High-voltage isolation switch state identification method and system based on multi-source data fusion

Abstract

The invention relates to the technical field of high-voltage isolation switch state recognition, in particular to a high-voltage isolation switch state recognition method and system based on multi-source data fusion, and the method comprises the steps: carrying out the filtering, denoising and linear transformation of input data through a data preprocessing module, and obtaining normalized multi-source data; respectively inputting the normalized data into a pre-trained single-mode uncertainty estimation network to obtain data distribution of each signal; fusing a plurality of data distributions through a Bayesian expert product distribution fusion method, and weighting features according to the uncertainty of each data distribution; sampling from the fused multi-modal distribution to obtain multi-modal fusion features used for isolating switch state identification; and inputting the multi-modal fusion features into a pre-trained classifier to obtain an isolation switch state identification result. The multi-source information fusion method has the beneficial effects that the multi-source information fusion method provided by the invention has good expansibility, and is convenient for coping with a complex and changeable real deployment environment.

Description

Technical Field

[0001] The present invention relates to the technical field of high - voltage disconnector state recognition, and particularly to a method and system for high - voltage disconnector state recognition based on multi - source data fusion. Background Technique

[0002] The high - voltage disconnector is a key device in the power system. Accurately identifying the health status of the disconnector is crucial for ensuring the safety and stability of the power system. Disconnectors are mostly installed outdoors and are easily affected by harsh working environments such as humidity, dust, large temperature differences, rain and snow. These factors can easily reduce the reliability of the disconnector's opening and closing. Manual inspection relies on the naked eye or experience to judge the health status of the disconnector, which has the disadvantages of strong subjectivity, long fault location time, late discovery of hidden dangers, many inspection times, and high labor costs. Existing condition monitoring technologies focus on using conventional monitoring means such as vibration analysis and acoustic emission monitoring. These methods are insufficient to timely detect all potential problems when dealing with complex power system operation data, especially in the performance evaluation under high load or extreme conditions.

[0003] With the development of smart grids, the complexity and dynamics of power systems are increasing continuously. A single information source often cannot comprehensively reflect the health status of equipment. Therefore, fusing multiple modal information has become a key means to improve the condition monitoring ability of power systems. By combining multiple device operation data, the status of disconnectors can be more comprehensively understood and predicted, thereby improving the accuracy of fault diagnosis and early warning. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is: in the existing high - voltage disconnector state monitoring technology, the data fusion stage lacks an estimation of data reliability and uncertainty, resulting in noise data and missing information affecting the accuracy of state recognition.

[0005] The above - mentioned technical problem is solved by the following technical solution: The present invention proposes a method for high - voltage disconnector state recognition based on multi - source data fusion, which includes inputting the multi - source data after normalization processing into a pre - trained single - modal uncertainty estimation network respectively to obtain the data distribution of each signal;

[0006] Fusing the data distributions of each signal through a distribution fusion method of Bayesian expert product, and weighting the data features according to the uncertainty of each data distribution to obtain multi - modal fusion data;

[0007] Sampling from the data distribution of the multi - modal fusion data to obtain multi - modal fusion features for disconnector state recognition;

[0008] By inputting the multi-modal fusion features into a pre-trained classifier, the recognition result of the disconnector state is obtained.

[0009] In a preferred embodiment of the method for identifying the state of a high-voltage disconnector based on multi-source data fusion in the present invention: the data distribution includes simultaneously modeling the embedding and uncertainty of data features;

[0010] Data features refer to the statistical information, patterns, and structures extracted from the original sensor data.

[0011] In a preferred embodiment of the method for identifying the state of a high-voltage disconnector based on multi-source data fusion in the present invention: weighting the data features includes

[0012] assigning higher weights to data features with low uncertainty;

[0013] assigning lower weights to data distributions with high uncertainty.

[0014] In a preferred embodiment of the method for identifying the state of a high-voltage disconnector based on multi-source data fusion in the present invention: multi-modal fusion features refer to the comprehensive feature representation obtained by fusing the data distributions of different modalities.

