A fault diagnosis method and system for an intelligent bus duct

Abstract

The application discloses a fault diagnosis method and system of an intelligent bus duct, relates to the technical field of fault diagnosis, and solves the technical problem that the prior art is difficult to capture the abnormality of the bus duct under dynamic operation conditions, such as characteristic drift caused by factors such as environmental temperature and humidity, long-term vibration accumulation, and harmonic component change, and false positives and false negatives are prone to occur. The application obtains multi-source real-time data of the bus duct, performs feature extraction on the multi-source real-time data based on a feature extraction technology to obtain feature data, constructs a digital twin model of the bus duct, determines a dynamic threshold corresponding to the feature data based on the digital twin model, and identifies and diagnoses the fault of the bus duct according to the feature data and the dynamic threshold.

Description

Technical Field

[0001] This application belongs to the field of fault diagnosis and relates to fault diagnosis technology for bus trunking, specifically a fault diagnosis method and system for intelligent bus trunking. Background Technology

[0002] Busbar trunking is an enclosed power distribution device composed of copper or aluminum conductive bars, insulating materials, and a metal shell. It is used to efficiently collect, distribute, and transmit high-current electrical energy. Replacing traditional cables with metal trunking, it serves as the main power supply line in high-rise buildings, factories, and other scenarios, offering advantages such as high current carrying capacity, low voltage drop, safety and reliability, and flexible tapping. Fault diagnosis of busbar trunking can significantly improve the safety and reliability of power distribution systems. As the main power distribution line in buildings and industries, busbar trunking is susceptible to serious problems such as overheating, edge aging, and loose joints. These faults can lead to local power outages, fires, or equipment damage. Regular or real-time fault diagnosis allows for precise identification of problems at their early stages, such as abnormal temperature rise, decreased insulation resistance, and increased contact resistance, thus enabling early fault identification and preventing the fault from escalating. At the same time, fault diagnosis can effectively reduce unplanned downtime, lower maintenance costs, and extend the service life of busbar trunking.

[0003] In existing technologies, fault diagnosis methods typically rely on simple thresholds, such as fixed temperature limits or current fluctuation ranges; or periodic manual inspections for judgment. These fixed threshold methods lack real-time performance and adaptability, making it difficult to capture anomalies in busbar trunking under dynamic operating conditions, such as characteristic drift caused by factors like ambient temperature and humidity, long-term vibration accumulation, and harmonic component changes, which can easily lead to false alarms and missed alarms. Manual inspections, on the other hand, are costly, inefficient, and highly subjective, failing to achieve real-time monitoring and early warning. This application provides a fault diagnosis method and system for intelligent bus trunking to solve the above-mentioned technical problems. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art; to this end, this application proposes a fault diagnosis method and system for intelligent bus trunking, which is used to solve the technical problem that the prior art is unable to capture the anomalies of bus trunking under dynamic operating conditions, such as characteristic drift caused by factors such as environmental temperature and humidity, long-term vibration accumulation, and harmonic component changes, which easily leads to false alarms and missed alarms.

[0005] To achieve the above objectives, the first aspect of this application provides a fault diagnosis method for intelligent busbar trunking, comprising: Acquire multi-source real-time data of the busbar trunking; Feature data is obtained by extracting features from multi-source real-time data based on feature extraction technology. Constructing a digital twin model of the busbar trunking; Determine the dynamic threshold corresponding to the feature data based on the digital twin model; Faults in busbar trunking are identified and diagnosed based on characteristic data and dynamic thresholds.

[0006] Preferably, the feature extraction based on feature extraction technology for multi-source real-time data includes: Retrieve real-time data from multiple sources; these real-time data include: real-time temperature data, real-time current data, real-time voltage data, real-time humidity data, real-time vibration data, and partial discharge signal data. The physical characteristics of the busbar trunking are analyzed based on real-time temperature, humidity, and vibration data; the electrical characteristics are analyzed based on real-time current, voltage, and partial discharge signal data. The physical characteristics include temperature, humidity, and vibration characteristics; the electrical characteristics include current, voltage, and discharge characteristics.

[0007] Preferably, the step of analyzing the physical characteristics of the busbar trunking based on real-time temperature data, real-time humidity data, and real-time vibration data includes: Real-time temperature data, real-time humidity data, and real-time vibration data are retrieved; the temperature characteristics of the busbar are analyzed based on a temperature change rate analysis function; the expression of the temperature change rate analysis function is: ;in, ; The first-order rate of change of real-time temperature; The data represents the real-time temperature data collected at time t. The sampling frequency time interval; Based on the humidity change rate formula Calculate the rate of change of humidity in real time; where, The real-time humidity data is collected at time t; the mean and variance of the humidity are calculated based on a set time sliding window; the formula for calculating the mean humidity is: The formula for calculating the variance of humidity is: ; The cumulative humidity hazard level of the busbar trunking is analyzed based on the cumulative humidity hazard analysis function; the expression of the cumulative humidity hazard analysis function is: The humidity change rate, mean and variance of humidity, and cumulative humidity hazard are integrated into a humidity characteristic; among which, This is the current analysis time. To monitor the start time; For time infinitesimal elements; The humidity safety threshold; Real-time vibration data over a set time period are integrated to obtain a vibration signal; the characteristics of the vibration signal are analyzed using wavelet transform to obtain vibration features; the expression for the wavelet transform is: ;in, For scale parameters, For translation parameters, For the mother wavelet function; It is a vibration signal.

