Multi-mode fusion switch cabinet partial discharge on-line monitoring system and method

Abstract

The invention provides a multi-mode fusion switch cabinet partial discharge on-line monitoring system and method, and the system comprises a space setting module which employs a preset space analysis algorithm, obtains an installation point position set according to the space structure of a switch cabinet, and carries out the installation of the installation point position set; the data acquisition module is used for acquiring data acquired by the ultrahigh frequency sensor, the ultrasonic sensor and the temperature sensor, preprocessing the data and then sending the data to the signal processing module; carrying out power frequency noise suppression on the ultrahigh frequency signal and the ultrasonic signal through a self-adaptive filter circuit, and extracting partial discharge characteristics from the filtered signal by adopting a wavelet transform algorithm; and the AI model fusing the convolutional neural network and the long-short-term memory network performs fusion analysis on the partial discharge characteristics and the temperature sensor data, realizes partial discharge mode identification and insulation degradation trend prediction, and outputs early warning information. According to the technical scheme provided by the invention, high-precision and anti-interference discharge detection can be realized.

Description

Technical Field

[0001] This document relates to the field of partial discharge monitoring technology for switchgear, and in particular to a multimodal fusion online monitoring system and method for partial discharge in switchgear. Background Technology

[0002] High-voltage switchgear in thermal power plants is a core piece of equipment in the power system. During long-term operation, factors such as temperature, humidity, and mechanical vibration can cause the insulation materials to deteriorate, leading to partial discharge (PD). Partial discharge is a precursor to equipment failure and requires online monitoring for early warning.

[0003] Most existing technologies are partial discharge detection systems based on ultrasonic sensors. Their technical features include: installing ultrasonic sensors on the inner wall of the switch cabinet to collect the sound wave signals generated by the discharge, using a threshold comparison method to determine the discharge intensity, triggering an alarm when the threshold is exceeded, and transmitting the data to the background monitoring system via wired connection.

[0004] The above technology has the following drawbacks: 1. Low sensitivity: Ultrasonic sensors are susceptible to mechanical vibration interference and cannot distinguish between discharge signals and background noise; 2. Incomplete coverage: single-point sensors cannot locate the discharge location, making it difficult to assess the severity of the discharge; 3. Data processing is lagging, relying on threshold alarms, lacking intelligent analysis, and unable to predict insulation degradation trends.

[0005] Therefore, there is an urgent need for a discharge monitoring method that can achieve high precision and is resistant to interference. Summary of the Invention

[0006] According to embodiments of the present invention, a multimodal fusion-based online monitoring system and method for partial discharge in switchgear is provided, aiming to solve the above-mentioned problems.

[0007] According to an embodiment of the present invention, a multi-modal fusion online monitoring system for partial discharge in switchgear is provided, comprising: The space setting module uses a pre-set space analysis algorithm to obtain a set of installation points based on the spatial structure of the switch cabinet, and installs ultra-high frequency sensors, ultrasonic sensors, and temperature sensors according to the set of installation points. The multi-source data acquisition module is used to acquire data from the ultra-high frequency sensor, ultrasonic sensor, and temperature sensor, and to send the preprocessed data to the signal processing module. The signal processing module is used to receive the preprocessed data sent by the multi-source data acquisition module, suppress power frequency noise of the ultra-high frequency signal and ultrasonic signal through an adaptive filtering circuit, and extract partial discharge features from the filtered signal using a wavelet transform algorithm. The intelligent analysis module receives the partial discharge characteristics and temperature sensor data, and performs fusion analysis on the partial discharge characteristics and temperature sensor data through an AI model that integrates convolutional neural networks and long short-term memory networks. This enables pattern recognition of partial discharge and prediction of insulation degradation trends, and outputs early warning information.

