Multi-mode power equipment safety detection device, system and method

Abstract

The invention provides a multi-mode power equipment safety detection device, system and method, and relates to the field of power detection. The multi-mode power equipment safety detection system comprises a power detection system control center, a system total architecture of the power detection system control center comprises a sensing layer module, a data processing layer module and a decision-making layer module, the sensing layer is the topmost layer of the system total architecture and has the function of a data acquisition module, and the data processing layer module has the function of a decision-making module. The device is composed of a plurality of sensors and is responsible for collecting multi-modal data of power equipment. According to the invention, intelligent operation and maintenance of the power system can be assisted, the reliability of the power system is significantly improved, the accuracy, timeliness and economy of detection are significantly improved through multi-dimensional information fusion, real-time intelligent analysis and global cooperative monitoring, and a core support is provided for safe and stable operation of the novel power system.

Description

Technical Field

[0001] The present invention relates to the field of power equipment detection, and particularly to a multi-modal power equipment safety detection device, system and method. Background Art

[0002] In today's society, electricity, as a crucial energy source, plays a fundamental role in the normal operation of the social economy. Power equipment, as a key component of the power system, covers all aspects such as power generation, transmission, transformation, distribution and consumption. Its safe and stable operation is directly related to the reliability and stability of power supply. Once a power equipment fails, the impacts are multi-faceted and extremely serious. After long-term development, traditional power equipment detection technologies have formed various mature methods, such as detection technologies based on electrical parameter measurement, which determine whether the equipment is operating normally by monitoring parameters such as current, voltage and power. These traditional detection technologies have many limitations. The detection information of a single modality is limited and it is difficult to comprehensively reflect the true operating state of the equipment. Moreover, their adaptability in complex environments is poor. Therefore, the present invention proposes a multi-modal power equipment safety detection system and detection method, which can greatly improve the efficiency and intelligent level of power equipment detection and effectively ensure the safe and stable operation of the power system. Summary of the Invention

[0003] (1) Technical Problems to be Solved

[0004] Aiming at the deficiencies of the prior art, the present invention provides a multi-modal power equipment safety detection device, system and method. Through the fusion analysis of multi-modal data such as vision, infrared and acoustics, problems such as overheating of transformer windings, partial discharge and core faults can be detected in a timely manner. The visual image recognition technology is used to detect whether there are cracks, oil leakage, etc. on the transformer shell. The infrared thermal imager monitors the oil temperature, winding temperature, etc. of the transformer. The acoustic sensor captures abnormal sounds during the operation of the transformer. The operating state of the transformer is comprehensively judged, potential faults are detected in advance, and the stable operation of the substation is guaranteed. Thus, the safety detection of power equipment can be realized in all directions.

[0005] (2) Technical Solutions

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A multi-modal power equipment safety detection system includes a control center for the power detection system. The overall system architecture of the control center for the power detection system includes three architecture levels: a perception layer module, a data processing layer module, and a decision-making layer module. The perception layer is the topmost layer of the overall system architecture, which includes a data acquisition module function and is composed of multiple sensors. It is responsible for collecting multi-modal data of power equipment and providing original information for subsequent data processing and analysis. It is composed of acquisition devices responsible for the safety detection of power equipment. The data processing layer is the information processing layer of the overall system architecture and is the core part of the multi-modal power equipment safety detection system. It performs operations such as preprocessing, feature extraction, and data fusion on the data collected by the perception layer, providing accurate and reliable data support for the decision-making layer. The data processing layer includes a data preprocessing function, a data feature extraction function, and a data fusion function. The decision-making layer is the information output layer of the overall power monitoring system architecture. Based on the fused data provided by the data processing layer, it makes judgments on the equipment status and fault diagnosis decisions and feeds back the results to the user, providing guidance for the operation and maintenance of power equipment. The decision-making layer includes a function of establishing a diagnosis model, a function of judging the equipment status, and a function of making fault diagnosis decisions;

[0007] Through the above technical solutions, the multi-modal power equipment safety detection system realizes comprehensive and accurate monitoring of the operating status of power equipment by integrating multiple detection technologies, timely discovers potential safety hazards, and the various levels cooperate with each other to jointly complete the power equipment safety detection task.

