Data storage method and device of power distribution network, storage medium and processor

Abstract

The invention discloses a data storage method and device of a power distribution network, a storage medium and a processor. The method comprises the following steps: acquiring an equipment image of a power distribution network; the equipment image is input into a target image recognition model to be analyzed, equipment parameters of the equipment image are obtained, and the target image recognition model is obtained through training of an equipment image sample of the power distribution network and a corresponding equipment parameter sample; classifying the equipment parameters to obtain a classification result; and returning the equipment image and the corresponding classification result to the server for storage. The technical problem that the data storage efficiency of the power distribution network is low is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data storage, and in particular to a data storage method, device, storage medium and processor for a distribution network. Background Art

[0002] At present, the distribution network is the direct platform for the distribution system to provide services to users, and the quality of its operation will have the most direct impact on the user's experience of electricity consumption. However, from the past to the present, in terms of the analysis and construction of the distribution network, the management of the distribution network has been negligent and the level of automation is poor. There are often discrepancies between the records of users, grid topology and ledgers, which seriously hinders the operation efficiency of the power grid.

[0003] In the related technology, the data storage problem of the distribution network is mainly carried out manually. However, in the manual data entry process, the human intervention coefficient is too large, resulting in low data entry efficiency. At the same time, there are problems such as input information errors, resulting in a high rate of missing records of the distribution network. Therefore, there is a technical problem of low data efficiency in the distribution network.

[0004] With respect to the above-mentioned technical problem of low data storage efficiency in distribution networks, no effective solution has been proposed so far. Summary of the invention

[0005] Embodiments of the present invention provide a data storage method, device, storage medium and processor for a distribution network, so as to at least solve the technical problem of low data storage efficiency in the distribution network.

[0006] According to one aspect of an embodiment of the present invention, a data storage method for a distribution network is provided. The method may include: obtaining a device image of the distribution network; inputting the device image into a target image recognition model for analysis to obtain device parameters of the device image, wherein the target image recognition model is trained by device image samples and corresponding device parameter samples of the distribution network; classifying the device parameters to obtain classification results; and transmitting the device image and the corresponding classification results back to a server for storage.

[0007] Optionally, a data storage method for a distribution network includes: obtaining historical equipment images of the distribution network; extracting equipment image samples and corresponding equipment parameter samples from the historical equipment images; and training an initial image recognition model using the equipment image samples and corresponding equipment parameter samples to obtain a target image recognition model.

[0008] Optionally, before inputting the device image into the target image recognition model for analysis to obtain device parameters of the device image, the method further includes: performing correction processing on the device image to obtain a corrected device image; and cutting out image characters from the corrected device image.

[0009] Optionally, the device image is input into a target image recognition model for analysis to obtain device parameters of the device image, including: extracting character features of the device image, wherein the character features are used to indicate basic information of the device image; and inputting the character features into a target image recognition model to obtain device parameters of the device image.

[0010] Optionally, inputting the character features into a target image recognition model to obtain device parameters of the device image includes: using the target image recognition model to recognize the character features to obtain the device parameters of the device image.

[0011] Optionally, the data storage method of the distribution network includes: generating a data form from at least one device image and at least one device parameter; and integrating the data form to obtain a device database of the distribution network.

[0012] According to another aspect of an embodiment of the present invention, a data storage device for a power distribution network is also provided. The device may include: an acquisition unit for acquiring a device image of the power distribution network; an analysis unit for inputting the device image into a target image recognition model for analysis to obtain device parameters of the device image, wherein the target image recognition model is trained by device image samples and corresponding device parameter samples of the power distribution network; a classification unit for classifying the device parameters to obtain classification results; and a storage unit for transmitting the device image and the corresponding classification results back to a server for storage.

[0013] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0014] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is used to run a program. When the program is run, the method of implementing each embodiment of the present invention is executed.

[0015] According to another aspect of an embodiment of the present invention, a computer program product is provided, including computer instructions, which implement the methods in various embodiments of the present invention when executed by a processor.

