Circuit breaker fault sensing method and system based on multi-source information fusion

Abstract

The invention discloses a circuit breaker fault sensing method and system based on multi-source information fusion, belongs to the technical field of power equipment state monitoring and fault diagnosis, and aims to solve the problems of high false report and missing report rate, insensitive early fault and incomplete diagnosis of a single monitoring means. Comprising the following steps: synchronously acquiring multi-modal data such as vibration, sound wave, infrared, current, displacement and the like through a multi-source sensor array; performing signal preprocessing and feature extraction; performing preliminary intelligent diagnosis by adopting a three-layer fusion model comprising a data layer, a feature layer and a decision layer; performing verification and confidence fusion on the preliminary result by utilizing an expert knowledge base to generate a final diagnosis conclusion; triggering early warning according to the confidence level; according to the method, a complete circuit breaker state sensing system is constructed by fusing various physical information such as mechanical vibration, acoustics, thermotics and electricity, mutual verification can be realized, the one-sidedness of a single information source is overcome, the accuracy of fault diagnosis is greatly improved, and false alarm and missing alarm are remarkably reduced.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and fault diagnosis technology, specifically relating to a circuit breaker fault perception method and system based on multi-source information fusion. Background Technology

[0002] As the most important control and protection equipment in a power system, the reliability of circuit breakers directly affects the safe and stable operation of the power grid. Statistics show that circuit breaker failure is one of the main causes of power grid accidents, with mechanical failures accounting for over 40%, insulation failures and conductive circuit failures accounting for approximately 30% and 20%, respectively.

[0003] Currently, both domestic and international condition monitoring of circuit breakers mainly employs analysis methods based on single parameters, commonly including: Vibration signal analysis method: This method involves collecting vibration signals during the opening and closing process using vibration sensors installed on the circuit breaker mechanism box, and analyzing their time-domain and frequency-domain characteristics to diagnose mechanical faults. However, this method is susceptible to environmental noise interference and has difficulty in accurately locating the faulty component.

[0004] Coil current analysis method: This method indirectly diagnoses problems such as core jamming and inter-turn short circuits by monitoring the current waveform characteristics (such as peak value and time parameters) of the opening and closing coils. However, this method cannot effectively detect mechanical faults that are not directly electrically related to the operating mechanism.

[0005] Infrared thermometry: This method uses an infrared thermal imager to detect abnormal heating in circuit breaker contacts, terminals, and other parts to diagnose faults such as poor contact. However, this method cannot detect early internal mechanical defects and is significantly affected by load current and ambient temperature.

[0006] Acoustic detection method: This method uses ultrasonic or acoustic emission sensors to detect abnormal noises generated during operation. However, sound wave signals attenuate significantly during propagation and are easily affected by ambient noise.

[0007] These single-source monitoring methods have the following inherent flaws: Each method can only reflect one aspect of the circuit breaker's condition, like "blind men touching an elephant," failing to fully grasp the overall health status of the equipment. One fault may trigger multiple physical phenomena, and one physical phenomenon may be caused by multiple faults, leading to unclear diagnostic conclusions. Single sensor signals are easily affected by factors such as on-site electromagnetic interference, mechanical vibration, and ambient temperature and humidity, generating numerous false alarms or masking the true fault characteristics. In the early stages of a fault, its characteristic signals are often very weak, making it difficult to reliably detect by a single sensor in strong background noise, thus missing the best opportunity for early warning. It is difficult to distinguish the specific type of fault (such as whether it is spring fatigue or connecting rod loosening) and its precise location, bringing difficulties to operation and maintenance.

[0008] Therefore, there is a need for circuit breaker fault perception methods and systems based on multi-source information fusion to solve the problems of high false alarm and false alarm rates, insensitivity to early faults, and incomplete diagnosis in existing technologies using single monitoring methods. Summary of the Invention

