A real-time state live detection method and system for electrical equipment

Abstract

The application relates to the technical field of electrical equipment detection, and discloses an electrical equipment real-time state live detection method and system, which comprises a non-contact multi-source sensing array, an edge computing unit, a high-frequency pulse excitation module, a self-adaptive mounting structure, an intelligent diagnosis platform and a self-powered power supply unit. When the electrical equipment state live detection is carried out, the infrared thermodynamic characteristics, ultraviolet corona intensity and ultrasonic discharge signals are collected by the non-contact multi-source sensing array, multi-parameter real-time sensing under the live condition is realized, the signal distortion problem caused by electromagnetic interference in traditional detection is overcome, and the precision of state parameter extraction is improved. Meanwhile, the multi-dimensional feature data are standardized by the edge computing unit, so that the system can eliminate the interference of environmental factors on the detection result, guarantee the consistency of the evaluation results under different working conditions, and reduce the state diagnosis error.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment detection, in particular to a real-time state live detection method and system for electrical equipment. BACKGROUND

[0002] The power grid develops rapidly, and the equipment is updated and replaced quickly. There are many maintenance work of equipment, and a large amount of loss will be caused if power is cut off during the maintenance process. Therefore, live maintenance should be used as much as possible in the maintenance of equipment. Live detection is the core technical means of power equipment condition-based maintenance, which is carried out in real time under the condition of live operation of equipment through infrared temperature measurement, X-ray detection, ultrasonic detection and other ways. The technology is an important part of the power equipment condition monitoring system, and together with online monitoring, it constitutes the technical pillar of smart grid transformation. Compared with traditional accident maintenance and periodic maintenance, live detection has the advantages of dynamic evaluation of equipment state and consideration of safety and economy.

[0003] At present, since the state detection of electrical equipment needs to be operated with power off, the traditional sensing system equipped cannot efficiently collect multi-source feature data under the condition of live equipment. When electromagnetic interference and signal attenuation occur during the detection process, the error of state parameter extraction will be large, and the accuracy of evaluation cannot be guaranteed.

[0004] Therefore, the present application provides a real-time state live detection method and system for electrical equipment to solve the above problems. SUMMARY

[0005] (I) Technical problems to be solved

[0006] In view of the defects of the prior art, the present application provides a real-time state live detection method and system for electrical equipment, which solves the problems proposed in the background art.

[0007] (II) Technical scheme

[0008] In order to achieve the above purpose, the present application provides the following technical scheme: a real-time state live detection method and system for electrical equipment, the method comprising the following steps:

[0009] S1, collecting multi-dimensional state characteristic quantities of live operation electrical equipment through a non-contact multi-source sensing array, and generating a real-time perception data set;

[0010] S2, pre-processing and feature extraction of the real-time perception data set based on an edge computing unit, and generating a standardized state feature vector;

[0011] S3, coupling a high-frequency pulse excitation module to the surface of the electrical equipment through a self-adaptive mounting structure, and generating an enhanced feature excitation signal;

[0012] S4, using the enhanced feature excitation signal to direct excitation of the equipment insulation defect area, synchronously collecting partial discharge response signals and generating a pulse response map;

[0013] S5, inputting the standardized state feature vector and the pulse response map into an intelligent diagnosis model for multi-source information fusion analysis, to generate an equipment health state evaluation index;

[0014] S6, when the equipment health state evaluation index exceeds a preset threshold, triggering a real-time warning instruction and generating a live maintenance decision scheme;

[0015] The non-contact multi-source sensing array includes an infrared thermal imaging module, an ultraviolet corona detection module, and an ultrasonic partial discharge module, and the self-adaptive mounting structure integrates a self-powered power supply unit and an insulation isolation device.

[0016] Preferably, the S1 includes the following steps:

[0017] S11, collecting device surface temperature field distribution data through a spatially distributed infrared sensor matrix to generate a thermodynamic feature vector;

[0018] S12, capturing corona discharge spectrum data using a solar blind ultraviolet detector to generate a corona intensity distribution map;

[0019] S13, collecting partial discharge acoustic signals based on a piezoelectric ultrasonic sensor array to generate a time-frequency domain discharge feature spectrum.

[0020] Preferably, the S2 includes the following steps:

[0021] S21, performing ambient temperature compensation correction on the thermodynamic feature vector to generate a temperature gradient change matrix;

[0022] S22, performing spatial registration processing on the corona intensity distribution map to generate a corona intensity spatiotemporal evolution model;

[0023] S23, performing wavelet denoising processing on the time-frequency domain discharge feature spectrum to extract a set of discharge pulse feature parameters.

