A method, system, device, and storage medium for identifying defects in power distribution network equipment.

Abstract

This invention discloses a method, system, device, and storage medium for defect identification of distribution network engineering equipment. The method includes: acquiring multi-dimensional feature data samples of the distribution network engineering equipment; training a constructed defect identification model using the multi-dimensional feature data samples; the training includes: initializing the weight states and branch structures of the weight connections in the defect identification model; updating the corresponding weight states and branch structures according to the number of times the weights are used during training; after training, if a weight has zero usage times, deleting that weight and its corresponding branch structure; otherwise, saving the data; and using the trained defect identification model to obtain abnormal diagnostic results for the distribution network engineering equipment based on its inspection information. A multi-set weight sharing mechanism is introduced, and a weight mapping dictionary is designed to avoid duplicate weight storage while enabling flexible addition, comparison, and retrieval of weights, thus ensuring model accuracy while improving model training efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology for power distribution networks, and in particular to a method, system, device, and storage medium for identifying defects in power distribution network engineering equipment. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Distribution networks are mostly concentrated in townships (towns, districts) and villages (communities). Compared with power transmission and transformation scenarios, the environment of distribution networks is more complex and diverse, and there is more interference information in the inspection information. In order to adapt to the complex environment, the form and location of the same power distribution equipment vary, which leads to significant differences in the same defect.

[0004] In addition, due to safety protection needs, various types of protection devices are installed in different parts of the power distribution network, and the similarity between different protection devices and between protection devices and other equipment is relatively high; however, in the power transmission and transformation samples, the defect characteristics of different equipment are quite different.

[0005] This shows that there are significant differences between the characteristics of distribution networks and those of power transmission and transformation, and the sample sizes are on different scales. Therefore, power transmission and transformation models and related strategies are not applicable to distribution networks and cannot be directly transferred to distribution networks for application.

[0006] The main equipment in power distribution network projects includes distribution boxes and cabinets for residential buildings, as well as transformers, high-voltage cabinets, and related equipment and accessories in the distribution rooms. Currently, the main target of power distribution network inspections is outdoor overhead lines, but the monitoring and inspection of the status of related indoor equipment mainly relies on regular manual inspections and maintenance. This is to achieve real-time diagnosis and analysis of defects in indoor power distribution network equipment.

[0007] Existing technologies include constructing fault probability models using historical equipment fault data and health status data, primarily addressing the difficulty in determining the parameters of these models; and establishing indexed databases for field data and matching real-time data to enable searching of field data in the power distribution network. Current identification methods mainly focus on the detection and judgment of power distribution network status data, determining whether there are defects in power distribution network equipment based on operational status data. This approach can effectively monitor the operational status of power distribution equipment, but it still has the following problems:

[0008] 1. Abnormal operating status data can only indicate that there is a problem with the current equipment, but it cannot effectively determine the location of the abnormality and the type of defect. Equipment maintenance personnel still need to conduct on-site inspection and maintenance.

[0009] 2. The construction process of models for judging the operation status of distribution networks and locating abnormal defects still mainly relies on database matching, which cannot effectively cope with the impact of factors such as sudden changes in the surrounding environment of distribution projects. Moreover, the model construction process requires a large amount of data support and continuous optimization and iteration of large-scale networks, resulting in a long model construction time. Summary of the Invention

[0010] To address the aforementioned issues, this invention proposes a method, system, equipment, and storage medium for identifying defects in distribution network engineering equipment. It proposes a multi-dimensional feature fusion method for identifying defects in distribution network engineering equipment, enabling precise location of abnormal equipment and defect types, while simultaneously improving robustness to the surrounding power distribution environment. Furthermore, it proposes a model weight sharing construction method, designing a weight mapping dictionary that avoids redundant weight storage during training while enabling flexible addition, comparison, and retrieval of weights, reducing the network model size, ensuring model accuracy, and improving model training efficiency.

