PD detection system based on multi-modal sensing fusion and working method and application thereof

Abstract

The invention relates to a PD detection system based on multi-mode sensing fusion and a working method and application thereof, and the system comprises a PD detection terminal which is used for collecting equipment environment information and partial discharge information, displaying a corresponding time domain waveform, a PRPD / PRPS atlas and an infrared thermal image atlas, and preliminarily determining an equipment state, a partial discharge type, a probability and an abnormal position; the remote cloud server is used for secondarily judging the equipment state, the partial discharge type, the probability and the abnormal position by adopting a multi-mode sensing fusion method, confirming a result, calculating a partial discharge risk index according to a dynamic partial discharge risk index formula, confirming a risk level and generating a disposal suggestion; and the mobile phone mobile terminal is used for interacting with the remote cloud server to complete a maintenance task. According to the method, the time domain waveform, the PRPD / PRPS spectrum, the infrared thermal image spectrum judgment and the multi-mode sensing fusion technology are combined, the equipment state, the partial discharge type, the probability and the abnormal position detection are realized, and the reliability and the safety of equipment operation are guaranteed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of partial discharge detection, and particularly relates to a PD detection system based on multi-modal sensor fusion and a working method and application thereof. BACKGROUND

[0002] As the main cause of insulation deterioration of power equipment, partial discharge has become a major threat to the safe operation of distribution systems, and its harm is manifested in insulation material damage, equipment life shortening, frequent safety accidents, and large hidden economic losses. The traditional pulse current method is easily disturbed by noise and is difficult to detect early weak discharge, and the existing system relies on a single signal, has a high misjudgment rate and cannot dynamically capture the evolution process of the discharge mode.

[0003] With the development of science and technology, new detection methods have appeared in partial discharge detection. For example, Chinese Patent CN119881561A discloses a converter transformer state detection method and platform based on PD monitoring, which mainly detects the converter transformer by using sound, light, electricity and heat signals, but does not consider the real-time of data, the influence of environmental temperature and humidity, and the multi-sensor collaborative optimization and intelligent analysis means. Therefore, how to ensure the real-time of data, the influence of environmental temperature and humidity, and the multi-sensor collaborative optimization to complete the judgment of partial discharge type, probability and abnormal position is an urgent technical problem to be solved. SUMMARY

[0004] The purpose of the present application is to provide a PD detection system based on multi-modal sensor fusion and a working method and application thereof, to solve the problem of low precision in partial discharge detection in the prior art, to realize accurate detection of equipment state, partial discharge type, probability and abnormal position, and to improve the reliability and safety of power equipment operation.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is: a PD detection system based on multi-modal sensor fusion, comprising:

[0006] A PD detection terminal is used to collect equipment environment information and partial discharge information, display corresponding time-domain waveforms, PRPD / PRPS graphs and infrared thermal image graphs, and preliminarily determine the equipment state, partial discharge type, probability and abnormal position. The PD detection terminal is connected to a remote cloud server through a 4G / 5G network to send the collected and processed information to the remote cloud server;

[0007] The remote cloud server is connected to the PD detection terminal and the mobile terminal via 4G / 5G networks respectively. It is used to receive information sent by the PD detection terminal, and to make secondary judgments on the equipment status, partial discharge type, probability and abnormal location using a multimodal sensor fusion method. The results are confirmed, the partial discharge risk index is calculated according to the dynamic partial discharge risk index formula, the risk level is confirmed, and disposal suggestions are generated and sent to the mobile terminal of the maintenance personnel.

[0008] The mobile terminal is used to interact with the remote cloud server, receive information sent by the remote cloud server, perform data parsing and display, and send detection and control commands to the remote cloud server to complete maintenance tasks.

[0009] Furthermore, the PD detection terminal includes a zero-cross detector, an ambient temperature probe, an ambient humidity probe, a UHF probe, an HFCT probe, a TEV probe, an AA / AE probe, a conditioning circuit, an AD circuit, an FPGA controller, an infrared module, and a tablet computer;

[0010] The zero-crossing detector is electrically connected to the FPGA controller and is used to receive a 220V AC power signal, which is then internally stepped down and zero-crossing compared before being sent to the FPGA controller in a pulse manner.

[0011] The ambient temperature probe, ambient humidity probe, UHF probe, HFCT probe, TEV probe, and AA / AE probe are electrically connected to the conditioning circuit to receive ambient temperature signals, ambient humidity signals, UHF signals, HFCT signals, TEV signals, and AA / AE signals, and amplify and filter the signals before sending them to the AD circuit.

[0012] The AD circuit is connected to the FPGA controller in parallel and is used to convert the analog signals from all probes after the conditioning circuit into digital signals and send them to the FPGA controller.

[0013] The FPGA controller is connected to the USB port of the tablet computer to receive digital signals from all probes and inserts a zero-phase stamp into the digital signals based on the pulse signals emitted by the zero-crossing detector before sending them to the tablet computer.

