Intelligent detection device and detection method for pole-mounted circuit breaker

Abstract

The present invention discloses an intelligent detection device and detection method for a pole-mounted circuit breaker, which relates to the field of intelligent detection technology. The device comprises: traversing the pole-mounted circuit breaker to arrange multiple sensors, constructing a data sensor network, and obtaining multiple data signals; performing fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals, and generating fault diagnosis results; performing health assessment, and obtaining a health score of the pole-mounted circuit breaker; performing fault prediction on the pole-mounted circuit breaker, and obtaining a multimodal prediction data set; performing coverage detection optimization on the operating parameter set, and formulating maintenance recommendations. The present invention solves the technical problems in the prior art of pole-mounted circuit breaker detection, such as single detection means, low fault diagnosis accuracy, lack of comprehensive evaluation of the health status of the equipment and forward-looking fault prediction, and achieves the technical effect of realizing comprehensive and accurate detection and maintenance of the pole-mounted circuit breaker, and effectively improving the reliability, safety and stability of the operation of the pole-mounted circuit breaker.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection technology, and in particular to an intelligent detection device and a detection method for a pole-mounted circuit breaker. Background Art

[0002] As a key device in the power distribution system, the pole-mounted circuit breaker bears the important responsibility of controlling and protecting the power lines. Its operating status is directly related to the stability and power supply reliability of the power system. In the context of the continuous development of modern power systems, pole-mounted circuit breakers are facing increasingly complex operating environments and higher performance requirements. However, traditional pole-mounted circuit breaker detection methods have many limitations. On the one hand, the detection means are relatively single, often relying on manual regular inspections or simple electrical parameter measurements, and it is difficult to obtain comprehensive and accurate equipment status information. For example, only by regularly observing the appearance and measuring a limited number of electrical parameters, it is impossible to deeply understand the health status of the mechanical structure inside the equipment, changes in insulation performance, and potential early fault hazards. On the other hand, fault diagnosis lacks accuracy and timeliness. Due to the lack of effective data analysis and processing methods, the diagnostic ability for complex faults is limited, and usually only post-processing can be performed after the fault occurs, and early warning and precise positioning of the fault cannot be achieved. In addition, the existing technology is not comprehensive enough for the health assessment of pole-mounted circuit breakers, lacks a systematic evaluation method and indicator system, and cannot quantitatively evaluate the overall health status of the equipment, and it is difficult to predict the remaining service life and potential failure risks of the equipment. At the same time, in terms of maintenance strategies, there is a lack of scientific and effective basis, and fixed-cycle maintenance methods are mostly adopted, which can easily lead to excessive or insufficient maintenance, which not only increases maintenance costs but also reduces equipment availability.

[0003] The existing technology has the following technical problems in the detection of pole-mounted circuit breakers: single detection means, low accuracy of fault diagnosis, lack of comprehensive assessment of equipment health status and forward-looking fault prediction. Summary of the invention

[0004] The present application provides an intelligent detection device and detection method for a pole-mounted circuit breaker, which is used to solve the technical problems in the prior art of pole-mounted circuit breaker detection, such as single detection means, low fault diagnosis accuracy, lack of comprehensive evaluation of equipment health status and forward-looking fault prediction.

[0005] In view of the above problems, the present application provides an intelligent detection device and a detection method for a pole-mounted circuit breaker.

[0006] In a first aspect of the present application, an intelligent detection device for a pole-mounted circuit breaker is provided, the device comprising:

[0007] A data signal acquisition module, the data signal acquisition module is used to traverse the pole-mounted circuit breaker to deploy multiple sensors, build a data sensor network, collect data from the pole-mounted circuit breaker through the data sensor network, and obtain multiple data signals; a fault diagnosis result generation module, the fault diagnosis result generation module is used to perform fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals and generate a fault diagnosis result; a health score acquisition module, the health score acquisition module is used to perform a health assessment according to the fault diagnosis result to obtain a health score of the pole-mounted circuit breaker; a multimodal prediction data set acquisition module, the multimodal prediction data set acquisition module is used to introduce an environmental parameter set, predict faults of the pole-mounted circuit breaker according to the health score combined with the environmental parameter set, and obtain a multimodal prediction data set; a maintenance suggestion formulation module, the maintenance suggestion formulation module is used to jointly analyze the multimodal prediction data set according to the multiple data signals, perform coverage detection optimization on the operating parameter set according to the joint analysis results, and formulate maintenance suggestions.

[0008] A second aspect of the present application provides an intelligent detection method for a pole-mounted circuit breaker, the method comprising:

[0009] Traverse the pole-mounted circuit breakers to deploy multiple sensors, build a data sensor network, collect data from the pole-mounted circuit breakers through the data sensor network, and obtain multiple data signals; perform fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals to generate a fault diagnosis result; perform health assessment according to the fault diagnosis result to obtain a health score of the pole-mounted circuit breaker; introduce an environmental parameter set, predict faults of the pole-mounted circuit breaker according to the health score combined with the environmental parameter set, and obtain a multimodal prediction data set; jointly analyze the multimodal prediction data set according to the multiple data signals, perform coverage detection optimization on the operating parameter set according to the joint analysis result, and formulate maintenance recommendations.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] Traverse the pole-mounted circuit breaker to deploy multiple sensors, build a data sensor network, and obtain multiple data signals; perform fault diagnosis on the pole-mounted circuit breaker based on the multiple data signals to generate fault diagnosis results; perform health assessment based on the fault diagnosis results to obtain the health score of the pole-mounted circuit breaker; introduce environmental parameter sets to predict faults on the pole-mounted circuit breaker to obtain a multimodal prediction data set; conduct joint analysis on the multimodal prediction data set, perform coverage detection optimization on the operating parameter set based on the joint analysis results, and formulate maintenance recommendations. The technical effect of achieving comprehensive and accurate detection and maintenance of pole-mounted circuit breakers and effectively improving the reliability, safety and stability of pole-mounted circuit breaker operation has been achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A schematic structural diagram of an intelligent detection device for a pole-mounted circuit breaker provided in an embodiment of the present application.

[0014] Figure 2 A schematic flow chart of an intelligent detection method for a pole-mounted circuit breaker provided in an embodiment of the present application.

[0015] Explanation of the reference numerals: data signal acquisition module 10 , fault diagnosis result generation module 20 , health score acquisition module 30 , multimodal prediction data set acquisition module 40 , maintenance suggestion formulation module 50 . DETAILED DESCRIPTION

[0016] The present application provides an intelligent detection device and a detection method for a pole-mounted circuit breaker, which is used to solve the technical problems in the prior art of pole-mounted circuit breaker detection, such as single detection means, low fault diagnosis accuracy, lack of comprehensive evaluation of equipment health status and forward-looking fault prediction.

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] Embodiment 1, as Figure 1 As shown, the present application provides an intelligent detection device for a pole-mounted circuit breaker, the device comprising:

[0019] The data signal acquisition module 10 is used to traverse the pole-mounted circuit breaker to deploy multiple sensors, build a data sensor network, collect data from the pole-mounted circuit breaker through the data sensor network, and obtain multiple data signals.

[0020] Specifically, first of all, a comprehensive and detailed traversal work will be carried out on the pole-mounted circuit breaker. The pole-mounted circuit breaker has a complex structure, including multiple key parts such as the incoming line end, the outgoing line end, the arc extinguishing chamber, the operating mechanism, and the insulating components. The technicians will carefully plan the layout of the sensors according to the functional characteristics of these components and the potential areas where failures may occur. For example, current transformers and voltage transformers are arranged near the incoming and outgoing lines to accurately measure the current and voltage signals passing through the circuit breaker. These signals are critical for monitoring the power transmission status and judging whether there are overloads, short circuits and other faults; temperature sensors and pressure sensors are installed around the arc extinguishing chamber, because high temperature and high pressure are generated during the arc extinguishing process. If the temperature or pressure changes abnormally, it indicates that the arc extinguishing performance is reduced or the arc extinguishing chamber is faulty; displacement sensors and vibration sensors are set at the key transmission parts of the operating mechanism. By monitoring the movement stroke and vibration of the mechanism, it can be timely discovered whether the operating mechanism has problems such as jamming and wear. After determining the sensor layout location, the appropriate type of sensor will be selected for installation. For current measurement, high-precision electromagnetic or electronic current transformers are selected to ensure that current signals under different working conditions can be accurately collected; for voltage measurement, capacitive or electromagnetic voltage transformers are used to meet the accuracy requirements of voltage measurement; temperature sensors can use thermocouples or thermistor sensors to quickly respond to temperature changes; pressure sensors can select appropriate piezoelectric or piezoresistive sensors according to the pressure range and accuracy requirements of the arc extinguishing chamber; displacement sensors can use grating or inductive sensors to achieve accurate measurement of the displacement of the operating mechanism; vibration sensors generally use acceleration sensors to detect the vibration amplitude and frequency of the operating mechanism. After all sensors are installed, a data sensing network is constructed. When the pole-mounted circuit breaker is put into operation, the various sensors in this network begin to collect data in real time, such as the current transformer continuously outputs the current size signal, and the temperature sensor continuously feeds back the temperature change signal of the arc extinguishing chamber. These signals from different sensors that reflect the different operating states of the pole-mounted circuit breaker together constitute multiple data signals, which provide the most original data basis for subsequent fault diagnosis, health assessment and other work.