[0008] Preferably, the control center for the power detection system is responsible for data acquisition of the perception layer by a multi-modal power equipment safety detection device. The acquisition devices include a visual acquisition module, an acoustic wave acquisition module, a lidar module, and a composite perception acquisition module. The devices of the visual acquisition module include multi-spectral imager acquisition, high-definition visual camera acquisition, and thermal imager acquisition. The devices of the acoustic wave acquisition module include ultrasonic detection devices and microphone arrays. The devices of the lidar module include 3D lidar and array millimeter waves. The devices of the composite perception acquisition module include a flexible sensor network, a quantum sensing device, and a multi-modal fusion sensing device;

[0009] Through the above technical solutions, the flexible sensor network in the composite perception acquisition device can be attached to the flexible piezoelectric film on the surface of the power equipment to real-time monitor partial discharge and temperature changes. The quantum sensing device can detect weak magnetic field changes using the quantum coherence effect to discover early faults such as winding deformation in advance.

[0010] Preferably, the data preprocessing function of the data processing module purifies and repairs the data collected by the data acquisition device. For the image data collected by the visual acquisition module, Gaussian filtering is used to perform weighted averaging on the image pixels, which can effectively remove Gaussian noise and make the image smoother. For the acoustic data collected by the acoustic acquisition module, the wavelet denoising method is used. The sound signal is decomposed into sub-signals of different frequencies through wavelet transform to process the frequency sub-band where the noise is located. After removing the noise, the wavelet inverse transform is performed to restore the pure sound signal. After denoising the image data and acoustic data, the data preprocessing synchronizes the multi-modal data through timestamps so that the data transfer error is less than 1 millisecond;

[0011] Through the above technical solution, for image data, problems such as blurring and noise may exist. Gaussian filtering can be used to remove image noise caused by light changes, sensor noise, etc., making the details of the device clearer and facilitating subsequent image recognition and analysis. For acoustic data, wavelet denoising can effectively remove the interference of background noise, improve the signal-to-noise ratio of the sound signal, make the partial discharge sound characteristics more obvious, and is conducive to accurately judging whether there is a partial discharge fault in the device.

[0012] Preferably, the data feature extraction function of the data processing module extracts key features that can reflect the operating state of the device from the preprocessed raw data. For visual feature image data, the Speeded-Up Robust Features (SURF) method is used to extract the target contour. For acoustic data with acoustic features, Mel Frequency Cepstral Coefficients (MFCC) are used to extract information such as the frequency, amplitude, and phase of the sound signal for sound recognition of device failures. For thermal imaging infrared feature data, the mean, variance, and change rate features of the temperature are extracted to judge whether there is an overheating fault in the device. For the feature data collected by composite perception, a cross-attention mechanism is used to integrate multi-modal features.

[0013] Preferably, the data fusion function of the data processing module organically integrates data of different modalities. The data fusion function directly performs fusion processing on the original sensor data, splices the image features and sound features to form a comprehensive feature vector, and fuses the decision results based on independent decisions of each modality data. Through Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), the multi-modal data is fused, and the decision results of image, acoustic wave, and thermal infrared temperature data are fused to obtain the overall operating state of the device;

[0014] Through the above technical solutions, when applying convolutional neural networks and recurrent neural networks to multi-modal data fusion, they can be combined with data processing models of other modalities. The thermal image data collected by infrared sensors can also be input into convolutional neural networks and recurrent neural networks. By sharing some convolutional layers, feature fusion of visual images and infrared images can be achieved, so as to more comprehensively judge the operating state of the device.

[0015] Preferably, the function of establishing a diagnostic model in the decision-making layer uses the artificial intelligence algorithm support vector machine (SVM) to establish a fault diagnosis model. The support vector machine effectively distinguishes data samples of different categories by finding an optimal classification hyperplane. Through a large number of sample trainings, the model can learn the data feature patterns in normal states and different fault states;

[0016] Through the above technical solutions, the electrical parameters, oil temperature, oil chromatographic analysis data of the transformer, and the data collected by multi-modal sensors can be used as inputs and input into the SVM model for training and classification.