[0008] According to an embodiment of the present invention, a multi-modal fusion method for online monitoring of partial discharge in switchgear is provided, comprising: S1. Using a pre-set spatial analysis algorithm, obtain a set of installation points based on the spatial structure of the switch cabinet, and install ultra-high frequency sensors, ultrasonic sensors, and temperature sensors according to the set of installation points. S2. Collect data from the ultra-high frequency sensor, ultrasonic sensor, and temperature sensor, and send the preprocessed data to the signal processing module. S3. The signal processing module receives the preprocessed data sent by the multi-source data acquisition module, performs power frequency noise suppression on the ultra-high frequency signal and ultrasonic signal through an adaptive filtering circuit, and extracts partial discharge features from the filtered signal using a wavelet transform algorithm. S4. Receive the partial discharge characteristics and temperature sensor data, and perform fusion analysis on the partial discharge characteristics and temperature sensor data through an AI model that integrates convolutional neural networks and long short-term memory networks to achieve pattern recognition of partial discharge and prediction of insulation degradation trends, and output early warning information.

[0009] The multimodal fusion switchgear partial discharge online monitoring system and method provided in this application achieves high-precision and anti-interference detection of partial discharge through spatial optimization of point layout, multi-source signal acquisition, adaptive noise suppression and intelligent fusion analysis. It effectively improves the ability to identify discharge modes and predict insulation degradation trends, and provides reliable technical support for switchgear condition assessment and fault early warning. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of a multimodal fusion-based online partial discharge monitoring system for switchgear according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the space setting module according to an embodiment of the present invention; Figure 3This is a flowchart of a multimodal fusion-based online monitoring method for partial discharge in switchgear, according to an embodiment of the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0013] System Implementation Examples According to embodiments of the present invention, a multi-modal fusion online monitoring system for partial discharge in switchgear is provided. Figure 1 This is a schematic diagram of a multimodal fusion-based online partial discharge monitoring system for switchgear according to an embodiment of the present invention. Figure 1 As shown, the multimodal fusion-based online monitoring system for partial discharge in switchgear according to an embodiment of the present invention specifically includes: The space setting module 10 uses a pre-set space analysis algorithm to obtain a set of installation points based on the spatial structure of the switch cabinet, and installs ultra-high frequency sensors, ultrasonic sensors and temperature sensors according to the set of installation points. Figure 2 This is a schematic diagram of the space setting module according to an embodiment of the present invention. Figure 2 As shown, the space setting module 10 specifically includes: The 3D space acquisition module 101 is used to establish a 3D mesh model of the internal space of the switchgear and define multiple key monitoring areas based on the physical structure of the switchgear. Specifically, the 3D mesh model is obtained by performing a 3D scan of the switchgear interior or by creating a 3D model based on CAD drawings, discretizing the internal space into cubic meshes of unit volume. The mesh resolution can be set according to accuracy requirements. The key monitoring areas include: busbar connections and the area around supporting insulators in the busbar compartment; the moving and stationary contact areas in the circuit breaker compartment; the cable termination area in the cable compartment; and the vicinity of the insulating partitions between air chambers. These areas are key locations where electric fields are concentrated or where partial discharge is easily caused by poor contact or insulation aging.

[0014] Candidate point generation unit 102, based on the three-dimensional mesh model and key monitoring area, generates corresponding candidate installation point sets for the UHF sensor and the ultrasonic sensor respectively; the candidate point generation unit 102 is specifically used for: For UHF sensors, by scanning all panels of the switchgear, non-metallic components and cabinet seams are identified, and the center points of these components or seams are used to form a set of UHF candidate points. Using image recognition technology or based on a preset 3D model database of the switchgear, the locations where electromagnetic wave signals can leak, such as observation windows, insulating basins, pressure relief channels, cabinet door seams, and bolt mounting holes, are automatically located, and the center coordinates of these locations are added to the set of candidate points.

[0015] For ultrasonic sensors, a set of spatial points that do not conflict with solid devices are uniformly generated within the internal space of each key monitoring area defined by the layout optimization unit to form a set of candidate ultrasonic points. Within the three-dimensional space of each key monitoring area, uniform grid points are generated at set spatial intervals. Then, a collision detection algorithm is used to remove points that overlap with solid device models such as busbars, insulators, and mechanisms. The remaining points are the candidate installation points for the ultrasonic sensors, ensuring that the sensors have a clear sound wave propagation path.