[0021] The fault diagnosis result generating module 20 is used to perform fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals and generate a fault diagnosis result.

[0022] Specifically, after receiving multiple data signals from the data signal acquisition module, the fault diagnosis result generation module immediately starts the fault diagnosis process. First, the data signals are subjected to multi-layer wavelet transform, which is like a fine spectrum analysis and feature extraction of the original signal. Through wavelet transform, the signal can be decomposed into components of different frequencies and scales, thereby obtaining multiple signal features, including signal frequency domain features and signal time domain features. In terms of frequency domain features, it can reveal the energy distribution of the signal in different frequency bands. For example, the current signal of the pole-mounted circuit breaker has a relatively stable energy distribution at certain specific frequencies during normal operation. If a fault occurs, such as partial discharge caused by internal insulation damage, additional energy peaks will be generated in the high frequency band. Through in-depth analysis of the signal spectrum data, these potential abnormal frequency features can be found, thereby providing important clues for fault diagnosis. In terms of time domain features, the focus is on the change law of the signal over time. For example, the action process of the operating mechanism has specific time series and waveform characteristics under normal circumstances. If there is a freeze or delay, it will show abnormal conditions such as waveform distortion and time delay in the time domain signal. By analyzing the signal timing data, these changes that do not conform to the normal operating mode can be captured. In order to achieve accurate fault diagnosis, this module uses machine learning technology combined with rich historical fault record logs to build a fault mode database. First, historical fault record logs are retrieved from multiple data sources of pole-mounted circuit breakers, such as equipment operation records, previous fault maintenance reports, laboratory test data, etc. Then these logs are labeled and multiple fault labels are generated for each record based on factors such as fault type, occurrence time, and severity. Then, cluster analysis is performed on the historical fault record logs according to these fault labels, and faults with similar characteristics are classified into one category, thereby determining multiple fault types. The long short-term memory network (LSTM) is used to perform deep learning on these determined fault types. LSTM has good memory ability and can learn the time series relationship and long-term dependency relationship in the fault data, so as to accurately identify the complex data patterns corresponding to different fault types. After deep learning, multiple fault modes are determined, and these fault modes are associated according to the fault labels, and then serialized according to the running time sequence, and finally a complete fault mode database is obtained. In actual diagnosis, the signal spectrum data and signal time series data obtained by wavelet transform are used as indexes to traverse and match in the fault mode database. Carefully compare each fault mode in the database. If a fault mode is found that is highly consistent with the current data, the pole-mounted circuit breaker is determined to be faulty and a fault data set is generated. In-depth analysis is performed on the data in the fault data set to determine the specific fault type, such as insulation fault, mechanical fault, electrical control fault, etc. Finally, the accurate fault type is added to the fault diagnosis result to provide a clear direction for subsequent repair and maintenance work.

[0023] The health score acquisition module 30 is used to perform a health assessment according to the fault diagnosis result to obtain a health score of the pole-mounted circuit breaker.

[0024] Specifically, the health status of the pole-mounted circuit breaker is quantitatively evaluated. When receiving the fault diagnosis results, it will comprehensively consider multiple factors to determine the health score. First, it will evaluate the severity of the fault type. For example, a serious insulation breakdown fault will pose a direct threat to the safe operation of the pole-mounted circuit breaker, and the corresponding weight of such faults is higher; while some minor problems such as loose external connections, although they also need attention, have relatively little impact on the overall operation and have a lower weight.

[0025] Secondly, consider the frequency of faults. If a fault occurs frequently, even if it is not serious, it may indicate a potential systemic problem with the pole-mounted circuit breaker, which will lead to a lower health score. For example, frequent operating mechanism jams suggest that there are hidden dangers such as wear or poor lubrication inside the mechanism. As the number of jams increases, the health score will gradually decrease.

[0026] In addition, the impact of the fault on the performance indicators of the pole-mounted circuit breaker will also be analyzed. For example, the increase in contact resistance caused by the fault may cause increased heat and power loss, which in turn affects the power transmission efficiency and the service life of the equipment. In this case, the health score will also be reduced accordingly. By comprehensively weighing these factors and using a specific scoring algorithm, the health status of the pole-mounted circuit breaker is converted into a specific health score value. This score can intuitively reflect the current health level of the pole-mounted circuit breaker and provide an important reference for subsequent fault prediction and maintenance decisions. A higher health score indicates that the equipment is in good condition and has a lower operating risk; a lower health score indicates that the equipment may have major problems and requires timely maintenance or overhaul.

[0027] The multimodal prediction data set acquisition module 40 is used to introduce an environmental parameter set, perform fault prediction on the pole-mounted circuit breaker according to the health score combined with the environmental parameter set, and obtain a multimodal prediction data set.

[0028] Specifically, the multimodal prediction data set acquisition module introduces an environmental parameter set, which adds key external factors to the fault prediction work. The environmental parameter set covers various key information in the operating environment of the pole-mounted circuit breaker, such as temperature, humidity, air pressure, light intensity, electromagnetic field intensity, etc. These environmental factors will have a significant impact on the performance and reliability of the pole-mounted circuit breaker. For example, high temperature environment may accelerate the aging of insulation materials, high humidity environment is easy to cause corrosion of electrical components, and strong electromagnetic field interference may affect the normal operation of the control circuit. This module predicts the fault of the pole-mounted circuit breaker according to the health score combined with the environmental parameter set. First, the health score is used as an important reference indicator, which reflects the current intrinsic health status of the pole-mounted circuit breaker. Combined with the environmental parameter set, the neural network algorithm is used to establish a fault prediction model. For example, in a high temperature and high humidity environment, if the health score of the pole-mounted circuit breaker is originally at a low level, according to the fault prediction model, the probability of insulation fault or electrical component corrosion fault will increase significantly. Through this fault prediction model, the operating status of the pole-mounted circuit breaker in the future is predicted to obtain a multimodal prediction data set. This data set may contain information on multiple modes, such as the probability of occurrence of different failure modes, the expected time range of failure, and the scope of possible impact of the failure. For example, the prediction results may show that in the next week, the probability of a short circuit failure of a pole-mounted circuit breaker due to insulation aging is 30%, and if a failure occurs, it may affect some downstream power users; or in the next month, due to environmental corrosion, the probability of poor contact failure in certain electrical connection parts is 15%, etc. These multi-modal prediction information provides a more comprehensive and forward-looking basis for subsequent maintenance decisions, helping operation and maintenance personnel to formulate reasonable maintenance plans in advance and reduce losses caused by sudden failures.

[0029] The maintenance suggestion formulation module 50 is used to jointly analyze the multimodal prediction data set according to the multiple data signals, perform coverage detection optimization on the operating parameter set according to the joint analysis results, and formulate maintenance suggestions.

[0030] Specifically, the maintenance recommendation formulation module is a key component of the intelligent detection system for pole-mounted circuit breakers. Its workflow begins with the joint analysis of multimodal prediction data sets and multiple data signals. It first accurately analyzes the data signals into two categories: electrical and mechanical signals, and then performs spatiotemporal alignment of the multimodal prediction data sets in turn to obtain electrical and mechanical alignment data sets. Then, data mapping is performed based on the operating parameter set to construct a data association network. Next, based on this network, the data set is subjected to transfer learning analysis, and multi-task joint learning is performed after assigning weight coefficients, arranging in descending order, and pruning to generate joint analysis results. Subsequently, the operating parameter set is divided into electrical and mechanical subspaces based on the joint analysis results, and an abnormal data signal is detected by coverage traversal. The abnormal location point is traced and cross-validated, and the risk is evaluated to obtain feedback data. Finally, the operating parameter set is optimized to generate an optimized set, and maintenance recommendations containing specific measures, time arrangements, and priorities are formulated to ensure the safe and efficient operation of the pole-mounted circuit breaker.

