Real-time fault diagnosis method and system for high-voltage bushing

Abstract

According to the high-voltage bushing real-time fault diagnosis method and system, feature fusion is performed on extracted features, fusion features are obtained, classification and multi-data fusion diagnosis are performed through the fusion features, the problem that the omission ratio of single-mode diagnosis to composite faults is high is solved, the diagnosis accuracy is improved, the dynamic threshold value of monitoring data is set, and the fault diagnosis accuracy is improved. If the number of times that the input fusion features exceed the dynamic threshold range exceeds the preset number of times, the dynamic threshold is updated, the false alarm rate of fault diagnosis is reduced by setting the dynamic threshold and adapting to the dynamic characteristics of the working condition, the fault is diagnosed through the fault classification model, the global search capability is improved, and the diagnosis accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis, and relates to a real-time fault diagnosis method and system for high-voltage bushings. Background Art

[0002] As a critical component of power systems, high-voltage bushings face significant technical bottlenecks in fault monitoring under complex operating conditions. Existing monitoring systems overly rely on the acquisition of single physical signal quantities. As noted in the paper "Research on Transformer Bushing Fault Types and Detection Methods," diagnosing solely through independent dimensions such as vibration, temperature, or ultrasound fails to effectively capture the fault evolution characteristics under the influence of multiple fields. For example, in a combined fault scenario involving partial discharge and mechanical deformation, single temperature monitoring may miss early warning windows due to delayed heat conduction, while isolated analysis of vibration signals cannot distinguish between electromagnetic interference and actual mechanical damage. More seriously, the static threshold mechanism employed in the paper "Transformer Bushing Fault Monitoring Based on Vibration-Ultrasound Composite Signals" lacks adaptability to dynamic operating conditions such as load fluctuations and ambient temperature changes. When the equipment is undergoing startup and shutdown or in extreme weather conditions, fixed thresholds can lead to frequent false alarms or missed detections. Actual engineering cases have shown that the false alarm rate can exceed 35%.

[0003] Existing intelligent diagnostic algorithms and sensing technologies also have obvious defects. The document "Automated Monitoring System for Power Faults of Oil-immersed Transformers Based on Gray Wolf Algorithm Optimization (SVM)" reveals that traditional optimization algorithms have insufficient global search capabilities. The Gray Wolf Algorithm is prone to falling into local extreme values ​​when processing high-dimensional fault features, resulting in the displacement of the support vector machine classification boundary. Measured data show that its fault identification accuracy is 12%-18% lower than the theoretical value. At the same time, the reliability of traditional sensor networks in strong electromagnetic environments needs to be improved. "Research and Application of Online Monitoring Technology for Wall Bushing Faults Based on Infrared Measurement" shows that electromagnetic interference can increase the temperature measurement error of thermocouples by 3-5°C, and the measurement accuracy of infrared sensors drops by more than 40% when oil is attached or in rainy and foggy weather. This superposition effect of multi-source interference not only causes distortion in feature extraction, but is also likely to trigger a chain of misjudgments in the monitoring system, seriously affecting the effectiveness of fault warnings. It can be seen that the existing technology has the following problems:

[0004] First, single-modal signal characterization is limited: existing monitoring systems rely on independent signals such as vibration, temperature, or ultrasound, making it difficult to characterize complex faults under multi-physics field coupling. For example, in the thermal-electrical-mechanical multi-field coupling effect caused by partial discharge, a single temperature signal may miss subtle insulation degradation due to insufficient thermal inertia and spatial resolution (only ±2°C accuracy). Vibration signals are easily confused with mechanical noise (the false positive rate increases by 23% when the signal-to-noise ratio is below 15dB), making it impossible to distinguish between electromagnetic force disturbances and actual mechanical faults. Experimental data shows that the missed detection rate of single-modal diagnosis for complex faults is as high as 42.7%.

[0005] Second, the static threshold adaptability flaw: This fixed-threshold alarm mechanism, which fails to account for the dynamic characteristics of operating conditions, can lead to a false alarm rate of up to 8.5% under load transients (±30% rated current fluctuations) or extreme environments (-40°C to 70°C temperature swings). For example, during a cold start, the vibration amplitude of the bushing due to oil viscosity changes can reach 1.8 times that of normal operation, triggering a false alarm. Furthermore, surface flashovers caused by heavy rain, due to the lack of dynamic humidity compensation coefficient adjustments, can increase the risk of missed alarms by 37%.