[0014] The infrared module is connected to the RJ45 port of the tablet computer and is used to receive infrared temperature matrix signals and convert them into digital signals to send to the tablet computer.

[0015] The tablet computer is used to receive ambient temperature, ambient humidity, UHF, HFCT, TEV, AA / AE signals and infrared temperature matrix signals. It displays the UHF, HFCT, TEV and AA / AE signals as time-domain waveforms and PRPD / PRPS maps, and the infrared temperature matrix signals as infrared thermal images. By incorporating ambient temperature and humidity factors, it determines the type and degree of partial discharge based on the characteristics of the time-domain waveforms and PRPD / PRPS maps, and determines the specific location of the partial discharge based on the infrared temperature matrix map. At the same time, it sends all data to a remote cloud server via a 4G / 5G network.

[0016] The present invention also provides a method for operating the above-mentioned PD detection system based on multimodal sensor fusion, comprising the following steps:

[0017] S1. The device status data is synchronously collected based on the pulse signal emitted by the zero-crossing detector through the ambient temperature probe, ambient humidity probe, UHF probe, HFCT probe, TEV probe, and AA / AE probe, and the infrared temperature matrix signal is collected through the infrared module.

[0018] S2. Display the corresponding time-domain waveform, PRPD / PRPS spectrum and infrared thermographic spectrum on the PD detection terminal. Based on the key features of the time-domain waveform, PRPD / PRPS spectrum and infrared thermographic spectrum and the discharge type discrimination rules, preliminarily determine the equipment status, partial discharge type, probability and abnormal location.

[0019] After adding a timestamp, the S3 PD detection terminal uploads ambient temperature, ambient humidity, UHF, HFCT, TEV, AA / AE signals, and infrared temperature matrix signals, as well as the preliminary determination of the type, probability, and location of partial discharge, to the remote cloud server.

[0020] S4 and the remote cloud server respectively use corresponding data processing algorithms to extract target features from UHF, HFCT, TEV, AA / AE signals and infrared temperature matrix signals;

[0021] S5. Construct an environmental interference correction factor matrix based on ambient temperature and humidity, and adjust the confidence weight of each probe data; adopt a multimodal cross-attention mechanism, using UHF signal as the reference query vector, infrared temperature matrix signal, HFCT current pulse signal, TEV voltage signal, and AA / AE acoustic signal as key value vectors, and calculate multidimensional feature association weights.

[0022] S6. Multidimensional features are weighted and fused according to attention weights. The fused temporal features are processed by LSTM or Transformer encoder. Spatial features are extracted by CNN and the device status, partial discharge type, probability and abnormal location are determined by a fully connected layer.

[0023] S7. When there are abnormal environmental conditions, abnormal data, or data conflicts from multiple probes, a dynamic strategy and conflict arbitration mechanism shall be adopted for determination.

[0024] S8. The remote cloud server combines the initial and secondary determinations of the device status, partial discharge type, probability, and abnormal location to confirm the device status, partial discharge type, probability, and abnormal location.

[0025] S9. The remote cloud server calculates the partial discharge risk index according to the dynamic partial discharge risk index formula, classifies the risk level according to the risk level classification standard, and generates disposal suggestions.

[0026] S10. The remote cloud server will send the final confirmed equipment status, partial discharge type, probability, abnormal location, risk index, risk level, and handling suggestions to the maintenance personnel's mobile terminal to complete the maintenance task.

[0027] Furthermore, in step S2, the key features include: UHF signal pulse clustering, spectral asymmetry, HFCT rise time, propagation delay, TEV signal voltage amplitude, AA / AE signal source azimuth deviation, energy accumulation, and steep temperature gradient of infrared temperature matrix signal.

[0028] The discharge type discrimination rules include: a single-peak distribution in the UHF signal spectrum indicates corona discharge; multiple continuous bursts indicate surface discharge; a short rise time and stable delay in the HFCT signal indicate internal insulation discharge; a TEV signal amplitude > 50mV indicates metal tip discharge or surface discharge; a sudden increase in acoustic emission energy and azimuth fluctuation in the AA / AE signal indicates mechanical loosening; and a hotspot temperature rise rate > 2℃ / s in the infrared temperature matrix signal indicates overheating discharge.

[0029] Furthermore, in step S4, for the UHF signal, the energy proportion and instantaneous frequency slope of the 300-2000MHz frequency band are extracted by short-time Fourier transform and wavelet packet decomposition.

[0030] For HFCT signals, the pulse rise time and propagation delay differences are calculated using db4 basis function wavelet front detection.

[0031] For TEV signals, an adaptive Kalman filter and peak statistics are used to generate a pulse amplitude distribution map after background noise suppression.