[0031] In a possible implementation, a fault diagnosis is performed on a pole-mounted circuit breaker according to the multiple data signals to generate a fault diagnosis result, and the device includes:

[0032] Perform multi-layer wavelet transform on the multiple data signals to obtain multiple signal features, where the multiple signal features include signal frequency domain features and signal time domain features.

[0033] The multiple data signals are analyzed according to the signal frequency domain characteristics to obtain signal spectrum data.

[0034] The multiple data signals are analyzed according to the signal time domain characteristics to obtain signal timing data.

[0035] A fault mode database is constructed by using machine learning combined with historical fault record logs. The signal spectrum data and the signal timing data are used as indexes. The fault mode database is traversed for matching. According to the matching results, it is determined whether there is a fault mode in the fault mode database that is consistent with the signal spectrum data and the signal timing data.

[0036] If there is a fault mode in the fault mode database that is consistent with the signal spectrum data and the signal timing data, it is considered that the pole-mounted circuit breaker has a fault, a fault data set is generated, a fault type is determined based on the fault data set, and the fault type is added to the fault diagnosis result.

[0037] Specifically, a multi-layer wavelet transform operation is performed on the acquired multiple data signals. Wavelet transform is a powerful signal processing technology. By convolving the wavelet basis functions of different scales and frequencies with the signal, the original signal is decomposed into multiple levels of detail signals and approximate signals, and then rich signal features are extracted from them. These features include signal frequency domain features and signal time domain features. The signal frequency domain features reveal information such as the energy distribution of the signal in the frequency domain, the composition of the frequency components, and the relative relationship between the frequency components; the signal time domain features reflect the amplitude, phase, period and other characteristics of the signal changing over time.

[0038] In-depth analysis of multiple data signals is carried out based on the signal frequency domain characteristics. Using spectrum analysis algorithms, such as fast Fourier transform (FFT) and its variants, the amplitude and phase of the signal at different frequency points are calculated to obtain the signal spectrum data. These spectrum data use frequency as the horizontal axis and amplitude or power as the vertical axis to intuitively display the frequency structure of the signal. For example, under normal operating conditions, the current signal spectrum of the pole-mounted circuit breaker has a specific shape and characteristic frequency components; when a fault occurs, such as an internal winding short circuit or insulation damage, new frequency components will appear in the spectrum or the amplitude of the original frequency components will change significantly, providing important clues for fault diagnosis.

[0039] At the same time, multiple data signals are analyzed based on the time domain characteristics of the signal. The time series analysis method is used to process the sampling points of the signal on the time axis, calculate the mean, variance, autocorrelation function, cross-correlation function, and extract time domain characteristic parameters such as peak value, valley value, rise time, fall time, pulse width, etc., so as to obtain the signal timing data. These timing data can reflect the changing trend and fluctuation of the signal at different times, as well as the time delay relationship with other signals. For example, by analyzing the timing data of the current or displacement signal when the operating mechanism is in action, it can be determined whether its action is smooth and whether there are abnormal conditions such as jamming or delay.

[0040] The fault mode database is constructed by combining machine learning technology with historical fault record logs. The historical fault record logs are collected from multiple data sources such as historical operation data, maintenance records, experimental test data, etc. of the pole-mounted circuit breaker. These logs contain relevant data signals of various fault conditions that occurred in the past. These logs are preprocessed, including data cleaning, denoising, normalization, etc., and then labeled to mark each fault event with a corresponding fault type label, such as "short circuit fault", "open circuit fault", "overheating fault", etc. According to these fault labels, the preprocessed historical fault record logs are clustered and analyzed, and data with similar characteristics are classified into one category, so as to determine multiple different fault types. The machine learning algorithms suitable for processing time series data, such as long short-term memory network (LSTM), are used to perform deep learning training on these determined fault types. The LSTM network can learn the long-term dependencies and time series patterns in the fault data, so as to accurately identify and distinguish the unique data patterns corresponding to different fault types and determine multiple fault modes. Finally, these fault modes are associated according to the fault labels, and serialized according to the chronological order or logical order of the fault occurrence to construct a complete fault mode database for subsequent rapid retrieval and matching. The signal spectrum data and signal time series data obtained through the above steps are used as indexes to perform comprehensive traversal matching in the constructed fault mode database. For each fault mode in the database, its similarity with the spectrum data and time series data of the current signal to be diagnosed is compared one by one. The method for calculating the similarity adopts multiple metrics such as Euclidean distance, cosine similarity, correlation coefficient, etc. The specific selection depends on the characteristics of the data and the application scenario. By traversing the entire database, it is determined based on the matching results whether there is a fault mode that is highly consistent with the current signal spectrum data and signal time series data.

[0041] If a fault mode consistent with the signal spectrum data and signal timing data is found in the fault mode database, then the pole-mounted circuit breaker is considered to have a fault. At this point, a fault data set containing all data signals related to the current fault is generated. This data set covers data from a period before the fault occurs to the time when the fault occurs and may continue to a period after the fault occurs, so as to comprehensively analyze the development process and characteristics of the fault. This fault data set is deeply analyzed, and the specific fault type is determined by using data mining technology, statistical analysis methods, and professional electrical and mechanical knowledge. For example, by analyzing the changing laws of signals such as current, voltage, temperature, and vibration in the fault data set, combined with the understanding of the internal structure and working principle of the pole-mounted circuit breaker, it is determined whether the fault is caused by damage to electrical components, wear of mechanical parts, degradation of insulation performance, or other reasons. Finally, the determined fault type is accurately added to the fault diagnosis results, providing clear guidance and basis for subsequent repair and maintenance work, helping operation and maintenance personnel to quickly locate the fault and take effective repair measures.

[0042] In one possible implementation, a failure mode database is constructed by combining machine learning with historical failure logs, and the device includes:

[0043] Data is retrieved according to multiple data sources of the pole-mounted circuit breaker to obtain the historical fault record log.

[0044] The historical fault record log is subjected to label processing to generate a plurality of fault labels, and the historical fault record log is subjected to cluster analysis according to the plurality of fault labels to determine a plurality of fault types.

[0045] A long short-term memory network is used to perform deep learning on the multiple fault types to determine multiple fault modes.

[0046] The multiple fault modes are associated according to the multiple fault labels, and the association results are serialized according to the running sequence to obtain the fault mode database.

[0047] Specifically, in order to build a comprehensive and accurate fault mode database, it is necessary to first retrieve data from multiple data sources of pole-mounted circuit breakers. These data sources cover the daily operation data records of pole-mounted circuit breakers, including electrical parameters such as current, voltage, power, etc. in different time periods, as well as data reflecting their mechanical operation status such as the number of switch actions and the stroke of the operating mechanism; equipment maintenance and overhaul records, which record in detail the time, maintenance content, and replacement of parts and other information of previous maintenance; fault repair reports, which clearly record the various fault phenomena, fault occurrence time, repair measures, and fault cause diagnosis and other key contents that have occurred; and may also include data obtained from various tests on pole-mounted circuit breakers in a laboratory environment. By integrating these data from different channels and in different forms, the original historical fault record log is formed. However, these original data often have problems such as inconsistent format, missing data, and noise interference. Therefore, a series of preprocessing operations are required, such as data cleaning to remove data records with obvious errors or duplications; data interpolation to supplement missing data points to ensure data continuity; data normalization to convert data of different dimensions to the same order of magnitude for subsequent analysis and processing. After these preprocessing steps, a relatively standardized, complete and analysis-suitable historical fault record log is obtained.

[0048] Labeling the preprocessed historical fault record log is one of the key steps in building a fault mode database. According to the nature, phenomenon, cause and degree of impact of the fault on the operation of the equipment, each historical fault record is assigned one or more fault labels with clear semantics. For example, if a fault is a short circuit fault caused by the aging of the insulation material inside the circuit breaker, it may be labeled as "insulation aging fault" and "short circuit fault". These labels can not only concisely summarize the core characteristics of the fault, but also provide an important basis for subsequent clustering analysis. Based on these fault labels, the clustering analysis algorithm is used to classify the historical fault record log. The clustering algorithm will classify records with similar fault labels or similar data features according to the similarity metric between the data. By continuously adjusting the clustering parameters and algorithm models, multiple relatively independent and representative fault types are finally determined. For example, different types of fault clusters such as "electrical control fault", "mechanical transmission fault" and "insulation fault" may be obtained, and the fault records in each cluster have high similarity in certain key features or fault modes.