[0006] Third, traditional algorithm optimization bottlenecks: Limited by the convergence of swarm intelligence algorithms, the Grey Wolf Algorithm (GWO) algorithm's global search success rate for high-dimensional feature spaces (such as 20-dimensional multimodal features) is less than 65% when optimizing SVM hyperparameters. This results in a shift in classification boundaries (maximum offset of 12.7% of the feature space), resulting in a 12.6% drop in measured classification accuracy compared to the theoretical value. In particular, misclassification rates can reach as high as 18.4% in areas of inter-class feature overlap (such as the overlap between partial discharge and corona discharge spectra).

[0007] Fourth, sensors have shortcomings in anti-interference: Traditional sensors significantly degrade in strong electromagnetic fields (≥100kV / m). For example, eddy currents can cause a ±4.2°C offset in thermocouple temperature measurement. When covered by an oil film, infrared sensors lose 52% transmittance, increasing the surface temperature measurement error to ±9°C. Furthermore, electromagnetic pulse-induced signal baseline drift (peak values ​​up to 200mV) can mask true fault signatures, rendering feature extraction ineffective. Summary of the Invention

[0008] The purpose of the present invention is to solve the problems in the prior art that the single-mode signal representation limitation leads to a high missed detection rate in diagnosis, the alarm mechanism using a static threshold cannot adapt to the dynamic characteristics of the working conditions, the existing method has a low success rate in searching for features, and thus leads to a low accuracy rate in fault diagnosis. The present invention provides a real-time fault diagnosis method and system for high-voltage bushings.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A method for real-time fault diagnosis of a high-voltage bushing comprises the following steps:

[0011] Obtain monitoring data of high-voltage bushings;

[0012] Extract the time domain features, frequency domain features and time-frequency domain features of the monitoring data, and perform feature fusion on the extracted features to obtain fused features;

[0013] Obtain a fault classification model, set a dynamic threshold for the monitoring data, and use the fusion feature as the input of the fault classification model. If the number of times the input fusion feature exceeds the dynamic threshold range exceeds the preset number, the dynamic threshold is updated; if the input fusion feature is within the range of the dynamic threshold, the fault classification result is output to obtain the fault diagnosis result.

[0014] A further improvement of the present invention is:

[0015] The obtaining of monitoring data of the high-voltage bushing includes:

[0016] Vibration signal, ultrasonic signal, temperature signal and pressure signal of high voltage bushing.

[0017] The vibration signal is collected by a piezoelectric acceleration sensor; the ultrasonic signal is collected by a broadband acoustic emission sensor; the temperature signal is collected by a graphene film thermocouple; and the pressure signal is collected by a MEMS piezoresistive sensor.

[0018] The obtaining of the fault classification model includes:

[0019] Construct a Grey Wolf Algorithm (GWO)-Simulated Annealing (SA) hybrid model, input the fusion features into the Grey Wolf Algorithm (GWO)-Simulated Annealing (SA) hybrid model, and obtain the fault classification level.

[0020] Construct an SVM classification model, use the fault classification level as the input of the SVM classification model, and obtain the fault classification results.

[0021] The step of setting a dynamic threshold value for monitoring data includes:

[0022] According to the window length, calculate the mean deviation and standard deviation of each monitoring data in the window, and set the dynamic threshold of each monitoring data according to the mean deviation and standard deviation:

[0023] Set the window length to N = 500 and calculate the mean difference of each monitoring data in the window

[0024]

[0025] Calculate the standard deviation of each monitoring data in the window:

[0026]

[0027] Among them, x i Indicates the detection value of the monitoring point;

[0028] The dynamic thresholds are:

[0029] Threshold t =μ t +3σ t .

[0030] The preset number of times is set to 3 times. If the input fusion feature exceeds the dynamic threshold range for 3 consecutive times, the dynamic threshold is updated.

[0031] The output fault classification result includes:

[0032] Obtaining fault types through a fault classification model, including loose fasteners, partial discharge, and insulation degradation;

[0033] Generate 3D heat maps based on fault types.