[0009] The purpose of this invention is to provide a circuit breaker fault perception method and system based on multi-source information fusion, so as to solve the problems mentioned in the background art.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a circuit breaker fault detection method based on multi-source information fusion, comprising the following steps: S100: Through a multi-source sensor array deployed on the circuit breaker, under unified clock synchronization, it collects multi-modal physical quantity data including vibration signal, sound wave signal, infrared thermal image signal, coil current signal and opening and closing displacement signal; S200. Perform signal preprocessing and feature extraction on the multimodal physical quantity data to obtain feature vectors corresponding to different data sources; the preprocessing includes wavelet denoising and time-frequency analysis for vibration and acoustic signals, image correction and temperature field feature extraction for infrared thermal imaging signals, and key point feature extraction for coil current and displacement signals. S300, Constructing a hierarchical information fusion model for intelligent diagnosis, including: Data layer fusion: weighted averaging or wavelet fusion of homogeneous multi-source signals to improve the signal-to-noise ratio; Feature layer fusion: Normalize and concatenate feature vectors from different data sources, and use a sparse autoencoder for nonlinear dimensionality reduction to obtain low-dimensional deep feature vectors. Decision-level fusion: The deep feature vectors are input into a pre-trained multi-class deep learning model to obtain a preliminary fault type probability distribution; S400. Based on a preset expert knowledge base, production rules and evidence theory are used to verify and fuse the probability distribution of the preliminary fault types to generate the final comprehensive fault diagnosis result and confidence level. S500: Trigger the corresponding early warning mechanism based on the confidence level and output a diagnostic report containing the fault type, location, severity, and maintenance recommendations.

[0011] It should be noted in the scheme that in step S200, For vibration signals, ensemble empirical mode decomposition is used to extract intrinsic mode functions, and the energy entropy and sample entropy of each component are calculated as features; For acoustic signals, sound pressure level, specific frequency band energy, and Mel frequency cepstral coefficients are extracted as features; For infrared thermal imaging signals, the highest temperature, average temperature, temperature gradient, and thermal image texture features are extracted as features. For the coil current signal, the start-up time, rise time, peak current, duration, and area under integral of the curve are extracted as features; For displacement signals, the closing / opening time, overtravel, average speed, maximum speed, and time parameters of specific travel stages are extracted as features.

[0012] It is worth noting that the multi-class deep learning model in step S300 is a long short-term memory network that incorporates an attention mechanism, used to capture long-range dependencies and key features in multi-source feature sequences.

[0013] Furthermore, it should be noted that the expert knowledge base in step S400 is stored in the form of production rules, with the rule form being "IF<condition>THEN<conclusion>WITH<confidence>", where the condition is a logical combination of multiple feature values; the rule confidence and the model output probability are fused using DS evidence theory or Bayesian networks.

[0014] In a preferred embodiment, the confidence level in step S500 is divided into three levels: A comprehensive confidence level greater than 0.9 triggers a high-level alarm, resulting in immediate alerts and work order dispatch. A comprehensive confidence level between 0.7 and 0.9 triggers a medium-level warning, prompting attention and the arrangement of inspections; A comprehensive confidence level below 0.7 is considered normal, and only data recording and trend analysis are performed.

[0015] This invention also provides the following technical solution: a circuit breaker fault detection system based on multi-source information fusion, used to implement any of the circuit breaker fault detection methods based on multi-source information fusion described above, comprising: The data acquisition layer includes a multi-source sensor array, signal conditioning circuit, and synchronous data acquisition card deployed on the circuit breaker body; The data transmission and storage layer uses industrial network and Internet of Things (IoT) technologies to upload data and uses time-series databases and relational databases for storage. The edge computing and data processing center is a computing device deployed locally in the substation. It includes a data preprocessing and feature extraction module, a hierarchical information fusion and diagnostic module, a built-in sparse autoencoder and ATT-LSTM classification model, an expert knowledge base and inference engine module, and a model management module. The application and presentation layer provides a web-based human-computer interaction interface for status visualization, alarm management, and report generation. A cloud-based collaborative platform is used for centralized management, big data analysis, and model optimization of multiple edge nodes.

[0016] In a preferred embodiment, the multi-source sensor array of the data acquisition layer includes a high-frequency response acceleration vibration sensor, an acoustic emission sensor and microphone array, an infrared thermal imager, a Rogowski coil or Hall current sensor, a laser displacement sensor or an LVDT displacement sensor.

[0017] In a preferred embodiment, the edge computing and data processing center is deployed in an industrial-grade edge computing gateway or server, enabling real-time data acquisition, processing, diagnosis, and local alarms without relying on the cloud.