[0049] After determining multiple fault types, a long short-term memory network (LSTM) is used to perform deep learning on these fault types. The LSTM network is a special recurrent neural network that is particularly suitable for processing data with time series characteristics, which is highly consistent with the operating data of the pole-mounted circuit breaker and the temporal characteristics of the fault occurrence process. The historical fault record data in each fault type is provided as input samples to the LSTM network for training. The network gradually builds a deep cognitive model for different fault types by learning the time series patterns in these data, the long-term dependencies between data features, and the laws of fault occurrence and development. During the training process, the network continuously adjusts its internal weight parameters to minimize the error between the predicted results and the actual fault labels. After repeated training of a large number of samples, the LSTM network can accurately identify and distinguish the complex data patterns corresponding to different fault types, thereby determining multiple fault modes that are closely related to the actual fault conditions. For example, for the "electrical control fault" type, the LSTM network may learn features such as the specific pattern of current fluctuations before the fault occurs and the abnormal change sequence of the control signal, thereby defining these feature combinations as a fault mode of this type of fault.

[0050] The multiple fault modes determined by the LSTM network are associated with the multiple fault labels generated previously. Each fault mode establishes a clear correspondence with the corresponding fault label to ensure that the classification and identification of the fault mode have a clear semantic interpretation. Then, these association results are serialized according to the actual operation sequence of the pole-mounted circuit breaker. This means that the fault modes are arranged and integrated according to the order of fault occurrence, the time stage in the fault development process, and the causal relationship or logical order between the fault modes. For example, for some faults, a slight abnormality of electrical parameters may occur first, followed by abnormal vibration of mechanical components, and finally a series of processes that cause the circuit breaker to trip. This time and logical sequence will be accurately reflected in the serialization process. Through such serialization processing, a complete fault mode database is constructed. This database not only contains rich and diverse fault mode information, but also reflects the dynamic process of fault occurrence and development and the inherent connection between different fault modes, providing a comprehensive, accurate and time-series reference basis for subsequent fault diagnosis, so that the current fault can be quickly located and matched with the historical fault mode in the database in the actual diagnosis process, improving the efficiency and accuracy of fault diagnosis.

[0051] In a possible implementation, the multimodal prediction data set is jointly analyzed according to the multiple data signals, and the apparatus includes:

[0052] Analyze and classify the multiple data signals to determine electrical signals and mechanical signals.

[0053] The multimodal prediction data set is time-space aligned according to the electrical signal to generate an electrical alignment data set.

[0054] The multimodal prediction data set is temporally and spatially aligned according to the mechanical signal to generate a mechanically aligned data set.

[0055] Data mapping is performed on the electrical alignment data set and the mechanical alignment data set based on the operating parameter set to obtain a data association network.

[0056] Based on the data association network, transfer learning analysis is performed on the multimodal prediction data set to generate a joint analysis result.

[0057] Specifically, in-depth analysis and classification work is carried out on multiple data signals from the data signal acquisition module. These data signals contain rich and diverse information, and electrical signals and mechanical signals are accurately distinguished from them through signal processing algorithms and technologies. Electrical signals mainly involve parameters related to the electrical performance of pole-mounted circuit breakers, such as current signals, which can reflect the amount of charge passing through the circuit breaker and the change of current, and are key to determining whether the circuit is overloaded, short-circuited or other electrical faults; voltage signals reflect the potential difference between two points in the circuit, and their stability and amplitude changes are closely related to the insulation performance of the equipment, power transmission efficiency, etc.; there are also signals such as power factor and harmonic content, which reveal the operating status of the electrical system from different angles. Mechanical signals focus on information related to the mechanical structure and moving parts of the pole-mounted circuit breaker. For example, the displacement signal of the operating mechanism accurately records the position changes of the operating mechanism during the switching action, which can be used to determine whether the mechanism is operating normally and whether the stroke meets the standards. The speed signal can reflect the speed of the operating mechanism. Abnormal speed changes may indicate wear, jamming or changes in spring force of mechanical parts. The vibration signal is also one of the important mechanical signals. By analyzing the frequency, amplitude, waveform and other characteristics of the vibration, problems such as looseness, imbalance and friction of mechanical parts can be detected.

[0058] After determining the electrical and mechanical signals, a spatiotemporal alignment operation is performed on the multimodal prediction data set. For the electrical signal part, the multimodal prediction data set is synchronously processed according to the time characteristics of the electrical signal to ensure that the electrical data collected from different sources or at different times are consistent on the time axis, for example, the current, voltage, power and other data at the same time can be accurately corresponded, so as to perform accurate joint analysis. At the same time, the electrical signals are aligned in space, taking into account the layout and connection relationship of the electrical components inside the pole-mounted circuit breaker, so that the relevant electrical parameters can be reasonably matched according to the actual physical position relationship, such as spatially correlating the current signals collected at different positions on the same phase line, so as to analyze the distribution of the current in the entire electrical circuit. After such a spatiotemporal alignment operation, an electrical alignment data set is generated, which provides a standardized and orderly electrical data foundation for subsequent precise analysis.

[0059] Similarly, for mechanical signals, the multimodal prediction data set is synchronously adjusted according to the time law of the mechanical signal, so that the displacement, speed, vibration and other signals of the operating mechanism can accurately correspond to the same action stage or operation time in time, thereby truly reflecting the dynamic changes of the mechanical structure during operation. In the spatial dimension, according to the structural composition and motion relationship of the mechanical parts of the pole-mounted circuit breaker, the mechanical signals are reasonably aligned in space, for example, the signals related to the mechanical parts at different positions on the same transmission chain are associated to fully understand the transmission process of mechanical force and the coordination of mechanical movement. Through such processing, a mechanical alignment data set is generated, which provides accurate data support for mechanical analysis.

[0060] Based on the operation parameter set of the pole-mounted circuit breaker, data mapping operations are performed on the electrical alignment data set and the mechanical alignment data set. The operation parameter set includes the design parameters of the pole-mounted circuit breaker (such as rated voltage, rated current, insulation level, etc.), operation history parameters (such as previous load conditions, operation time, number of operations, etc.), and physical structure parameters of the equipment (such as contact spacing, spring stiffness, etc.). By establishing a complex mathematical model and mapping relationship, the electrical parameters such as current and voltage in the electrical alignment data set are associated with the relevant parameters in the operation parameter set, and the mechanical parameters such as displacement and speed in the mechanical alignment data set are also matched with the operation parameter set. For example, based on electrical theory and mechanical motion principles, a relationship model between current and contact resistance and spring pressure is established, as well as a relationship model between displacement and operating mechanism travel and load force. Through this data mapping, data from the two fields of electrical and mechanical can be organically combined to obtain a data association network. This data association network clearly shows the intrinsic connection between electrical parameters and mechanical parameters, as well as their mutual influence on the overall operating status of the pole-mounted circuit breaker. For example, excessive current may cause overheating of the contacts, which in turn affects the performance of the operating mechanism. This causal relationship is reflected in the data association network.

[0061] Based on the constructed data association network, transfer learning analysis is performed on the multimodal prediction data set. Transfer learning uses existing data knowledge (in this case, the electrical-mechanical parameter relationship contained in the data association network) to help analyze new data (multimodal prediction data set). During the analysis process, weight coefficients are first assigned to the electrical alignment data set and the mechanical alignment data set, and corresponding weight values ​​are assigned to different parameters according to the importance of each parameter in the fault prediction and operation status assessment of the pole-mounted circuit breaker. For example, a higher weight is assigned to key parameters that directly affect the safe operation of the equipment (such as the current and voltage of the main circuit), while a lower weight is assigned to some auxiliary parameters that have a relatively small impact on the overall operation. Then, according to the weight coefficient allocation results, the data set is sorted in descending order, and priority is given to those parameter data with higher weights and greater impact on the operation of the equipment, and the initial descending list of multimodal data is obtained. Next, based on the data association network, the features of the predetermined last-place ratio threshold (such as 10% or 20%, etc., which can be set according to the actual situation) in the initial descending list of multimodal data are pruned to remove those data features that contribute less to fault prediction and analysis, and a descending list of multimodal data is obtained. Doing so can not only reduce the data dimension, reduce the amount of calculation and storage space, improve the analysis efficiency, but also avoid the accuracy of the analysis results affected by the interference of too much irrelevant data. Finally, according to the descending list of multimodal data combined with the multimodal prediction data set, multi-task joint learning is carried out, and various electrical and mechanical factors are comprehensively considered to generate joint analysis results. This joint analysis result can accurately identify the potential fault risk points, fault development trends and correlations between different fault modes of the pole-mounted circuit breaker. For example, it predicts the risk of poor electrical contact due to long-term mechanical wear, which may lead to short-circuit faults, and provides corresponding risk probability and time range information, providing a scientific and comprehensive basis for subsequent maintenance decisions.