[0015] In a preferred embodiment of the method for identifying the state of a high-voltage disconnector based on multi-source data fusion in the present invention: the multi-modal fusion features are sourced from multi-source data;

[0016] Multi-source data includes current signals, vibration signals, angular displacement signals, visible light images, and infrared light images.

[0017] In a preferred embodiment of the method for identifying the state of a high-voltage disconnector based on multi-source data fusion in the present invention: the current signal reflects the current change of the disconnector during electrical operation;

[0018] The vibration signal captures the vibration information generated by the disconnector during mechanical operation;

[0019] The angular displacement signal records the dynamic information of the angle change during the operation of the disconnector;

[0020] The visible light image obtains the appearance state of the disconnector through optical imaging technology, including the integrity and position information of components;

[0021] The infrared light image uses infrared imaging technology to obtain the thermal distribution information of the disconnector for detecting potential thermal faults.

[0022] In a preferred embodiment of the method for identifying the state of a high-voltage disconnector based on multi-source data fusion in the present invention: the disconnector state recognition result includes the probability values of various faults;

[0023] Obtain the evaluation result of the health status of the disconnector through the probability value.

[0024] To solve the above technical problems, the present invention also provides the following technical solutions: A high-voltage disconnector status recognition system based on multi-source data fusion, which includes a data acquisition module for collecting the operating status data of the high-voltage disconnector from a variety of sensors;

[0025] A data preprocessing module for filtering, denoising, and linearly transforming the collected multi-source disconnector status data to obtain normalized multi-source data, providing high-quality and standardized data input for subsequent single-modal feature extraction and multi-modal fusion;

[0026] A multi-modal data fusion module for fusing the uncertainty distributions of different modalities and dynamically adjusting the weights of each modality.

[0027] A computer device includes a memory and a processor. The memory stores a computer program. It is characterized in that when the processor executes the computer program, it implements the content of an intelligent early warning system for the oil pressure device of a hydropower station governor.

[0028] A computer-readable storage medium stores a computer program. It is characterized in that when the computer program is executed by a processor, it implements the content of an intelligent early warning system for the oil pressure device of a hydropower station governor.

[0029] The beneficial effects of the present invention: By collecting monitoring data from different sources, constructing a single-modal uncertainty distribution estimation network, obtaining the distribution representation of each modality, guiding multi-modal distribution fusion through uncertainty modeling, and finally performing multi-classification to output the health status of the disconnector. By learning the reliability of data from different sources and dynamically assigning weights to different data, it overcomes unreliable problems such as modal noise and semantic ambiguity to achieve more efficient information fusion; in addition, the multi-source information fusion method proposed by the present invention has good scalability. When the number of modalities increases or decreases, it is not necessary to retrain all network parameters. Only need to add or reduce the single-modal learning branches, and the remaining part remains unchanged, which is convenient for dealing with complex and changeable real-world deployment environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings in the following description only relate to some embodiments of the present invention and do not limit the present invention.

[0031] Figure 1 Shows the overall flowchart of the technical solution of the high-voltage disconnector status recognition method based on multi-source data fusion.

[0032] Figure 2 It shows a schematic diagram of an uncertain distribution estimation network for the state recognition method of high-voltage disconnectors based on multi-source data fusion. Detailed implementation manners

[0033] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below in conjunction with the detailed implementation manners and the accompanying drawings.

[0034] The terms used in the present invention are those general terms that are currently widely used in the art in consideration of the functions of the present invention. However, these terms may change according to the intentions of those of ordinary skill in the art, precedents, or new technologies in the art. In addition, specific terms may be selected by the applicant, and in this case, their detailed meanings will be described in the detailed description of the present invention. Therefore, the terms used in the specification should not be understood as simple names, but based on the meanings of the terms and the overall description of the present invention.