[0017] Preferably, when the decision-making layer inputs new device data into the diagnostic model for the device state judgment function, it can accurately judge whether the device is in a fault state and the type of the fault. Combining the operation history and maintenance records of the device, using a fault assessment algorithm, the severity of the fault is quantitatively evaluated. The fault diagnosis decision of the decision-making layer determines a reasonable warning threshold according to the operation parameters and historical fault data of the power equipment, and provides reasonable operation and maintenance suggestions for users according to the fault diagnosis results;

[0018] Through the above technical solutions, when a fault is detected in the device, the decision-making layer needs to further determine the severity and scope of the fault. According to the fault type and relevant parameters of the device, combined with the operation history and maintenance records of the device, at the same time, the decision-making layer also needs to consider the impact of the fault on other devices in the power system, such as whether it will cause voltage fluctuations, power imbalances, etc., in order to take corresponding measures to ensure the safe and stable operation of the power system. The decision-making layer can also be linked with the dispatching system of the power system. According to the fault situation of the device, the operation mode of the power system is adjusted, and power dispatching is optimized to ensure the continuity and reliability of power supply. When a fault is detected in a certain transmission line, the decision-making layer can promptly notify the dispatching system to transfer the load to other lines, and at the same time arrange maintenance personnel to repair the fault line as soon as possible to reduce the power outage time and losses.

[0019] Working principle: The steps of the power equipment safety detection method of this multi-modal power equipment safety detection system are as follows:

[0020] S1: Device deployment and debugging, select appropriate data acquisition devices for installation according to the scenario;

[0021] S2: Collect multi-dimensional data of power equipment, where the collected data includes visual data, acoustic data, thermal infrared temperature data, and composite data;

[0022] S3: Perform data preprocessing and feature extraction on the multi-dimensional data collected by the equipment;

[0023] S4: Perform multi-modal algorithm fusion and data fusion on the data with extracted features, including data layer fusion, feature layer fusion, and decision layer fusion;

[0024] S5: State analysis and fault warning;

[0025] S6: Make a diagnostic decision plan for the safety state of power equipment.

[0026] (III) Beneficial effects

[0027] The present invention provides a multi-modal safety detection device, system, and method for power equipment. It has the following beneficial effects:

[0028] The present invention provides a multi-modal safety detection system for power equipment. During the detection of power equipment, it can comprehensively monitor equipment such as transformers, high-voltage switchgears, and busbars in the substation. Through the fusion analysis of multi-modal data such as vision, infrared, and acoustics, problems such as overheating of transformer windings, partial discharge, and core faults can be detected in a timely manner. The visual image recognition technology is used to detect whether there are cracks, oil leakage, etc. on the transformer shell. The infrared thermal imager monitors the oil temperature, winding temperature, etc. of the transformer. The acoustic sensor captures abnormal sounds during the operation of the transformer, comprehensively judges the operation state of the transformer, discovers potential faults in advance, and ensures the stable operation of the substation. Thus, it can realize all-round safety detection of power equipment, help the power system achieve intelligent operation and maintenance, significantly improve the reliability of the power system, and significantly enhance the accuracy, timeliness, and economy of detection through multi-dimensional information fusion, real-time intelligent analysis, and global collaborative monitoring, providing core support for the safe and stable operation of the new power system. Description of the drawings

[0029] Figure 1 It is a schematic diagram of the system structure of the multi-modal safety detection system for power equipment of the present invention;

[0030] Figure 2 It is a structural diagram of the device modules of the multi-modal safety detection device for power equipment of the present invention;

[0031] Figure 3 It is a flowchart of the safety detection method of the multi-modal safety detection system for power equipment of the present invention. Detailed implementation manners

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. In the description of this application, it should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments of the present application. For the convenience of description, the dimensions of the various parts shown in the drawings are not drawn in actual proportional relationships. Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and devices should be regarded as part of the authorization specification. In all the examples shown and discussed here, any specific value should be construed as merely exemplary and not as a limitation. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0033] Embodiment 1:

