genuine mechanical physical fault such as a broken operating mechanism link or a deformed spindle. However, if the system continues to use the statistical dynamic threshold calculated based on a conventional time window, the characteristic values ​​under extreme cold conditions will easily exceed this conventional dynamic threshold due to the low mean and small standard deviation of the data within the conventional window before the temperature drop. This will cause the system to issue a large number of false fault alarm messages that do not reflect reality. At the same time, if the system continues to include these distorted data under extreme cold conditions in the conventional time window for rolling statistics, it will cause serious statistical data pollution, significantly widening the statistical standard deviation of historical features. This will cause the conventional dynamic threshold after the temperature recovers to become abnormally broad, thereby greatly reducing the system's diagnostic sensitivity for real minor mechanical wear faults.

[0150] To overcome the contradiction between the statistical threshold algorithm based on conventional time windows and the nonlinear baseline shift caused by phase change of physical media due to extreme weather conditions in outdoor equipment, this embodiment also includes an environmental phase change baseline adaptive migration mechanism in the fault diagnosis and trend prediction linkage module.

[0151] Specifically, the execution process of the environmental phase change baseline adaptive migration mechanism includes: obtaining the real-time ambient temperature of the environment where the circuit breaker is located, and calculating the temperature change rate per unit time.

[0152] In the actual hardware configuration, the multi-source data acquisition and preprocessing module continuously acquires real-time ambient temperature data at a set sampling frequency through a high-precision digital temperature sensor array attached to the circuit breaker housing or built into the control board. The microprocessor inside the fault diagnosis and trend prediction linkage module receives this time-series temperature data and calculates the rate of temperature change per unit time using a first-order difference algorithm or a polynomial fitting and differentiation algorithm. This rate of temperature change is calculated using the following formula:

[0153] ;

[0154] in, The rate of temperature change per unit time. The real-time ambient temperature is obtained at the previously set sampling time. This represents the time interval between the current sampling time and the previous set sampling time. In practical engineering applications, the rate of temperature change per unit time objectively reflects the severity of environmental heat loss and is a core quantitative indicator for determining whether a sudden cold wave will occur.

[0155] Optionally, the internal logic discrimination unit of the system performs conditional judgment based on the acquired temperature data. In response to the real-time ambient temperature being lower than the set phase transition critical temperature, or the temperature change rate being greater than the set abrupt change rate threshold, an environmental phase transition isolation signal is generated.

[0156] The phase transition critical temperature set here is based on the thermodynamic performance parameters of the specific type of grease added at the time of the circuit breaker's manufacture. Each type of industrial grease has a specific pour point and freezing point. When the ambient temperature approaches or falls below this freezing point, the base oil molecules inside the grease crystallize and precipitate, causing the grease to lose its fluidity and transform into a near-solid phase transition state. Simultaneously, the set abrupt change rate threshold is used to identify extreme weather conditions with rapid temperature drops in a short period. This is because if the rate of thermal expansion and contraction of the metal mechanism housing does not match the contraction rate of the internal connecting rods, it can also generate significant additional jamming stress.

[0157] The logical decision expression executed by the system is represented by the following formula:

[0158] ;

[0159] in, The generated environmental phase change isolation signal has a logical value of true when any of the logical conditions within the parentheses are met. The set critical temperature for phase transition; It is the absolute value of the rate of temperature change per unit time; The threshold for the rate of change of mutation is set. This is a logical OR operator. When the condition is met, the system immediately generates a high-level ambient phase-change isolation signal to trigger subsequent statistical isolation and baseline reconstruction actions.

[0160] It is also important to note that during the effective period of receiving the environmental phase change isolation signal, the routine rolling statistical update of the historical fault assessment index within a specific time window should be suspended in order to cut off the contamination of the normal sample model by abnormal environmental data.

[0161] Under normal, stable operation, the system's specific time window is a first-in, first-out (FIFO) circular queue data structure. New fault assessment indices are continuously added over time, while the oldest data is removed, maintaining a rolling sample library reflecting the recent state. However, during the period when the environmental phase change isolation signal is in a logical true state, the system actively triggers a write protection mechanism. The controller freezes the enqueue and dequeue operations of this circular queue, ensuring that the sample set within the specific time window remains in a pure state as before the cold wave. This isolation action has significant practical implications; it effectively prevents characteristic mutation data caused by the environment from being mixed into the regular statistical samples, and prevents the standard deviation of the regular dynamic threshold from being abnormally widened. This ensures that once the cold wave recedes and the weather returns to normal, the system can immediately and seamlessly switch back to the highly sensitive normal diagnostic mode.