[0024] Preferably, the S3 includes the following steps:

[0025] S31, attaching a high-frequency pulse excitation module to the device surface through a magnetic type adjustable support, the magnetic type adjustable support including a three-dimensional angle adjustment mechanism and a pressure feedback device;

[0026] S32, automatically adjusting the inclination angle of the excitation probe based on the structural features of the equipment, so that the angle between the pulse emission direction and the normal line of the device surface is less than 5°;

[0027] S33, adjust the output impedance to 50±2Ω through the impedance matching circuit, and ensure that the signal transmission efficiency is greater than 95%.

[0028] Preferably, the working parameters of the high-frequency pulse excitation module are:

[0029] Pulse frequency range: 1MHz-10MHz adjustable;

[0030] Output voltage amplitude: 0.1kV-5kV programmable;

[0031] Pulse rise time: less than 10ns;

[0032] Duty cycle adjustment range: 1%-50%

[0033] Preferably, the S4 comprises the following steps:

[0034] S41, use a high-speed data acquisition card to synchronously collect the pulse response signal, and the sampling rate is greater than 200MSPS;

[0035] S42, locate the insulation defect position through the time domain reflection algorithm, and generate a defect spatial coordinate set;

[0036] S43, extract the discharge repetition frequency spectrum, single discharge energy distribution, phase resolution discharge spectrum and pulse sequence correlation matrix based on the pulse waveform characteristics.

[0037] Preferably, the S5 comprises the following steps:

[0038] S51, construct a deep belief network model, and the input layer contains 12 feature channels corresponding to the temperature gradient change rate, corona intensity growth rate, discharge pulse repetition rate, discharge energy entropy value, pulse waveform distortion degree, phase distribution symmetry and other fusion characteristic quantities;

[0039] S52, calculate the weight coefficient of each channel through a feature importance weighting algorithm, and the weight update formula is:

[0040] ;

[0041] Wherein is the weight coefficient of the i-th feature channel at the iteration number t, is the updated weight coefficient of the i-th feature channel at the iteration number t+1, the learning rate , is the partial derivative of the loss function L with respect to weight, is the loss function, is the feature channel index, is the iteration number;

[0042] S53, Output Device Health Status Assessment Index The formula for its calculation is:

[0043] ;

[0044] in As an index for assessing the health status of equipment, For feature index, For the total number of features, For the first The fusion feature quantity of each feature, For the first The degradation factor of a feature.

[0045] Preferably, step S6 includes the following steps:

[0046] S61. Establish a three-tiered early warning mechanism: the primary early warning threshold is 0.7. <0.8, the intermediate warning threshold is 0.6< <0.7, the emergency warning threshold is <0.6;

[0047] S62. Generate differentiated maintenance plans based on the warning level: a primary warning triggers a 72-hour maintenance plan, a medium warning triggers a 24-hour maintenance plan, and an emergency warning triggers an immediate power outage maintenance command.

[0048] Preferably, the system includes:

[0049] The non-contact multi-source sensor array uses an infrared thermal imaging module to collect infrared thermodynamic feature data of the device surface, an ultraviolet corona detection module to capture corona discharge spectrum data, and an ultrasonic partial discharge module to collect partial discharge acoustic signals to generate a real-time sensing dataset.

[0050] The edge computing unit preprocesses the real-time sensing dataset through a signal preprocessing circuit and uses a feature extraction coprocessor to extract multi-dimensional features and generate a standardized state feature vector.

[0051] The high-frequency pulse excitation module generates high-frequency pulse signals through a programmable pulse generator, outputs the signals to the intelligent diagnostic platform for analysis, and uses an impedance matching network for impedance matching to generate an enhanced characteristic excitation signal.

[0052] The adaptive installation structure features a magnetically adjustable bracket to adjust the installation position and angle, and monitors the contact pressure through a pressure feedback device to couple the high-frequency pulse excitation module to the surface of the electrical equipment.

[0053] The intelligent diagnostic platform runs a multi-source information fusion analysis algorithm, receives standardized state feature vectors and impulse response spectra, and generates equipment health status assessment indices.

[0054] The self-powered power supply unit obtains energy from a power frequency magnetic field through an inductive power coil, and stores the energy in a super capacitor energy storage device to supply power to each unit of the system.