[0034] As Figures 1 - 3 shown, the embodiment of the present invention provides a multi-modal power equipment safety detection device, system and method, including a power detection system control center, characterized in that: the overall system architecture of the power detection system control center includes three architecture levels: a perception layer module, a data processing layer module, and a decision-making layer module. The perception layer is the top layer of the overall system architecture, including a data acquisition module function, which is composed of a variety of sensors and is responsible for collecting multi-modal data of power equipment, providing raw information for subsequent data processing and analysis, and is composed of acquisition devices responsible for power equipment safety detection. The data processing layer is the information processing layer of the overall system architecture and is the core part of the multi-modal power equipment safety detection system. It performs operations such as preprocessing, feature extraction, and data fusion on the data collected by the perception layer, providing accurate and reliable data support for the decision-making layer. The data processing layer includes a data preprocessing function, a data feature extraction function, and a data fusion function. The decision-making layer is the information output layer of the overall power monitoring system architecture. Based on the fusion data provided by the data processing layer, it makes device status judgments and fault diagnosis decisions and feeds back the results to the user, providing guidance for the operation and maintenance of power equipment. The decision-making layer includes a diagnostic model establishment function, a device status judgment function, and a fault diagnosis decision function. The multi-modal power equipment safety detection system realizes comprehensive and accurate monitoring of the operating status of power equipment by integrating a variety of detection technologies, timely discovers potential safety hazards, and the various levels cooperate with each other to jointly complete the power equipment safety detection task;

[0035] The control center of the power detection system is responsible for data acquisition of the perception layer by the multi-modal power equipment safety detection device. The acquisition devices include a visual acquisition module, an acoustic wave acquisition module, a lidar module, and a composite perception acquisition module. The devices of the visual acquisition module include multi-spectral imager acquisition, high-definition visual camera acquisition, and thermal imager acquisition. The devices of the acoustic wave acquisition module include ultrasonic detection devices and microphone arrays. The devices of the lidar module include 3D lidar and array millimeter waves. The devices of the composite perception acquisition module include a flexible sensor network, a quantum sensing device, and a multi-modal fusion sensing device. The flexible sensor network in the composite perception acquisition device can be attached to the flexible piezoelectric film on the surface of the power equipment to monitor partial discharge and temperature changes in real time. The quantum sensing device can detect weak magnetic field changes using the quantum coherence effect to detect early faults such as winding deformation in advance.

[0036] The data preprocessing function of the data processing module purifies and repairs the data collected by the data acquisition device. For the image data collected by the visual acquisition module, Gaussian filtering is used to perform weighted averaging on the image pixels, which can effectively remove Gaussian noise and make the image smoother. For the acoustic data collected by the acoustic acquisition module, the wavelet denoising method is adopted. The sound signal is decomposed into sub-signals of different frequencies through wavelet transform to process the frequency sub-band where the noise is located. After removing the noise, the wavelet inverse transform is performed to restore the pure sound signal. After denoising the image data and acoustic data, the data preprocessing synchronizes the multi-modal data through timestamps so that the data transfer error is less than 1 millisecond. For image data, there may be problems such as blurring and noise. Using Gaussian filtering can remove image noise caused by light changes, sensor noise, etc., making the details of the device clearer and facilitating subsequent image recognition and analysis. For acoustic data, wavelet denoising can effectively remove the interference of background noise, improve the signal-to-noise ratio of the sound signal, and make the partial discharge sound characteristics more obvious, which is conducive to accurately judging whether there is a partial discharge fault in the device. The data feature extraction function of the data processing module extracts key features that can reflect the operating state of the device from the preprocessed original data. For visual feature image data, the Speeded Up Robust Features (SURF) method is used to extract the target contour. For acoustic feature acoustic data, Mel Frequency Cepstral Coefficients (MFCC) are used to extract information such as the frequency, amplitude, and phase of the sound signal for sound recognition of device faults. For thermal imaging infrared feature data, the mean, variance, and change rate features of the temperature are extracted to judge whether there is an overheating fault in the device. For the feature data collected by composite perception, the cross-attention mechanism is used to integrate multi-modal features. The data fusion function of the data processing module organically integrates data of different modalities. The data fusion function directly performs fusion processing on the original sensor data, splices the image features and sound features to form a comprehensive feature vector, and fuses the decision results based on independent decisions of each modality data. Through the Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), multi-modal data is fused, and the decision results of image, acoustic, and thermal infrared temperature data are fused to obtain the overall operating state of the device. When applying the convolutional neural network and recurrent neural network to multi-modal data fusion, they can be combined with data processing models of other modalities. The thermal image data collected by the infrared sensor can also be input into the convolutional neural network and recurrent neural network, and through sharing some convolutional layers, feature fusion of visual images and infrared images can be achieved, so as to more comprehensively judge the operating state of the device;