[0008] Preferably, the analysis of the electrical characteristics of the busbar trunking based on real-time current data, real-time voltage data, and partial discharge signal data includes: Real-time current data, real-time voltage data, and partial discharge signal data are retrieved; the real-time current data within a set time period are integrated to obtain a current signal; the current is processed using a Fast Fourier Transform (FFT) to obtain frequency domain characteristics; the expression for the FFT is: , ;in, It is a current signal; The number of points in the Fourier transform; For discrete frequency indexes in the frequency domain; It is the rotation factor; The total harmonic distortion (THD) of a current signal is analyzed based on a total harmonic distortion (THD) analysis function; the expression for the THD analysis function is as follows: ;in, Let be the amplitude of the kth harmonic, and its expression is: ; The fundamental amplitude; Voltage imbalance is analyzed based on a voltage imbalance analysis function to obtain voltage characteristics; the discharge characteristics of the busbar trunking are analyzed based on partial discharge signals; the expression of the voltage imbalance analysis function is as follows: ;in ; , and These are the real-time voltage values ​​of the three phases A, B, and C of the busbar.

[0009] Preferably, the analysis of the discharge characteristics of the busbar trunking based on partial discharge signals includes: Partial discharge signals are retrieved and integrated within a set time period to obtain a discharge analysis sequence; The discharge analysis model is invoked, and the discharge analysis sequence is input into the discharge analysis model to obtain the discharge characteristics of the busbar; the discharge analysis model is constructed based on an artificial intelligence model.

[0010] Preferably, the discharge analysis model is constructed based on an artificial intelligence model, including: Select a model framework and deep learning algorithm from the artificial intelligence model library; build a model based on the model framework and deep learning algorithm to obtain the initial model; Obtain the standard dataset; the standard dataset includes standard input data consistent with the content attributes of the discharge analysis sequence, and standard output data consistent with the content attributes of the discharge characteristics; The standard dataset is divided into a training set, a validation set, and a test set according to a set ratio; the initial model is trained using the training set; the internal parameters of the initial model are adjusted using the validation set; and the trained initial model is tested using the test set to obtain test metrics. Obtain the standard range of indicators; compare the test indicators with the standard range of indicators; if all test indicators are within the standard range of indicators, mark the initial model as the discharge analysis model; otherwise, rebuild and train the initial model again.

[0011] Preferably, the construction of the digital twin model of the busbar trunking includes: Obtain the physical structure parameters, material parameters, and electrical parameters of the busbar trunking; use 3D modeling software to construct the geometric structure model of the busbar trunking based on its physical structure parameters; The material and electrical parameters of the busbar are imported into the geometric model of the busbar to establish a mathematical model of the busbar; the historical data of the busbar is transmitted to the mathematical model of the busbar for simulation to obtain a digital twin model of the busbar.

[0012] Preferably, the step of determining the dynamic threshold corresponding to the feature data based on the digital twin model includes: Obtain normal historical data; retrieve feature data for a specified time period; input the historical data into the digital twin model to obtain the baseline value corresponding to the feature data; Calculate the difference between the feature data and the benchmark value to obtain the residual; then use the formula... Recursively calculate the mean of the residuals; based on the formula The residual variance is recursively calculated; where, Forgetting factor, Let be the residual at time t; Let be the mean residual at time t-1; The dynamic expansion coefficient of the threshold is calculated based on the dynamic coefficient analysis function; the expression for the dynamic coefficient analysis is: ;in, The length of the sliding window; Let i be the value of the i-th residual sequence; The average value is calculated for the residuals within the window; Based on formula Calculate the final dynamic threshold limit, and determine the dynamic threshold range based on the dynamic threshold limit; where, ; This is the baseline coefficient.

[0013] Preferably, the step of identifying and diagnosing busbar faults based on feature data and dynamic thresholds includes: Retrieve feature data and corresponding dynamic thresholds; compare the feature data with the corresponding dynamic thresholds. When all feature data are within the corresponding dynamic threshold range, a fault alarm signal is generated; otherwise, the fault identification and diagnosis of the bus trunking continues.

[0014] The second aspect of this application provides a fault diagnosis system for intelligent bus trunking, including: a data acquisition module, a model building module, and a fault diagnosis module; The data acquisition module is used to acquire multi-source real-time data of the busbar trunking; The model building module is used to extract features from multi-source real-time data based on feature extraction technology to obtain feature data; and to build a digital twin model of the busbar. The fault diagnosis module is used to determine the dynamic threshold corresponding to the feature data based on the digital twin model; and to identify and diagnose the faults of the bus trunking based on the feature data and the dynamic threshold.

[0015] Compared with the prior art, the beneficial effects of this application are: 1. This application employs multi-source real-time data collaborative analysis, establishing a more comprehensive dynamic feature description system for busbar trunking status through multi-dimensional extraction of physical and electrical characteristics. These features, compared to single indicators, can more sensitively reflect the evolution trend of potential faults. Furthermore, by constructing a digital twin model, the operating characteristics of the busbar trunking under normal conditions are simulated, and adaptive thresholds for feature data are calculated in real time based on the dynamic expansion coefficient method. This allows diagnostic criteria to follow changes in operating conditions and environment, significantly reducing misjudgments of thresholds caused by external interference or equipment aging. Therefore, this method can achieve early and high-precision fault warnings, avoiding missed detections caused by the poor adaptability of traditional fixed thresholds, improving the intelligence level and response speed of the monitoring system, and providing strong technical support for timely detection of potential hazards such as overheating, partial discharge, and loose joints, preventing the escalation of faults.

[0016] 2. This application integrates artificial intelligence models and data-driven models, forming a high degree of simulation and prediction capability for the overall state of the busbar trunking. On the one hand, it uses artificial intelligence methods to intelligently identify and extract features from partial discharge signals, overcoming the limitations of traditional discharge analysis methods that require extensive experience and manual parameter setting. On the other hand, through continuous learning of historical data and operating scenarios using a digital twin model, it continuously optimizes dynamic thresholds, forming a closed-loop learning feedback mechanism, thereby enhancing the adaptive evolution capability of the entire diagnostic system. This intelligent diagnosis not only enables early and accurate fault detection but also significantly reduces reliance on regular manual inspections and on-site maintenance, thus significantly reducing maintenance manpower costs and unplanned downtime. The long-term stable operation of the system helps to extend the overall service life of the busbar trunking, ensuring the highly reliable and safe continuous operation of the power supply and distribution backbone, and has important practical significance for preventing serious accidents such as fires caused by busbar trunking faults, possessing good economic and social benefits. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram illustrating the overall steps of the method described in this application; Figure 2 This is a schematic diagram showing the connection of the system modules in this application. Detailed Implementation

[0019] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] Example 1: Please see Figure 1 The first aspect of this application provides a fault diagnosis method for intelligent busbar trunking, including: S110. Obtain multi-source real-time data of the busbar trunking.