[0035] Embodiment 1

[0036] As Figures 1 to 2 , this embodiment provides a state recognition method for high-voltage disconnectors based on multi-source data fusion, including:

[0037] Input the multi-source data after normalization processing into the pre-trained single-modal uncertainty estimation network respectively to obtain the data distribution of each signal;

[0038] It should be noted that this solution takes into account the uncertainty of the data and uses the uncertainty distribution estimation network to model each modal information, that is, input the normalized data into the pre-trained single-modal uncertainty estimation network respectively. The specific steps are as follows:

[0039] First, represent the feature z embedded by the signal x in the latent space as a Gaussian distribution, that is, p(z|x) = N(μ, σ 2 ), where the statistics μ and σ are obtained through a multi-layer perceptron:

[0040]

[0041] Among them, μ m represents the most likely feature embedding; σ m represents the uncertainty of the data, and the larger the variance value, the higher the uncertainty; and are multi-layer perceptrons used to obtain the two statistics.

[0042] Furthermore, sample from the unimodal distribution. To prevent direct sampling from making it impossible to compute gradients during model training, use the reparameterization trick to enable gradient propagation. For each modality, sample random noise $\epsilon$ from the normal distribution $N(0, 1)$, and then generate $z$ m As the sampling result:

[0043] $z$ m $ = \mu$ m $+ \epsilon\sigma$ m , $\epsilon \sim N(0, 1)$

[0044] Furthermore, to enable the estimated distribution to effectively capture modality-specific intrinsic information, use the reconstruction loss to constrain it to minimize the difference between the unimodal signal $x$ and the modality embedding $z$ m First, use an MLP as the decoder to reconstruct the unimodal signal:

[0045] $z$ m ' $ = \text{MLP}(z$ m )

[0046] Second, use the mean squared error (MSE) as the reconstruction loss, and its formula is as follows:

[0047]

[0048] Furthermore, to prevent degradation to a deterministic distribution due to too small variance, introduce the Kullback-Leibler divergence as a regularization term to constrain the estimated distribution $N(\mu, \sigma$ 2 ) to approximate the normal distribution $N(0, 1)$:

[0049]

[0050] Through these two constraints, the reconstruction loss and the KL divergence, we estimate the uncertainty distribution of each modality's data, thus providing a strong basis for subsequent multimodal data fusion.

[0051] Fuse the data distributions of each signal through the distribution fusion method of the product of Bayesian experts, and weight the data features according to the uncertainty of each data distribution to obtain multimodal fusion data;

[0052] Sample from the data distribution of the multimodal fusion data to obtain multimodal fusion features for isolating switch state recognition;

[0053] Input the multimodal fusion features into a pre-trained classifier to obtain the isolating switch state recognition result.

[0054] Furthermore, the data distribution includes jointly modeling the embedding and uncertainty of data features;

[0055] Data features refer to statistical information, patterns, and structures extracted from raw sensor data.

[0056] It should be noted that data features can effectively represent the internal attributes and states of data, reflecting the health status, potential faults, and the impact of the operating environment of the disconnector.

[0057] Furthermore, weighting the data features includes

[0058] assigning higher weights to data features with low uncertainty;

[0059] assigning lower weights to data distributions with high uncertainty.

[0060] Furthermore, multi-modal fusion features refer to the comprehensive feature representation obtained by fusing data distributions of different modalities.

[0061] Furthermore, the source of multi-modal fusion features is multi-source data;

[0062] Multi-source data includes current signals, vibration signals, angular displacement signals, visible light images, and infrared light images.

[0063] Furthermore, the current signal reflects the current change of the disconnector during electrical operation;

[0064] The vibration signal captures the vibration information generated by the disconnector during mechanical operation;

[0065] The angular displacement signal records the dynamic information of the angle change during the operation of the disconnector;

[0066] The visible light image obtains the appearance state of the disconnector through optical imaging technology, including the integrity and position information of components;

[0067] The infrared light image obtains the thermal distribution information of the disconnector using infrared imaging technology for detecting potential thermal faults. It should be noted that the multi-modal fusion network refers to fusing the obtained n data distributions into a unified distribution.