[0162] For example, since conventional dynamic thresholds are not applicable to the current extreme cold phase transition conditions, the system must find an alternative benchmark that matches the current extreme climate characteristics. Therefore, the system performs a source tracing operation across a specific time window, retrieving and matching historical extreme climate cycles that match the current real-time ambient temperature and climate characteristics from the full lifecycle operational database module, and extracting the historical characteristic reference mean and historical characteristic reference standard deviation within that historical extreme climate cycle.

[0163] The full lifecycle operation database module stores all multi-dimensional historical data of the circuit breaker since it was put into operation in the distribution network, including complete operation records for each spring, summer, autumn, and winter. The system initiates a multi-dimensional similarity search algorithm within this vast historical database, with the matching criteria expressed by the following formula:

[0164] ;

[0165] in, The database module stores historical environmental temperature records throughout its entire lifecycle. The current real-time ambient temperature, This is the preset temperature matching tolerance range.

[0166] By traversing and comparing data, the system identifies time intervals during its historical operation where the equipment was subjected to the same ambient temperature conditions, defining these intervals as historical extreme climate cycles. Subsequently, the system performs statistical calculations on all fault assessment index samples generated within these historical extreme climate cycles, extracting the expected value and dispersion indicators for this interval. The extraction and calculation process uses the following formula:

[0167] ;

[0168] ;

[0169] in, The extracted historical feature reference mean, The extracted historical feature reference standard deviation, This represents the total number of valid samples within historical extreme climate cycles with similar frequencies. This is the sample of the i-th historical fault assessment index within a historical extreme climate cycle with the same frequency.

[0170] Through this time-traveling, cross-cycle retrieval, the system obtains a reasonable reference benchmark that is strictly aligned with the current severe cold state in terms of meteorological environment.

[0171] like Figure 7 This demonstrates the process of environmental phase transition isolation and cross-cycle aging compensation baseline adaptive migration under extreme cold wave climate conditions. In the figure, the left vertical axis represents the ambient temperature in degrees Celsius, the right vertical axis represents the feature evaluation values ​​and diagnostic thresholds, and the horizontal axis represents the running time in hours.

[0172] The blue solid line in the graph represents the real-time ambient temperature curve, which dropped from 5 degrees Celsius to -25 degrees Celsius between hours 40 and 50. The green solid line represents the core characteristic assessment value, which climbed from around 45 to around 65 during the cooling period.

[0173] The thick solid red stepped line in the diagram represents the diagnostic threshold. Before the temperature drop, the red threshold was stable around 50. When the system detected a temperature change and triggered an environmental phase change isolation signal, the red threshold paused its regular updates. After the system retrieved historical benchmarks and compensated for this with aging increments, the threshold stepped and shifted to 75 during the extreme cold period. This threshold adjustment range covered characteristic evaluation values ​​rising to 65 without triggering any over-limit alarms.

[0174] The multi-axis data graph shows that when the system responds to rapid changes in ambient temperature, it suppresses false fault alarms by adjusting the dynamic threshold, and resumes normal diagnosis after the temperature warms up to 90 hours later, adapting to the threshold setting requirements under outdoor climate change.

[0175] It is also important to note that directly using the historical average under similar temperatures from last year as the current benchmark still has technical limitations. This is because, although the meteorological conditions have aligned, the circuit breaker mechanism has undergone hundreds of opening and closing operations in the past year, and its inherent aging, such as mechanical wear and spring fatigue, has irreversibly increased. If last year's historical baseline is directly applied, the incremental aging over the past year will be ignored, resulting in a lower reconstruction threshold and a still high risk of false alarms.

[0176] To solve this technical challenge, the system retrieves the dynamic aging coefficient of the equipment based on the current equipment status, uses the dynamic aging coefficient to perform annual aging increment compensation on the historical characteristic reference mean, and calculates and generates the expected baseline for aging correction.