[0055] Preferably, it specifically comprises:

[0056] The self-adaptive mounting structure comprises a carbon fiber insulated connecting rod with a withstand voltage level greater than 35kV / cm, a universal joint adjusting mechanism with an angle adjusting range of ±30°, a contact pressure sensor with a range of 0-50N adjustable, and a radio frequency shielding shell with a shielding effectiveness greater than 80dB.

[0057] The self-powered power supply unit is implemented by a Rogowski coil for inductive power, and outputs a power greater than 5W under a 1A power frequency current.

[0058] The intelligent diagnosis platform comprises a real-time analysis module based on FPGA with a processing delay less than 10ms, a cloud deep training module, and a mobile terminal interaction interface.

[0059] (Three) beneficial effects

[0060] Compared with the prior art, the present application provides a real-time state live detection method and system for electrical equipment, which has the following beneficial effects:

[0061] 1. In the present application, when the live detection of the state of the electrical equipment is carried out, the infrared thermodynamic characteristics, ultraviolet corona intensity and ultrasonic discharge signals are collected by the non-contact multi-source sensing array fusion, realizing the real-time sensing of multiple parameters under the live condition, overcoming the signal distortion problem caused by electromagnetic interference in traditional detection, improving the accuracy of state parameter extraction; at the same time, the multi-dimensional feature data is standardized by the edge computing unit, so that the system can eliminate the interference of environmental factors on the detection result, ensure the consistency of the evaluation result under different working conditions, and reduce the state diagnosis error.

[0062] 2. In the present application, when the sensor is deployed, the contact angle and pressure of the excitation probe and the surface of the equipment are adjusted in real time by the synergistic action of the magnetic attraction type adjustable support and the pressure feedback device, and the angle between the pulse emission direction and the normal line of the equipment is controlled; when the detection position deviates, the system automatically triggers three-dimensional posture correction according to the impedance matching feedback, so that the detection module can be coupled with zero deviation on the surface of the complex structure equipment, solving the detection failure problem caused by the traditional fixed installation method, and ensuring the accuracy and reliability of the data acquisition position.

[0063] 3. In the present application, when performing equipment health state evaluation, the temperature gradient change, corona intensity growth and discharge pulse characteristic parameters are analyzed by intelligent diagnosis model for multi-source information fusion, which automatically identifies the weak characteristics of insulation aging and local discharge hidden defects; at the same time, based on the dynamic weight optimization mechanism, the priority of key characteristic quantities is calibrated, so that the system can quantify the degradation degree of equipment health state, avoid the maintenance decision deviation caused by single parameter misjudgment, improve the predictability and accuracy of power supply system safety control, and realize the technical leap from passive repair to active protection. BRIEF DESCRIPTION OF DRAWINGS

[0064] Fig. 1 A flowchart of the real-time state live detection method of the electrical equipment according to the present application;

[0065] Fig. 2 A framework diagram of the real-time state live detection system of the electrical equipment according to the present application. DETAILED DESCRIPTION

[0066] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0067] Specific embodiments: please refer to Figs. 1-2 A real-time state live detection method and system of electrical equipment, the method comprising the following steps:

[0068] S1, collecting multi-dimensional state characteristic quantities of the live electrical equipment by a non-contact multi-source sensing array, and generating a real-time sensing data set;

[0069] S2, pre-processing and feature extraction of the real-time sensing data set based on an edge computing unit, and generating a standardized state feature vector;

[0070] S3, coupling a high-frequency pulse excitation module to the surface of the electrical equipment through a self-adaptive mounting structure, and generating an enhanced feature excitation signal;

[0071] S4, using the enhanced feature excitation signal to perform directional excitation on the equipment insulation defect area, synchronously collecting local discharge response signals and generating a pulse response map;

[0072] S5, inputting the standardized state feature vector and the pulse response map into an intelligent diagnosis model for multi-source information fusion analysis, and generating an equipment health state evaluation index;

[0073] S6, triggering a real-time early warning instruction and generating a live maintenance decision scheme when the device health state evaluation index exceeds a preset threshold value;

[0074] The non-contact multi-source sensing array comprises an infrared thermal imaging module, an ultraviolet corona detection module, and an ultrasonic partial discharge module.