[0068] Specifically, the joint distribution of multiple distributions is obtained through the Bayesian Product-of-Experts. Among them, the weights of each single-modal distribution are defined, and the formula is as follows:

[0069]

[0070] Among them, μ f is μ1, μ2,..., μ nThe weighted sum is such that the larger the variance of the data, the smaller the contribution weight. In the above way, the information of each modal distribution will be dynamically incorporated into the multi-modal fusion distribution. In addition, the Bayesian product of experts model does not need to introduce additional learnable parameters, and only calculates the multi-modal fusion distribution through the parameters learned by the single-modal network. This means that the network has good scalability, that is, when the number of modalities increases, only the new uncertainty estimation network needs to be trained separately and then incorporated into the original single-modal estimation network, without retraining the entire network.

[0071] Furthermore, sample from the multi-modal distribution to obtain the multi-modal fusion feature:

[0072] z f = μ f + ∈σ f , ∈~N(0,1)

[0073] Send the multi-modal feature into the classifier to obtain the recognition result:

[0074]

[0075] In the model training stage, use Cross-Entropy Loss to optimize the parameters:

[0076]

[0077] Combining the two losses mentioned in the single-modal uncertainty estimation network, the total loss can be summarized as:

[0078] L all = L CE + λ1L Rec + λ2L KL

[0079] Among them, λ1 and λ2 are adjustable balance coefficients.

[0080] Furthermore, the disconnector status recognition result includes the probability values of various faults;

[0081] Obtain the evaluation result of the health status of the disconnector through the probability values.

[0082] In summary, the present invention acquires monitoring data from different sources, constructs a single-modal uncertainty distribution estimation network to obtain the distribution representation of each modality, guides multi-modal distribution fusion through uncertainty modeling, and finally performs multi-classification to output the health state of the disconnector. By learning the reliability of different source data and dynamically assigning weights to different data, the problems of unreliable factors such as modal noise and semantic ambiguity are overcome to achieve more efficient information fusion. In addition, the multi-source information fusion method proposed by the present invention has good scalability. When the number of modalities increases or decreases, it is not necessary to retrain all network parameters. Only the single-modal learning branches need to be increased or decreased, and the remaining parts remain unchanged, which is convenient for dealing with complex and changeable real-world deployment environments.

[0083] Embodiment 2

[0084] As Figures 1 to 2 , this embodiment provides a method for identifying the state of a high-voltage disconnector based on multi-source data fusion, including a data acquisition module for acquiring the operating state data of the high-voltage disconnector from multiple sensors;

[0085] It should be noted that the data acquisition module includes a current sensor for acquiring the current signal of the disconnector;

[0086] a vibration sensor for detecting the vibration information during the operation of the disconnector;

[0087] an angular displacement sensor for measuring the change in the operating angle of the disconnector;

[0088] a visible light camera for taking the appearance image of the disconnector;

[0089] an infrared thermal imager for obtaining the thermal distribution image of the disconnector;

[0090] a data acquisition card or a data acquisition terminal for converting the sensor signal into a digital signal and transmitting it to the processing unit.

[0091] Its input signals are current signals, vibration signals, angular displacement signals, visible light images, infrared light images, etc.; the output signal is the original multi-source data.

[0092] A data preprocessing module for filtering, denoising, and linearly transforming the acquired multi-source disconnector state data to obtain normalized multi-source data, providing high-quality and standardized data input for subsequent single-modal feature extraction and multi-modal fusion;

[0093] It should be noted that the data preprocessing module includes an industrial computer and a storage device, and its input signal is the original multi-source data; the output signal is the normalized multi-source data.

[0094] The multimodal data fusion module is used to fuse the uncertainty distributions of different modalities and dynamically adjust the weights of each modality.

[0095] It should be noted that the input signal of the multimodal data fusion module is the Gaussian distribution (mean and variance) of each modality; the output signal is the fused multimodal distribution.