[0016] In an embodiment of the present invention, a device image of a distribution network is acquired; the device image is input into a target image recognition model for analysis to obtain device parameters of the device image, wherein the target image recognition model is obtained by training with device image samples of the distribution network and corresponding device parameter samples; the device parameters are classified to obtain classification results; and the device image and the corresponding classification results are transmitted back to a server for storage. That is to say, by inputting the device image of the distribution network into a target image recognition model for analysis to obtain device parameters of the device image, the device image and its corresponding device parameters are transmitted back to a server for storage. Since the present invention can automatically process the device image, it avoids the shortcomings of information loss caused by manual entry of device information, thereby solving the technical problem of low data storage efficiency of the distribution network, and further achieving the technical effect of improving the data storage efficiency of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 is a flow chart of a data storage method for a distribution network according to an embodiment of the present invention;

[0019] Figure 2 is a flow chart of a current low-voltage distribution network equipment verification method according to an embodiment of the present invention;

[0020] Figure 3 is a flow chart of a method for automatically identifying a low-voltage distribution network ledger according to an embodiment of the present invention;

[0021] Figure 4 is a flowchart of an image processing according to an embodiment of the present invention;

[0022] Figure 5 is a schematic diagram of a basic structure of a DERT neural network according to an embodiment of the present invention;

[0023] Figure 6 is a schematic diagram of a data storage device for a distribution network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, functional component or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, functional components or devices.

[0026] Example 1

[0027] According to an embodiment of the present invention, an embodiment of a data storage method for a distribution network is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] Figure 1 is a flow chart of a data storage method for a distribution network according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0029] Step S101, obtaining a device image of a distribution network.

[0030] In the technical solution provided in the above step S101 of the present invention, an image of equipment in the distribution network is obtained.

[0031] In this embodiment, the device image of the power distribution network is obtained. For example, the device image of the power distribution network is photographed by a photographing device. This is only an example and does not limit the specific method of obtaining the device image of the power distribution network.

[0032] Step S102: input the device image into the target image recognition model for analysis to obtain device parameters of the device image.

[0033] In the technical solution provided in the above step S102 of the present invention, the target image recognition model is obtained by training with equipment image samples and corresponding equipment parameter samples of the power distribution network.

[0034] In this embodiment, after acquiring the device image of the power distribution network according to step S101, the device image is input into the target image recognition model for analysis to obtain the device parameters of the device image. For example, the target image recognition model may be a neural network, which is only an illustrative example and does not limit the specific content of the target image recognition model.

[0035] Optionally, by extracting character features from the device image, wherein the character features are used to indicate basic information of the device image, for example, the character features may be rated current, which is only an illustrative example and does not limit the specific content of the character features.

[0036] Optionally, the character features are input into the target image recognition model to obtain the device parameters of the device image. For example, the character features are recognized using a neural network model to obtain the device parameters of the device image. This is only an illustrative example and does not limit the specific method of obtaining the device parameters of the device image.

[0037] Step S103, classify the device parameters to obtain classification results.

[0038] In the technical solution provided in the above step S103 of the present invention, the device parameters are classified to obtain the classification results.

[0039] In this embodiment, after obtaining the device parameters of the device image in step S102, the device parameters are classified to obtain classification results. For example, the device parameters may be the rated current of the circuit breaker, the long delay setting current of the circuit breaker, the rated insulation voltage of the circuit breaker, the rated impulse withstand voltage, etc. This is only an example and does not limit the specific content of the device parameters.

[0040] Optionally, categorizing device parameters into groups can help administrators more easily manage large numbers of devices and quickly locate and resolve issues.

[0041] Step S104, the device image and the corresponding classification result are transmitted back to the server for storage.

[0042] In the technical solution provided in the above step S104 of the present invention, after the classification result is obtained in step S103, the device image and the corresponding classification result are transmitted back to the server for storage.

[0043] In this embodiment, the device image and the corresponding classification result are transmitted back to the server for storage by means of command feedback, wherein command feedback refers to returning the execution result or processing status to the party sending the command.

[0044] Optionally, command feedback can replace manual entry, avoiding the shortcomings of manual entry such as lack of accuracy in information transmission, and can achieve rapid and accurate identification of equipment ledgers and summarize ledger information.

[0045] It should be noted that the above embodiments may be executed by a data storage device of a power distribution network.