[0177] The core of this aging incremental compensation lies in quantifying the mechanical attenuation over the span from the historical data acquisition time to the current time. The system first synchronously retrieves the historical dynamic aging coefficients of equipment corresponding to historical extreme weather cycles from the database. Then, it combines these with the current equipment dynamic aging coefficients to calculate and generate the expected baseline for aging correction. This calculation process uses the following formula:

[0178] ;

[0179] in, To calculate the generated aging correction expected baseline, The historical feature reference mean extracted above, The dynamic aging coefficient of the equipment at the current moment is calculated by comprehensively considering its manufacturing age, operating conditions, number of maintenance visits, and health status. The dynamic aging coefficient of historical equipment recorded during extreme climate cycles of the same historical frequency. This is a preset aging drift conversion factor, which is used to map the relative growth rate of the aging factor to the physical growth rate of the fault assessment index baseline.

[0180] Through the aforementioned aging increment compensation mechanism, a completely new theoretical benchmark is constructed. This expected baseline for aging correction includes both the reasonable resistance bias caused by the phase change of lubricating grease due to extreme cold weather and the normal mechanical wear bias that increases after a year of wear. This baseline best reflects the expected value of the actual health status of the equipment under the dual constraints of extreme climate and its own aging.

[0181] It is also important to note that after obtaining an accurate baseline for aging correction, the system begins to reconstruct the final threshold used for fault determination. The system adds the product of the aging correction baseline and the historical characteristic reference standard deviation to reconstruct the extreme phase transition dynamic threshold used to replace the conventional dynamic threshold.

[0182] The reconstruction process of this extreme phase transition dynamic threshold is described by the following formula:

[0183] ;

[0184] in, The reconstructed extreme phase transition dynamic threshold, The baseline for aging correction, which includes environmental alignment and aging compensation, is as follows. For preset adjustment coefficients, This is the historical reference standard deviation extracted from historical extreme climate cycles with similar frequencies. The rationale for retaining the historical standard deviation here is that the data fluctuation and dispersion characteristics of the same equipment under similar extremely cold conditions usually exhibit a high degree of consistency and inheritance.

[0185] Finally, the system uses extreme phase transition dynamic thresholds to re-execute the over-limit comparison of the currently extracted core features, outputs the corresponding confirmed fault judgment result or normal offset judgment result, and automatically resumes the regular rolling statistical update of a specific time window after the environmental phase transition isolation signal is released.

[0186] During the effective period of the environmental isolation signal, the system's data comparison module blocks the input channel of the conventional threshold and instead receives the extreme phase transition dynamic threshold calculated above. If the currently calculated core feature mapping value is still significantly greater than this multi-reconstructed extreme phase transition dynamic threshold, it indicates that the current current distortion amplitude has far exceeded the range that can be explained simply by the freezing of grease due to extreme cold and normal historical aging. The system determines that the equipment has indeed experienced a confirmed fault, such as a broken mechanical bearing or severe deformation of a connecting rod, and outputs this result as the final comprehensive equipment status conclusion. Conversely, if the feature value is within the envelope range of the extreme phase transition dynamic threshold, the system determines that the current value change is merely a normal offset caused by extreme cold weather and does not output a fault alarm, thus effectively suppressing the occurrence of false alarms.

[0187] As weather conditions improve, once the real-time ambient temperature rises and stabilizes above the set critical phase transition temperature, and the rate of temperature change returns to a more gradual level, the system's judgment logic condition no longer holds, the environmental phase transition isolation signal flips to a false condition and is released. At this point, the system automatically cancels the statistical write protection command, unfreezes the regular circular queue, and resumes using recently uncontaminated data for regular rolling statistical updates within the set time window. This achieves seamless integration and adaptive smooth transition of the diagnostic mechanism between extreme and normal climates.

[0188] Analysis of existing technologies reveals two distinct shortcomings in traditional online monitoring and mechanical fault diagnosis methods for circuit breakers when it comes to threshold setting. One approach uses static, factory-fixed thresholds, which are ill-suited for long-term mechanical aging and sudden, dramatic weather changes, easily leading to widespread false alarms across the entire network during winter. The other approach, while incorporating statistically adaptive thresholds that adjust alarm limits by calculating the mean and variance of recent data, suffers from a severe disconnect between recent stable data and sudden extreme cold weather. Allowing abrupt data fluctuations into the statistical model may temporarily prevent alarms, but it permanently compromises the sensitivity of subsequent diagnostic models, causing the system to become chronically sluggish after weather conditions return to normal, thus missing potentially fatal faults.