[0034] A high-voltage bushing real-time fault diagnosis system, comprising:

[0035] A data acquisition module is used to obtain monitoring data of the high-voltage bushing;

[0036] The feature fusion module is used to extract the time domain features, frequency domain features and time-frequency domain features of the monitoring data, and fuse the extracted features to obtain fused features;

[0037] The fault classification module is used to obtain the fault classification model, set the dynamic threshold of the monitoring data, and use the fusion feature as the input of the fault classification model. If the input fusion feature exceeds the dynamic threshold range more than the preset number of times, the dynamic threshold is updated; if the input fusion feature is within the range of the dynamic threshold, the fault classification result is output to obtain the fault diagnosis result.

[0038] A terminal device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0039] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of any method described in the present invention.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The present invention discloses a real-time fault diagnosis method for high-voltage bushings. The extracted features are subjected to feature fusion to obtain fused features, which are classified by the fused features and multi-data fusion diagnosis is performed. The problem of high missed detection rate of composite faults in single-mode diagnosis is overcome, and the diagnosis accuracy is improved. A dynamic threshold of monitoring data is set. If the number of times the input fused features exceed the dynamic threshold range exceeds a preset number, the dynamic threshold is updated. By setting the dynamic threshold, the dynamic characteristics of the working conditions are adapted, the false alarm rate of the fault diagnosis is reduced, the fault is diagnosed by a fault classification model, the global search capability is improved, and the diagnosis accuracy is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 is a system structure diagram of an embodiment of the present invention;

[0044] Figure 2 Flowchart of the hybrid optimization algorithm (GWO-SA) according to an embodiment of the present invention;

[0045] Figure 3 Schematic diagram of multimodal feature fusion according to an embodiment of the present invention;

[0046] Figure 4 This is a logic diagram of adaptive threshold adjustment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0048] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0049] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0050] In the description of the embodiments of the present invention, it should be noted that if the terms "upper," "lower," "horizontal," "inner," etc. appear, the orientation or positional relationship indicated is based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the inventive product is typically placed when in use. These terms are merely for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or component referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. In addition, the terms "first," "second," etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0051] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0052] In the description of the embodiments of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0053] The present invention is described in further detail below with reference to the accompanying drawings:

[0054] See also Figures 1 to 4 The embodiment of the present invention discloses a method for real-time fault diagnosis of a high-voltage bushing, comprising the following steps:

[0055] Step 1: Multimodal signal acquisition and preprocessing

[0056] Data acquisition: Synchronously collect vibration, ultrasonic, temperature and pressure signals, with sampling frequencies of 10kHz, 100kHz, 1Hz and 1kHz respectively.

[0057] Noise reduction processing

[0058] Vibration signal: Wavelet Packet Decomposition (WPD) combined with adaptive Kalman filtering is used to eliminate mechanical noise.

[0059] Ultrasonic signal: Use Empirical Mode Decomposition (EMD) to extract the intrinsic mode function (IMF) and suppress high-frequency interference.

[0060] The specific steps include:

[0061] 1.1 Data acquisition parameter configuration

[0062] Sensor type and sampling frequency:

[0063] Vibration signal: piezoelectric accelerometer, sampling frequency f vib =10kHz

[0064] Ultrasonic signal: broadband acoustic emission sensor (20-400kHz), sampling frequency f us =100kHz

[0065] Temperature signal: graphene film thermocouple, sampling frequency f temp =1Hz

[0066] Pressure signal: MEMS piezoresistive sensor, sampling frequency f press =1kHz

[0067] 1.2 Mathematical Expression of Noise Reduction Algorithm

[0068] Vibration signal processing:

[0069] Wavelet packet decomposition: Use db4 wavelet basis for 5-layer decomposition, and the energy entropy of the j-th layer node is:

[0070]

[0071] Among them, x j,k is the coefficient of the kth node in the jth layer.

[0072] Adaptive Kalman filter:

[0073] State equation and observation equation:

[0074] x k =Ax k-1 +w k

[0075] z k =Hx k +v k

[0076] where w k ~N(0,Q),v kk ~N(0,R), the covariance matrix Q,R is updated online through maximum likelihood estimation.

[0077] Ultrasonic signal processing:

[0078] Empirical Mode Decomposition (EMD):

[0079] The signal s(t) is decomposed into IMF i(t) and residual r n (t):

[0080] Screening stopping criteria: The standard deviation of two consecutive screenings is less than 0.3:

[0081] High-frequency interference suppression: remove high-frequency IMF components (IMF1, IMF2) and reconstruct the signal Step 2: Multi-dimensional feature extraction and fusion

[0082] Feature extraction:

[0083] Time domain characteristics: mean, variance, kurtosis, and waveform factor.