[0046] In this embodiment, the device image of the distribution network is obtained; the device image is input into the target image recognition model for analysis to obtain the device parameters of the device image, wherein the target image recognition model is obtained by training the device image samples and corresponding device parameter samples of the distribution network; the device parameters are classified to obtain the classification results; the device image and the corresponding classification results are transmitted back to the server for storage. In other words, the device image of the distribution network is input into the target image recognition model for analysis to obtain the device parameters of the device image, and then the device image and its corresponding device parameters are transmitted back to the server for storage. Since the present invention can automatically process the device image, it avoids the shortcomings of information loss caused by manual entry of device information, thereby solving the technical problem of low data storage efficiency of the distribution network, and further achieving the technical effect of improving the data storage efficiency of the distribution network.

[0047] The above method of this embodiment is further introduced below.

[0048] As an optional implementation method, the data storage method of the distribution network also includes: obtaining historical equipment images of the distribution network; extracting equipment image samples and corresponding equipment parameter samples from the historical equipment images; using the equipment image samples and the corresponding equipment parameter samples to train an initial image recognition model to obtain a target image recognition model.

[0049] In this embodiment, historical equipment images of the power distribution network are acquired; and equipment image samples and corresponding equipment parameter samples are extracted from the historical equipment images.

[0050] Optionally, after obtaining the device image samples and the corresponding device parameter samples, the device image samples and the corresponding device parameter samples are input into the initial image recognition model for training to obtain the target image recognition model. For example, the target image recognition model may be a DERT neural network (Detection Transformer), which is only an illustrative example and does not limit the specific content of the target image recognition model.

[0051] Optionally, the accuracy of identifying device parameters can be improved by performing AI training on a large number of images through image acquisition and DERT's neural network, and adding attention to key locations in the images.

[0052] As an optional implementation mode, before inputting the device image into the target image recognition model for analysis to obtain the device parameters of the device image, the method also includes: correcting the device image to obtain a corrected device image; and cutting out image characters from the corrected device image.

[0053] In this embodiment, the device image is corrected to obtain a corrected device image.

[0054] Optionally, correction processing can eliminate distortion, blur, distortion and other problems in the image, improve the clarity and detail display of the image, and make the image quality better.

[0055] Optionally, after obtaining the correction device image, the image characters are cut out from the correction device image.

[0056] As an optional implementation method, the device image is input into a target image recognition model for analysis to obtain device parameters of the device image, including: extracting character features of the device image, wherein the character features are used to indicate basic information of the device image; inputting the character features into the target image recognition model to obtain device parameters of the device image.

[0057] In this embodiment, character features of the device image are extracted, and then the character features are input into a target image recognition model to obtain device parameters of the device image.

[0058] Optionally, by identifying equipment parameters, an equipment file can be established to achieve tracking and management of the equipment, which is beneficial to the maintenance and care of the equipment.

[0059] As an optional implementation manner, inputting character features into a target image recognition model to obtain device parameters of the device image includes: using the target image recognition model to recognize character features to obtain device parameters of the device image.

[0060] In this embodiment, the target image recognition model is used to recognize the character features to obtain the device parameters of the device image. The target image recognition model may be a DERT neural network model.

[0061] Optionally, the target image recognition model can be used to automatically identify character features in the device image, eliminating the need to manually input device parameters one by one, saving time and labor costs.

[0062] As an optional implementation manner, the data storage method of the distribution network also includes: generating a data form by combining at least one device image and at least one device parameter; and integrating the data form to obtain a device database of the distribution network.

[0063] In this embodiment, at least one device image and at least one device parameter are automatically generated into a data form, and the data form is integrated to establish a device database of the distribution network, that is, to establish a basic ledger database.

[0064] Optionally, all equipment information in the distribution network is stored in a database for easy unified management and viewing.

[0065] It should be noted that the above embodiments may be executed by a data storage device of a power distribution network.

[0066] In this embodiment, the device image of the distribution network is obtained; the device image is input into the target image recognition model for analysis to obtain the device parameters of the device image, wherein the target image recognition model is obtained by training the device image samples and corresponding device parameter samples of the distribution network; the device parameters are classified to obtain the classification results; the device image and the corresponding classification results are transmitted back to the server for storage. In other words, the device image of the distribution network is input into the target image recognition model for analysis to obtain the device parameters of the device image, and then the device image and its corresponding device parameters are transmitted back to the server for storage. Since the present invention can automatically process the device image, it avoids the shortcomings of information loss caused by manual entry of device information, thereby solving the technical problem of low data storage efficiency of the distribution network, and further achieving the technical effect of improving the data storage efficiency of the distribution network.