[0189] Compared to the inherent limitations of the existing technologies, the technical solution described in this embodiment proposes an adaptive migration mechanism for environmental phase change baselines, specifically addressing the complex and ever-changing outdoor service environment. This solution does not blindly rely on recent data or static parameters. Instead, it decisively implements statistical isolation measures at critical junctures where sudden weather changes may trigger phase changes in the physical medium, effectively preventing abnormal environmental data from contaminating the core statistical model. More importantly, this solution leverages the deep value of full lifecycle data, employing cross-cycle historical retrieval and matching, and cleverly utilizing the equipment's dynamic aging coefficient for cross-year aging increment compensation, to reconstruct an extreme phase change dynamic threshold that highly conforms to the expectations of physical reality.

[0190] This complete data migration and compensation logic enables the diagnostic system to achieve intelligent and professional diagnostic assessments when facing severe external environmental shocks such as extreme cold and heat. It accurately isolates environmental interference and reveals the true mechanical and physical condition of the equipment. The overall solution not only substantially avoids widespread false alarms of faults under extreme weather conditions, saving a significant amount of unnecessary on-site line inspection costs, but also perfectly preserves the system's high-sensitivity flaw detection capabilities under normal operating conditions. This provides a solid and accurate intelligent decision-making guarantee for the high-reliability operation of the core control nodes of the distribution network under all-weather and all-season conditions.

[0191] Example 4:

[0192] like Figure 8 As shown, this embodiment provides a mechanical fault diagnosis method for magnetically controlled pole-mounted circuit breakers based on coil current characteristics. The technical logic of this method corresponds to the aforementioned system architecture, and is mainly deployed on the server side of the distribution network automation master station system in the form of software algorithms or executable programs, or run in the edge computing unit of the distribution automation feeder terminal (FTU) installed nearby below the pole-mounted circuit breaker.

[0193] Existing technologies for monitoring the mechanical condition of circuit breakers, if employing a single, fixed, static judgment standard, are prone to generating frequent false alarms when faced with outdoor temperature and pressure fluctuations and natural aging of equipment. Relaxing the standard, on the other hand, can easily lead to missed detections of genuine early mechanical jamming or wear hazards. To address this technical challenge of balancing high sensitivity with a high false alarm rate, this embodiment provides a multi-level filtering and dynamic closed-loop verification digital diagnostic process.

[0194] Specifically, this embodiment provides a mechanical fault diagnosis method for a magnetically controlled pole-mounted circuit breaker based on coil current characteristics, comprising the following steps:

[0195] First, the coil current characteristic parameters of the magnetically controlled pole-mounted circuit breaker are obtained, and environmental decoupling is performed based on the correlation between the coil current characteristic parameters, real-time voltage, and ambient temperature to extract the core features.

[0196] In practical applications of power distribution networks, data acquisition relies on supporting hardware sensor networks. Typically, a high-frequency through-hole Hall effect current sensor is connected in series in the secondary control circuit of the circuit breaker to capture the transient current waveform sequence at the moment the circuit breaker performs opening or closing operations. From this, characteristic parameters of the coil current, reflecting mechanical characteristics, such as start-up time, peak current amplitude, and current trough, are extracted. However, long outdoor feeder lines and large load fluctuations can cause the real-time voltage of the operating power supply to drop or fluctuate. Simultaneously, changes in ambient temperature due to day-night cycles and seasonal changes directly alter the DC resistance of the coil's copper conductors. These normal fluctuations in external physical conditions can cause deviations in the coil current characteristic parameters that are remarkably similar to those observed during mechanical faults.

[0197] Therefore, the system synchronously collects real-time voltage and ambient temperature data through voltage transformers and temperature probes. Utilizing pre-established correlations based on factory calibration data, such as multiple linear regression, it calculates the appropriate current bias under the current temperature and pressure conditions. By subtracting this appropriate bias from the actual measured characteristic parameters, the system completes environmental decoupling. This step is essentially a physical noise reduction process, objectively eliminating the contamination of diagnostic data by meteorological conditions and power grid fluctuations, ultimately extracting core features that purely reflect mechanical conditions such as increased friction in the operating mechanism, deformation of the transmission shaft, or fatigue of the buffer springs.