[0062] Based on the operation parameter set of the pole-mounted circuit breaker, data mapping operations are performed on the electrical alignment data set and the mechanical alignment data set. The operation parameter set includes the design parameters of the pole-mounted circuit breaker (such as rated voltage, rated current, insulation level, etc.), operation history parameters (such as previous load conditions, operation time, number of operations, etc.), and physical structure parameters of the equipment (such as contact spacing, spring stiffness, etc.). By establishing a mapping relationship, the electrical parameters such as current and voltage in the electrical alignment data set are associated with the relevant parameters in the operation parameter set, and the mechanical parameters such as displacement and speed in the mechanical alignment data set are also matched with the operation parameter set. For example, based on electrical theory and mechanical motion principles, a relationship model between current and contact resistance and spring pressure is established, as well as a relationship model between displacement and operating mechanism stroke and load force. Through this data mapping, data from the two fields of electrical and mechanical can be organically combined to obtain a data association network. This data association network clearly shows the inherent connection between electrical parameters and mechanical parameters, as well as their mutual influence relationship with the overall operation status of the pole-mounted circuit breaker. For example, excessive current may cause overheating of the contact, thereby affecting the performance of the operating mechanism. This causal relationship is reflected in the data association network.

[0063] Based on the constructed data association network, transfer learning analysis is performed on the multimodal prediction data set. Transfer learning uses existing data knowledge (i.e., the electrical-mechanical parameter relationship contained in the data association network) to help analyze new data (multimodal prediction data set). During the analysis process, weight coefficients are first assigned to the electrical alignment data set and the mechanical alignment data set, and corresponding weight values ​​are assigned to different parameters according to the importance of each parameter in the fault prediction and operation status assessment of the pole-mounted circuit breaker. For example, a higher weight is assigned to the key parameters that directly affect the safe operation of the equipment (such as the current and voltage of the main circuit), while a lower weight is assigned to some auxiliary parameters that have a relatively small impact on the overall operation. Then, the data set is sorted in descending order according to the weight coefficient allocation result, and priority is given to those parameter data with higher weights and greater impact on the operation of the equipment, and the initial descending list of multimodal data is obtained. Next, based on the data association network, the features of the predetermined last-place ratio threshold (such as 10% or 20%, etc., which can be set according to the actual situation) in the initial descending list of multimodal data are pruned to remove those data features that contribute less to fault prediction and analysis, and a descending list of multimodal data is obtained. Doing so can not only reduce the data dimension, reduce the amount of calculation and storage space, improve the analysis efficiency, but also avoid the accuracy of the analysis results affected by the interference of too much irrelevant data. Finally, according to the descending list of multimodal data combined with the multimodal prediction data set, multi-task joint learning is carried out, and various electrical and mechanical factors are comprehensively considered to generate joint analysis results. This joint analysis result can accurately identify the potential fault risk points, fault development trends and correlations between different fault modes of the pole-mounted circuit breaker. For example, it predicts the risk of poor electrical contact due to long-term mechanical wear, which may lead to short-circuit faults, and provides corresponding risk probability and time range information, providing a scientific and comprehensive basis for subsequent maintenance decisions.

[0064] In a possible implementation, a transfer learning analysis is performed on the multimodal prediction data set based on the data association network to generate a joint analysis result, and the device includes:

[0065] Weight coefficients are assigned to the electrical alignment data set and the mechanical alignment data set to obtain a feature weight assignment result.

[0066] The electrical alignment data set and the mechanical alignment data set are arranged in descending order according to the feature weight distribution result to obtain an initial descending list of multimodal data.

[0067] Based on the data association network, features of a predetermined last ratio threshold in the initial descending list of multimodal data are pruned to obtain a descending list of multimodal data.

[0068] Multi-task joint learning is performed according to the descending list of the multimodal data in combination with the multimodal prediction data set to generate the joint analysis result.

[0069] Specifically, the weight coefficient allocation for the electrical alignment data set and the mechanical alignment data set is a key step. This process requires comprehensive consideration of many factors to determine the relative importance of each data feature in reflecting the operating status of the pole-mounted circuit breaker and predicting faults. The allocation is based on professional knowledge in the electrical and mechanical fields and the analysis experience of long-term operation data of pole-mounted circuit breakers. For example, for the features directly related to power transmission in the electrical alignment data set, such as the amplitude and phase of the main circuit current, the stability of the voltage, etc., a higher weight coefficient will be assigned, because the abnormal changes of these features are often closely related to serious electrical faults (such as short circuits and overloads); while for some auxiliary electrical features, such as small current signals in the control circuit, relatively low weight coefficients may be assigned. Similarly, in the mechanical alignment data set, the key action parameters of the operating mechanism, such as the speed of closing and opening, the accuracy of the displacement, etc., will be given a larger weight, because they directly affect the normal operation and reliability of the circuit breaker; while some relatively minor mechanical vibration features have lower weights. Through such a detailed weight coefficient allocation, a feature weight allocation result that can accurately reflect the importance of each feature is obtained.

[0070] According to the feature weight distribution results, the electrical alignment data set and the mechanical alignment data set are arranged in descending order, aiming to highlight the key feature data that have a greater impact on the operating status of the pole-mounted circuit breaker, so that subsequent analysis can give priority to these important information. The features in each data set are sorted from large to small according to their corresponding weight coefficients to form an initial descending list of multimodal data. In this list, the features at the top are the most critical factors in fault prediction and operating status assessment. For example, in terms of electricity, it may be the peak value of current, the fluctuation range of voltage, etc. In terms of mechanics, it may be the maximum stroke deviation of the operating mechanism, the vibration peak of key parts, etc. This sorting method helps to quickly locate and focus on the factors that are most likely to affect the normal operation of the pole-mounted circuit breaker in the subsequent analysis process, thereby improving the efficiency and accuracy of the analysis.

[0071] Based on the previously constructed data association network, the features of the predetermined last-place ratio threshold (e.g., set to 10% or 20%, etc., determined according to actual conditions and data analysis experience) in the initial descending list of multimodal data are pruned. The data association network contains the intrinsic relationship between electrical and mechanical parameters and their association with the overall operating status of the pole-mounted circuit breaker. By referring to these relationships, the data features in the initial descending list that are relatively unimportant and contribute less to fault prediction and analysis are identified. These features may be caused by measurement errors, environmental interference, or factors that have little effect on the main fault mode in the current analysis scenario. These features are removed from the list to obtain a descending list of multimodal data. The purpose of pruning is to reduce the data dimension, reduce the subsequent calculation amount and storage space requirements, and avoid affecting the accuracy of the analysis results due to interference from too much irrelevant data. For example, if an electrical feature exists in the data collection, but the data association network analysis finds that its correlation with the main fault mode is extremely low and its weight coefficient is also small, then it will be deleted during the pruning process.

[0072] Finally, multi-task joint learning is performed according to the descending list of multimodal data combined with the multimodal prediction data set to generate joint analysis results. Multi-task joint learning can make full use of the key feature information that has been screened in the descending list of multimodal data, as well as various fault prediction related data contained in the multimodal prediction data set (such as the probability distribution of different fault modes, the time prediction of fault occurrence, etc.). In the learning process, a variety of electrical and mechanical factors are comprehensively considered, and a neural network model is established to explore the deep relationship and potential laws between the data. For example, the model can learn how the change of electrical parameters affects the wear of mechanical parts, thereby affecting the failure probability of the entire pole-mounted circuit breaker; or how the abnormal vibration of mechanical parts generates feedback in the electrical system, resulting in a decrease in electrical performance. Through this multi-task joint learning, the joint analysis results finally generated can accurately identify the potential fault risk points, fault development trends, and the mutual influence relationship between different fault modes of the pole-mounted circuit breaker. For example, the joint analysis results may show that long-term mechanical wear has caused changes in the motion characteristics of the operating mechanism, which in turn has caused poor contact of the electrical contacts, and there is a high probability of causing a short circuit failure in the future. It also provides corresponding risk quantification indicators (such as a failure probability of 30%, an expected failure time of the next two months, etc.), providing a scientific and comprehensive basis for the subsequent formulation of targeted maintenance recommendations.

[0073] In one possible implementation, the operation parameter set is optimized for coverage detection based on the joint analysis results, and maintenance recommendations are formulated. The device includes:

[0074] The operating parameter set is divided according to the electrical signal and the mechanical signal to determine a plurality of operating parameter subspaces.