[0021] In this embodiment, multi-source real-time data refers to a collection of various sensor data that can reflect the current operating status and environmental influencing factors of the busbar trunking. Specifically, data acquisition can be achieved through a sensor network deployed at key nodes of the busbar trunking (such as joints, conductor surfaces, and outer shells). Data types include, but are not limited to, real-time temperature, current, voltage, humidity, vibration, and partial discharge signals. It should be understood that although various data types are listed here, in other embodiments, some data or other types of sensor data (such as infrared thermography, gas concentration, etc.) can be selected according to actual monitoring needs, as long as these data can provide effective information for subsequent status assessment. By acquiring multi-source heterogeneous data, the operating status of the busbar trunking can be comprehensively perceived from multiple dimensions such as heat, electricity, force, and environment, overcoming the shortcomings of insufficient characterization capabilities of single indicators and laying a solid data foundation for subsequent fault diagnosis.

[0022] S120. Based on feature extraction technology, feature extraction is performed on multi-source real-time data to obtain feature data.

[0023] Because raw data often contains a lot of noise and has high dimensionality, direct use for diagnosis is inefficient. Therefore, feature extraction techniques are needed to transform it into more characterizing data about the equipment's health status. Specifically, feature extraction can include signal preprocessing, time-domain / frequency-domain transformation, and statistical calculations, converting the original waveforms or numerical sequences into indicators with clear physical meaning, such as temperature change rate, harmonic distortion rate, and vibration energy spectrum. Compared to the raw data, this feature data not only significantly reduces the amount of data but is also more sensitive to fault symptoms. The core function of this step is to extract key information from massive monitoring data, achieving data dimensionality reduction and value concentration, enabling subsequent diagnostic models to focus on the essential characteristics that truly reflect the equipment's condition, thereby improving diagnostic accuracy and computational efficiency.

[0024] S130, Construct a digital twin model of the busbar trunking.

[0025] A digital twin model is a high-fidelity mapping of the physical entity of a busbar trunking system in virtual space. Its construction process integrates multi-source information, including the busbar trunking's geometry, material properties, electrical parameters, and historical operating data. Specifically, a geometric model can be created using 3D modeling software, and its physical behavior characteristics can be assigned by combining finite element simulation or multiphysics coupling algorithms. The model parameters are then calibrated using historical data to accurately simulate the theoretical response of the busbar trunking system under different operating conditions. This model can not only reproduce the static structure of the equipment but also dynamically extrapolate its normal state baseline values ​​under specific load and environmental conditions. By introducing a digital twin model, this application provides a dynamically updated "health benchmark" for fault diagnosis based on individual equipment differences and aging levels, solving the problem that traditional methods rely on general empirical thresholds and cannot adapt to specific equipment characteristics.

[0026] S140. Determine the dynamic threshold corresponding to the feature data based on the digital twin model.

[0027] This is the key step that distinguishes this application from existing technologies. Traditional fault diagnosis typically uses fixed alarm thresholds, which are difficult to adapt to the characteristic drift of busbars caused by changes in ambient temperature, load fluctuations, or long-term aging, easily leading to false alarms or missed alarms. In this embodiment, however, the dynamic threshold is calculated in real time based on the baseline value output by the digital twin model, combined with the deviation distribution law between real-time feature data and the baseline value. This means that the threshold boundary is no longer a static straight line, but a dynamic curve that can automatically expand and contract with changes in operating conditions. For example, when the load increases and causes an overall temperature rise, the dynamic threshold will shift upward accordingly to accommodate normal temperature rise trends; while when the environment is stable, the threshold will tighten to improve sensitivity to minor anomalies. This adaptive mechanism fundamentally eliminates the misleading influence of external interference on diagnostic results and significantly improves the robustness of the system in complex dynamic environments.

[0028] S150. Identify and diagnose busbar faults based on feature data and dynamic thresholds.

[0029] After feature extraction and dynamic threshold generation, the system compares real-time feature data with the current dynamic threshold range to determine whether the busbar trunking is in an abnormal state. Specifically, when the feature data exceeds the normal fluctuation range defined by the dynamic threshold, the system determines that a potential fault exists and triggers a corresponding diagnostic response; conversely, if the feature data remains within the dynamic threshold range, the equipment is considered to be operating normally, and the system continues to monitor it. This judgment logic fully utilizes the adaptive characteristics of the aforementioned dynamic threshold, ensuring that an alarm is only issued when the feature deviates from the normal trajectory to a statistically significant level, effectively filtering out false signals caused by environmental noise or normal operating condition switching. Through the collaborative work of the entire process described above, this embodiment achieves early detection and accurate identification of faults in intelligent busbar trunking, providing reliable technical support for ensuring the safe and stable operation of the power supply and distribution system.

[0030] Example 2: In this embodiment, the process of feature extraction from multi-source real-time data based on feature extraction technology is further refined to construct a multi-dimensional state characterization system. Specifically, multi-source real-time data is retrieved, including real-time temperature data, real-time current data, real-time voltage data, real-time humidity data, real-time vibration data, and partial discharge signal data. The physical characteristics of the busbar trunking are analyzed based on the real-time temperature data, real-time humidity data, and real-time vibration data. The electrical characteristics of the busbar trunking are analyzed based on the real-time current data, real-time voltage data, and partial discharge signal data. The physical characteristics include temperature characteristics, humidity characteristics, and vibration characteristics; the electrical characteristics include current characteristics, voltage characteristics, and discharge characteristics. This approach of dividing multi-source heterogeneous data into two dimensions, physical and electrical, is not a simple classification but a decoupled design based on the fault mechanism of the busbar trunking. Physical characteristics mainly reflect the thermodynamic state and mechanical structural integrity of the equipment, such as overheating, moisture, or loose bolts; while electrical characteristics focus on characterizing electromagnetic performance and insulation status, such as harmonic pollution, three-phase imbalance, or internal discharge. By using dual-dimensional collaborative analysis, misjudgments caused by abnormalities in a single indicator can be effectively avoided. For example, it can distinguish between overall temperature rise caused by high ambient temperature and local overheating caused by poor contact, thus laying a solid data foundation for subsequent high-precision diagnosis.