[0032] For AA / AE signals, a probability contour map of the sound source location is constructed by estimating the direction of arrival and integrating the energy accumulation.

[0033] For the infrared temperature matrix signal, the coordinates of the hot spot center are located and the temperature rise rate is calculated through the thermal diffusion model and gradient descent optimization.

[0034] Further, in step S5, the expression for the environmental interference correction factor matrix is:

[0035]

[0036] Among them, W env W represents the adjusted weights. UHF W represents the UHF weights. HFCT W represents the HFCT weights. TEV W represents the TEV weight. AA / AE W represents the AA / AE weights. IR Indicates infrared weighting;

[0037] The multimodal cross-attention mechanism constructs a temperature-discharge correlated attention model, aligns the hotspot coordinates of the infrared temperature matrix signal with the UHF signal pulse timing, calculates the spatial-temporal collaborative weights, and uses an acoustic-optical joint attention branch to capture the causal relationship between acoustic emission events of the AE / AE probe and infrared temperature abrupt changes in parallel through a multi-head attention mechanism.

[0038] Further, in step S6, the weighted fusion specifically includes: performing tensor fusion on the infrared temperature matrix signal and the UHF signal to generate a discharge-temperature correlation spectrum; when a local temperature >150℃ is detected, activating the overheat protection protocol; and preferentially fusing the acoustic emission positioning data of the AA / AE probe with the hotspot coordinates of the infrared temperature matrix signal.

[0039] Furthermore, in step S7, the dynamic strategy and conflict arbitration mechanism include: when humidity > 85%, reducing the TEV weight and increasing the AA / AE weight and infrared weight; when infrared temperature rise > 10℃ / s, enabling UHF+HFCT to use the TDOA algorithm for joint positioning; when acoustic-electric signal orientation deviation > 20°, resetting the top 3 arbitrations based on attention weight, prioritizing the adoption of UHF, infrared, and HFCT signals; when multiple probe data loss ≥ 2 channels, initiating a degradation mode, relying only on UHF+infrared signals, and relaxing the confidence threshold to 0.75.

[0040] Further, in step S9, the formula for the dynamic partial discharge risk index is:

[0041]

[0042]

[0043] Among them, P i For the i-th type of partial discharge, W i T represents the weight corresponding to the i-th type of partial discharge. 实际 T represents the actual ambient temperature of the equipment. 额定 The rated ambient temperature of the equipment, H实际 R and T represent the actual ambient humidity of the equipment. effect H effect For intermediate calculation variables;

[0044] The risk level classification standard is: partial discharge risk index R. 动态 The interval [1, +∞) is defined as high risk, and the partial discharge risk index R is... 动态 The interval [0.6, 1) is defined as medium risk, and the rest are defined as low risk;

[0045] The proposed measures include: routine inspections for low-risk cases, daily monitoring combined with UHF / TEV positioning of the power supply for medium-risk cases, and immediate shutdown and maintenance for high-risk cases, using HFCT + infrared retesting.

[0046] This invention also provides applications of the aforementioned PD detection system and its working method based on multimodal sensor fusion. The PD detection system and its working method are applied to the detection of substations and overhead lines. When applied to the detection of substations, it enables partial discharge inspection and online monitoring of related equipment, including switchgear and transformers, within the substation. When applied to the detection of overhead lines, it enables temperature detection and partial discharge detection of related devices, including conductor joints and insulators, on the overhead lines.

[0047] Compared with existing technologies, this invention has the following advantages: Addressing the problems of existing partial discharge detection technologies such as reliance on single signals, poor environmental adaptability, and static risk assessment, this invention provides a PD detection system and its working method based on multimodal sensor fusion. Through innovative designs such as multimodal sensor fusion, environmental interference correction, dynamic arbitration, and intelligent decision-making, it significantly improves detection accuracy, reliability, and operational efficiency. The system integrates UHF, HFCT, TEV, AA / AE probes, and an infrared module through a PD detection terminal, simultaneously acquiring multi-dimensional electrical, acoustic, and thermal signals. The remote cloud server employs a multimodal cross-attention mechanism, using the UHF signal as a reference, and combining key-value vectors such as the infrared temperature matrix and HFCT current pulses to calculate spatiotemporal collaborative weights, achieving deep fusion of multimodal features. Combined with key features of PRPD / PRPS maps, it significantly improves the accuracy of partial discharge type identification and reduces the risk of misjudgment from single-modal detection. Simultaneously, it innovatively constructs an environmental interference correction factor matrix, dynamically adjusting the confidence weights of each sensor based on environmental temperature and humidity to ensure the reliability of data acquisition in complex environments. Furthermore, it designs dynamic strategies and conflict arbitration mechanisms to enhance the system's continuous operation capability under complex conditions. Furthermore, by integrating discharge type weights, environmental parameters, and equipment ratings through the dynamic partial discharge risk index formula, the risk level is quantified and disposal suggestions are automatically generated, optimizing the operation and maintenance response process and improving inspection efficiency. The PD detection terminal adopts a modular design, supporting rapid migration to equipment of different voltage levels. Through 4G / 5G, it achieves "terminal-cloud-mobile terminal" collaboration, forming a "collection-analysis-decision" closed loop, reducing system deployment costs, and adapting to multiple scenarios such as power substations and overhead lines. It effectively solves the problems of low accuracy, weak adaptability, and insufficient efficiency of existing technologies, providing an intelligent solution for power equipment monitoring. Attached Figure Description