[0198] Subsequently, based on the core features and the dynamic aging coefficient of the equipment, the probability of failure is predicted.

[0199] In the asset management practice of power equipment, circuit breakers with different service lifespans exhibit objective differences in their tolerance to changes in the same amplitude characteristics. The system retrieves the commissioning year, historical short-circuit current interruption count, and inspection and overhaul records of equipment with a specific ledger number from the Power Production Management System (PMS) database. Based on this, a dynamic aging coefficient characterizing the current physical aging degree of the equipment is calculated. The decoupled and extracted core features are then input into a forward prediction model such as a support vector machine or deep neural network. After outputting an initial risk value, the model incorporates the aforementioned dynamic aging coefficient for mathematical correction, appropriately amplifying the risk weight of older equipment according to its decay law. This processing makes the predicted failure probability more consistent with the objective laws of life-cycle wear and tear on industrial equipment.

[0200] Next, in response to the failure probability being greater than or equal to a preset probability threshold, a verification process is triggered.

[0201] The preset probability threshold is typically set based on the importance of the circuit breaker's line (for example, lines supplying power to hospitals or other primary loads have a lower threshold to maintain high alertness). When the fault occurrence probability calculated by the system reaches or exceeds this preset numerical boundary, it indicates that the algorithm model has detected early signs of potential problems within the operating mechanism. To effectively avoid false alarms caused by single sensor sampling errors or strong electromagnetic transient interference in the field, this method incorporates a status interception mechanism. At this point, the system temporarily refrains from issuing a confirmed alarm command to the dispatch center. Instead, it marks the equipment status as a suspected stage and uses this as a trigger to initiate a subsequent rigorous closed-loop verification process.

[0202] Furthermore, in the verification process, the weight parameters of the core features corresponding to the fault type in the prediction model are increased, and the fault judgment is re-executed in combination with the dynamic threshold generated based on the historical fault assessment index, so as to output the comprehensive status conclusion of the equipment.

[0203] This step is the core verification mechanism for solving the problem of high false alarm rate in this method. After the verification process is started, if the prediction model initially tends to judge it as a tripping jam fault, the system will programmatically increase the weight parameters of features that are strongly correlated with the mechanism of this type of fault, such as the total action time or the energy characteristics of a specific frequency band, at the algorithm level, thereby artificially enhancing the sensitivity and focusing ability of the prediction model to this specific suspected fault.

[0204] Meanwhile, the system abandons the static factory judgment criteria shared by all circuit breakers, and instead retrieves the historical fault assessment index accumulated by the circuit breaker during its recent stable operation, i.e., within a specific historical time window, from the database. By calculating the statistical mean and standard deviation of these historical data, the system constructs a dynamic threshold that closely reflects the current normal aging state of the equipment. Subsequently, the system cross-compares and re-evaluates the scores mapped by the weighted core features with the equipment-specific dynamic threshold. If the distortion of the current features still significantly exceeds the reasonable statistical upper limit of the dynamic threshold, the system excludes the possibility of normal aging and occasional interference, diagnoses it as real mechanical physical damage, and outputs a comprehensive equipment status conclusion including fault category and severity based on this logic.

[0205] Finally, based on the overall equipment status conclusions and preset constraints, a predictive maintenance decision scheme is generated, which includes maintenance timing and resource allocation plans.

[0206] In actual power system repair and maintenance operations, simple fault diagnosis is insufficient to directly guide production; planning must be combined with the actual boundaries of power grid operation. The system connects to the Distribution Network Management System (OMS) to obtain the current allowable load transfer and outage maintenance windows, the availability of personnel for the current maintenance team, and the inventory matching degree of corresponding circuit breaker spare parts in nearby material warehouses, such as operating mechanism assemblies and drive coils. These real-world factors are used as preset constraints. Subsequently, the system employs a multi-objective optimization algorithm to find the optimal solution among reducing the risk of equipment operating with defects, reducing power outage duration, and controlling maintenance manpower and material costs, automatically generating a highly executable predictive maintenance decision plan. This plan clarifies the recommended on-site maintenance timing and the required resource allocation scheme, and is directly pushed to the mobile work terminals of front-line maintenance personnel through a data interface.