[0016] The coverage evaluation unit 103 is used to construct coverage evaluation functions for UHF candidate points and ultrasonic candidate points respectively, and to obtain the detection effect of candidate installation points on the key monitoring area. The coverage assessment unit is specifically used for: The evaluation of the coverage of UHF candidate points specifically includes: for any grid point in the three-dimensional mesh model, if there exists a straight path from that grid point to an UHF candidate point without metal obstruction, then the candidate point is determined to cover that grid point; this process is implemented using a ray casting algorithm in three-dimensional space to simulate the straight-line propagation characteristics of electromagnetic waves. If the ray is not blocked by metal components inside the switch cabinet, it is considered to be covered.

[0017] The evaluation of the coverage of ultrasonic candidate points specifically includes: for any grid point in the three-dimensional mesh model, calculating the Euclidean distance between it and an ultrasonic candidate point, and calculating the coverage based on the distance value using a preset attenuation function, and when the distance exceeds the preset maximum effective detection distance, the coverage is zero.

[0018] The installation point set acquisition unit 104 is used to select a predetermined number of final installation points from the UHF candidate point set and the ultrasonic candidate point set, respectively, with the goal of maximizing the total joint coverage of all key monitoring areas, using an optimization algorithm. Specifically, the installation point set acquisition unit 104 is used to: employ a particle swarm optimization algorithm, using maximizing the total joint coverage as the fitness function, to iteratively solve for the optimal combination of installation points for the UHF sensor and the ultrasonic sensor. The total joint coverage is defined as: for all grid points within the key monitoring areas, if a point is covered by any of the finally selected sensors, it is considered covered; the total joint coverage is the percentage of covered grid points out of the total number of grid points. The optimization objective is to maximize this percentage.

[0019] The multi-source data acquisition module 12 is used to acquire data from the ultra-high frequency sensor, ultrasonic sensor and temperature sensor, and send the data to the signal processing module after preprocessing; the preprocessing includes signal amplification, preliminary filtering to remove obvious outliers, and analog-to-digital conversion.

[0020] Signal processing module 14 is used to receive preprocessed data sent by the multi-source data acquisition module, suppress power frequency noise of UHF signal and ultrasonic signal through adaptive filtering circuit, and extract partial discharge features from the filtered signal using wavelet transform algorithm. The signal processing module 14 is specifically used for: The adaptive filtering circuit employs an adaptive noise cancellation technique based on the LMS algorithm to suppress power frequency interference of specific frequencies from ultra-high frequency signals and ultrasonic signals. Its core principle is to provide a reference signal related to power frequency noise, and dynamically adjust the filter weights through the LMS algorithm to subtract the optimal estimate of the noise component from the original signal.

[0021] For the noise-suppressed signal, the wavelet packet transform algorithm is used to extract the energy characteristics of the signal in different frequency bands, which are used as the partial discharge characteristics.

[0022] The intelligent analysis module 16 receives the partial discharge characteristics and temperature sensor data, and performs fusion analysis on the partial discharge characteristics and temperature sensor data through an AI model that integrates convolutional neural networks and long short-term memory networks, so as to realize the pattern recognition of partial discharge and the prediction of insulation degradation trend, and output early warning information.

[0023] The intelligent analysis module 16 is specifically used for: The partial discharge features are input into a pre-trained convolutional neural network model, which outputs the pattern recognition results of partial discharge. The patterns include at least internal insulation discharge, surface discharge, corona discharge, and suspension discharge. The CNN model is trained with a large amount of known types of partial discharge sample data and can automatically learn the abstract patterns corresponding to different discharge patterns in the feature vector.

[0024] The time-series data, composed of the pattern recognition results and temperature sensor data, is input into a Long Short-Term Memory (LSTM) network model to predict the future trend of insulation condition and generate different levels of early warning information based on the prediction results. The LSTM network can memorize historical states and predict the rate of insulation degradation and risk level over a future period by analyzing the changing trends of discharge mode-temperature time-series data, thereby achieving predictive maintenance. Early warning levels can be divided into "Attention," "Warning," and "Severe" based on the predicted discharge intensity growth rate and temperature rise trend.