[0075] S1 comprises the following steps:

[0076] S11, collecting device surface temperature field distribution data through a spatially distributed infrared sensor matrix to generate a thermodynamic feature vector;

[0077] S12, capturing corona discharge spectral data using a solar-blind ultraviolet detector to generate a corona intensity distribution map;

[0078] S13, collecting partial discharge acoustic signals based on a piezoelectric ultrasonic sensor array to generate a time-frequency domain discharge characteristic spectrum.

[0079] S2 comprises the following steps:

[0080] S21, performing environmental temperature compensation correction on the thermodynamic feature vector to generate a temperature gradient change matrix;

[0081] S22, performing spatial registration processing on the corona intensity distribution map to generate a corona intensity spatio-temporal evolution model;

[0082] S23, performing wavelet denoising processing on the time-frequency domain discharge characteristic spectrum to extract a set of discharge pulse feature parameters, wherein the pulse energy density The calculation formula is:

[0083] ;

[0084] wherein is the pulse energy density, is the sampling time window, is the time domain discharge signal function, is the time variable, is the time differential variable.

[0085] S3 comprises the following steps:

[0086] S31, attaching a high-frequency pulse excitation module to the device surface through a magnetic type adjustable support, the magnetic type adjustable support comprising a three-dimensional angle adjustment mechanism and a pressure feedback device;

[0087] S32, automatically adjusting the inclination angle of the excitation probe based on the device structure characteristics, so that the angle between the pulse emission direction and the normal line of the device surface is less than 5°;

[0088] S33, adjust the output impedance to 50±2Ω through the impedance matching circuit, and ensure that the signal transmission efficiency is greater than 95%.

[0089] The working parameters of the high-frequency pulse excitation module are:

[0090] Pulse frequency range: 1MHz-10MHz adjustable;

[0091] Output voltage amplitude: 0.1kV-5kV programmable;

[0092] Pulse rise time: less than 10ns;

[0093] Duty cycle adjustment range: 1%-50%;

[0094] And meet the waveform fidelity constraints:

[0095] ;

[0096] Wherein is the allowable voltage fluctuation amplitude, is the set output voltage peak value, is the actual working frequency, is the circuit characteristic frequency.

[0097] S4 includes the following steps:

[0098] S41, use a high-speed data acquisition card to synchronously collect the pulse response signal, and the sampling rate is greater than 200MSPS;

[0099] S42, locate the insulation defect position through the time domain reflection algorithm, and generate a defect space coordinate set;

[0100] S43, based on the pulse waveform characteristics, extract the discharge repetition frequency spectrum, single discharge energy distribution, phase resolution discharge spectrum and pulse sequence correlation matrix, and the calculation formula is:

[0101] ;

[0102] Wherein is the adjacent pulse sequence correlation index, is the sampling voltage value of the group pulse at the same time, is the arithmetic mean of the pulse sequence and , and is the number of sampling points in a single pulse period.

[0103] S5 includes the following steps:

[0104] S51, construct a deep belief network model, the input layer contains 12 feature channels corresponding to the temperature gradient change rate, the corona intensity growth rate, the discharge pulse repetition rate, the discharge energy entropy value, the pulse waveform distortion degree, the phase distribution symmetry and other fusion feature quantities;

[0105] S52, calculate the weight coefficient of each channel through the feature importance weighting algorithm, and the weight updating formula is:

[0106] ;

[0107] Among them is the weight coefficient of the i-th feature channel at the iteration number t, is the updated weight coefficient of the i-th feature channel at the iteration number t+1, the learning rate , is the partial derivative of the loss function L with respect to weight, is the loss function, is the feature channel index, is the iteration number;

[0108] S53, output the equipment health state evaluation index , the calculation formula is:

[0109] ;

[0110] Among them is the equipment health state evaluation index, is the feature index, is the total number of features, is the fusion feature quantity of the i-th feature, is the degradation factor of the i-th feature. S6 includes the following steps:

[0111] S61, establish a three-level early warning mechanism and a warning confidence factor:

[0112] ;

[0113] The primary warning threshold is: and

[0114] ; The intermediate warning threshold is: and

[0115] ; The emergency warning threshold is: ,

[0116] ; ;

[0117] ​wherein is a pre-warning confidence, is a standard deviation of the health index H, is a mean value of the health index H, is a decay coefficient, is a device health status evaluation index.

[0118] S62, generating a differentiated maintenance scheme according to the pre-warning level: the primary pre-warning triggers a 72h maintenance plan, the intermediate pre-warning triggers a 24h maintenance plan, and the emergency pre-warning triggers an immediate power-off maintenance instruction.