[0207] The mechanical fault diagnosis method provided in this embodiment tightly integrates data processing, model prediction, dynamic verification, and business decision-making. In actual operation, environmental decoupling is achieved through temperature-pressure correlation equations in the early stage to ensure the purity of the original input signal; aging correction is introduced in the middle stage, and an interception trigger mechanism is established to avoid overreaction to minor fluctuations; in the later verification process, feature weighting and cross-judgment of equipment-specific dynamic thresholds effectively distinguish the baseline shift caused by meteorological interference and natural aging from actual mechanical wear. In actual distribution network automation scenarios, the overall solution successfully resolves the technical conflict that high diagnostic sensitivity is often accompanied by high false alarm rates, ensuring the objectivity and accuracy of output decisions. It can effectively guide power companies to carry out efficient predictive condition maintenance, optimize asset allocation, and thus improve the overall operational reliability of distribution network equipment.

[0208] Example 5:

[0209] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0210] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0211] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0212] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0213] The memory 103 stores a computer program corresponding to a mechanical fault diagnosis method for a magnetically controlled pole-mounted circuit breaker based on coil current characteristics according to the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0214] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0215] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A mechanical fault diagnosis system for a magnetically controlled pole-mounted circuit breaker based on coil current characteristics, characterized in that, include: The feature extraction and analysis module is used to obtain the coil current characteristic parameters of the magnetically controlled pole-mounted circuit breaker, and to perform environmental decoupling based on the correlation between the coil current characteristic parameters, real-time voltage and ambient temperature to extract the core features. The fault diagnosis and trend prediction module is used to predict the probability of fault occurrence based on the core features and the dynamic aging coefficient of the equipment. The verification process is triggered in response to the failure occurrence probability being greater than or equal to a preset probability threshold. In the verification process, the weight parameters of the core features corresponding to the fault type in the prediction model are increased, and the fault judgment is re-executed in combination with the dynamic threshold generated based on the historical fault assessment index to output the comprehensive status conclusion of the equipment. The predictive maintenance decision engine module is used to generate a predictive maintenance decision scheme that includes maintenance timing and resource allocation scheme based on the overall equipment status conclusion and preset constraints. In the fault diagnosis and trend prediction module, the calculation process of the equipment dynamic aging coefficient includes: Obtain the correlation coefficients for the manufacturing age, operating conditions, number of maintenance operations, and health status of the magnetically controlled pole-mounted circuit breaker. The dynamic aging coefficient of the equipment is obtained by multiplying the correlation coefficient of the manufacturing age, the correlation coefficient of the operating condition, the correlation coefficient of the number of maintenance, and the correlation coefficient of the health by their respective preset weighting factors and then summing them. A correction coefficient is generated based on the dynamic aging coefficient of the equipment, and the preliminary calculated probability of failure is adjusted using the correction coefficient. The fault diagnosis and trend prediction module is also equipped with a weight reconstruction mechanism; the specific execution process of the weight reconstruction mechanism includes: The circuit breaker's coil interruption current is collected in real time during the interruption operation, and the peak value of the single interruption current and the square integral value of the current during the corresponding interruption time are calculated. In response to the peak value of the single interruption current being greater than or equal to a set current damage threshold, or the integral value of the square of the current being greater than a set energy mutation threshold, an extreme operating condition signal is generated. Upon receiving the extreme operating condition signal, the calculation rules for the dynamic aging coefficient of the equipment are adjusted: the preset weight factor corresponding to the operating condition correlation coefficient is set as the first extreme value weight, and the preset weight factor corresponding to the manufacturing age correlation coefficient and the maintenance frequency correlation coefficient is set as the second extreme value weight. The dynamic aging coefficient of the equipment is recalculated based on the adjusted extreme value weights. Wherein, the value of the first extreme value weight is greater than the value of the second extreme value weight.