[0031] For the extraction of physical features, this embodiment employs an algorithm system that integrates rate of change analysis and cumulative effect quantification. First, real-time temperature data, real-time humidity data, and real-time vibration data are retrieved; then, the temperature characteristics of the busbar are analyzed based on the temperature rate of change analysis function; the expression for the temperature rate of change analysis function is: ; in, ; The first-order rate of change of real-time temperature; The data represents the real-time temperature data collected at time t. The sampling frequency time interval; Here, a second rate of change is introduced. The concept of temperature rise acceleration is crucial for distinguishing between steady-state thermal equilibrium and sudden thermal runaway. When the busbar is under normal load fluctuations, the temperature may rise, but the rate of change is relatively stable. It remains at a low level; however, when loose connections or signs of an impending short circuit occur, the temperature rise often exhibits an exponential acceleration trend. It will increase significantly. In specific implementation, the sampling frequency time interval The response speed can be set according to the sensor response speed, for example, preferably 1s or 5s, which can capture transient thermal shocks and smooth high-frequency measurement noise through differential operation.

[0032] Secondly, in order to more accurately assess the long-term impact of ambient humidity on insulation performance, a humidity change rate formula was used. Calculate the rate of change of humidity in real time; where, The real-time humidity data is collected at time t; the mean and variance of the humidity are calculated based on a set time sliding window; the formula for calculating the mean humidity is: The formula for calculating the variance of humidity is: Based on this, the cumulative humidity hazard level of the busbar trunking is analyzed using a cumulative humidity hazard analysis function; the expression for the cumulative humidity hazard analysis function is: ; The humidity change rate, mean and variance of humidity, and cumulative humidity hazard are integrated into a humidity characteristic; among which, This is the current analysis time. To monitor the start time; For time infinitesimal elements; This refers to a humidity safety threshold; traditional humidity monitoring only focuses on whether instantaneous values ​​exceed the limit, easily overlooking the chronic corrosion of insulation materials caused by prolonged moderate humidity. In this embodiment... By performing integral calculations, the humidity exceeding the safe threshold was quantified. The product effect of the duration and the degree of exceedance. For example, setting a humidity safety threshold. If the humidity remains at 85%RH for 4 hours, the cumulative risk is far greater than if the humidity briefly reaches 90%RH for only 1 minute. This cumulative effect more accurately reflects the risk of insulation aging and effectively filters out momentary humidity interference caused by opening and closing doors or breathing.

[0033] Furthermore, for stability monitoring of mechanical structures, real-time vibration data over a set time period are integrated to obtain vibration signals; wavelet transform is used to analyze the characteristics of the vibration signals to obtain vibration features; the expression for wavelet transform is: ; in, For scale parameters, For translation parameters, For the mother wavelet function; The vibration signal is typically non-stationary, containing numerous transient impact components (such as impacts caused by loose bolts). Traditional Fourier transforms cannot simultaneously provide time and frequency information, while wavelet transforms, by adjusting the width of the analysis window through a scaling parameter, achieve multi-resolution analysis of low-frequency trends and high-frequency transients. Translation parameters ensure accurate time positioning. In practical applications, tightly supported wavelets such as db4 or sym8 can be selected as the mother wavelet function to better match the waveform characteristics of mechanical impact signals, thereby accurately extracting the specific frequency band energy characterizing the loosening of connectors.

[0034] Regarding the extraction of electrical features, this embodiment focuses on the quantitative characterization of power quality and insulation status. Specifically, it retrieves real-time current data, real-time voltage data, and partial discharge signal data; integrates the real-time current data within a set time period to obtain a current signal; and processes the current using a Fast Fourier Transform (FFT) to obtain frequency domain features. The expression for the FFT is: , ; in, It is a current signal; The number of points in the Fourier transform; For discrete frequency indexes in the frequency domain; The number of Fourier transform points directly determines the frequency resolution. Considering that busbar harmonic analysis typically focuses on the 50th harmonic (2500Hz) and real-time calculation is also necessary, N is preferably set to 1024 or 2048 points. Through frequency domain transformation, the mixed current waveform in the time domain can be decomposed into the amplitude and phase of each harmonic, providing a basis for subsequent harmonic distortion calculation.

[0035] Furthermore, the total harmonic distortion (THD) of the current signal is analyzed based on the THD analysis function; the expression for the THD analysis function is: ; in, Let be the amplitude of the kth harmonic, and its expression is: ; The fundamental frequency amplitude is denoted as . Total harmonic distortion (THD) is a key indicator of power quality and a significant cause of additional eddy current losses and heat generation in busbar trunking. The harmonic order K in the formula should cover the distribution of major harmonic energy; in engineering practice, it is typically taken up to the 50th order or dynamically adjusted based on actual load characteristics (such as characteristic harmonics generated by frequency converters and rectifiers). High THD values ​​not only indicate grid pollution but also suggest that the busbar trunking may be subjected to thermal stress exceeding its rated design, requiring comprehensive assessment in conjunction with temperature characteristics.