[0011] To achieve the above objectives, the present invention adopts the following technical solution:

[0012] In a first aspect, the present invention provides a method for identifying defects in power distribution network engineering equipment, comprising:

[0013] Obtain multi-dimensional feature data samples of power distribution network engineering equipment;

[0014] The defect discrimination model is trained using multidimensional feature data samples. The training includes: initializing the weight states and branch structures of the weight connections in the defect discrimination model; updating the corresponding weight states and branch structures according to the number of times the weights are used during the training process; and deleting the weights and their corresponding branch structures if there are weights with zero usage times after training is completed, otherwise saving the data.

[0015] The inspection information of the distribution network equipment is used to obtain the abnormal diagnosis results of the distribution network equipment by adopting the trained defect discrimination model.

[0016] As an alternative implementation, the weight state is initialized to zero, and when the weight is used for the nth time during training, the weight state is incremented by n.

[0017] As an alternative implementation, if the branch structure of the weights changes during training and the changed branch structure is not in the initial branch structure, then it is added.

[0018] As an alternative implementation, if the branch structure of the weights appears repeatedly during training, the branch structure is not saved repeatedly, and the already saved parameters are used directly for connection.

[0019] As an alternative implementation, if the weight is not the default weight, the weight, weight status and corresponding branch structure are added first, and then the weight is updated and shared.

[0020] As an alternative implementation, the multidimensional feature data samples include historical visible light images, historical sound data, and historical temperature data of the device, and are used to construct and train defect discrimination models for visible light images, sound data, and temperature data, respectively.

[0021] As an optional implementation, the inspection information includes visible light images, sound, and temperature of the equipment. Based on these, defect discrimination models for visible light images, sound, and temperature are used to diagnose abnormalities in the equipment's operating status. If at least one abnormality exists, the type of equipment defect is determined, and visible light images of the equipment are collected again. The collected visible light images are then analyzed a second time to locate the defect.

[0022] Secondly, the present invention provides a system for identifying defects in power distribution network engineering equipment, comprising:

[0023] The data acquisition module is configured to acquire multi-dimensional feature data samples of power distribution network engineering equipment;

[0024] The training module is configured to train the constructed defect discrimination model using multidimensional feature data samples; the training includes: initializing the weight states and branch structures of the weight connections in the defect discrimination model; updating the corresponding weight states and branch structures according to the number of times the weights are used during the training process; and after training is completed, if there are weights with zero usage times, deleting the weights and their corresponding branch structures, otherwise saving them.

[0025] The discrimination module is configured to use the inspection information of the distribution network equipment and the trained defect discrimination model to obtain the abnormal diagnosis results of the distribution network equipment.

[0026] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0027] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] This invention innovatively proposes a model weight sharing construction method. Unlike existing model construction processes that repeatedly save and retrieve weights and their corresponding branch structures in different parts of the model, this invention introduces a multi-set weight sharing mechanism. It adopts a weight comparison and sharing method and designs a weight mapping dictionary. This not only avoids the repeated saving of weights but also enables flexible addition, comparison, and retrieval of weights. It realizes the replacement of irrelevant weights and the sharing of weight projections in the model, thereby simplifying the network structure and the number of weights during model training, ensuring model accuracy while improving model training efficiency.

[0030] This invention innovatively designs a method for judging defects in power distribution network equipment by fusing multi-dimensional features, which enables precise location of abnormal equipment and defect types, while improving the robustness of the model to the surrounding power distribution environment.

[0031] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0032] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0033] Figure 1 This is a flowchart of the equipment defect identification method for power distribution network engineering provided in Embodiment 1 of the present invention;

[0034] Figure 2 This is a schematic diagram of weighted connections provided in Embodiment 1 of the present invention;

[0035] Figure 3 This is a schematic diagram of the weight sharing mechanism provided in Embodiment 1 of the present invention. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0039] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0040] Example 1

[0041] This embodiment provides a method for identifying defects in distribution network engineering equipment, such as... Figure 1 As shown, it includes:

[0042] Obtain multi-dimensional feature data samples of power distribution network engineering equipment;

[0043] The defect discrimination model is trained using multidimensional feature data samples. The training includes: initializing the weight states and branch structures of the weight connections in the defect discrimination model; updating the corresponding weight states and branch structures according to the number of times the weights are used during the training process; and deleting the weights and their corresponding branch structures if there are weights with zero usage times after training is completed, otherwise saving the data.