[0048] Figure 1 This is an implementation architecture diagram of a PD detection system based on multimodal sensor fusion provided in an embodiment of the present invention;

[0049] Figure 2 This is a flowchart illustrating the working method of the PD detection system based on multimodal sensor fusion provided in this embodiment of the invention. Detailed Implementation

[0050] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0051] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0052] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0053] like Figure 1 As shown, this embodiment provides a PD detection system based on multimodal sensor fusion, including: a PD detection terminal, a remote cloud server, and a mobile terminal.

[0054] The PD detection terminal is used to collect equipment environmental information and partial discharge information, display the corresponding time-domain waveform, PRPD / PRPS spectrum and infrared thermal image spectrum, and preliminarily determine the equipment status, partial discharge type, probability and abnormal location; the PD detection terminal is connected to a remote cloud server through a 4G / 5G network to send the collected and processed information to the remote cloud server.

[0055] The remote cloud server is connected to the PD detection terminal and the mobile terminal via 4G / 5G networks. It is used to receive information sent by the PD detection terminal, use a multimodal sensor fusion method to make secondary determinations on the device status, partial discharge type, probability and abnormal location, and confirm the results. It calculates the partial discharge risk index according to the dynamic partial discharge risk index formula, confirms the risk level, generates handling suggestions, and sends them to the mobile terminal of the maintenance personnel.

[0056] The mobile terminal is used to interact with the remote cloud server, receive information sent by the remote cloud server, perform data parsing and display, and send detection and control commands to the remote cloud server to complete maintenance tasks.

[0057] This invention utilizes a PD detection terminal, a remote cloud server, and a mobile terminal to implement a PD detection system based on multimodal sensor fusion. This system combines traditional time-domain waveform, PRPD / PRPS spectrum, infrared thermal image spectrum determination, and multimodal sensor fusion technology to detect device status, partial discharge type, probability, and abnormal location, ensuring the safety and stability of device operation. Through design, simulation, and verification, a modular product is formed, enabling rapid portability between different platforms and accelerating the product development process.

[0058] The PD detection terminal includes a zero-cross detector, an ambient temperature probe, an ambient humidity probe, a UHF probe, an HFCT probe, a TEV probe, an AA / AE probe, a conditioning circuit, an AD circuit, an FPGA controller, an infrared module, and a tablet computer.

[0059] The zero-crossing detector is electrically connected to the FPGA controller and is used to receive a 220V AC power signal, which is then internally stepped down and zero-crossing compared before being sent to the FPGA controller in a pulse manner.

[0060] The ambient temperature probe, ambient humidity probe, UHF probe, HFCT probe, TEV probe, and AA / AE probe are electrically connected to the conditioning circuit to receive ambient temperature signals, ambient humidity signals, UHF signals, HFCT signals, TEV signals, and AA / AE signals, and then amplify and filter the signals before sending them to the AD circuit.

[0061] The AD circuit is connected in parallel to the FPGA controller and is used to convert the analog signals from all probes after the conditioning circuit into digital signals and send them to the FPGA controller.

[0062] The FPGA controller is connected to the USB port of the tablet computer to receive digital signals from all probes and inserts a zero-phase stamp into the digital signals based on the pulse signal emitted by the zero-crossing detector before sending them to the tablet computer.

[0063] The infrared module is connected to the RJ45 port of the tablet computer and is used to receive infrared temperature matrix signals and convert them into digital signals to be sent to the tablet computer.

[0064] The tablet computer is used to receive ambient temperature, ambient humidity, UHF, HFCT, TEV, AA / AE signals and infrared temperature matrix signals. It displays the UHF, HFCT, TEV and AA / AE signals as time-domain waveforms and PRPD / PRPS maps, and the infrared temperature matrix signals as infrared thermal images. By incorporating ambient temperature and humidity factors, it determines the type and degree of partial discharge based on the characteristics of the time-domain waveforms and PRPD / PRPS maps, and determines the specific location of the partial discharge based on the infrared temperature matrix map. At the same time, it sends all data to a remote cloud server via a 4G / 5G network.

[0065] The infrared module is an M640F manufactured by Wuhan Gewu Youxin, and the FPGA controller hardware is a Cyclone IV EP4CE10F17C8, which was compiled using Quartus II_13.0 software.