[0075] According to the joint analysis result, the multiple operating parameter subspaces are subjected to coverage traversal detection to generate coverage traversal results, wherein the coverage traversal detection results include electrical abnormality data signals and mechanical abnormality data signals.

[0076] The source is traced according to the electrical abnormality data signal to determine the first abnormality location, and the source is traced according to the mechanical abnormality data signal to determine the second abnormality location.

[0077] The first anomaly location and the second anomaly location are cross-validated to determine a plurality of anomaly location points, wherein the plurality of anomaly location points include a bidirectional anomaly location point and a unidirectional anomaly location point.

[0078] An abnormal risk assessment is performed on the bidirectional abnormal positioning point and the unidirectional abnormal positioning point to obtain abnormal operation risk feedback data.

[0079] The operation parameter set is optimized according to the abnormal operation risk feedback data to generate an operation optimization parameter set, and the maintenance suggestion is formulated according to the operation optimization parameter set.

[0080] Specifically, based on the characteristics of electrical signals and mechanical signals, the operating parameter set of the pole-mounted circuit breaker is divided in detail to determine multiple operating parameter subspaces. The operating parameter subspace related to electrical signals includes parameters that directly reflect electrical performance, such as rated voltage, rated current, power factor, and electrical insulation resistance; while the operating parameter subspace related to mechanical signals covers parameters that reflect mechanical structure and motion state, such as the stroke, speed, acceleration, elastic coefficient of the spring, and connection gap of mechanical parts of the operating mechanism. This division method helps to analyze and manage different types of operating parameters in a more targeted manner, laying the foundation for subsequent detection and optimization work.

[0081] Based on the joint analysis results, a comprehensive coverage traversal test is carried out on these divided multiple operating parameter subspaces. During the detection process, each indicator and threshold in the joint analysis results are compared one by one with the actual parameter values ​​in each operating parameter subspace. For the electrical parameter subspace, check whether the current exceeds the normal range (such as overload), whether the voltage is stable, whether the power factor meets the standard, and whether the electrical insulation performance is good. If it is found that it is inconsistent with the normal operating state, it will be recorded as an electrical abnormal data signal; for the mechanical parameter subspace, monitor whether the action stroke of the operating mechanism is accurate, whether the movement speed is normal, whether there is abnormal vibration or excessive wear of the mechanical parts, etc. Once an abnormality occurs, it will be determined as a mechanical abnormal data signal. Through such coverage traversal detection, a coverage traversal result containing electrical abnormal data signals and mechanical abnormal data signals is generated, which provides important clues for accurately finding the root cause of the fault.

[0082] According to the detected electrical abnormality data signal, a traceability analysis is performed to determine the first abnormal location. For example, if the current is found to increase abnormally, by analyzing the connection relationship of each component in the electrical circuit, the current transmission path and the electrical characteristics of the components, the source that may cause the current abnormality is gradually traced back, such as a loose electrical connector that causes an increase in contact resistance, and then the electrical connector is determined to be the first abnormal location. Similarly, based on the mechanical abnormality data signal, the mechanical structure and moving parts are checked to determine the second abnormal location. For example, if abnormal vibration of the operating mechanism is detected, the specific components or parts that cause the abnormal vibration are found by analyzing the mechanical transmission chain, the force conditions of each component and the kinematic relationship, such as wear of a gear that causes poor meshing, and the location of the gear is determined as the second abnormal location.

[0083] The first abnormal location is cross-validated with the second abnormal location to determine multiple abnormal location points. If there are both electrical and mechanical abnormal signs at a certain location, such as a connection part with both poor electrical contact (electrical abnormality) and abnormal vibration caused by looseness (mechanical abnormality), the point is determined as a bidirectional abnormal location point; if there is an abnormality only in one of the electrical or mechanical aspects, it is a unidirectional abnormal location point. A comprehensive abnormal risk assessment is conducted for these bidirectional abnormal location points and unidirectional abnormal location points. Factors to consider include the severity of the abnormality (such as whether it will immediately cause equipment failure or cause serious damage to the equipment), the development trend (whether it will gradually deteriorate or may remain stable), and the scope of influence on the operation of the entire pole-mounted circuit breaker system (whether it will affect other related equipment or the entire power distribution network). By comprehensively evaluating these factors, each abnormal location point is assigned a corresponding risk level, and abnormal operation risk feedback data is generated. These data intuitively reflect the risk status of each abnormal location point and provide a key basis for subsequent optimization decisions.

[0084] According to the abnormal operation risk feedback data, the operation parameter set is optimized and adjusted to generate an optimized operation parameter set. For abnormal positioning points with higher risks, targeted optimization measures are taken. For example, if the abnormal parameters of an electrical component lead to a higher risk, its operating voltage, current limit or replacement of the component will be adjusted to restore it to the normal operating parameter range; for problems with mechanical parts, such as adjusting the stroke parameters of the operating mechanism, replacing worn parts or reinforcing the mechanical structure, the relevant mechanical operation parameters are optimized. By optimizing each abnormal positioning point, the entire operation parameter set is updated and improved to form an optimized operation parameter set.

[0085] Finally, detailed maintenance recommendations are made based on the operation optimization parameter set. The maintenance recommendations include specific maintenance operations, such as replacement or repair steps for electrical components at electrical abnormality locations, and adjustment or replacement plans for mechanical components at mechanical abnormality locations; clarify the time schedule for maintenance work, determine whether emergency maintenance requires immediate shutdown or can be handled during regular inspections; and set maintenance work priorities, giving priority to abnormal locations with high risks and large impacts. These maintenance recommendations provide comprehensive, scientific, and targeted guidance for operation and maintenance personnel, helping to ensure that pole-mounted circuit breakers can operate continuously, stably, and efficiently, and ensure safe and reliable power supply to the power system.

[0086] In a possible implementation, the operating parameter set is optimized according to the abnormal operation risk feedback data to generate an optimized operating parameter set, and the device includes:

[0087] According to the abnormal operation risk feedback data, the operation parameter set is traversed to extract the i-th operation parameter, where i is an integer greater than 0.

[0088] Perform fitness analysis on the i-th operating parameter to obtain an i-th fitness value, wherein the i-th fitness value corresponds to the i-th operating parameter.

[0089] Determine whether the i-th fitness value is greater than or equal to the i+1-th fitness value. If so, add the i+1-th operating parameter to the eliminated data group. If so, add the i-th operating parameter to the eliminated data group. There is a corresponding relationship between the i+1-th operating parameter and the i+1-th fitness value.

[0090] When i satisfies the taboo table update cycle, the i-th fitness value or the i-th fitness value is input into the taboo table for updating, and it is determined whether the number of taboo table updates meets the preset number of updates. If the number of taboo table updates meets the preset number of updates, the taboo table update value is extracted based on the taboo table, and the i-th operating parameter or the i+1-th operating parameter is set as the operating optimization parameter set according to the taboo table update value.

[0091] Specifically, the operating parameter set is traversed to extract the i-th operating parameter (where i is an integer greater than 0). This process requires a systematic and detailed combing of the operating parameter set, which covers many key operating parameters of the pole-mounted circuit breaker, such as the rated voltage, rated current, power factor, phase angle, resistance value, capacitance value, inductance value, etc. in terms of electrical aspects. These parameters reflect the basic characteristics of the electrical performance of the equipment; the mechanical aspects include the stroke, speed, acceleration, closing time, opening time, elastic coefficient of the spring, tightness of the mechanical connection, wear degree of parts and other parameters of the operating mechanism, which reflect the important information of the mechanical structure and motion state of the equipment. During the traversal process, starting from the first operating parameter (i.e., when i=1), they are extracted one by one as the i-th operating parameter. This extraction action is not just a simple acquisition of parameter values, but also involves the identification and recording of the location of the parameter, its type (electrical or mechanical), and its relationship with other parameters. For example, when a current parameter is extracted, it is necessary to clarify whether it is a current in the main circuit or the control circuit, its normal value range, and its physical relationship with related parameters such as voltage and resistance. As the traversal proceeds, the value of i increases, and subsequent operating parameters are extracted in turn, thereby ensuring comprehensive coverage of the entire operating parameter set. Each extracted i-th operating parameter will serve as the basis for subsequent fitness analysis, comparative screening, and final optimization of the operating parameter set, providing key data support for accurate evaluation and optimization of the operating status of the pole-mounted circuit breaker.