[0018] As a preferred implementation, the cloud-based collaborative platform can receive data from each edge node, use the collected data to retrain and optimize the diagnostic model, and send the optimized model parameters to the edge nodes for updates, thereby achieving self-evolution of diagnostic capabilities.

[0019] Compared with the prior art, the circuit breaker fault detection method and system based on multi-source information fusion provided by the present invention has at least the following beneficial effects: (1) This invention integrates various physical information such as mechanical vibration, acoustics, thermal and electrical information to construct a complete circuit breaker status perception system, which can mutually verify each other and overcome the one-sidedness of a single information source, greatly improving the accuracy of fault diagnosis and significantly reducing false alarms and missed alarms; and the multi-source information fusion technology can capture the weak and hidden features in the early stage of the fault and enhance them in the fusion process.

[0020] (2) This invention effectively suppresses the influence of various random noises and interferences on site by using a layered fusion model, especially the fusion of the data layer and the feature layer, thereby improving the robustness and reliability of the system. It also integrates monitoring, diagnosis, early warning and management functions into one, and introduces expert knowledge and adaptive learning mechanisms, making the system not only a monitoring tool, but also a constantly evolving intelligent diagnostic expert system that can provide direct and powerful support for operation and maintenance decisions. Attached Figure Description

[0021] Figure 1 This is a flowchart of a circuit breaker fault detection method based on multi-source information fusion according to the present invention. Figure 2 This is a structural block diagram of a circuit breaker fault sensing system based on multi-source information fusion according to the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to embodiments.

[0023] See also Figure 1 This invention provides a circuit breaker fault detection method based on multi-source information fusion, comprising the following steps: Step S100: Synchronous Acquisition of Multi-Source Heterogeneous Data By deploying a multi-source sensor array at key parts of the circuit breaker and triggering it with a unified clock source, multi-modal physical quantity data characterizing the circuit breaker's operating status are collected at high frequency and synchronously. The multi-modal physical quantity data includes at least: vibration signals, acoustic signals (including ultrasonic and audible sounds), infrared thermal image sequences, opening and closing coil current signals, and linear displacement signals of moving contacts. At the same time, the three-phase current, voltage, and ambient temperature and humidity during circuit breaker operation are recorded as auxiliary criteria.

[0024] Step S200: Signal preprocessing and feature extraction Targeted preprocessing and feature extraction are performed on the various raw signals acquired in step S100: For vibration signals: an adaptive wavelet threshold denoising method is used to eliminate environmental vibration interference. Then, intrinsic mode functions are extracted through ensemble empirical mode decomposition. The energy entropy, sample entropy, and time domain statistics (root mean square, kurtosis, peak value) of each IMF component are calculated as feature vectors F_vib.

[0025] For acoustic signals: First, perform spectrum analysis and bandpass filtering to highlight the characteristics of the operating sound, and then extract the sound pressure level, specific frequency band energy, and Mel frequency cepstral coefficients as the feature vector F_audio.

[0026] For infrared thermal imaging sequences: the images are processed using a non-uniformity correction algorithm based on reference temperature, and then the highest temperature, average temperature, temperature gradient, relative temperature difference, and thermal texture features of the key areas of the circuit breaker body are extracted as feature vectors F_thermal.

[0027] For coil current signals: the differential method is used to accurately identify key feature points such as the start and end points, peak points, and plateau regions of the current waveform, and the start time, rise time, peak current, duration, and curve integral area are extracted as feature vector F_current.

[0028] For displacement signals: the first and second derivatives of the displacement curve are obtained to obtain the velocity and acceleration curves, and the closing time, opening time, overtravel, average speed, maximum speed and time parameters of a specific travel stage are extracted as the feature vector F_displacement.

[0029] Step S300: Layered Information Fusion and Intelligent Diagnosis Construct a three-layer information fusion model for comprehensive diagnosis: Data layer fusion: Weighted averaging or wavelet fusion is performed on signals with the same physical meaning and sampling rate (such as signals from multiple vibration sensors) to improve the signal-to-noise ratio. Weighted averaging fusion distributes weights inversely proportional to sensor accuracy for signals from multiple sensors of the same type (such as two vibration sensors), using the following formula:

[0030] in , Let be the accuracy error of the i-th sensor; Wavelet fusion uses the Haar wavelet basis, with a decomposition layer of 3. The low-frequency components are taken as the mean and the high-frequency components are taken as the maximum absolute value, which improves the signal-to-noise ratio by ≥15dB.