[0037] The function of establishing a diagnostic model in the decision-making layer uses the artificial intelligence algorithm support vector machine (SVM) to establish a fault diagnosis model. The support vector machine effectively differentiates data samples of different categories by finding an optimal classification hyperplane. Through a large number of sample trainings, the model can learn the data feature patterns in normal states and different fault states. The electrical parameters of the transformer, oil temperature, oil chromatographic analysis data, and data collected by multi-modal sensors can be used as inputs and input into the SVM model for training and classification. When new device data is input into the diagnostic model in the decision-making layer, the device status judgment function can accurately determine whether the device is in a fault state and the type of the fault. Combining the operation history and maintenance records of the device, and using a fault assessment algorithm, the severity of the fault is quantitatively evaluated. The fault diagnosis decision in the decision-making layer determines a reasonable early warning threshold based on the operation parameters and historical fault data of the power device, and provides reasonable operation and maintenance suggestions for users according to the fault diagnosis results. When a fault is detected in the device, the decision-making layer needs to further determine the severity and impact range of the fault. According to the fault type and relevant parameters of the device, combined with the operation history and maintenance records of the device, at the same time, the decision-making layer also needs to consider the impact of the fault on other devices in the power system, such as whether it will cause voltage fluctuations, power imbalances, etc., in order to take corresponding measures to ensure the safe and stable operation of the power system. The decision-making layer can also be linked with the dispatching system of the power system. According to the fault situation of the device, the operation mode of the power system is adjusted, and the power dispatching is optimized to ensure the continuity and reliability of power supply. When a fault is detected in a certain transmission line, the decision-making layer can timely notify the dispatching system to transfer the load to other lines, and at the same time arrange maintenance personnel to repair the faulty line as soon as possible to reduce the power outage time and losses.

[0038] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multimodal power equipment safety detection system, including a control center of the power detection system, characterized in that: The overall system architecture of the power detection system control center includes three architecture levels: the perception layer module, the data processing layer module, and the decision-making layer module. The perception layer is the topmost layer of the overall system architecture, which includes the function of the data acquisition module. It consists of a variety of sensors and is responsible for collecting multi-modal data of power equipment, providing the original information for subsequent data processing and analysis. It is composed of acquisition devices responsible for the safety detection of power equipment. The data processing layer is the information processing layer of the overall system architecture and is the core part of the multi-modal power equipment safety detection system. It performs operations such as preprocessing, feature extraction, and data fusion on the data collected by the perception layer, providing accurate and reliable data support for the decision-making layer. The data processing layer includes the functions of data preprocessing, data feature extraction, and data fusion. The decision-making layer is the information output layer of the overall power detection system architecture. Based on the fused data provided by the data processing layer, it makes decisions on equipment status judgment and fault diagnosis, and feeds back the results to the user, providing guidance for the operation and maintenance of power equipment. The decision-making layer includes the functions of establishing a diagnostic model, equipment status judgment, and fault diagnosis decision-making.