[0066] like Figure 2 As shown, this embodiment provides a working method for the above-mentioned PD detection system based on multimodal sensor fusion, including the following steps:

[0067] S1. The device status data is synchronously collected based on the pulse signal emitted by the zero-crossing detector through the ambient temperature probe, ambient humidity probe, UHF probe, HFCT probe, TEV probe, and AA / AE probe, and the infrared temperature matrix signal is collected through the infrared module.

[0068] S2. Display the corresponding time-domain waveform, PRPD / PRPS spectrum and infrared thermographic spectrum on the PD detection terminal. Based on the key features of the time-domain waveform, PRPD / PRPS spectrum and infrared thermographic spectrum and the discharge type discrimination rules, preliminarily determine the equipment status, partial discharge type, probability and abnormal location.

[0069] After adding a timestamp, the S3 PD detection terminal uploads ambient temperature, ambient humidity, UHF, HFCT, TEV, AA / AE signals, and infrared temperature matrix signals, as well as the preliminary determination of the type, probability, and location of partial discharge, to the remote cloud server.

[0070] S4 and the remote cloud server respectively use corresponding data processing algorithms to extract target features from UHF, HFCT, TEV, AA / AE signals and infrared temperature matrix signals;

[0071] S5. Construct an environmental interference correction factor matrix based on ambient temperature and humidity, and adjust the confidence weight of each probe data; adopt a multimodal cross-attention mechanism, using UHF signal as the reference query vector, infrared temperature matrix signal, HFCT current pulse signal, TEV voltage signal, and AA / AE acoustic signal as key value vectors, and calculate multidimensional feature association weights.

[0072] S6. Multidimensional features are weighted and fused according to attention weights. The fused temporal features are processed by LSTM or Transformer encoder. Spatial features are extracted by CNN and the device status, partial discharge type, probability and abnormal location are determined by a fully connected layer.

[0073] S7. When there are abnormal environmental conditions, abnormal data, or data conflicts from multiple probes, a dynamic strategy and conflict arbitration mechanism shall be adopted for determination.

[0074] S8. The remote cloud server combines the initial and secondary determinations of the device status, partial discharge type, probability, and abnormal location to confirm the device status, partial discharge type, probability, and abnormal location.

[0075] S9. The remote cloud server calculates the partial discharge risk index according to the dynamic partial discharge risk index formula, classifies the risk level according to the risk level classification standard, and generates disposal suggestions.

[0076] S10. The remote cloud server will send the final confirmed equipment status, partial discharge type, probability, abnormal location, risk index, risk level, and handling suggestions to the maintenance personnel's mobile terminal to complete the maintenance task.

[0077] In step S2, the key features include: UHF signal pulse clustering, spectral asymmetry, HFCT rise time, propagation delay, TEV signal voltage amplitude, AA / AE signal source azimuth deviation, energy accumulation, and steep temperature gradient of infrared temperature matrix signal.

[0078] The discharge type discrimination rules include: a single-peak distribution in the UHF signal spectrum indicates corona discharge; multiple continuous bursts indicate surface discharge; a short rise time and stable delay in the HFCT signal indicate internal insulation discharge; a TEV signal amplitude > 50mV indicates metal tip discharge or surface discharge; a sudden increase in acoustic emission energy and azimuth fluctuation in the AA / AE signal indicates mechanical loosening; and a hotspot temperature rise rate > 2℃ / s in the infrared temperature matrix signal indicates overheating discharge.

[0079] In step S4, for UHF signals, the energy proportion and instantaneous frequency slope of the 300-2000MHz frequency band are extracted by short-time Fourier transform and wavelet packet decomposition.

[0080] For HFCT signals, the pulse rise time and propagation delay differences are calculated using db4 basis function wavelet front detection.

[0081] For TEV signals, an adaptive Kalman filter and peak statistics are used to generate a pulse amplitude distribution map after background noise suppression.

[0082] For AA / AE signals, a probability contour map of the sound source location is constructed by estimating the direction of arrival and integrating the energy accumulation.

[0083] For the infrared temperature matrix signal, the coordinates of the hot spot center are located and the temperature rise rate is calculated through the thermal diffusion model and gradient descent optimization.

[0084] In step S5, the expression for the environmental disturbance correction factor matrix is:

[0085]

[0086] Among them, W env W represents the adjusted weights. UHF W represents the UHF weights. HFCT W represents the HFCT weights. TEV W represents the TEV weight. AA / AE W represents the AA / AE weights. IR Indicates infrared weights.

[0087] The multimodal cross-attention mechanism constructs a temperature-discharge correlated attention model, aligns the hotspot coordinates of the infrared temperature matrix signal with the UHF signal pulse timing, calculates the spatial-temporal collaborative weights, and uses an acoustic-optical joint attention branch to capture the causal relationship between acoustic emission events of the AE / AE probe and infrared temperature abrupt changes in parallel through a multi-head attention mechanism.