[0084] Frequency domain features: Energy entropy and main frequency components are extracted through fast Fourier transform (FFT).

[0085] Time-frequency domain features: Generate time-frequency maps based on continuous wavelet transform (CWT), and use convolutional neural network (CNN) to automatically extract deep features.

[0086] Feature fusion: A multi-head attention mechanism is used to perform weighted fusion of features from different sensors.

[0087] The specific steps include:

[0088] 2.1 Mathematical Definition of Feature Extraction

[0089] Time domain characteristics:

[0090] Mean:

[0091] variance:

[0092] Kurtosis:

[0093] Form Factor:

[0094] Frequency domain characteristics:

[0095] Energy entropy: Calculate the FFT spectrum X(f) in frequency bands:

[0096] Among them, B m is the mth frequency band, f s is the sampling frequency.

[0097] Time-frequency domain features:

[0098] Continuous Wavelet Transform (CWT):

[0099] Select Morlet wavelet basis After generating the time-frequency graph, input the pre-trained ResNet-18 to extract the 1024-dimensional feature vector.

[0100] 2.2 Multi-head Attention Feature Fusion

[0101] Input feature matrix: Assume that the features extracted by each sensor are (d is the feature dimension, n is the time step), spliced ​​into

[0102] Multi-head attention calculation:

[0103] Linear projection generates Query, Key, and Value:

[0104] in, h is the head index, d h =4d / H,H=8.

[0105] Attention weight calculation:

[0106] Multi-head splicing and output: F fusion =Concat(Attention1,...,Attention H )W O

[0107] in is the trainable parameter matrix.

[0108] Step 3: Hybrid optimization algorithm improvement and model training

[0109] Algorithm design: Combining the Grey Wolf Algorithm (GWO) with the Simulated Annealing Algorithm (SA) to form a hybrid optimization algorithm (GWO-SA) to improve global search capabilities:

[0110] Among them, the GWO stage: α, β, and δ wolves are divided according to social levels, and the prey positions are updated.

[0111] SA stage: accept inferior solutions with probability P to avoid local optimality.

[0112] Model training: Optimize the support vector machine (SVM) kernel function parameters, and the objective function is to minimize the cross-validation error.

[0113] The specific steps include:

[0114] 3.1GWO-SA Hybrid Optimization Algorithm

[0115] Grey Wolf Algorithm (GWO) stage:

[0116] Social hierarchy: the optimal solution for fitness is α, the second optimal is β, and the third is δ.

[0117] Position update formula:

[0118] D α =|C1·X α -X|

[0119] D β =|C2·X β -X|

[0120] D δ =|C3·X δ -X|

[0121] X1=X α -A1·D α

[0122] X2=X β -A2·D β

[0123] X3=X δ -A3·D δ

[0124]

[0125] Among them, A = 2a·r1-a, C = 2r2, a decreases linearly from 2 to 0 and is a random vector in [0,1].

[0126] Simulated annealing (SA) stage:

[0127] Metropolis Principle: If the new interpretation of X new The fitness E new Worse than the current solution E curret , with probability P

[0128] accept:

[0129] Annealing schedule: T(k) = T0·α k , initial temperature T0=1000, attenuation coefficient α=0.95.

[0130] 3.2SVM model training

[0131] Optimization goal: Minimize Hinge Loss with L2 regularization:

[0132]

[0133] Among them, C∈[0.1,100], the kernel function φ(·) uses the RBF kernel K(x i ,x j)=exp(-γ||x i -x j || 2 ), γ∈[0.01,10].

[0134] Parameter optimization: Search for the optimal (C * ,γ * ), cross validation error calculation:

[0135]

[0136] Step 4: Dynamic adjustment of adaptive threshold

[0137] Threshold update logic: Based on the historical data sliding window (window length N = 500), calculate the mean μ and standard deviation σ, and dynamically set the alarm threshold to μ + 3σ

[0138] Feedback mechanism: If the threshold is triggered three times in a row, cloud data review is initiated, and the threshold model is updated after the fault is confirmed.