[0036] Simultaneously, the voltage imbalance degree is analyzed based on the voltage imbalance analysis function to obtain voltage characteristics; the discharge characteristics of the busbar trunking are analyzed based on the partial discharge signal; the expression of the voltage imbalance analysis function is: ; in ; , and The values ​​represent the real-time voltage values ​​of phases A, B, and C of the busbar trunking. Voltage imbalance reflects the symmetry of the three-phase power supply system. Severe voltage imbalance can generate negative sequence current in the busbar trunking, leading to additional heating and mechanical vibration in the conductors, and accelerating insulation aging. The engineering significance of this indicator lies in its ability to help distinguish between single-phase overload and systemic power supply problems, providing crucial clues for fault tracing. It should be understood that the above-mentioned algorithms for extracting physical and electrical characteristics are only one preferred implementation method. Other equivalent mathematical transformations or statistical analysis methods can be used to achieve the same function without departing from the concept of this application.

[0037] Example 3: In this embodiment, deep feature mining using artificial intelligence technology is employed to target the weak and nonlinear partial discharge signals caused by insulation defects within the busbar trunking. Specifically, the process of analyzing the discharge characteristics of the busbar trunking based on the partial discharge signals includes: retrieving the partial discharge signals, integrating the partial discharge signals over a set time period to obtain a discharge analysis sequence; and calling the discharge analysis model by inputting the discharge analysis sequence into the model to obtain the discharge characteristics of the busbar trunking. The discharge analysis model is constructed based on an artificial intelligence model. Unlike traditional signal processing methods such as Fourier transform or wavelet transform used in Embodiment 2, the artificial intelligence model introduced in this embodiment can automatically learn high-dimensional, unstructured implicit patterns in the partial discharge signals, effectively overcoming the limitations of traditional methods that rely on manual experience to set thresholds or feature parameters. The discharge analysis sequence here is not the original time-domain waveform directly input, but rather a pre-processed structured data representation. For example, the discharge signal within the power frequency cycle can be mapped to a phase-amplitude-frequency (PRPD) spectrum, or a fixed length of time-series sampling points can be extracted and normalized to adapt to the input tensor dimension of the artificial intelligence model. This preprocessing method preserves the fingerprint characteristics of the discharge signal in the phase distribution, enabling the model to distinguish different fault types such as air gap discharge, surface flashover, or metal particle vibration, thereby outputting a multi-dimensional discharge feature vector that includes the confidence level and severity of the fault category.

[0038] To ensure the reliability and accuracy of the aforementioned discharge analysis model in practical engineering applications and to prevent it from becoming an unexplainable black box, this application further clarifies the entire process of model construction and verification. Specifically, the steps for constructing the discharge analysis model based on an artificial intelligence model include: selecting a model framework and deep learning algorithm from an artificial intelligence model library; and constructing the model based on the model framework and deep learning algorithm to obtain an initial model. In specific implementations, considering that partial discharge signals have both image-like spectral features and temporal evolution patterns, the model framework can preferably be a hybrid architecture of convolutional neural networks (CNN) and long short-term memory networks (LSTM), or an attention mechanism model such as Transformer, to simultaneously capture spatial morphology and temporal dependence. The deep learning algorithm can be selected for gradient updates based on computational resources and real-time requirements, using backpropagation algorithms and their variants (such as the Adam optimizer). It should be understood that although specific model architectures are listed here, in other implementations, lightweight networks (such as MobileNet) or more complex deep residual networks can also be selected based on the computing power conditions of the busbar monitoring terminal, as long as they have the ability to extract high-order features from the discharge analysis sequence.

[0039] Furthermore, a standard dataset is obtained. This standard dataset includes standard input data consistent with the content attributes of the discharge analysis sequence, and standard output data consistent with the content attributes of the discharge characteristics. The standard dataset is the cornerstone of model performance, and its construction must be closely aligned with the specific object of the busbar trunking. Specifically, the standard input data should cover various typical discharge samples accumulated in laboratory simulations and field measurements of the busbar trunking (such as insulator air gaps, tip discharges at joints, and floating potential discharges in the casing) as well as interference noise samples (such as inverter interference and corona interference). The standard output data consists of fault type labels and quantitative indicators after expert annotation or high-precision instrument calibration. By dividing the dataset into training, validation, and test sets according to a set ratio (e.g., 7:2:1 or 8:1:1), the initial model is trained using the training set, the internal parameters of the initial model are adjusted using the validation set, and the trained initial model is tested using the test set to obtain test indicators. In this process, the validation set is used to monitor overfitting and guide the fine-tuning of hyperparameters (such as learning rate and regularization coefficient), while the test set serves as a completely independent data source for objectively evaluating the model's generalization ability.

[0040] Finally, the standard range of the indicators is obtained; the test indicators are compared with the standard range; if all test indicators are within the standard range, the initial model is marked as the discharge analysis model; otherwise, the initial model is rebuilt and trained. The test indicators here include not only the overall accuracy, but also, and more importantly, the recall and false positive rates for key fault types. For example, for severe discharge types that may lead to breakdown accidents, the lower limit of the standard range for recall can be set above 95%, while the overall false positive rate should be controlled below 1%. If the test results do not meet the standards, the system will trigger iterative optimization strategies, including but not limited to expanding scarce fault samples, adjusting the weights of the loss function to increase attention to minority class samples, or replacing the backbone network with one that has better representational capabilities, until the model performance meets the preset engineering acceptance standards. Through this rigorous closed-loop verification mechanism, this embodiment transforms the application of artificial intelligence technology from an abstract concept into a quantifiable and reproducible technical means, ensuring the credibility of the discharge feature analysis results. Combining the physical and electrical characteristics in Example 2, the AI ​​discharge characteristics provided in this example form a complementary evidence chain driven by both mechanism and data, significantly improving the comprehensive diagnostic capability for complex insulation faults in busbar trunking.

[0041] Example 4: In this embodiment, the process of constructing a digital twin model of the busbar in step S130 and determining the dynamic threshold corresponding to the feature data based on the digital twin model in step S140 are further detailed. This is the core link in realizing adaptive fault diagnosis, aiming to solve the technical problem that traditional fixed thresholds cannot adapt to the drift of operating conditions through the deep integration of mechanism models and data statistical characteristics.