[0031] Feature layer fusion: The heterogeneous feature vectors F_vib, F_audio, F_thermal, F_current, and F_displacement obtained in step S200 are normalized and then concatenated into a high-dimensional joint feature vector F_combined. A deep feature extraction network based on a sparse autoencoder is used to perform nonlinear dimensionality reduction on this high-dimensional vector to obtain a low-dimensional dense deep feature vector F_deep, which preserves the essential features of the original multi-source information to the greatest extent.

[0032] The normalization process uses z-score standardization, and the formula is: ,in , These are the mean and standard deviation of the historical baseline, respectively.

[0033] Decision layer fusion: The deep feature vector F_deep is input into a pre-trained multi-class deep learning model (such as the long short-term memory network ATT-LSTM with attention mechanism). The model outputs a preliminary fault diagnosis result, namely the probability distribution of various preset fault types P_preliminary=[p1,p2,...,pn].

[0034] Step S400: Knowledge-based decision optimization and validation Establish a circuit breaker fault knowledge base containing expert rules and historical cases; the expert rules are represented by production rules, for example: "IF (vibration signal energy in the 1000-2000Hz frequency band increases by 10dB compared to the historical baseline) AND (closing time is extended by more than 15%) THEN (diagnosis result is 'operating mechanism jamming') WITH (CF=0.85)"; the system matches and infers the preliminary diagnosis result of step S300 with the rules in the knowledge base, uses DS evidence theory or Bayesian network to correct and fuse the preliminary probabilities, calculates the final comprehensive confidence level CF_final, and generates a diagnosis report containing fault type, location, severity and confidence level.

[0035] Specifically, the DS evidence theory fusion includes: Definition of evidence body: Evidence E1 = the preliminary probability P_preliminary output by ATT-LSTM, and evidence E2 = the confidence level obtained by matching the knowledge base rules (if R1 is matched, the confidence level of E2 is 0.85); Basic probability assignment (BPA): For E1, , is the i-th type of fault. For E2, , match the corresponding rule fault, , is the universal set, indicating uncertainty; Fusion rule: Adopt the Yager conflict handling method (to avoid the fusion failure caused by high-conflict evidence), and the formula is: , which is enabled when the conflict coefficient k > 0.7; Final confidence level CF_final: Take the maximum value of the belief function Bel(A_i) of each fault category after fusion, .

[0036] Bayesian network verification includes: Network structure: The nodes include 10 feature nodes such as "vibration energy entropy", "contact temperature", "coil current peak", etc., and 6 fault nodes. The edges represent causal relationships (such as "low coil current peak" → "coil turn-to-turn short circuit"); Prior probability: Obtained by statistical analysis from the historical case library (such as the prior probability of "coil turn-to-turn short circuit" = the proportion of this fault in the cases = 12%); Posterior probability calculation: Update the posterior probability through Gibbs sampling. If the deviation between the fault probability output by the Bayesian network and the D-S fusion result > 10%, then re-match the knowledge base rules (to avoid the deviation of a single algorithm).

[0037] Step S500: Result output and warning action According to the diagnosis result and confidence level, trigger different warning mechanisms: If CF_final > 0.9, immediately trigger a high-level alarm, alarm through various methods such as sound and light, SMS, system pop-up window, etc., and generate a detailed maintenance suggestion report.

[0038] If 0.7 < CF_final ≤ 0.9, trigger an intermediate warning, prompt the operation and maintenance personnel to pay key attention and arrange inspections.

[0039] If CF_final ≤ 0.7, it is regarded as normal or slight change in the state, and the result is stored in the database for trend analysis.

[0040] All diagnosis results, original data and feature data are stored in the historical database for model iteration optimization and fault traceability.

[0041] See also Figure 2 This invention provides a circuit breaker fault detection system based on multi-source information fusion, used to implement the aforementioned circuit breaker fault detection method based on multi-source information fusion. The system includes a data acquisition layer, a data transmission and storage layer, an edge computing and data processing center, an application and display layer, and a cloud collaboration platform, as detailed below: The data acquisition layer includes a multi-source sensor array and a signal conditioning and acquisition unit. The multi-source sensor array includes, but is not limited to: A high-frequency piezoelectric acceleration vibration sensor (≥25kHz) is installed on the mechanism box and base.