[0044] The inspection information of the distribution network equipment is used to obtain the abnormal diagnosis results of the distribution network equipment by adopting the trained defect discrimination model.

[0045] In this embodiment, during the construction of the defect discrimination model, scalar normalization calculation is performed on the data of the sampling points during the training sampling process. A multi-set weight sharing method mechanism is introduced to realize the substitution of irrelevant weights and the sharing of weight projection in the model, thereby realizing the rapid and efficient construction of the defect discrimination model of power distribution network equipment.

[0046] In this embodiment, scalar normalization calculation includes:

[0047] Multidimensional feature data are merged into a multidimensional matrix. A scalar normalization operator is used to normalize the multidimensional matrix, converting vectors into scalar representations and mapping all real numbers to (0, 1). This mapping confines the data to the range of (0, 1), eliminating the adverse effects of outlier data, reducing the order-of-magnitude differences between different data, ensuring the balance of sample distribution, and significantly reducing the representation dimension of the dataset while fully preserving the dataset information, thus effectively accelerating model construction time.

[0048] The scalar normalization operator is represented as:

[0049]

[0050] Where z is a real number, representing the feature data in the multidimensional matrix.

[0051] In this embodiment, as Figure 2-3 As shown, the multi-group weight sharing mechanism includes:

[0052] The design incorporates a weight mapping dictionary. During model training, it selectively deletes, updates, and shares default weights and their connected branch structures. Weights that match the training characteristics are retained and further updated and iterated. New weights are added, weights not present during training are removed, existing weights are repeatedly shared in each layer of the network, and related network branch structures of weights not used during training are pruned and deleted. This establishes the most concise mapping dictionary between model structure and weights, avoiding the repeated saving of the same weight parameter and ensuring that each weight is saved only once when retaining parameters.

[0053] The main implementation method is as follows:

[0054] (1) If the weight is the default weight, the weight operations include deletion, update, and sharing; specifically: let the weight number be a, and its corresponding connection branch structure be represented as a:{b i},a:{b i} is the weight mapping dictionary. The weight state is initialized to a = 0. When weight a is used for the nth time during training, the weight state is incremented by n, at which point a = n.

[0055] (2) When the corresponding branch structure of a changes, if the changed branch structure is not in b i In the middle, add it to b i In the dictionary;

[0056] (3) During the training process, if the branch structure corresponding to a appears repeatedly, it is not necessary to save the branch structure repeatedly. The saved parameters can be used directly for connection.

[0057] (4) When the entire training is completed, if the weight state of a is always 0, it proves that a has not been used. Then, a and its related branch structure are deleted from the model parameters. If it is not 0, then it can be saved.

[0058] (5) If the weight is not the default weight and is generated randomly, add the weight, weight status and related branch structure, and update and share the weight according to steps (1)-(4).

[0059] Existing model building processes involve repeatedly saving and retrieving weights and their corresponding branch structures at different parts of the model. In contrast, this embodiment adopts a weight comparison and sharing method and designs a weight mapping dictionary. This avoids the repeated saving of weights and enables flexible addition, comparison, and retrieval of weights. As a result, the network structure and the number of weights are simplified during model training, ensuring model accuracy while improving model training efficiency.

[0060] In this embodiment, the determination and location of abnormal types of distribution network engineering equipment are realized through comprehensive analysis of multi-dimensional hierarchical information feature fusion. This includes: multi-dimensional inspection information of distribution network engineering equipment, including visible light images, sound, and temperature of the equipment, thereby generating multi-dimensional feature data of the distribution network engineering equipment; using the above-trained defect discrimination model, the equipment temperature and sound are detected in real time to diagnose abnormal defects of the distribution network engineering equipment; and then, combined with the visible light images of the equipment, the location of the abnormal defects is realized.