2. The multimodal power equipment safety detection system according to claim 1, wherein: The multi-modal power equipment safety detection device is responsible for the data acquisition of the perception layer in the power detection system control center. The acquisition devices include a visual acquisition module, an acoustic wave acquisition module, a lidar module, and a composite perception acquisition module. The devices of the visual acquisition module include multi-spectral imager acquisition, high-definition visual camera acquisition, and thermal imager acquisition. The devices of the acoustic wave acquisition module include ultrasonic detection devices and microphone arrays. The devices of the lidar module include 3D lidar and array millimeter waves. The devices of the composite perception acquisition module include flexible sensor networks, quantum sensing devices, and multi-modal fusion sensing devices.

3. A multimodal power equipment safety detection system according to claim 1, characterized in that: The data preprocessing function of the data processing module purifies and repairs the data collected by the data acquisition device. For the image data collected by the visual acquisition module, Gaussian filtering is used to perform weighted averaging on the image pixels, which can effectively remove Gaussian noise and make the image smoother. For the acoustic wave data collected by the acoustic wave acquisition module, the wavelet denoising method is used. The sound signal is decomposed into sub-signals of different frequencies through wavelet transform to process the frequency sub-band where the noise is located. After removing the noise, the wavelet inverse transform is performed to restore the pure sound signal. After denoising the image data and acoustic wave data, the data preprocessing synchronizes the multi-modal data through timestamps so that the data transfer error is less than 1 millisecond.

4. A multimodal power equipment safety detection system according to claim 1, characterized in that: The data feature extraction function of the data processing module extracts key features that can reflect the operating state of the device from the preprocessed original data. For visual feature image data, the Speeded Up Robust Features (SURF) method is used to extract the target contour. For acoustic feature sound wave data, the Mel Frequency Cepstral Coefficients (MFCC) are used to extract information such as the frequency, amplitude, and phase of the sound signal for sound recognition of device faults. For thermal imaging infrared feature data, the mean, variance, and change rate features of the temperature are extracted to determine whether the device has overheating faults. For the feature data collected by multi-modal perception, the cross-attention mechanism is used to integrate multi-modal features.

5. A multimodal power equipment safety detection system according to claim 1, characterized in that: The data fusion function of the data processing module organically integrates data of different modalities. The data fusion function directly performs fusion processing on the original sensor data, splices the image features and sound features to form a comprehensive feature vector, fuses the decision results based on independent decisions made for each modality data, and fuses the decision results of image, sound wave, and thermal infrared temperature data through Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) to obtain the overall operating state of the device.

6. A multimodal power equipment safety detection system according to claim 1, characterized in that: The function of establishing a diagnostic model in the decision-making layer uses the artificial intelligence algorithm Support Vector Machine (SVM) to establish a fault diagnosis model. The support vector machine effectively distinguishes data samples of different categories by finding an optimal classification hyperplane. Through a large number of sample trainings, the model can learn the data feature patterns in the normal state and different fault states.

7. A multimodal power equipment safety detection system according to claim 1, characterized in that: When new device data is input into the diagnostic model in the device state judgment function of the decision-making layer, it can accurately judge whether the device is in a fault state and the type of the fault. Combining the operation history and maintenance records of the device, using the fault assessment algorithm, it quantitatively evaluates the severity of the fault. The fault diagnosis decision in the decision-making layer determines a reasonable warning threshold based on the operating parameters and historical fault data of the power device, and provides reasonable operation and maintenance suggestions for users according to the fault diagnosis results.

8. A multimodal power equipment safety detection system according to claim 1, characterized in that: The steps of the power device safety detection method of this multi-modal power device safety detection system are as follows: S1: Device deployment and debugging, select and install appropriate data acquisition devices according to the scenario. S2: Conduct multi-dimensional data acquisition on the power device, and the collected data includes visual data, sound wave data, thermal infrared temperature data, and composite data. S3: Perform data preprocessing and feature extraction on the multi-dimensional data collected by the device. S4: Perform multi-modal algorithm fusion and data fusion on the data with extracted features, including data layer fusion, feature layer fusion, and decision layer fusion. S5: State analysis and fault warning. S6: Make a diagnostic decision plan for the safety state of the power device.

Images