[0042] Ultrasonic sensors and microphone arrays are used for acoustic emission and audible sound acquisition.

[0043] An uncooled focal plane infrared thermal imager is aimed at the circuit breaker's break point and conductive connection parts.

[0044] A high-precision Rogowski coil or Hall current sensor is fitted into the opening and closing coil circuit.

[0045] Laser displacement sensors or linear variable differential transformers are used to monitor the travel of the moving contact.

[0046] Signal conditioning and acquisition unit: including anti-aliasing filter, amplifier, synchronous clock generator and high-resolution data acquisition card, to ensure synchronous acquisition and high-quality digitization of multi-channel signals.

[0047] The data transmission and storage layer uses industrial Ethernet and / or wireless IoT technologies (such as 5GCPE, LoRaWAN) to upload the collected data to edge computing nodes or cloud platforms; the data transmission protocol supports industrial standard protocols such as MQTT and OPCUA, and has data encryption and disconnection resumption functions; it is equipped with a time-series database and a relational database to store massive amounts of raw data, feature data and diagnostic results.

[0048] The edge computing and data processing center can be deployed in the edge computing gateway or local server within the substation, and includes the following core software modules: Data preprocessing and feature extraction module: used to execute the algorithm of step S200 of the circuit breaker fault perception method based on multi-source information fusion.

[0049] Layered information fusion diagnostic module: embeds a pre-trained fusion model (SAE, ATT-LSTM, etc.) and executes step S300.

[0050] Expert knowledge base and inference engine module: Stores rules and cases, and performs logical reasoning in step S400.

[0051] Model Management Module: Responsible for online model updates, version control, and performance evaluation.

[0052] The application and presentation layer provides a web-based graphical human-computer interaction interface that supports access from PCs and mobile devices. Its functions include: real-time data dashboards, health status visualization (such as health index dashboards), fault alarm center, historical data query and analysis, automatic generation and export of diagnostic reports, and operation and maintenance work order management.

[0053] As a central node, the cloud-based collaborative platform receives data from multiple edge nodes, performs cross-site circuit breaker performance big data comparative analysis and in-depth mining; utilizes the collected massive data to centrally train and iteratively optimize the diagnostic model, and pushes the optimized model to each edge node to achieve the system's self-evolution.

[0054] In summary, this invention constructs a complete circuit breaker status perception system by integrating multiple physical information sources, including mechanical vibration, acoustics, thermal, and electrical information. This system can mutually verify each other, overcome the limitations of single information sources, greatly improve the accuracy of fault diagnosis, and significantly reduce false alarms and missed alarms. Furthermore, the multi-source information fusion technology can capture weak and hidden features in the early stages of faults and enhance them during the fusion process. Through a hierarchical fusion model, especially the fusion of the data layer and the feature layer, the influence of various random noises and interferences in the field is effectively suppressed, improving the robustness and reliability of the system. Moreover, the system integrates monitoring, diagnosis, early warning, and management functions and introduces expert knowledge and adaptive learning mechanisms, making the system not only a monitoring tool but also a continuously evolving intelligent diagnostic expert system that can provide direct and powerful support for operation and maintenance decisions.

[0055] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0056] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A circuit breaker fault detection method based on multi-source information fusion, characterized in that: Includes the following steps: S100: Through a multi-source sensor array deployed on the circuit breaker, under unified clock synchronization, it collects multi-modal physical quantity data including vibration signal, sound wave signal, infrared thermal image signal, coil current signal and opening and closing displacement signal; S200. Perform signal preprocessing and feature extraction on the multimodal physical quantity data to obtain feature vectors corresponding to different data sources; the preprocessing includes wavelet denoising and time-frequency analysis for vibration and acoustic signals, image correction and temperature field feature extraction for infrared thermal imaging signals, and key point feature extraction for coil current and displacement signals. S300, Constructing a hierarchical information fusion model for intelligent diagnosis, including: Data layer fusion: weighted averaging or wavelet fusion of homogeneous multi-source signals to improve the signal-to-noise ratio; Feature layer fusion: Normalize and concatenate feature vectors from different data sources, and use a sparse autoencoder for nonlinear dimensionality reduction to obtain low-dimensional deep feature vectors. Decision-level fusion: The deep feature vectors are input into a pre-trained multi-class deep learning model to obtain a preliminary fault type probability distribution; S400. Based on a preset expert knowledge base, production rules and evidence theory are used to verify and fuse the probability distribution of the preliminary fault types to generate the final comprehensive fault diagnosis result and confidence level. S500: Trigger the corresponding early warning mechanism based on the confidence level and output a diagnostic report containing the fault type, location, severity, and maintenance recommendations.