[0025] Furthermore, the multimodal fusion online monitoring system for partial discharge in switchgear according to embodiments of the present invention also includes: The visualization and interaction module is used to display the partial discharge pattern recognition results of the switchgear in real time in the form of a web page. The visualization interface further integrates and displays a 3D model of the switchgear, and dynamically marks the sensor location, real-time data, partial discharge type, warning level and trend curve on the model, providing a human-machine interface for users to query historical data and confirm alarms.

[0026] Furthermore, the multi-source data acquisition module sends the preprocessed data to the signal processing module via the LoRa wireless communication network.

[0027] The multimodal fusion switchgear partial discharge online monitoring system and method provided in this application achieves high-precision and anti-interference detection of partial discharge through spatial optimization of point layout, multi-source signal acquisition, adaptive noise suppression and intelligent fusion analysis. It effectively improves the ability to identify discharge modes and predict insulation degradation trends, and provides reliable technical support for switchgear condition assessment and fault early warning.

[0028] Method Implementation Examples According to embodiments of the present invention, a multi-modal fusion method for online monitoring of partial discharge in switchgear is provided. Figure 3 This is a flowchart of the multimodal fusion-based online monitoring method for partial discharge in switchgear according to an embodiment of the present invention. Figure 3 As shown, the multimodal fusion-based online monitoring method for partial discharge in switchgear according to an embodiment of the present invention specifically includes: S1. Using a pre-set spatial analysis algorithm, obtain a set of installation points based on the spatial structure of the switch cabinet, and install ultra-high frequency sensors, ultrasonic sensors, and temperature sensors according to the set of installation points. S2. Collect data from the ultra-high frequency sensor, ultrasonic sensor, and temperature sensor, and send the preprocessed data to the signal processing module. S3. The signal processing module receives the preprocessed data sent by the multi-source data acquisition module, performs power frequency noise suppression on the ultra-high frequency signal and ultrasonic signal through an adaptive filtering circuit, and extracts partial discharge features from the filtered signal using a wavelet transform algorithm. S4. Receive the partial discharge characteristics and temperature sensor data, and perform fusion analysis on the partial discharge characteristics and temperature sensor data through an AI model that integrates convolutional neural networks and long short-term memory networks to achieve pattern recognition of partial discharge and prediction of insulation degradation trends, and output early warning information.

[0029] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-modal fusion switch cabinet partial discharge online monitoring system, characterized in that The method comprises the following steps: The spatial setting module adopts a pre-set spatial analysis algorithm to obtain an installation point set according to the spatial structure of the switch cabinet, and installs a UHF sensor, an ultrasonic sensor and a temperature sensor according to the installation point set; The multi-source data acquisition module is used for acquiring data obtained by the UHF sensor, the ultrasonic sensor and the temperature sensor, and sending the pre-processed data to the signal processing module; The signal processing module is used for receiving the pre-processed data sent by the multi-source data acquisition module, suppressing power frequency noise of the UHF signal and the ultrasonic signal through an adaptive filter circuit, and extracting partial discharge characteristics from the filtered signal through a wavelet transform algorithm; The intelligent analysis module receives the partial discharge characteristics and the temperature sensor data, and realizes pattern recognition of partial discharge and prediction of insulation deterioration trend through fusion analysis of the partial discharge characteristics and the temperature sensor data by an AI model fusing a convolutional neural network and a long short-term memory network, and outputs early warning information.