[0042] Firstly, the construction of a digital twin model for the busbar trunking involves: acquiring the busbar trunking's physical structural parameters, material parameters, and electrical parameters; and using 3D modeling software to construct a geometric structural model of the busbar trunking based on these physical parameters. Specifically, the physical structural parameters encompass geometric information such as conductor cross-sectional dimensions, insulation layer thickness, shell shape, and assembly gaps, which form the basis for establishing a high-fidelity model. Material parameters include the conductor's resistivity, heat capacity, thermal conductivity, and the dielectric constant of the insulating material, determining the model's physical response characteristics. Electrical parameters involve operating indicators such as rated current and impedance spectrum. Based on this, the busbar trunking's material and electrical parameters are imported into its geometric structural model to establish a mathematical model; historical data of the busbar trunking is then transferred to the mathematical model for simulation, resulting in a digital twin model of the busbar trunking. This step represents a leap from static geometry to dynamic behavior. By inputting historical operating data (such as load curves and environmental temperature and humidity records from the past year) into the mathematical model for reverse simulation and parameter calibration, the deviation between the theoretical model and the actual equipment can be corrected. This allows the generated digital twin model to not only possess theoretical physical laws but also accurately reflect the aging state and operating habits of the specific busbar trunking, thus laying the foundation for providing accurate benchmark values ​​in the future.

[0043] Furthermore, the process of determining the dynamic threshold corresponding to the feature data based on the aforementioned digital twin model is not a simple numerical comparison, but a complex computational process involving residual analysis, statistical recursion, and adaptive adjustment. Specifically, it includes the following steps: Acquire normal historical data; retrieve characteristic data for a set time period; input the historical data into the digital twin model to obtain the baseline value corresponding to the characteristic data. Here, the baseline value represents the theoretically expected "healthy state" value of the busbar under the current operating conditions. For example, when the real-time load is 1000A and the ambient temperature is 30℃, the baseline value for the connector temperature output by the digital twin model might be 65℃, rather than a fixed alarm upper limit.

[0044] Subsequently, the difference between the feature data and the baseline value is calculated to obtain the residual. The physical meaning of the residual is that it removes the state changes caused by normal factors such as load variations and environmental fluctuations, and only retains the deviation components that may be caused by faults or anomalies. If the equipment is operating normally, the residual should appear as random white noise with a mean of zero; if there is a potential fault, the residual sequence will show a trend of deviation or abnormal impulses.

[0045] To accurately capture the statistical characteristics of the residuals and suppress measurement noise, this embodiment employs an exponentially weighted moving average (EWMA) strategy to update the mean and variance of the residuals in real time. Specifically, this is achieved through the formula... Recursively calculate the mean of the residuals; based on the formula The residual variance is recursively calculated; where, Forgetting factor, Let be the residual at time t; Let be the mean residual at time t-1; the forgetting factor determines the balance between the algorithm's memory length of historical data and its sensitivity to new data. In specific implementations, the preferred value range for is 0.90 to 0.99. For example, when set to , it means that the current statistic is mainly determined by historical information from approximately 20 sampling points in the past. This effectively smooths out transient interference and tracks the slow drift of device status in a timely manner. Compared to the fixed-window moving average, this recursive method has a smaller computational load and is more suitable for real-time processing in embedded systems.

[0046] More importantly, to address non-Gaussian disturbances that may occur in the busbar operating environment (such as the instantaneous impact caused by the start-up and shutdown of large equipment), this embodiment introduces an adaptive adjustment mechanism based on kurtosis. The dynamic expansion coefficient of the threshold is calculated according to the dynamic coefficient analysis function; the expression for the dynamic coefficient analysis is: ;in, The length of the sliding window; Let i be the value of the i-th residual sequence; The residuals within the window are calculated as the average; kurtosis It is a statistic that measures the kurtosis of a probability distribution. For a standard normal distribution, its kurtosis is approximately 3. When the busbar is in stable operation, the residual distribution is close to normal. The residual distribution remains at a low level; however, when the system is subjected to strong external disturbances or early internal failures, the residual distribution exhibits a "heavy-tailed" phenomenon, leading to... The value increases significantly. The choice of the sliding window length W should balance the real-time performance of the calculation with statistical stability, and it can usually be set to 30 to 60 sampling periods.

[0047] Based on formula Calculate the final dynamic threshold limit, and determine the dynamic threshold range based on the dynamic threshold limit; where, ; This is the baseline coefficient.

[0048] here This is the dynamic expansion coefficient, which is no longer the constant 3 in the fixed 3σ principle of traditional SPC control charts, but a variable that changes in real time with the data distribution pattern. As a benchmark coefficient, it can be preset according to the safety level requirements of the busbar trunking, for example, a value of 2.5 or 3.0. The formula uses... The design of taking the fourth root has significant engineering implications: on the one hand, it ensures... The threshold monotonically increases with kurtosis, while utilizing the saturation characteristic of the root function to prevent excessive threshold expansion and missed detections under extreme disturbances. Through this mechanism, when operating conditions are stable, Smaller The corresponding contraction narrows the dynamic threshold range, thereby improving the detection sensitivity for minute fault characteristics; when a strong disturbance is detected, causing... During a surge, Automatic amplification widens the threshold boundary to encompass normal transient fluctuations and avoid false alarms. The final generated... This forms a dynamic safety envelope that can "breathe" with changes in environment and load, fundamentally solving the problem of poor adaptability of fixed thresholds under complex working conditions, and providing a highly reliable criterion for subsequent fault identification.