[0119] The system comprises:

[0120] A non-contact multi-source sensing array, which collects device surface infrared thermodynamic characteristic data by using an infrared thermal imaging module, captures corona discharge spectrum data by using an ultraviolet corona detection module, and collects partial discharge sound signals by using an ultrasonic partial discharge module to generate a real-time perception data set;

[0121] An edge computing unit, which pre-processes the real-time perception data set by using a signal pre-processing circuit, and extracts multi-dimensional features by using a feature extraction coprocessor to generate a standardized state feature vector;

[0122] A high-frequency pulse excitation module, which generates a high-frequency pulse signal by using a programmable pulse generator, outputs the signal to an intelligent diagnosis platform for analysis, and performs impedance matching by using an impedance matching network to generate an enhanced feature excitation signal;

[0123] An adaptive mounting structure, which has a magnetic adjustable support to adjust the mounting position and angle, and monitors the contact pressure by using a pressure feedback device to couple the high-frequency pulse excitation module to the surface of the electrical equipment;

[0124] An intelligent diagnosis platform, which runs a multi-source information fusion analysis algorithm, receives the standardized state feature vector and the pulse response atlas, and generates a device health status evaluation index;

[0125] A self-powered power supply unit, which obtains energy from a power frequency magnetic field by using an inductive power coil, and stores the energy by using a super capacitor energy storage device to supply power to each unit of the system.

[0126] Specifically, it comprises:

[0127] The adaptive mounting structure includes a carbon fiber insulated connecting rod with a withstand voltage level greater than 35kV / cm, a universal joint adjustment mechanism with an angle adjustment range of ±30°, a contact pressure sensor with a range of 0-50N adjustable, and a radio frequency shielding shell with a shielding effectiveness greater than 80dB;

[0128] The self-powered power supply unit is implemented by using a Rogowski coil for inductive power, which outputs a power greater than 5W under a 1A power frequency current, and the output power satisfies:

[0129] ;

[0130] wherein is the output power, is the energy conversion efficiency, is the power frequency magnetic field strength, is the effective cross-sectional area of the induction coil, is the power frequency, is the number of turns of the coil;

[0131] The intelligent diagnosis platform comprises a real-time analysis module based on FPGA, a cloud deep training module, and a mobile terminal interactive interface, and the processing delay is less than 10 ms.

[0132] The operation steps of the real-time state live detection method and system of the electrical equipment are as follows:

[0133] Step one, multi-source signal cooperative acquisition and preprocessing

[0134] The surface temperature field distribution of the live equipment is scanned in real time by a spatial distributed infrared sensor matrix to generate a dynamic thermodynamic feature vector. A solar-blind ultraviolet detector is used to capture the corona discharge spectrum intensity to construct a corona intensity space-time evolution model. A piezoelectric ultrasonic sensor array is used to collect local discharge sound signals, and a wavelet denoising process is used to extract the time-frequency domain discharge characteristic spectrum. The three-source data are gathered to an edge computing unit through a low-delay wireless transmission protocol. The thermodynamic features are corrected by environmental temperature compensation to eliminate environmental interference. The corona intensity distribution map is subjected to spatial registration processing to ensure space-time consistency. Finally, the standardized state feature vector is output, the signal distortion problem caused by electromagnetic noise in traditional detection is solved, and multi-physical quantity synchronous perception under live conditions is realized.

[0135] Step two, adaptive excitation and defect feature enhancement

[0136] A three-dimensional angle adjustment mechanism based on a magnetic attraction type adjustable support and a pressure feedback device are used. A high-frequency pulse excitation module is used, the working frequency is adjustable at 1-10 MHz, the output voltage is programmable at 0.1-5 kV, and the module is adaptively attached to the surface of the equipment. The probe inclination angle is adjusted in real time to make the angle between the pulse emission direction and the equipment normal less than 5°. The output impedance is stabilized at 50±2Ω through an impedance matching circuit to ensure that the signal transmission efficiency is greater than 95%. The excitation signal is used to directionally excite the insulation defect area. A 200MS / s high-speed data acquisition card is used to synchronously capture the local discharge response. The time-domain reflection algorithm is used to locate the defect spatial coordinates. The discharge repetition frequency spectrum, single discharge energy distribution, and pulse sequence correlation index are extracted. The similarity of adjacent pulse waveforms is calculated to enhance the weak feature capture ability of hidden defects.