2. The circuit breaker fault detection method based on multi-source information fusion according to claim 1, characterized in that: In step S200 For vibration signals, ensemble empirical mode decomposition is used to extract intrinsic mode functions, and the energy entropy and sample entropy of each component are calculated as features; For acoustic signals, sound pressure level, specific frequency band energy, and Mel frequency cepstral coefficients are extracted as features; For infrared thermal imaging signals, the highest temperature, average temperature, temperature gradient, and thermal image texture features are extracted as features. For the coil current signal, the start-up time, rise time, peak current, duration, and area under integral of the curve are extracted as features; For displacement signals, the closing / opening time, overtravel, average speed, maximum speed, and time parameters of specific travel stages are extracted as features.

3. The circuit breaker fault detection method based on multi-source information fusion according to claim 1, characterized in that: The multi-class deep learning model in step S300 is a long short-term memory network that incorporates an attention mechanism, used to capture long-range dependencies and key features in multi-source feature sequences.

4. The circuit breaker fault detection method based on multi-source information fusion according to claim 1, characterized in that: The expert knowledge base in step S400 is stored in the form of production rules, with the rule form being "IF<condition>THEN<conclusion>WITH<confidence level>", where the condition is a logical combination of multiple feature values; the rule confidence level and the model output probability are fused using DS evidence theory or Bayesian network.

5. The circuit breaker fault detection method based on multi-source information fusion according to claim 1, characterized in that: In step S500, the confidence level is divided into three levels: A comprehensive confidence level greater than 0.9 triggers a high-level alarm, resulting in immediate alerts and work order dispatch. A comprehensive confidence level between 0.7 and 0.9 triggers a medium-level warning, prompting attention and the arrangement of inspections; A comprehensive confidence level below 0.7 is considered normal, and only data recording and trend analysis are performed.

6. A circuit breaker fault detection system based on multi-source information fusion, used to implement the circuit breaker fault detection method based on multi-source information fusion as described in any one of claims 1-5, characterized in that: include: The data acquisition layer includes a multi-source sensor array, signal conditioning circuit, and synchronous data acquisition card deployed on the circuit breaker body; The data transmission and storage layer uses industrial network and Internet of Things (IoT) technologies to upload data and uses time-series databases and relational databases for storage. The edge computing and data processing center is a computing device deployed locally in the substation. It includes a data preprocessing and feature extraction module, a hierarchical information fusion and diagnostic module, a built-in sparse autoencoder and ATT-LSTM classification model, an expert knowledge base and inference engine module, and a model management module. The application and presentation layer provides a web-based human-computer interaction interface for status visualization, alarm management, and report generation. A cloud-based collaborative platform is used for centralized management, big data analysis, and model optimization of multiple edge nodes.

7. A circuit breaker fault detection system based on multi-source information fusion according to claim 6, characterized in that: The multi-source sensor array of the data acquisition layer includes a high-frequency response acceleration vibration sensor, an acoustic emission sensor and microphone array, an infrared thermal imager, a Rogowski coil or Hall current sensor, a laser displacement sensor or an LVDT displacement sensor.

8. A circuit breaker fault detection system based on multi-source information fusion according to claim 6, characterized in that: The edge computing and data processing center is deployed in an industrial-grade edge computing gateway or server, and can complete real-time data acquisition, processing, diagnosis and local alarm without relying on the cloud.

9. A circuit breaker fault detection system based on multi-source information fusion according to claim 6, characterized in that: The cloud-based collaborative platform can receive data from each edge node, use the collected data to retrain and optimize the diagnostic model, and send the optimized model parameters to the edge nodes for updates, thereby achieving self-evolution of diagnostic capabilities.

Images