[0049] Example 5: In this embodiment, the specific execution logic for identifying and diagnosing busbar faults based on feature data and dynamic thresholds in step S150 is described in detail. Specifically, feature data and corresponding dynamic thresholds are retrieved; the feature data is compared with the corresponding dynamic thresholds; when the feature data exceeds the range of the corresponding dynamic thresholds, a fault alarm signal is generated; otherwise, the identification and diagnosis of busbar faults continues. It should be noted that the judgment logic here is based on the "over-limit triggering" principle, that is, the system only determines an anomaly when the real-time monitored feature data breaks through the dynamic safety envelope generated by embodiment 4. The fundamental reason for this logic setting is that the dynamic threshold UCL(t) used in this application is not a traditional fixed constant, but an adaptive boundary that has integrated the digital twin reference value and real-time operating condition fluctuation characteristics. In other words, the characteristic drift caused by normal load increases and decreases, ambient temperature changes, or equipment aging has been encompassed and digested by the adaptive adjustment mechanism of the dynamic threshold. Therefore, data fluctuations within the dynamic threshold range are considered normal responses of the equipment under current operating conditions and require no intervention. Only when the characteristic data exceeds this adaptively corrected reasonable range does it indicate a sudden failure or severe degradation trend that cannot be explained by normal operating conditions. At this point, triggering an alarm is accurate and necessary. This corrected logic effectively avoids the proliferation of false alarms caused by confusing "normal fluctuations" with "abnormal failures," ensuring the accuracy of operation and maintenance decisions.

[0050] In practical implementation, the comparison between feature data and dynamic thresholds can be processed in parallel across multiple dimensions. For example, for temperature features, it is necessary to simultaneously check whether its real-time value exceeds the upper limit of UCL(t) or falls below the lower limit (for certain special cooling fault scenarios); for humidity accumulation risk DH(t), the main focus is on whether it exceeds the upper limit threshold. When any key feature data exceeds its corresponding dynamic threshold range, the system generates a fault alarm signal. To further improve the practicality and refined management level of the diagnostic system, as a preferred implementation, the fault alarm signal can also be graded according to the degree or duration of the feature data exceeding the dynamic threshold. For example, when the feature data first exceeds the dynamic threshold but falls back within a preset buffer time, or when the exceedance is less than a preset secondary threshold, a "warning signal" can be generated to prompt maintenance personnel to pay attention and arrange planned inspections; while when the feature data continuously exceeds the dynamic threshold and the exceedance is large, or when multiple features exceed the limit simultaneously, an "emergency alarm signal" is generated, indicating that immediate power outage or emergency protection measures should be taken. This grading mechanism retains sensitivity to early minor faults while preventing overreaction due to transient interference, achieving a balance between safety and economy.

[0051] Furthermore, when the feature data does not exceed the corresponding dynamic threshold range, the system is not in a static state, but continuously identifies and diagnoses busbar trunking faults. This means that the entire process of data acquisition, feature extraction, digital twin simulation, and dynamic threshold updates is executed cyclically within a set sampling period. During continuous monitoring, the system continuously uses the latest measured data to calibrate the digital twin model parameters and updates the residual statistics and dynamic expansion coefficients in real time, ensuring that the dynamic threshold always closely follows the actual operating trajectory of the equipment. It should be understood that although this embodiment uses "exceeding the threshold and triggering an alarm" as the core judgment logic, in other embodiments, a trend prediction algorithm can be combined to issue an early warning when the feature data has not yet exceeded the limit but shows an accelerating trend approaching the threshold. These are all reasonable extensions of the technical concept of this application. Through the above-mentioned rigorous logical closed loop, this embodiment ensures that the intelligent busbar trunking fault diagnosis system can both keenly capture real faults and robustly filter normal interference in complex and ever-changing actual operating environments, truly achieving intelligent and highly reliable status monitoring.

[0052] Example 6: Please see Figure 2 The second aspect of this application provides a fault diagnosis system for intelligent bus trunking, including: a data acquisition module, a model building module, and a fault diagnosis module; The data acquisition module is used to acquire multi-source real-time data of the busbar trunking; The model building module is used to extract features from multi-source real-time data based on feature extraction technology to obtain feature data; and to build a digital twin model of the busbar. The fault diagnosis module is used to determine the dynamic threshold corresponding to the feature data based on the digital twin model; and to identify and diagnose the faults of the bus trunking based on the feature data and the dynamic threshold.

[0053] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0054] The working principle of this application is as follows: This application acquires multi-source real-time data of the busbar trunking; extracts features from the multi-source real-time data based on feature extraction technology to obtain feature data; constructs a digital twin model of the busbar trunking; determines the dynamic threshold corresponding to the feature data based on the digital twin model; and identifies and diagnoses faults in the busbar trunking based on the feature data and the dynamic threshold.

[0055] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A fault diagnosis method for intelligent busbar trunking, characterized in that, include: Acquire multi-source real-time data of the busbar trunking; Feature data is obtained by extracting features from multi-source real-time data based on feature extraction technology. Constructing a digital twin model of the busbar trunking; Determine the dynamic threshold corresponding to the feature data based on the digital twin model; Faults in busbar trunking are identified and diagnosed based on characteristic data and dynamic thresholds.

2. The fault diagnosis method for an intelligent busbar trunking according to claim 1, characterized in that, The feature extraction based on feature extraction technology for multi-source real-time data includes: Retrieve real-time data from multiple sources; these real-time data include: real-time temperature data, real-time current data, real-time voltage data, real-time humidity data, real-time vibration data, and partial discharge signal data. The physical characteristics of the busbar trunking are analyzed based on real-time temperature, humidity, and vibration data; the electrical characteristics are analyzed based on real-time current, voltage, and partial discharge signal data. The physical characteristics include temperature, humidity, and vibration characteristics; the electrical characteristics include current, voltage, and discharge characteristics.