[0088] In step S6, the weighted fusion specifically includes: performing tensor fusion on the infrared temperature matrix signal and the UHF signal to generate a discharge-temperature correlation spectrum; when a local temperature >150℃ is detected, activating the overheat protection protocol; and preferentially fusing the acoustic emission positioning data of the AA / AE probe with the hotspot coordinates of the infrared temperature matrix signal.

[0089] In step S7, the dynamic strategy and conflict arbitration mechanism include: when humidity > 85%, reducing the TEV weight and increasing the AA / AE weight and infrared weight; when infrared temperature rise > 10℃ / s, enabling UHF+HFCT to use the TDOA algorithm for joint positioning; when acoustic-electric signal orientation deviation > 20°, resetting the top 3 arbitrations based on attention weight, prioritizing the adoption of UHF, infrared, and HFCT signals; when multiple probe data loss ≥ 2 channels, starting the degradation mode, relying only on UHF+infrared signals, and relaxing the confidence threshold to 0.75.

[0090] In step S9, the formula for the dynamic partial discharge risk index is:

[0091]

[0092]

[0093] Among them, P i For the i-th type of partial discharge, W i T represents the weight corresponding to the i-th type of partial discharge. 实际 T represents the actual ambient temperature of the equipment. 额定 The rated ambient temperature of the equipment, H 实际 R and T represent the actual ambient humidity of the equipment. effect H effect For intermediate calculation variables.

[0094] The risk level classification standard is: partial discharge risk index R. 动态 The interval [1, +∞) is defined as high risk, and the partial discharge risk index R is... 动态 The range [0.6, 1) is defined as medium risk, and the rest are defined as low risk.

[0095] The proposed measures include: routine inspections for low-risk cases, daily monitoring combined with UHF / TEV positioning of the power supply for medium-risk cases, and immediate shutdown and maintenance for high-risk cases, using HFCT + infrared retesting.

[0096] This invention synchronously collects equipment status data based on pulse signals emitted by a zero-crossing detector using ambient temperature probes, ambient humidity probes, UHF probes, HFCT probes, TEV probes, and AA / AE probes. An infrared module collects infrared temperature matrix signals. Based on key features of time-domain waveforms, PRPD / PRPS spectra, and infrared thermal images, and discharge type discrimination rules, the device status, partial discharge type, probability, and abnormal location are initially determined. A remote cloud server uses corresponding data processing algorithms to extract target features. An environmental interference correction factor matrix formula is constructed based on ambient temperature and humidity, adjusting the confidence weights of each sensor's data. A multimodal cross-attention mechanism is used to calculate the multi-dimensional feature association weights. Multimodal features are weighted and fused according to the attention weights. A second determination of the device status, partial discharge type, probability, and abnormal location is made by comparing with a preset model. The partial discharge risk is calculated using a dynamic partial discharge risk index formula. Risk levels are classified according to risk level classification standards, and treatment suggestions are generated. Finally, the data is sent to the mobile terminal of maintenance personnel to complete the maintenance task.

[0097] The PD detection system based on multimodal sensor fusion in this embodiment is implemented using standard C language and a modular design approach. It uses a PD detection terminal, a remote cloud server, and a mobile terminal to achieve partial discharge detection. It combines traditional time-domain waveform, PRPD / PRPS spectrum, infrared thermal image spectrum determination, and multimodal sensor fusion technology to achieve detection of device status, partial discharge type, probability, and abnormal location, thus ensuring the safety and stability of device operation.

[0098] The PD detection system and its working method based on multimodal sensor fusion provided by this invention can be applied to the detection of substations and overhead lines. When applied to substations, it enables partial discharge inspection and online monitoring of related equipment, including switchgear and transformers. When applied to overhead lines, it enables temperature detection and partial discharge detection of related devices, including conductor joints and insulators.

[0099] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0100] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0101] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0102] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A PD detection system based on multimodal sensor fusion, characterized in that, include: The PD detection terminal is used to collect equipment environmental information and partial discharge information, display the corresponding time-domain waveform, PRPD / PRPS spectrum and infrared thermal image spectrum, and preliminarily determine the equipment status, partial discharge type, probability and abnormal location; the PD detection terminal is connected to a remote cloud server through a 4G / 5G network to send the collected and processed information to the remote cloud server. The remote cloud server is connected to the PD detection terminal and the mobile terminal via 4G / 5G networks respectively. It is used to receive information sent by the PD detection terminal, and to make secondary judgments on the equipment status, partial discharge type, probability and abnormal location using a multimodal sensor fusion method. The results are confirmed, the partial discharge risk index is calculated according to the dynamic partial discharge risk index formula, the risk level is confirmed, and disposal suggestions are generated and sent to the mobile terminal of the maintenance personnel. The mobile terminal is used to interact with the remote cloud server, receive information sent by the remote cloud server, perform data parsing and display, and send detection and control commands to the remote cloud server to complete maintenance tasks.