[0092] After the extraction of the i-th operating parameter is completed, the fitness analysis is carried out to obtain the corresponding i-th fitness value, and there is a close correspondence between the two. The fitness analysis aims to comprehensively and deeply evaluate the influence of the i-th operating parameter on the operating state of the pole-mounted circuit breaker and its advantages and disadvantages. For the i-th operating parameter of the electrical type, such as the rated voltage parameter, if the actual operating voltage deviates too much from the rated voltage, whether it is too high or too low, it may cause equipment insulation damage, abnormal heating of electrical components, increased power loss and other problems, thereby reducing the reliability and service life of the equipment. Therefore, in the fitness analysis, according to the degree of voltage deviation from the rated value, the impact on other electrical parameters (such as current, power factor, etc.) and the impact on the overall operating stability of the equipment, a value reflecting its fitness, that is, the i-th fitness value, is calculated through a specific mathematical model and algorithm. If the voltage is always stable near the rated value and the impact on other parameters is within a reasonable range, then the fitness value of the voltage parameter will be higher; on the contrary, if the voltage fluctuates frequently and exceeds the safety range, its fitness value will be lower. For the i-th operating parameter of the mechanical type, take the stroke of the operating mechanism as an example. If the stroke of the operating mechanism does not meet the design requirements, the circuit breaker may not be closed or opened normally, affecting the normal power supply of the power system. In the fitness analysis, the size of the stroke deviation, the impact on the contact pressure of the contact, mechanical wear and operation time, etc. will be considered. If the stroke deviation is extremely small, the impact on the mechanical performance and operation sequence of the equipment is slight, and its fitness value will be higher; if the stroke deviation is too large, it will cause a series of mechanical performance problems, such as increased contact wear and operation jamming, and its fitness value will be lower. Through such a detailed analysis of the characteristics of different types of operating parameters, comprehensively considering factors such as its stability within the normal operating range, its synergy with other related parameters, and its contribution to the overall performance and reliability of the equipment, using professional analysis methods and calculation models, we can finally obtain the i-th fitness value that can accurately quantify the pros and cons of the i-th operating parameter, providing a key basis for the subsequent selection of the optimal combination among many operating parameters.

[0093] Next, compare the i-th fitness value with the i+1-th fitness value. If the i-th fitness value is greater than or equal to the i+1-th fitness value, this means that the i+1-th operating parameter performs worse than the i-th operating parameter under the current evaluation system, so the i+1-th operating parameter is added to the eliminated data group; if the i-th fitness value is less than the i+1-th fitness value, the i-th operating parameter is added to the eliminated data group. Through this pair-by-pair comparison and screening method, those operating parameters that contribute little to the optimization of equipment operation or may even have an adverse effect are gradually eliminated, thereby streamlining the operating parameter set and focusing on more valuable parameter combinations.

[0094] In this process, when i meets the taboo table update cycle (the taboo table update cycle is a pre-set parameter selection number or stage threshold used to control the update timing of the taboo table), the i-th fitness value or the i+1-th fitness value will be entered into the taboo table for update. The taboo table is a mechanism for recording and avoiding repeated selection of parameter combinations or parameter values ​​that have been judged to be poor. By updating the taboo table, it can be prevented from falling into the local optimal solution and being unable to jump out in the subsequent parameter selection process. After each update of the taboo table, it is necessary to determine whether the number of taboo table updates meets the preset number of updates. If the number of taboo table updates meets the preset number of updates, it means that enough iterations and explorations have been carried out. At this time, the taboo table update value is extracted based on the taboo table. This taboo table update value is an indication of a relatively good parameter value or parameter combination after a series of screening and optimization processes. Finally, the i-th operating parameter or the i+1-th operating parameter is set as the operating optimization parameter set according to the taboo table update value, so as to obtain the optimized operating parameter combination, which provides a more reasonable parameter configuration for the stable operation of the pole-mounted circuit breaker and improves the reliability and performance of the equipment.

[0095] Embodiment 2 is based on the same inventive concept as the intelligent detection device for pole-mounted circuit breakers in the above-mentioned embodiments. Figure 2 As shown, the present application provides an intelligent detection method for a pole-mounted circuit breaker, and the method and device embodiments in the present application are based on the same inventive concept. The method includes:

[0096] Step S100: traverse the pole-mounted circuit breakers to deploy multiple sensors, build a data sensor network, collect data from the pole-mounted circuit breakers through the data sensor network, and obtain multiple data signals.

[0097] Step S200: performing fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals to generate a fault diagnosis result.

[0098] Step S300: Perform health assessment according to the fault diagnosis result to obtain a health score of the pole-mounted circuit breaker.

[0099] Step S400: introducing an environmental parameter set, performing fault prediction on the pole-mounted circuit breaker according to the health score combined with the environmental parameter set, and obtaining a multi-modal prediction data set.

[0100] Step S500: jointly analyzing the multimodal prediction data set according to the multiple data signals, performing coverage detection optimization on the operating parameter set according to the joint analysis result, and formulating maintenance suggestions.

[0101] Furthermore, step S200 further includes:

[0102] Step S210: performing multi-layer wavelet transform on the multiple data signals to obtain multiple signal features, wherein the multiple signal features include signal frequency domain features and signal time domain features.

[0103] Step S220: Analyze the multiple data signals according to the signal frequency domain characteristics to obtain signal spectrum data.

[0104] Step S230: Analyze the multiple data signals according to the signal time domain characteristics to obtain signal timing data.

[0105] Step S240: Use machine learning combined with historical fault record logs to build a fault mode database, use the signal spectrum data and the signal timing data as indexes, traverse the fault mode database for matching, and determine whether there is a fault mode in the fault mode database that is consistent with the signal spectrum data and the signal timing data based on the matching results.

[0106] Step S250: If there is a fault mode in the fault mode database that is consistent with the signal spectrum data and the signal timing data, it is considered that there is a fault in the pole-mounted circuit breaker, a fault data set is generated, a fault type is determined according to the fault data set, and the fault type is added to the fault diagnosis result.

[0107] Furthermore, step S240 further includes:

[0108] Step S241: retrieve data according to multiple data sources of the pole-mounted circuit breaker to obtain the historical fault record log.

[0109] Step S242: labeling the historical fault record log to generate a plurality of fault labels, clustering the historical fault record log according to the plurality of fault labels, and determining a plurality of fault types.

[0110] Step S243: Use a long short-term memory network to perform deep learning on the multiple fault types to determine multiple fault modes.

[0111] Step S244: Associating the multiple fault modes according to the multiple fault labels, serializing the association results according to the running sequence, and obtaining the fault mode database.

[0112] Furthermore, step S500 further includes:

[0113] Step S510: Analyze and classify the multiple data signals to determine electrical signals and mechanical signals.

[0114] Step S520: performing spatiotemporal alignment on the multimodal prediction data set according to the electrical signal to generate an electrical alignment data set.

[0115] Step S530: performing spatiotemporal alignment on the multimodal prediction dataset according to the mechanical signal to generate a mechanically aligned dataset.

[0116] Step S540: performing data mapping on the electrical alignment data set and the mechanical alignment data set based on the operating parameter set to obtain a data association network.

[0117] Step S550: performing transfer learning analysis on the multimodal prediction data set based on the data association network to generate a joint analysis result.

[0118] Furthermore, step S550 further includes:

[0119] Step S551: performing weight coefficient assignment on the electrical alignment data set and the mechanical alignment data set to obtain a feature weight assignment result.

[0120] Step S552: Arrange the electrical alignment data set and the mechanical alignment data set in descending order according to the feature weight distribution result to obtain an initial descending list of multimodal data.

[0121] Step S553: ​​Pruning the features of the predetermined last ratio threshold in the initial descending list of the multimodal data based on the data association network to obtain a descending list of the multimodal data.

[0122] Step S554: performing multi-task joint learning in combination with the multimodal prediction data set according to the multimodal data descending list to generate the joint analysis result.

[0123] Furthermore, step S500 further includes:

[0124] Step S560: Divide the operating parameter set according to the electrical signal and the mechanical signal to determine a plurality of operating parameter subspaces; perform coverage traversal detection on the plurality of operating parameter subspaces according to the joint analysis result to generate a coverage traversal result, wherein the coverage traversal detection result includes an electrical abnormal data signal and a mechanical abnormal data signal; perform source tracing according to the electrical abnormal data signal to determine a first abnormal location, and perform source tracing according to the mechanical abnormal data signal to determine a second abnormal location; cross-validate the first abnormal location with the second abnormal location to determine a plurality of abnormal location points, wherein the plurality of abnormal location points include a bidirectional abnormal location point and a unidirectional abnormal location point; perform abnormal risk assessment on the bidirectional abnormal location point and the unidirectional abnormal location point to obtain abnormal operation risk feedback data; optimize the operating parameter set according to the abnormal operation risk feedback data to generate an operation optimization parameter set, and formulate the maintenance recommendation according to the operation optimization parameter set.