3. The fault diagnosis method for an intelligent busbar trunking according to claim 2, characterized in that, The analysis of the physical characteristics of the busbar trunking based on real-time temperature data, real-time humidity data, and real-time vibration data includes: Real-time temperature data, real-time humidity data, and real-time vibration data are retrieved; the temperature characteristics of the busbar are analyzed based on a temperature change rate analysis function; the expression of the temperature change rate analysis function is: ;in, ; The first-order rate of change of real-time temperature; The data represents the real-time temperature data collected at time t. The sampling frequency time interval; Based on the humidity change rate formula Calculate the rate of change of humidity in real time; where, The real-time humidity data is collected at time t; the mean and variance of the humidity are calculated based on a set time sliding window; the formula for calculating the mean humidity is: The formula for calculating the variance of humidity is: ; The cumulative humidity hazard level of the busbar trunking is analyzed based on the cumulative humidity hazard analysis function; the expression of the cumulative humidity hazard analysis function is: The humidity change rate, mean and variance of humidity, and cumulative humidity hazard are integrated into a humidity characteristic; among which, This is the current analysis time. To monitor the start time; For time infinitesimal elements; The humidity safety threshold; Real-time vibration data over a set time period are integrated to obtain a vibration signal; the characteristics of the vibration signal are analyzed using wavelet transform to obtain vibration features; the expression for the wavelet transform is: ;in, For scale parameters, For translation parameters, For the mother wavelet function; It is a vibration signal.

4. The fault diagnosis method for an intelligent busbar trunking according to claim 2, characterized in that, The analysis of the electrical characteristics of the busbar trunking based on real-time current data, real-time voltage data, and partial discharge signal data includes: Real-time current data, real-time voltage data, and partial discharge signal data are retrieved; the real-time current data within a set time period are integrated to obtain a current signal; the current is processed using a Fast Fourier Transform (FFT) to obtain frequency domain characteristics; the expression for the FFT is: , ;in, It is a current signal; The number of points in the Fourier transform; For discrete frequency indexes in the frequency domain; It is the rotation factor; The total harmonic distortion (THD) of a current signal is analyzed based on a total harmonic distortion (THD) analysis function; the expression for the THD analysis function is as follows: ;in, Let be the amplitude of the kth harmonic, and its expression is: ; The fundamental amplitude; Voltage imbalance is analyzed based on a voltage imbalance analysis function to obtain voltage characteristics; the discharge characteristics of the busbar trunking are analyzed based on partial discharge signals; the expression of the voltage imbalance analysis function is as follows: ;in ; , and These are the real-time voltage values ​​of the three phases A, B, and C of the busbar.

5. The fault diagnosis method for an intelligent busbar trunking according to claim 4, characterized in that, The analysis of the discharge characteristics of the busbar trunking based on partial discharge signals includes: Partial discharge signals are retrieved and integrated within a set time period to obtain a discharge analysis sequence; The discharge analysis model is invoked, and the discharge analysis sequence is input into the discharge analysis model to obtain the discharge characteristics of the busbar; the discharge analysis model is constructed based on an artificial intelligence model.

6. The fault diagnosis method for an intelligent busbar trunking according to claim 5, characterized in that, The discharge analysis model is built based on an artificial intelligence model and includes: Select a model framework and deep learning algorithm from the artificial intelligence model library; build a model based on the model framework and deep learning algorithm to obtain the initial model; Obtain the standard dataset; the standard dataset includes standard input data consistent with the content attributes of the discharge analysis sequence, and standard output data consistent with the content attributes of the discharge characteristics; The standard dataset is divided into a training set, a validation set, and a test set according to a set ratio; the initial model is trained using the training set; the internal parameters of the initial model are adjusted using the validation set; and the trained initial model is tested using the test set to obtain test metrics. Obtain the standard range of indicators; compare the test indicators with the standard range of indicators; if all test indicators are within the standard range of indicators, mark the initial model as the discharge analysis model; otherwise, rebuild and train the initial model again.

7. The fault diagnosis method for an intelligent busbar trunking according to claim 1, characterized in that, The construction of the digital twin model of the busbar trunking includes: Obtain the physical structure parameters, material parameters, and electrical parameters of the busbar trunking; use 3D modeling software to construct the geometric structure model of the busbar trunking based on its physical structure parameters; The material and electrical parameters of the busbar are imported into the geometric model of the busbar to establish a mathematical model of the busbar; the historical data of the busbar is transmitted to the mathematical model of the busbar for simulation to obtain a digital twin model of the busbar.

8. The fault diagnosis method for an intelligent busbar trunking according to claim 1, characterized in that, The determination of the dynamic threshold corresponding to the feature data based on the digital twin model includes: Obtain normal historical data; retrieve feature data for a specified time period; input the historical data into the digital twin model to obtain the baseline value corresponding to the feature data; Calculate the difference between the feature data and the benchmark value to obtain the residual; then use the formula... Recursively calculate the mean of the residuals; based on the formula The residual variance is recursively calculated; where, Forgetting factor, Let be the residual at time t; Let be the mean residual at time t-1; The dynamic expansion coefficient of the threshold is calculated based on the dynamic coefficient analysis function; the expression for the dynamic coefficient analysis is: ;in, The length of the sliding window; Let i be the value of the i-th residual sequence; The residuals within the window are calculated as the average value; Based on formula Calculate the final dynamic threshold limit, and determine the dynamic threshold range based on the dynamic threshold limit; where, ; This is the baseline coefficient.

9. The fault diagnosis method for an intelligent busbar trunking according to claim 1, characterized in that, The method of identifying and diagnosing busbar faults based on feature data and dynamic thresholds includes: Retrieve feature data and corresponding dynamic thresholds; compare the feature data with the corresponding dynamic thresholds. When all feature data are within the corresponding dynamic threshold range, a fault alarm signal is generated; otherwise, the fault identification and diagnosis of the bus trunking continues.

10. A fault diagnosis system for an intelligent busbar trunking, applied to the fault diagnosis method for an intelligent busbar trunking as described in any one of claims 1-9, characterized in that, include: Data acquisition module, model building module, and fault diagnosis module; The data acquisition module is used to acquire multi-source real-time data of the busbar trunking; The model building module is used to extract features from multi-source real-time data based on feature extraction technology to obtain feature data; and to build a digital twin model of the busbar. The fault diagnosis module is used to determine the dynamic threshold corresponding to the feature data based on the digital twin model; and to identify and diagnose the faults of the bus trunking based on the feature data and the dynamic threshold.

Images