[0061] Specifically:

[0062] (1) The multi-dimensional feature data samples of the power distribution network equipment include: historical visible light images of the equipment, historical sound information of the equipment, and historical temperature information of the equipment, which are used to train the constructed defect discrimination model;

[0063] Among them, the visible light history images of the equipment that reflect its appearance and shape are obtained, including the normal visible light history images of the equipment and the abnormal visible light history images of the equipment, and a defect discrimination model of the visible light images of the equipment is constructed.

[0064] Acquire historical device sound data, including historical data of normal device sound and historical data of abnormal device sound, and construct a defect discrimination model for device sound.

[0065] Acquire historical equipment temperature data, including historical data of normal equipment temperature and historical data of abnormal equipment temperature, and construct a defect discrimination model for equipment temperature.

[0066] In the construction and training process of the aforementioned defect discrimination model, a multi-set weight sharing method mechanism is introduced, and a weight mapping dictionary is designed. During the training process, the repeated saving of weights is avoided, and the flexible addition, comparison and calling of weights are realized, thereby reducing the size of the network model, ensuring model accuracy, and improving model training efficiency.

[0067] (2) Obtain real-time inspection information of power distribution network equipment, including:

[0068] Acquire device sound information and use a trained device sound defect discrimination model to determine whether there are any abnormalities in the device's operating sound;

[0069] The equipment temperature information is obtained, and the trained equipment temperature defect discrimination model is used to determine whether the current equipment temperature change trend and range are within the normal range, thereby determining whether the equipment status is normal.

[0070] The visible light image of the equipment is acquired to obtain the appearance and morphology information of the power distribution network equipment in the current environment. The defect discrimination model of the trained visible light image of the equipment is used to determine whether there is any abnormality in the current equipment morphology.

[0071] This embodiment analyzes information such as visible light, equipment sound, and equipment temperature. Different data are used to analyze different operating states of the equipment, and multi-dimensional feature data is constructed from dimensions such as appearance, internal temperature, and sound performance. By analyzing the equipment temperature and operating sound status, it is determined whether the equipment has malfunctioned or has defects. Through visible light image analysis, abnormal equipment shape and the location of malfunctions and defects are accurately identified and located.

[0072] In this embodiment, the most commonly used community distribution transformer and distribution cabinet in distribution network engineering equipment are taken as examples. The main defect of the distribution transformer is abnormal transformer grounding, and the main defect of the distribution cabinet is poor contact of the distribution cabinet disconnect switch.

[0073] Acquire visible light images of transformer grounding normal and abnormal, and distribution cabinet disconnectors normal and poor contact; acquire historical sound data of transformers and distribution cabinets during normal and abnormal operation; acquire historical temperature data of transformers and distribution cabinets during normal and abnormal operation.

[0074] Then, based on the acquired equipment operation status, the visible light data, operating sound change status data, and temperature change status data of the transformer and distribution cabinet are scalar normalized into a unified multi-dimensional feature matrix, and then fed into the trained defect discrimination model for data detection. When none of the three are abnormal, it indicates that the equipment is operating normally. If any of the three is abnormal, visible light images must be collected again and analyzed a second time to locate the type and location of equipment defects.

[0075] Example 2

[0076] This embodiment provides a power distribution network equipment defect identification system, including:

[0077] The data acquisition module is configured to acquire multi-dimensional feature data samples of power distribution network engineering equipment;

[0078] The training module is configured to train the constructed defect discrimination model using multidimensional feature data samples; the training includes: initializing the weight states and branch structures of the weight connections in the defect discrimination model; updating the corresponding weight states and branch structures according to the number of times the weights are used during the training process; and after training is completed, if there are weights with zero usage times, deleting the weights and their corresponding branch structures, otherwise saving them.

[0079] The discrimination module is configured to use the inspection information of the distribution network equipment and the trained defect discrimination model to obtain the abnormal diagnosis results of the distribution network equipment.