[0137] Step three, multi-source fusion diagnosis and dynamic early warning decision

[0138] The deep belief network model receives 12-dimensional standardized feature vectors, including temperature gradient change rate, corona intensity growth rate, discharge pulse repetition rate and pulse response spectrum, dynamically optimizes the weight coefficients of each channel through the feature importance weighting algorithm, and outputs the equipment health state evaluation index H. When the H value is lower than the preset threshold, a three-level early warning mechanism is triggered: combined with the early warning confidence factor Execute hierarchical response: H≥0.7 and >0.85 triggers a 72-hour maintenance plan, 0.6≤H<0.7 and >0.75 starts 24-hour emergency maintenance, H<0.6, ≤0.6, immediately execute power-off maintenance, realize the quantification and active protection of equipment state degradation process.

[0139] Step four, system cooperative control and safety protection

[0140] The self-powered power supply unit induces power frequency magnetic field energy through a Rogowski coil, and outputs power greater than 5W, driving the carbon fiber insulated connecting rod, with a withstand voltage level greater than 35kV / cm, supporting the sensing-stimulating module; the intelligent diagnosis platform realizes real-time analysis of less than 10ms relying on the FPGA chip, and the cloud training module continuously updates the multi-source information fusion algorithm; the radio frequency shielding shell: shielding effectiveness greater than 80dB, external electromagnetic interference suppression, universal joint adjusting mechanism: angle range ±30° and contact pressure sensor: range 0-50N, constitute double safety monitoring, guarantee the zero safety accident in the process of live detection, form the full closed loop technology system from data acquisition, feature enhancement, intelligent diagnosis to safety maintenance.

[0141] It should be noted that in this text, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "including a…" does not exclude the presence of additional identical elements in the process, method, article or equipment including the element.

[0142] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for real-time condition live detection of electrical equipment, characterized in that: The method comprises the following steps: S1, collecting multi-dimensional state characteristic quantities of the electrically charged operating electrical equipment through a non-contact multi-source sensing array to generate a real-time perception data set; S2, pre-processing and feature extraction on the real-time perception data set based on an edge computing unit to generate a standardized state feature vector; S3, coupling a high-frequency pulse excitation module to the surface of the electrical equipment through an adaptive mounting structure to generate an enhanced feature excitation signal; S4, using the enhanced feature excitation signal to perform directional excitation on the insulation defect area of the equipment, synchronously collecting a partial discharge response signal and generating a pulse response map; S5, inputting the standardized state feature vector and the pulse response map into an intelligent diagnosis model for multi-source information fusion analysis to generate an equipment health state evaluation index; S6, when the equipment health state evaluation index exceeds a preset threshold, triggering a real-time warning instruction and generating an on-line maintenance decision scheme; Wherein, the non-contact multi-source sensing array comprises an infrared thermal imaging module, an ultraviolet corona detection module and an ultrasonic partial discharge module, and the adaptive mounting structure integrates a self-powered power supply unit and an insulation isolation device.

2. The method of claim 1, wherein: The S1 comprises the following steps: S11, collecting device surface temperature field distribution data through a spatially distributed infrared sensor matrix to generate a thermodynamic feature vector; S12, capturing corona discharge spectrum data using a solar blind ultraviolet detector to generate a corona intensity distribution map; S13, collecting partial discharge acoustic signals based on a piezoelectric ultrasonic sensor array to generate a time-frequency domain discharge feature spectrum.

3. The method of claim 2, wherein: The S2 comprises the following steps: S21, performing ambient temperature compensation correction on the thermodynamic feature vector to generate a temperature gradient change matrix; S22, performing spatial registration processing on the corona intensity distribution map to generate a corona intensity spatio-temporal evolution model; S23, performing wavelet denoising processing on the time-frequency domain discharge feature spectrum to extract a set of discharge pulse feature parameters.

4. The method of claim 1, wherein: The S3 comprises the following steps: S31, attaching a high-frequency pulse excitation module to the surface of the equipment through a magnetic type adjustable support, the magnetic type adjustable support comprising a three-dimensional angle adjusting mechanism and a pressure feedback device; S32, automatically adjusting the inclination angle of the excitation probe based on the structural features of the equipment, so that the angle between the pulse emission direction and the normal line of the equipment surface is less than 5°; S33, adjusting the output impedance to 50±2Ω through an impedance matching circuit to ensure that the signal transmission efficiency is greater than 95%.