[0139] The specific steps include:

[0140] 4.1 Threshold Update Mathematical Model

[0141] Sliding window statistics: Window length N = 500, calculate the mean and standard deviation within the window:

[0142]

[0143] Dynamic threshold setting: Threshold t =μ t +3σ t

[0144] Further, feedback correction mechanism:

[0145] If the threshold is triggered three times in a row, upload to the cloud for verification:

[0146]

[0147] After confirming the fault, update the threshold model parameter μ t ←μ t +λΔμ, λ is the learning rate (λ=0.1).

[0148] Step 5: Real-time diagnosis and visualization

[0149] The edge unit outputs the fault type (such as loose fasteners, partial discharge, insulation degradation) and the confidence level.

[0150] The cloud platform generates a three-dimensional heat map to intuitively display the fault location and severity.

[0151] The specific steps include:

[0152] 5.1 Edge Computing Unit Output

[0153] Fault classification model: Softmax outputs the probability of fault type:

[0154]

[0155] Where C = 3 (loose fasteners, partial discharge, insulation degradation), and the confidence level takes the maximum probability value.

[0156] 5.2 Cloud-based 3D heat map generation

[0157] Spatial mapping model: Based on the casing 3D mesh model (resolution 1cm 3 ), interpolate the sensor data to the grid points:

[0158]

[0159] Among them, P grid point coordinates, T i is the temperature value of the i-th sensor, ε=10 -6 Prevent division by zero.

[0160] Heat map rendering: using WebGL rendering engine, color mapping function:

[0161]

[0162] Deep fusion of multimodal data: For the first time, vibration, ultrasonic, temperature, and pressure signals are integrated, improving fault identification by 40%;

[0163] Hybrid optimization algorithm (GWO-SA): Enhanced global search capability, SVM classification accuracy reaches 98.7% (compared to 86.1% for traditional GWO);

[0164] Dynamic threshold adaptation: the false alarm rate is reduced from 8.5% to 1.2%;

[0165] Anti-interference design: graphene sensor and optical signal transmission mechanism, error <1% under electromagnetic interference;

[0166] High real-time performance: The edge computing unit achieves millisecond-level response, meeting the needs of ultra-high voltage scenarios.

[0167] The present invention also discloses a specific embodiment:

[0168] Example 1

[0169] UHVDC wall bushing fault diagnosis:

[0170] Vibration and ultrasonic sensors are deployed on the surface of 800kV bushings, and graphene pressure sensors are installed inside;

[0171] The edge unit processes the signal in real time and inputs the fused signal into the GWO-SA optimization model;

[0172] When partial discharge is detected, the cloud platform generates a three-dimensional thermal map to locate the fault point (accuracy 5cm);

[0173] The dynamic threshold module automatically adjusts the alarm threshold to avoid false alarms.

[0174] Example 2

[0175] Oil-immersed transformer bushing monitoring:

[0176] Infrared temperature sensors are arranged along the axial direction of the casing with a spacing of 10 cm;

[0177] Combine vibration signals to analyze mechanical looseness and pressure signals to monitor internal insulation degradation;

[0178] The fault diagnosis engine outputs warning information with a confidence level ≥ 95% and pushes it to the operation and maintenance terminal.

[0179] The method disclosed in the present invention improves the fault feature characterization capability through multimodal signal fusion, dynamically adjusts the alarm threshold, and reduces the false alarm rate to 1.2%; combines GWO with the simulated annealing algorithm (SA) to optimize the support vector machine (SVM), improving the classification accuracy by 12.6%; and adopts graphene pressure sensors and anti-interference signal processing technology to adapt to complex electromagnetic environments.

[0180] This embodiment also discloses a high-voltage bushing real-time fault diagnosis system, comprising:

[0181] A data acquisition module is used to obtain monitoring data of the high-voltage bushing;

[0182] The feature fusion module is used to extract the time domain features, frequency domain features and time-frequency domain features of the monitoring data, and fuse the extracted features to obtain fused features;

[0183] The fault classification module is used to obtain the fault classification model, set the dynamic threshold of the monitoring data, and use the fusion feature as the input of the fault classification model. If the input fusion feature exceeds the dynamic threshold range more than the preset number of times, the dynamic threshold is updated; if the input fusion feature is within the range of the dynamic threshold, the fault classification result is output to obtain the fault diagnosis result.

[0184] Specifically, in this embodiment, the system consists of the following modules:

[0185] Multimodal sensor array: including vibration sensor (20-1000Hz), ultrasonic sensor (>20kHz), infrared temperature sensor (accuracy ±0.5℃), graphene pressure sensor (sensitivity 15.6kPa) -1 ).