2. The system according to claim 1, characterized in that, It also includes a multi-source data acquisition and preprocessing module, which is used for: Collect the coil's original current, ambient temperature, and real-time voltage data; Low-pass filtering, cyclic redundancy check (CRC) verification, and sliding window outlier filtering are performed sequentially, followed by data normalization to obtain preprocessed data, which is then used by the feature extraction and analysis module to obtain the coil current characteristic parameters. The data normalization process includes: subtracting the minimum value of the corresponding class of data from the filtered individual data, and dividing the difference by the difference between the maximum and minimum values ​​of the corresponding class of data to obtain the normalized data.

3. The system according to claim 1, characterized in that, In the feature extraction and analysis module, the correlation relationship is a multivariate linear relationship, and the process of environmental decoupling includes: The estimated value of the coil current characteristic parameter is calculated by summing the product of the real-time voltage and the first corresponding coefficient, the product of the ambient temperature and the second corresponding coefficient, and the third corresponding coefficient. The estimated value is subtracted from the actual obtained coil current characteristic parameters to extract the core features after eliminating environmental interference.

4. The system according to claim 1, characterized in that, The process by which the feature extraction and analysis module extracts the core features specifically includes: Calculate the corrected instantaneous power, current rate of change, and temperature rate of change, including total harmonic distortion, and use them as candidate features; Calculate the Pearson correlation coefficient between each of the candidate features and the preset fault assessment index; Select candidate features whose absolute value of the Pearson correlation coefficient is greater than or equal to a set correlation coefficient threshold, and use them as the core features.

5. The system according to claim 1, characterized in that, In the fault diagnosis and trend prediction module, the process of re-executing fault judgment based on a dynamic threshold generated from historical fault assessment indices specifically includes: The historical fault assessment indices within a continuously set time window are statistically analyzed, and the mean and standard deviation of the fault assessment indices are calculated. The dynamic threshold is calculated by adding the product of the preset adjustment coefficient and the standard deviation of the fault assessment index to the mean of the fault assessment index. When the value corresponding to the currently extracted core feature is greater than the dynamic threshold, it is determined to be a confirmed fault, and the confirmed result is used as the overall status conclusion of the device.

6. The system according to claim 1, characterized in that, The process by which the predictive maintenance decision engine module generates predictive maintenance decision schemes includes: The following constraints are extracted: fault risk assessment value, equipment importance level weight, maintenance window period, and resource matching degree. Construct a multi-objective optimization function, the input variables of which include the reduction in failure risk after the implementation of the solution, the amount of operation and maintenance costs, and the expected power outage time; Multiple alternative maintenance schemes are substituted into the multi-objective optimization function to calculate a comprehensive score, and the scheme with the highest comprehensive score is output as the final predictive maintenance decision scheme.

7. The system according to claim 1, characterized in that, The system is deployed using an architecture that includes both edge devices and a cloud platform; The edge device is equipped with a lightweight prediction model for performing real-time data acquisition, preprocessing, and local preliminary calculation of the probability of the fault occurrence. The cloud platform is equipped with a complete trend prediction model and the predictive maintenance decision engine module. In response to the edge terminal calculating that the probability of the fault occurrence is greater than or equal to the preset probability threshold, the data of the edge terminal is triggered to be uploaded to the cloud platform in real time, and the cloud platform performs high-precision prediction and the verification process.

8. The system according to claim 5, characterized in that, The fault diagnosis and trend prediction module is also equipped with a baseline migration mechanism; the specific execution process of the baseline migration mechanism includes: Obtain the real-time ambient temperature of the environment where the circuit breaker is located, and calculate the rate of temperature change per unit time. An environmental isolation signal is generated in response to the real-time ambient temperature being lower than a set phase transition critical temperature, or the temperature change rate being greater than a set abrupt change rate threshold. During the period when the environmental isolation signal is received, the use of the historical fault assessment index within the currently set time window to update the dynamic threshold is suspended; The historical operating intervals in which the difference between the historical temperature and the real-time ambient temperature is within a preset range are retrieved from the full life cycle operation database module, and the historical mean and historical standard deviation within the historical operating intervals are extracted. The historical average is incrementally compensated using the current dynamic aging coefficient of the equipment to calculate and generate a corrected baseline; The modified baseline is added to the product of a preset scaling factor and the historical standard deviation to generate an extreme phase transition dynamic threshold. The extreme phase transition dynamic threshold is then used to replace the original dynamic threshold to perform the fault judgment.

Images