[0080] It should be noted that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0081] In further embodiments, the following is also provided:

[0082] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0083] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0084] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0085] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0086] The method in Example 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0087] Those skilled in the art will recognize that the units, i.e., algorithm steps, of the various examples described in connection with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0088] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for identifying defects in power distribution network equipment, characterized in that, include: Obtain multi-dimensional feature data samples of power distribution network engineering equipment; The defect discrimination model is trained using multidimensional feature data samples. The training includes: initializing the weight states and branch structures of the weight connections in the defect discrimination model; updating the corresponding weight states and branch structures according to the number of times the weights are used during the training process; and deleting the weights and their corresponding branch structures if there are weights with zero usage times after training is completed, otherwise saving the data. The inspection information of the distribution network equipment is used to obtain the abnormal diagnosis results of the distribution network equipment by adopting the trained defect discrimination model. Multiple weight sharing mechanisms are introduced during the training sampling process, including: designing a weight mapping dictionary; selectively deleting, updating, and sharing the default weights and their connected branch structures in the model during training; retaining weights that meet the training characteristics and performing further updates and iterations; adding newly appearing weights; deleting weights that did not appear during training; repeatedly sharing existing weights in each layer of the network; and pruning and deleting the related network branch structures of weights that were not used during training. If the branch structure of a weight appears repeatedly during training, the branch structure is not saved repeatedly, and the already saved parameters are used directly for connection.

2. The method for identifying defects in power distribution network equipment as described in claim 1, characterized in that, The weight state is initialized to zero. When the weight is used for the nth time during training, the weight state is incremented by n.

3. The method for identifying defects in power distribution network equipment as described in claim 1, characterized in that, When the branch structure of the weights changes during training, if the changed branch structure is not in the initial branch structure, it will be added.

4. The method for identifying defects in power distribution network equipment as described in claim 1, characterized in that, If the weight is not the default weight, first add the weight, the weight status and the corresponding branch structure, and then update and share the weight.

5. The method for identifying defects in power distribution network equipment as described in claim 1, characterized in that, The multidimensional feature data samples include historical visible light images, historical sound data, and historical temperature data of the equipment. Based on these data, defect discrimination models for visible light images, sound data, and temperature data are constructed and trained.

6. The method for identifying defects in power distribution network equipment as described in claim 1, characterized in that, The inspection information includes visible light images, sound, and temperature of the equipment. Based on these, defect discrimination models for visible light images, sound, and temperature are used to diagnose abnormalities in the equipment's operating status. If at least one abnormality exists, the type of equipment defect is determined, and visible light images of the equipment are collected again. The collected visible light images are then analyzed a second time to locate the defect.

7. A defect identification system for power distribution network engineering equipment, characterized in that, include: The data acquisition module is configured to acquire multi-dimensional feature data samples of power distribution network engineering equipment; The training module is configured to train the constructed defect discrimination model using multidimensional feature data samples; the training includes: initializing the weight states and branch structures of the weight connections in the defect discrimination model; updating the corresponding weight states and branch structures according to the number of times the weights are used during the training process; and after training is completed, if there are weights with zero usage times, deleting the weights and their corresponding branch structures, otherwise saving them. The discrimination module is configured to use the inspection information of the distribution network equipment and the trained defect discrimination model to obtain the abnormal diagnosis results of the distribution network equipment. Multiple weight sharing mechanisms are introduced during the training sampling process, including: designing a weight mapping dictionary; selectively deleting, updating, and sharing the default weights and their connected branch structures in the model during training; retaining weights that meet the training characteristics and performing further updates and iterations; adding newly appearing weights; deleting weights that did not appear during training; repeatedly sharing existing weights in each layer of the network; and pruning and deleting the related network branch structures of weights that were not used during training. If the branch structure of a weight appears repeatedly during training, the branch structure is not saved repeatedly, and the already saved parameters are used directly for connection.

8. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-6.

Images