2. The PD detection system based on multimodal sensor fusion according to claim 1, characterized in that, The PD detection terminal includes a zero-cross detector, an ambient temperature probe, an ambient humidity probe, a UHF probe, an HFCT probe, a TEV probe, an AA / AE probe, a conditioning circuit, an AD circuit, an FPGA controller, an infrared module, and a tablet computer. The zero-crossing detector is electrically connected to the FPGA controller and is used to receive a 220V AC power signal, which is then internally stepped down and zero-crossing compared before being sent to the FPGA controller in a pulse manner. The ambient temperature probe, ambient humidity probe, UHF probe, HFCT probe, TEV probe, and AA / AE probe are electrically connected to the conditioning circuit to receive ambient temperature signals, ambient humidity signals, UHF signals, HFCT signals, TEV signals, and AA / AE signals, and amplify and filter the signals before sending them to the AD circuit. The AD circuit is connected to the FPGA controller in parallel and is used to convert the analog signals from all probes after the conditioning circuit into digital signals and send them to the FPGA controller. The FPGA controller is connected to the USB port of the tablet computer to receive digital signals from all probes and inserts a zero-phase stamp into the digital signals based on the pulse signals emitted by the zero-crossing detector before sending them to the tablet computer. The infrared module is connected to the RJ45 port of the tablet computer and is used to receive infrared temperature matrix signals and convert them into digital signals to send to the tablet computer. The tablet computer is used to receive ambient temperature, ambient humidity, UHF, HFCT, TEV, AA / AE signals and infrared temperature matrix signals. It displays the UHF, HFCT, TEV and AA / AE signals as time-domain waveforms and PRPD / PRPS maps, and the infrared temperature matrix signals as infrared thermal images. By incorporating ambient temperature and humidity factors, it determines the type and degree of partial discharge based on the characteristics of the time-domain waveforms and PRPD / PRPS maps, and determines the specific location of the partial discharge based on the infrared temperature matrix map. At the same time, it sends all data to a remote cloud server via a 4G / 5G network.

3. The working method of the PD detection system based on multimodal sensor fusion according to claim 2, characterized in that, Includes the following steps: S1. The device status data is synchronously collected based on the pulse signal emitted by the zero-crossing detector through the ambient temperature probe, ambient humidity probe, UHF probe, HFCT probe, TEV probe, and AA / AE probe, and the infrared temperature matrix signal is collected through the infrared module. S2. Display the corresponding time-domain waveform, PRPD / PRPS spectrum and infrared thermographic spectrum on the PD detection terminal. Based on the key features of the time-domain waveform, PRPD / PRPS spectrum and infrared thermographic spectrum and the discharge type discrimination rules, preliminarily determine the equipment status, partial discharge type, probability and abnormal location. After adding a timestamp, the S3 PD detection terminal uploads ambient temperature, ambient humidity, UHF, HFCT, TEV, AA / AE signals, and infrared temperature matrix signals, as well as the preliminary determination of the type, probability, and location of partial discharge, to the remote cloud server. S4 and the remote cloud server respectively use corresponding data processing algorithms to extract target features from UHF, HFCT, TEV, AA / AE signals and infrared temperature matrix signals; S5. Construct an environmental interference correction factor matrix based on ambient temperature and humidity, and adjust the confidence weight of each probe data; adopt a multimodal cross-attention mechanism, using UHF signal as the reference query vector, infrared temperature matrix signal, HFCT current pulse signal, TEV voltage signal, and AA / AE acoustic signal as key value vectors, and calculate multidimensional feature association weights. S6. Multidimensional features are weighted and fused according to attention weights. The fused temporal features are processed by LSTM or Transformer encoder. Spatial features are extracted by CNN and the device status, partial discharge type, probability and abnormal location are determined by a fully connected layer. S7. When there are abnormal environmental conditions, abnormal data, or data conflicts from multiple probes, a dynamic strategy and conflict arbitration mechanism shall be adopted for determination. S8. The remote cloud server combines the initial and secondary determinations of the device status, partial discharge type, probability, and abnormal location to confirm the device status, partial discharge type, probability, and abnormal location. S9. The remote cloud server calculates the partial discharge risk index according to the dynamic partial discharge risk index formula, classifies the risk level according to the risk level classification standard, and generates disposal suggestions. S10. The remote cloud server will send the final confirmed equipment status, partial discharge type, probability, abnormal location, risk index, risk level, and handling suggestions to the maintenance personnel's mobile terminal to complete the maintenance task.