[0067] Example 2

[0068] The technical solution of the embodiment of the present invention is illustrated below in conjunction with preferred implementation modes.

[0069] At present, the low-voltage distribution network, also known as the distribution transformer area, is the direct platform for the distribution system to provide services to users. The quality of its operation will have the most direct impact on the user's experience of electricity consumption. However, from the past to the present, the analysis and construction of the distribution network have mainly focused on the medium-voltage distribution network, ignoring the low-voltage distribution network, resulting in negligence in the management of the low-voltage distribution network, poor automation level, and frequent discrepancies between users, grid topology and ledgers, which seriously hinders the operation efficiency of the power grid.

[0070] In the related technologies, the problem of low-voltage distribution network ledger storage is mainly addressed by manual data entry. However, during the manual data entry process, the human intervention coefficient is too large, resulting in low data entry efficiency. At the same time, there are problems such as incorrect data entry, resulting in a high rate of missing data in the low-voltage distribution network ledger. Therefore, there is a technical problem of a high missing rate in the low-voltage distribution network ledger. At present, no effective solution has been proposed for the above-mentioned technical problem of a high missing rate in the low-voltage distribution network ledger.

[0071] However, an embodiment of the present invention proposes a method for automatically identifying low-voltage distribution network ledgers. A large number of historical low-voltage distribution network equipment images are trained through image acquisition and neural networks, an attention mechanism is added to the part of the image about equipment parameters, the equipment images are classified, the image parameters of the equipment images are identified, and the image parameters are returned to the database background in the form of command feedback, and a data form is automatically generated. Since the present invention can automatically realize the entry of low-voltage distribution network ledgers, the low efficiency of manual entry is avoided, thereby solving the technical problem of low efficiency of low-voltage distribution network ledger entry, and further achieving the technical effect of improving the efficiency of low-voltage distribution network ledger entry.

[0072] The embodiments of the present invention are further described below.

[0073] Figure 2 1 is a flow chart of a current low-voltage distribution network equipment verification method according to an embodiment of the present invention, the process comprising the following steps:

[0074] Step 201, manual on-site verification.

[0075] In this embodiment, the topology of the blank phase of the low-voltage power distribution network is manually checked on site.

[0076] Optionally, the topology of the blank stage refers to the missing information or data in the low-voltage distribution network equipment ledger, which needs to be filled by manual on-site verification to ensure the integrity and accuracy of the distribution network equipment information. Because the missing rate is large, it is impossible to rely solely on the existing ledger to solve the problem, and on-site verification is required to verify and improve the topological structure information of the distribution network equipment.

[0077] Alternatively, due to the high missing rate and high blankness of the low-voltage distribution network equipment ledger, the on-site verification link cannot be fundamentally solved, and the topology of the blank stage still requires manual on-site verification.

[0078] Step 202: Manually take photos and record.

[0079] In this embodiment, the staff takes photos and records the low-voltage distribution network equipment.

[0080] Step 203: Manually upload the archive.

[0081] In this embodiment, the image taken in step S202 is uploaded to the server for archiving.

[0082] Step 204, manual entry.

[0083] In this embodiment, the low-voltage distribution network equipment ledger is manually entered.

[0084] Figure 31 is a flow chart of a method for automatically identifying a low-voltage distribution network account according to an embodiment of the present invention, the process comprising the following steps:

[0085] Step 301: image acquisition.

[0086] In this embodiment, images of low-voltage distribution network equipment are collected.

[0087] Step 302: image processing.

[0088] In this embodiment, grayscale conversion and other processing are performed on the collected image.

[0089] Step 303: image recognition.

[0090] In this embodiment, recognition is performed on the processed image.

[0091] Step 304: output the result and generate a data form.

[0092] In this embodiment, the recognition result is output, and a data form is generated according to the recognition result.

[0093] Figure 4 1 is a flowchart of an image processing according to an embodiment of the present invention, the process includes the following steps:

[0094] Step 401: grayscale the image.

[0095] In this embodiment, the acquired image is grayed out.

[0096] Alternatively, image grayscale is the process of converting a color image into a grayscale image. In a grayscale image, each pixel has only one grayscale value, which represents the brightness of that pixel. Grayscale can help simplify the image processing process, reduce the amount of calculation, and make the image easier to analyze and process.