[0125] Furthermore, step S560 further includes:

[0126] Step S561: According to the abnormal operation risk feedback data, traverse the operation parameter set to extract the i-th operation parameter, where i is an integer greater than 0.

[0127] Step S562: Perform fitness analysis on the i-th operating parameter to obtain an i-th fitness value, where the i-th fitness value corresponds to the i-th operating parameter.

[0128] Step S563: Determine whether the i-th fitness value is greater than or equal to the i+1-th fitness value. If it is greater than or equal to, add the i+1-th operating parameter to the eliminated data group. If it is less than, add the i-th operating parameter to the eliminated data group. There is a corresponding relationship between the i+1-th operating parameter and the i+1-th fitness value.

[0129] Step S564: When i satisfies the taboo table update cycle, the i-th fitness value or the i-th fitness value is input into the taboo table for updating, and it is determined whether the number of taboo table updates meets the preset number of updates. If the number of taboo table updates meets the preset number of updates, the taboo table update value is extracted based on the taboo table, and the i-th operating parameter or the i+1-th operating parameter is set as the operating optimization parameter set according to the taboo table update value.

[0130] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0131] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0132] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An intelligent detection device for a pole-mounted circuit breaker, characterized in that: The device comprises: A data signal acquisition module, the data signal acquisition module is used to traverse the pole-mounted circuit breaker to deploy multiple sensors, build a data sensor network, collect data from the pole-mounted circuit breaker through the data sensor network, and obtain multiple data signals; A fault diagnosis result generating module, the fault diagnosis result generating module is used to perform fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals and generate a fault diagnosis result; A health score acquisition module, the health score acquisition module is used to perform a health assessment according to the fault diagnosis result to obtain a health score of the pole-mounted circuit breaker; A multimodal prediction data set acquisition module, the multimodal prediction data set acquisition module is used to introduce an environmental parameter set, perform fault prediction on the pole-mounted circuit breaker according to the health score combined with the environmental parameter set, and obtain a multimodal prediction data set; A maintenance suggestion formulation module, the maintenance suggestion formulation module is used to jointly analyze the multimodal prediction data set according to the multiple data signals, perform coverage detection optimization on the operating parameter set according to the joint analysis result, and formulate maintenance suggestions, including: The apparatus for jointly analyzing the multimodal prediction data set according to the multiple data signals comprises: Analyze and classify the multiple data signals to determine electrical signals and mechanical signals; Performing spatiotemporal alignment of the multimodal prediction dataset according to the electrical signal to generate an electrical alignment dataset; Performing spatiotemporal alignment on the multimodal prediction dataset according to the mechanical signal to generate a mechanically aligned dataset; Performing data mapping on the electrical alignment data set and the mechanical alignment data set based on the operating parameter set to obtain a data association network; Performing transfer learning analysis on the multimodal prediction data set based on the data association network to generate a joint analysis result; Dividing the operating parameter set according to the electrical signal and the mechanical signal to determine a plurality of operating parameter subspaces; Performing coverage traversal detection on the multiple operating parameter subspaces according to the joint analysis result to generate a coverage traversal result, wherein the coverage traversal result includes an electrical abnormality data signal and a mechanical abnormality data signal; Tracing the source according to the electrical abnormality data signal to determine the first abnormality location, and tracing the source according to the mechanical abnormality data signal to determine the second abnormality location; Cross-validate the first abnormality location and the second abnormality location to determine a plurality of abnormality location points, wherein the plurality of abnormality location points include bidirectional abnormality location points and unidirectional abnormality location points; Performing abnormal risk assessment on the bidirectional abnormal positioning point and the unidirectional abnormal positioning point to obtain abnormal operation risk feedback data; The operation parameter set is optimized according to the abnormal operation risk feedback data to generate an operation optimization parameter set, and the maintenance suggestion is formulated according to the operation optimization parameter set.

2. The intelligent detection device for a pole mounted circuit breaker according to claim 1, characterized in that: The device performs fault diagnosis on the pole-mounted circuit breaker according to the multiple data signals and generates a fault diagnosis result, and comprises: Performing multi-layer wavelet transform on the multiple data signals to obtain multiple signal features, wherein the multiple signal features include signal frequency domain features and signal time domain features; Analyze the multiple data signals according to the signal frequency domain characteristics to obtain signal spectrum data; Analyze the multiple data signals according to the signal time domain characteristics to obtain signal timing data; A fault mode database is constructed by using machine learning combined with historical fault record logs, the signal spectrum data and the signal time series data are used as indexes, the fault mode database is traversed for matching, and it is determined whether there is a fault mode in the fault mode database that is consistent with the signal spectrum data and the signal time series data according to the matching result; If there is a fault mode in the fault mode database that is consistent with the signal spectrum data and the signal timing data, it is considered that the pole-mounted circuit breaker has a fault, a fault data set is generated, a fault type is determined based on the fault data set, and the fault type is added to the fault diagnosis result.

3. The intelligent detection device for a pole mounted circuit breaker according to claim 2, characterized in that: The failure mode database is constructed by combining machine learning with historical failure logs. The device includes: Retrieving data from multiple data sources of the pole-mounted circuit breaker to obtain the historical fault record log; Labeling the historical fault record log to generate multiple fault labels, clustering the historical fault record log according to the multiple fault labels to determine multiple fault types; Using a long short-term memory network to perform deep learning on the multiple fault types to determine multiple fault modes; The multiple fault modes are associated according to the multiple fault labels, and the association results are serialized according to the running sequence to obtain the fault mode database.

4. The intelligent detection device for a pole mounted circuit breaker according to claim 1, characterized in that: Based on the data association network, transfer learning analysis is performed on the multimodal prediction data set to generate a joint analysis result, and the device includes: Performing weight coefficient allocation on the electrical alignment data set and the mechanical alignment data set to obtain a feature weight allocation result; Arranging the electrical alignment data set and the mechanical alignment data set in descending order according to the feature weight distribution result to obtain an initial descending list of multimodal data; Based on the data association network, pruning is performed on the features of the predetermined last ratio threshold in the initial descending list of the multimodal data to obtain a descending list of the multimodal data; Multi-task joint learning is performed according to the descending list of the multimodal data in combination with the multimodal prediction data set to generate the joint analysis result.

5. The intelligent detection device for a pole mounted circuit breaker according to claim 1, characterized in that: The operation parameter set is optimized according to the abnormal operation risk feedback data to generate an operation optimization parameter set, and the device includes: According to the abnormal operation risk feedback data, traverse the operation parameter set to extract the i-th operation parameter, where i is an integer greater than 0; Performing fitness analysis on the i-th operating parameter to obtain an i-th fitness value, wherein the i-th fitness value corresponds to the i-th operating parameter; Determine whether the i-th fitness value is greater than or equal to the i+1-th fitness value, if it is greater than or equal to, add the i+1-th operating parameter to the eliminated data group, if it is less than, add the i-th operating parameter to the eliminated data group, and there is a corresponding relationship between the i+1-th operating parameter and the i+1-th fitness value; When i satisfies the taboo table update cycle, the i-th fitness value or the i-th fitness value is input into the taboo table for updating, and it is determined whether the number of taboo table updates meets the preset number of updates. If the number of taboo table updates meets the preset number of updates, the taboo table update value is extracted based on the taboo table, and the i-th operating parameter or the i+1-th operating parameter is set as the operating optimization parameter set according to the taboo table update value.

6. An intelligent detection method for a pole-mounted circuit breaker, characterized in that: The method is used to implement the intelligent detection device for a pole-mounted circuit breaker according to any one of claims 1 to 5, and the method comprises: Traversing the pole-mounted circuit breakers to deploy multiple sensors, construct a data sensing network, collect data from the pole-mounted circuit breakers through the data sensing network, and obtain multiple data signals; Performing fault diagnosis on the pole-mounted circuit breaker according to the plurality of data signals to generate a fault diagnosis result; Perform health assessment according to the fault diagnosis result to obtain a health score of the pole-mounted circuit breaker; An environmental parameter set is introduced, and fault prediction of the pole-mounted circuit breaker is performed according to the health score combined with the environmental parameter set to obtain a multi-modal prediction data set; The multimodal prediction data set is jointly analyzed according to the multiple data signals, and coverage detection optimization is performed on the operating parameter set according to the joint analysis result to formulate maintenance suggestions.

Images