[0096] In summary, the present invention collects monitoring data from different sources, constructs a single-modal uncertainty distribution estimation network, obtains the distribution representation of each modality, guides the multimodal distribution fusion through uncertainty modeling, and finally performs multi-classification to output the health status of the disconnector. By learning the reliability of data from different sources and dynamically assigning weights to different data, it overcomes unreliable problems such as modal noise and semantic ambiguity to achieve more efficient information fusion; in addition, the multi-source information fusion method proposed by the present invention has good scalability. When the number of modalities increases or decreases, it is not necessary to retrain all network parameters. Only need to increase or decrease the single-modal learning branches, and the remaining part remains unchanged, which is convenient for dealing with complex and changeable actual deployment environments.

[0097] Embodiment 3

[0098] This embodiment is the third embodiment of the present invention. The difference from the previous two embodiments is:

[0099] If the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0100] The logic and / or steps represented in the flowchart or otherwise described herein can be considered, for example, a definable sequence list of executable instructions for implementing a logical function, which can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0101] More specific examples (nonexhaustive list) of computer-readable media include the following: an electrical connection portion (electronic device) having one or more wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0102] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0103] 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 preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying the state of a high-voltage disconnector based on multi-source data fusion, characterized in that: including Inputting the normalized multi-source data into a pre-trained single-modal uncertainty estimation network respectively to obtain the data distributions of each type of signal; Fusing the data distributions of each type of signal by the distribution fusion method of the product of Bayesian experts, and weighting the data features according to the uncertainty of each data distribution to obtain multi-modal fusion data; Sampling from the data distribution of the multi-modal fusion data to obtain multi-modal fusion features for isolating switch state recognition; Inputting the multi-modal fusion features into a pre-trained classifier to obtain the isolating switch state recognition result.

2. The method for identifying the state of a high-voltage disconnector based on multi-source data fusion according to claim 1, wherein: The data distribution includes jointly modeling the embedding and uncertainty of data features; The data features refer to the statistical information, patterns, and structures extracted from the original sensor data.

3. The method for identifying the state of a high-voltage disconnector based on multi-source data fusion according to claim 2, characterized in that: The weighting of the data features includes Assigning higher weights to data features with low uncertainty; Assigning lower weights to data distributions with high uncertainty.

4. The method for identifying the state of a high-voltage disconnector based on multi-source data fusion according to claim 3, characterized in that: The multi-modal fusion features refer to the comprehensive feature representations obtained by fusing the data distributions of different modalities.

5. The method for identifying the state of a high-voltage disconnector based on multi-source data fusion according to claim 4, wherein: The source of the multi-modal fusion features is the multi-source data; The multi-source data includes current signals, vibration signals, angular displacement signals, visible light images, and infrared light images.

6. The method for identifying the state of a high-voltage disconnector based on multi-source data fusion according to claim 5, wherein: The current signal reflects the current change of the isolating switch during electrical operation; The vibration signal captures the vibration information generated by the isolating switch during mechanical operation; The angular displacement signal records the dynamic information of the angle change during the operation of the isolating switch; The visible light image obtains the appearance state of the isolating switch through optical imaging technology, including the integrity and position information of components; The infrared light image obtains the thermal distribution information of the isolating switch using infrared imaging technology for detecting potential thermal faults.

7. The method for identifying the state of a high-voltage disconnector based on multi-source data fusion according to any one of claims 4 to 6, characterized in that: The isolating switch state recognition result includes the probability values of various types of faults; Obtaining the evaluation result of the health state of the isolating switch through the probability values.

8. A system adopting a method for identifying the state of a high-voltage disconnector based on multi-source data fusion as described in any one of claims 1 to 7, characterized in that: including A data acquisition module for collecting the operation state data of a high-voltage isolating switch from multiple sensors; A data preprocessing module for filtering, denoising, and linearly transforming the collected multi-source isolating switch state data to obtain normalized multi-source data, providing high-quality and standardized data input for subsequent single-modal feature extraction and multi-modal fusion; A multi-modal data fusion module for fusing the uncertainty distributions of different modalities and dynamically adjusting the weights of each modality.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the content of a system for intelligent warning of a pressure oil device of a hydropower station governor described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the content of a system for intelligent warning of a pressure oil device of a hydropower station governor described in any one of claims 1 to 7.

Images