[0097] There are many ways to grayscale, the most commonly used method is to take the weighted average of the red, green and blue channels to calculate the grayscale value. The specific calculation formula is shown below:

[0098] Gray value = 0.299 × R + 0.587 × G + 0.114 × B

[0099] Among them, R, G, and B represent the pixel values ​​of the red, green, and blue channels respectively. Through the formula, each pixel in the color image can be converted into a grayscale value to obtain a grayscale image.

[0100] Optionally, the grayscale image no longer contains color information, but still retains the brightness information of the image, making the image easier to understand and process visually.

[0101] Step 402: binarization of the image.

[0102] In this embodiment, the grayscaled image is binarized.

[0103] Alternatively, image binarization is the process of converting the pixel values ​​in an image into only two values ​​(usually black and white). This processing method can simplify, segment, and extract specific information from an image.

[0104] Optionally, common image binarization methods include global threshold method, adaptive threshold method and histogram-based method. Among them, the global threshold method compares the pixel value of the entire image with a fixed threshold, and sets the pixels greater than the threshold to white, and the pixels less than the threshold to black. The adaptive threshold method determines the threshold of each pixel according to the local characteristics of the image, which can better handle images with uneven lighting or complex backgrounds.

[0105] Optionally, the applications of image binarization include character recognition, target detection, image segmentation and other fields, which can help us extract information from images more conveniently and perform subsequent processing and analysis.

[0106] Step 403: image denoising.

[0107] In this embodiment, denoising is performed on the binarized image.

[0108] Alternatively, image denoising refers to removing noise interference from an image through a series of methods and techniques to improve the quality and clarity of the image.

[0109] Optionally, common image denoising methods include: Mean filtering: Smoothing the image by calculating the average grayscale value of the neighborhood around the pixel to reduce the impact of noise. Median filtering: Taking the median of the neighborhood around the pixel as the value of the pixel, it has a good effect on removing noise points. Gaussian filtering: Using the Gaussian function to perform weighted averaging on the weights of the neighborhood around the pixel can effectively reduce high-frequency noise. Wavelet denoising: Using wavelet transform to decompose the image into frequency domain signals of different scales, and denoising is achieved by removing low-frequency and high-frequency noise components. Total variation denoising: Smoothing the image by minimizing the total variation of the image can effectively remove noise and maintain image details. Fourier transform denoising: Converting the image to the frequency domain for denoising, and removing noise frequency components by filtering.

[0110] Optionally, different image denoising methods are suitable for different types of noise and image characteristics, and a suitable method can be selected for denoising processing according to specific circumstances.

[0111] Step 404, tilt correction.

[0112] In this embodiment, the denoised image is subjected to tilt correction.

[0113] Optionally, image tilt correction is an image processing technique used to correct the tilt or rotation in an image to make the image more visually upright and accurate. This technique can be applied to various scenarios, such as scanning documents, photographing buildings, or photographing objects.

[0114] Optionally, image tilt correction is usually achieved through the following steps: Detecting the tilt: Using image processing algorithms or techniques to detect the tilt in the image. Calculating the correction angle: Based on the detected tilt, calculating the angle to be corrected. Correcting: Applying techniques such as rotation or affine transformation to correct the image to make it upright. Crop the edges: Crop the edges of the corrected image to remove excess blank parts. Outputting the corrected image: Outputting the corrected image and saving it or applying it to the required scene.

[0115] Optionally, through image tilt correction technology, the quality and accuracy of the image can be improved, making it more suitable for various application scenarios.

[0116] Step 405: character cutting.

[0117] In this embodiment, character cutting is performed on the image after the tilt correction.

[0118] Optionally, image character segmentation refers to separating characters or words in an image to facilitate text recognition or other related operations.

[0119] Optionally, character cutting generally includes the following steps: Character positioning: using image processing algorithms to find the position and boundaries of characters in the image and determine the size and shape of the characters. Character segmentation: segmenting characters according to their positions and boundaries, separating the characters from the image to form separate character images. Character recognition: performing text recognition on the segmented character images, identifying the content of the characters and performing subsequent processing.

[0120] Optionally, commonly used image character segmentation methods include pixel-based segmentation, connected region-based segmentation, edge detection-based segmentation, etc. These methods can be processed by selecting a suitable method according to specific needs and application scenarios.

[0121] Step 406: normalize the image.

[0122] In this embodiment, the cropped image is normalized.