2. The system of claim 1, wherein, The spatial setting module specifically comprises: A three-dimensional space acquisition module is used for establishing a three-dimensional grid model of the internal space of the switch cabinet, and defining a plurality of key monitoring areas based on the physical structure of the switch cabinet; A candidate point generation unit generates a corresponding candidate installation point set for the UHF sensor and the ultrasonic sensor based on the three-dimensional grid model and the key monitoring areas; An coverage evaluation unit is used for constructing a coverage evaluation function for the UHF candidate point and the ultrasonic candidate point respectively, and obtaining the detection effect of the candidate installation point on the key monitoring area; An installation point set acquisition unit is used for selecting a set number of final installation points from the UHF candidate point set and the ultrasonic candidate point set respectively by using an optimization algorithm with the goal of maximizing the total joint coverage of all key monitoring areas.

3. The system of claim 2, wherein, The candidate point generation unit is specifically used for: For the UHF sensor, scanning all panels of the switch cabinet, identifying non-metallic components and cabinet joints, and constructing a UHF candidate point set with the center points of these components or gaps; For the ultrasonic sensor, uniformly generating a group of space points that do not conflict with solid devices in the internal space of each key monitoring area defined by the layout optimization unit to form an ultrasonic candidate point set.

4. The system of claim 2, wherein, The coverage evaluation unit is specifically used for: Evaluating the coverage of the UHF candidate point, specifically including: for any grid point in the three-dimensional grid model, if there is a straight line path from the grid point to a UHF candidate point without metal obstruction, it is determined that the candidate point covers the grid point; Evaluating the coverage of the ultrasonic candidate point, specifically including: for any grid point in the three-dimensional grid model, calculating the Euclidean distance between it and an ultrasonic candidate point, and calculating the coverage through a pre-set attenuation function according to the distance value, and when the distance exceeds a pre-set maximum effective detection distance, the coverage is zero.

5. The system of claim 1, wherein, The signal processing module is specifically used for: Through the adaptive filter circuit, using an adaptive noise cancellation technology based on the LMS algorithm to suppress power frequency interference of a specific frequency from the UHF signal and the ultrasonic signal; The energy features of the signals in different frequency bands are extracted from the signals after noise suppression by using a wavelet packet transform algorithm, as the partial discharge features.

6. The system of claim 1, wherein, The intelligent analysis module is specifically used for: inputting the partial discharge features into a pre-trained convolutional neural network model, outputting a mode recognition result of the partial discharge, and the mode at least including internal insulation discharge, surface discharge, corona discharge and suspension discharge; inputting the mode recognition result and a time sequence formed by the temperature sensor data into a long short-term memory network model to predict a future development trend of the insulation state, and generating early warning information of different levels based on the prediction result.

7. The system of claim 2, wherein, The installation point set acquisition unit is specifically used for: using a particle swarm optimization algorithm to maximize the total joint coverage as a fitness function, and iteratively solving the optimal installation point combination of the ultra-high frequency sensor and the ultrasonic sensor.

8. The system of claim 1, wherein, The system further comprises: a visual interaction module for displaying the partial discharge mode recognition result of the switch cabinet in the form of a Web page in real time.

9. The system of claim 1, wherein, The multi-source data acquisition module sends the preprocessed data to the signal processing module through a LoRa wireless communication network.

10. A multi-modal fusion switch cabinet partial discharge online monitoring method, characterized in that, Comprise: S1, using a pre-set spatial analysis algorithm, acquiring an installation point set according to the spatial structure of the switch cabinet, and installing the ultra-high frequency sensor, the ultrasonic sensor and the temperature sensor according to the installation point set; S2, acquiring the data obtained by the ultra-high frequency sensor, the ultrasonic sensor and the temperature sensor, and sending the above data to the signal processing module after preprocessing; S3, the signal processing module receives the preprocessed data sent by the multi-source data acquisition module, suppresses the power frequency noise of the ultra-high frequency signal and the ultrasonic signal through an adaptive filter circuit, and extracts the partial discharge features from the filtered signals by using a wavelet transform algorithm; S4, receiving the partial discharge features and the temperature sensor data, and fusing and analyzing the partial discharge features and the temperature sensor data by using an AI model of a convolutional neural network and a long short-term memory network, realizing the mode recognition of the partial discharge and the insulation deterioration trend prediction, and outputting the early warning information.

Images