4. The working method of the PD detection system based on multimodal sensor fusion according to claim 3, characterized in that, In step S2, the key features include: UHF signal pulse clustering, spectral asymmetry, HFCT rise time, propagation delay, TEV signal voltage amplitude, AA / AE signal source azimuth deviation, energy accumulation, and steep temperature gradient of infrared temperature matrix signal. The discharge type discrimination rules include: a single-peak distribution in the UHF signal spectrum indicates corona discharge; multiple continuous bursts indicate surface discharge; a short rise time and stable delay in the HFCT signal indicate internal insulation discharge; a TEV signal amplitude > 50mV indicates metal tip discharge or surface discharge; a sudden increase in acoustic emission energy and azimuth fluctuation in the AA / AE signal indicates mechanical loosening; and a hotspot temperature rise rate > 2℃ / s in the infrared temperature matrix signal indicates overheating discharge.

5. The working method of the PD detection system based on multimodal sensor fusion according to claim 3, characterized in that, In step S4, for UHF signals, the energy proportion and instantaneous frequency slope of the 300-2000MHz frequency band are extracted by short-time Fourier transform and wavelet packet decomposition. For HFCT signals, the pulse rise time and propagation delay differences are calculated using db4 basis function wavelet front detection. For TEV signals, an adaptive Kalman filter and peak statistics are used to generate a pulse amplitude distribution map after background noise suppression. For AA / AE signals, a probability contour map of the sound source location is constructed by estimating the direction of arrival and integrating the energy accumulation. For the infrared temperature matrix signal, the coordinates of the hot spot center are located and the temperature rise rate is calculated through the thermal diffusion model and gradient descent optimization.

6. The working method of the PD detection system based on multimodal sensor fusion according to claim 3, characterized in that, In step S5, the expression for the environmental disturbance correction factor matrix is: Among them, W env W represents the adjusted weights. UHF W represents the UHF weights. HFCT W represents the HFCT weights. TEV W represents the TEV weight. AA / AE W represents the AA / AE weights. IR Indicates infrared weighting; The multimodal cross-attention mechanism constructs a temperature-discharge correlated attention model, aligns the hotspot coordinates of the infrared temperature matrix signal with the UHF signal pulse timing, calculates the spatial-temporal collaborative weights, and uses an acoustic-optical joint attention branch to capture the causal relationship between acoustic emission events of the AE / AE probe and infrared temperature abrupt changes in parallel through a multi-head attention mechanism.

7. The working method of the PD detection system based on multimodal sensor fusion according to claim 3, characterized in that, In step S6, the weighted fusion specifically includes: performing tensor fusion on the infrared temperature matrix signal and the UHF signal to generate a discharge-temperature correlation spectrum; when a local temperature >150℃ is detected, activating the overheat protection protocol; and preferentially fusing the acoustic emission positioning data of the AA / AE probe with the hotspot coordinates of the infrared temperature matrix signal.

8. The working method of the PD detection system based on multimodal sensor fusion according to claim 3, characterized in that, In step S7, the dynamic strategy and conflict arbitration mechanism include: when humidity > 85%, reducing the TEV weight and increasing the AA / AE weight and infrared weight; when infrared temperature rise > 10℃ / s, enabling UHF+HFCT to use the TDOA algorithm for joint positioning; when acoustic-electric signal orientation deviation > 20°, resetting the top 3 arbitrations based on attention weight, prioritizing the adoption of UHF, infrared, and HFCT signals; when multiple probe data loss ≥ 2 channels, starting the degradation mode, relying only on UHF+infrared signals, and relaxing the confidence threshold to 0.

75.

9. The working method of the PD detection system based on multimodal sensor fusion according to claim 3, characterized in that, In step S9, the formula for the dynamic partial discharge risk index is: Among them, P i For the i-th type of partial discharge, W i T represents the weight corresponding to the i-th type of partial discharge. 实际 T represents the actual ambient temperature of the equipment. 额定 The rated ambient temperature of the equipment, H 实际 R and T represent the actual ambient humidity of the equipment. effect H effect For intermediate calculation variables; The risk level classification standard is: partial discharge risk index R. 动态 The interval [1, +∞) is defined as high risk, and the partial discharge risk index R is... 动态 The interval [0.6, 1) is defined as medium risk, and the rest are defined as low risk; The proposed measures include: routine inspections for low-risk cases, daily monitoring combined with UHF / TEV positioning of the power supply for medium-risk cases, and immediate shutdown and maintenance for high-risk cases, using HFCT + infrared retesting.

10. The application of the PD detection system and its working method based on multimodal sensor fusion according to any one of claims 1-9, characterized in that, The PD detection system and its working method are applied to the detection of substations and overhead lines. When applied to substations, it enables partial discharge inspection and online monitoring of related equipment, including switchgear and transformers. When applied to overhead lines, it enables temperature detection and partial discharge detection of related devices, including conductor joints and insulators, on the overhead lines.

Images