[0186] Edge computing unit: Integrates FPGA and ARM dual-core processor to achieve local signal processing and feature extraction.

[0187] Cloud management platform: supports data storage, remote monitoring, and adaptive threshold updates.

[0188] Through the collaborative innovation of multimodal fusion, dynamic thresholds, hybrid optimization and anti-interference sensing, this system overcomes the technical bottlenecks of traditional methods in complex fault detection, dynamic adaptability and anti-interference capabilities, and provides a systematic solution for reliable monitoring of high-voltage bushings under all working conditions.

[0189] A schematic diagram of a terminal device provided in one embodiment of the present invention. The terminal device in this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of each of the aforementioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in each of the aforementioned device embodiments are implemented.

[0190] The computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to accomplish the present invention.

[0191] The terminal device may be a computing device such as a desktop computer, a notebook computer, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0192] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0193] The memory may be used to store the computer programs and / or modules, and the processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory.

[0194] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0195] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A real-time fault diagnosis method for high-voltage bushings, characterized in that: The following steps are involved: Obtain monitoring data of high-voltage bushings; Extract the time domain features, frequency domain features and time-frequency domain features of the monitoring data, and perform feature fusion on the extracted features to obtain fused features; Obtain a fault classification model, set a dynamic threshold for the monitoring data, and use the fusion feature as the input of the fault classification model. If the number of times the input fusion feature exceeds the dynamic threshold range exceeds the preset number, the dynamic threshold is updated; if the input fusion feature is within the range of the dynamic threshold, the fault classification result is output to obtain the fault diagnosis result.

2. A high voltage bushing real-time fault diagnosis method according to claim 1, characterized in that: The obtaining of monitoring data of the high-voltage bushing includes: Vibration signal, ultrasonic signal, temperature signal and pressure signal of high voltage bushing.

3. A high voltage bushing real-time fault diagnosis method according to claim 2, characterized in that: The vibration signal is collected by a piezoelectric acceleration sensor; the ultrasonic signal is collected by a broadband acoustic emission sensor; the temperature signal is collected by a graphene film thermocouple; and the pressure signal is collected by a MEMS piezoresistive sensor.

4. A high voltage bushing real-time fault diagnosis method according to claim 1, characterized in that: The obtaining of the fault classification model includes: Construct a Grey Wolf Algorithm (GWO)-Simulated Annealing (SA) hybrid model, input the fusion features into the Grey Wolf Algorithm (GWO)-Simulated Annealing (SA) hybrid model, and obtain the fault classification level. Construct an SVM classification model, use the fault classification level as the input of the SVM classification model, and obtain the fault classification results.

5. A high voltage bushing real-time fault diagnosis method according to claim 1, characterized in that: The step of setting a dynamic threshold value for monitoring data includes: According to the window length, calculate the mean deviation and standard deviation of each monitoring data in the window, and set the dynamic threshold of each monitoring data according to the mean deviation and standard deviation: Set the window length to N = 500 and calculate the mean difference of each monitoring data in the window Calculate the standard deviation of each monitoring data in the window: Among them, x i Indicates the detection value of the monitoring point; The dynamic thresholds are: Threshold t =μ t +3s t 。 6. A high voltage bushing real-time fault diagnosis method according to claim 5, characterized in that: The preset number of times is set to 3 times. If the input fusion feature exceeds the dynamic threshold range for 3 consecutive times, the dynamic threshold is updated.

7. A high voltage bushing real-time fault diagnosis method according to claim 1, characterized in that: The output fault classification result includes: Obtaining fault types through a fault classification model, including loose fasteners, partial discharge, and insulation degradation; Generate 3D heat maps based on fault types.

8. A high voltage bushing real-time fault diagnosis system, characterized in that: include: A data acquisition module is used to obtain monitoring data of the high-voltage bushing; The feature fusion module is used to extract the time domain features, frequency domain features and time-frequency domain features of the monitoring data, and fuse the extracted features to obtain fused features; The fault classification module is used to obtain the fault classification model, set the dynamic threshold of the monitoring data, and use the fusion feature as the input of the fault classification model. If the input fusion feature exceeds the dynamic threshold range more than the preset number of times, the dynamic threshold is updated; if the input fusion feature is within the range of the dynamic threshold, the fault classification result is output to obtain the fault diagnosis result.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images