5. The method of claim 1, wherein: The working parameters of the high-frequency pulse excitation module are: Pulse frequency range: 1MHz-10MHz adjustable; Output voltage amplitude: 0.1kV-5kV programmable; Pulse rise time: less than 10ns; Duty cycle adjustment range: 1%-50%.

6. The method of claim 1, wherein: The S4 comprises the following steps: S41, synchronously collecting pulse response signals using a high-speed data acquisition card with a sampling rate greater than 200MSPS; S42, positioning the insulation defect position through a time domain reflection algorithm to generate a set of defect spatial coordinates; S43, extracting a discharge repetition frequency spectrum, a single discharge energy distribution, a phase-resolved discharge spectrum and a pulse sequence correlation matrix based on pulse waveform features.

7. The method of claim 1, wherein: The S5 comprises the following steps: S51, a deep belief network model is constructed, and the input layer includes 12 feature channels corresponding to the temperature gradient change rate, the corona intensity growth rate, the discharge pulse repetition rate, the discharge energy entropy value, the pulse waveform distortion degree, the phase distribution symmetry, and other fused feature quantities; S52, each channel weight coefficient is calculated through a feature importance weighting algorithm, and the weight updating formula is: ； wherein is the weight coefficient of the i-th feature channel at iteration number t, is the updated weight coefficient of the i-th feature channel at iteration number t+1, learning rate , is the partial derivative of the loss function L with respect to the weight, is the loss function, is the feature channel index, is the iteration number; S53, outputting the device health state evaluation index , the calculation formula of which is: ； wherein is a device health state evaluation index, is a feature index, is a total number of features, is a fusion feature quantity of the th feature, is a degradation factor of the th feature.

8. The method of claim 7, wherein the method further comprises: The S6 includes the following steps: S61、establish three-level early warning mechanism: the primary warning threshold is 0.7 <0.8, the intermediate warning threshold is 0.6 <0.7, the emergency warning threshold is <0.6; S62, a differentiated maintenance scheme is generated according to the early warning level: the primary early warning triggers a 72h maintenance plan, the intermediate early warning triggers a 24h maintenance plan, and the emergency early warning triggers an immediate power-off maintenance instruction.

9. An electrical equipment real-time condition on-line detection system for implementing the electrical equipment real-time condition on-line detection method of any one of claims 1-8, characterized in that, The system comprises: A non-contact multi-source sensing array acquires infrared thermodynamic characteristic data of the equipment surface by using an infrared thermal imaging module, captures corona discharge spectrum data by using an ultraviolet corona detection module, and acquires local discharge sound signals by using an ultrasonic local discharge module to generate a real-time sensing data set; An edge computing unit pre-processes the real-time sensing data set through a signal pre-processing circuit, extracts multi-dimensional features by using a feature extraction coprocessor, and generates a standardized state feature vector; A high-frequency pulse excitation module generates a high-frequency pulse signal by using a programmable pulse generator, outputs the signal to an intelligent diagnosis platform for analysis, and generates an enhanced feature excitation signal by using an impedance matching network for impedance matching; An adaptive mounting structure has a magnetic adjustable support to adjust the mounting position and angle, and a pressure feedback device to monitor the contact pressure, so as to couple the high-frequency pulse excitation module to the surface of the electrical equipment; An intelligent diagnosis platform runs a multi-source information fusion analysis algorithm, receives the standardized state feature vector and the pulse response spectrum, and generates a device health state evaluation index; A self-powered power supply unit acquires energy from a power frequency magnetic field by using an inductive power coil, and stores the energy by using a super capacitor energy storage device to supply power to each unit of the system.

10. The real-time condition live detection system for electrical equipment according to claim 9, characterized in that: Specifically, the adaptive mounting structure includes a carbon fiber insulated connecting rod with a withstand voltage level greater than 35kV / cm, a universal joint adjusting mechanism with an angle adjusting range of ±30°, a contact pressure sensor with a adjustable range of 0-50N, and a radio frequency shielding shell with a shielding effectiveness greater than 80dB; The self-powered power supply unit is implemented by using a Rogowski coil for inductive power acquisition, and outputs a power greater than 5W under a 1A power frequency current; The intelligent diagnosis platform includes an FPGA-based real-time analysis module with a processing delay less than 10ms, a cloud deep training module, and a mobile terminal interaction interface. ​

Images