[0123] Optionally, image normalization refers to scaling the pixel values ​​of an image to a specific range for better image processing and analysis. Common image normalization methods include minimum-maximum normalization and Z-score normalization.

[0124] Optionally, min-max normalization linearly maps the pixel values ​​of the image to a specified range, typically [0, 1].

[0125] Optionally, Z-score normalization converts the pixel values ​​of the image into a normal distribution with a mean of 0 and a standard deviation of 1. The specific calculation formula is as follows:

[0126] Optionally, by normalizing the image, the pixel value range of the image can be made consistent, which facilitates subsequent processing and analysis and can also reduce problems caused by different pixel value ranges.

[0127] Figure 5 is a schematic diagram of the basic structure of a DERT neural network according to an embodiment of the present invention, which is used to process natural language processing tasks. Different from the traditional Transformer model, the DERT network introduces a data enhancement mechanism to improve the performance and generalization ability of the model.

[0128] Input Embedding refers to the process of converting input data into a fixed-length vector representation. In neural networks, input embedding is often used to convert original discrete data (such as words, categories, etc.) into continuous vectors so that neural networks can better process and learn these data. Through input embedding, neural networks can map input data into a low-dimensional continuous vector space, thereby better capturing the relationship and semantic information between data.

[0129] Output Embedding is a technique used in neural networks to convert the final output into a vector representation. In neural networks, the output is usually a vector that represents the model's prediction of the input. In some tasks, we need to represent the output vector as a more meaningful embedding vector, which is the role of Output Embedding.

[0130] Optionally, Output Embedding can help the neural network learn better representations, making the output easier to understand and apply. By using output embedding technology, we can map the output vector into a more semantic space, making the model's predictions more accurate and explainable.

[0131] Multi-Head Attention is an attention mechanism used in neural networks to process sequence data. In traditional attention mechanisms, the model calculates an attention distribution to determine the importance of each position in the input sequence. In Multi-Head Attention, the model calculates multiple different attention distributions at the same time and then merges them together. The benefit of this is that it allows the model to learn different aspects of attention, thereby improving the performance and generalization ability of the model.

[0132] In Multi-Head Attention, the input sequence is processed by multiple different attention heads, and each head learns different focus points. Finally, the outputs of these heads are concatenated together and a linear transformation is performed to obtain the final output. This allows the model to focus on the input sequence from different perspectives and better capture the information in the sequence.

[0133] Add&Norm is a neural network layer structure that implements residual connections and layer normalization in the network. In Add&Norm, the input is first added to the output of the network through a residual connection, and then the added result is layer normalized. This can help the network learn and optimize more easily, improve training speed and model performance. The Add&Norm structure is often used in deep neural networks, especially in Transformer models and other self-attention models.

[0134] Feed Forward refers to the process of passing data from the input layer to the hidden layer and then to the output layer. In this process, data can only flow in one direction, from the input layer to the output layer, and there will be no loops or feedback. The feed forward process of a neural network usually includes the following steps: The input layer receives the raw data and passes the data to the hidden layer. The hidden layer performs a weighted summation on the received data and applies an activation function to generate the output. The output layer receives the output of the hidden layer and performs a similar weighted summation and activation function. The final output result is used for prediction or classification.

[0135] When training a neural network, the feedforward process is the process of performing matrix multiplication and activation function operations between input data and weights. By continuously adjusting the weights and biases of the network, the network can generate more accurate output results during the feedforward process. The feedforward process is one of the most basic computational processes in a neural network and is also the basis for realizing the prediction and classification functions of a neural network.

[0136] Alternatively, Masked Multi-Head Attention is an attention mechanism commonly used in neural networks to process sequence data. In this attention mechanism, the input sequence is divided into multiple heads, each head calculates a set of attention weights, and then these weights are weighted summed to obtain the final representation.

[0137] Optionally, Masked Multi-Head Attention introduces a mask to limit the model to only consider information before the current position when calculating the attention weight, thereby preventing the model from seeing future information when predicting. This is very important for processing time dependencies in sequence data and can help the model better capture long-range dependencies in the sequence.

[0138] The role of Positional Encoding in neural networks is to help the model learn the position information of words in the input sequence. In natural language processing tasks, the order of words in the input sequence is usually very important, and traditional neural networks cannot directly capture the position information of words. Therefore, the introduction of Positional Encoding can provide the model with information about the position of words in the sequence, thereby helping the model better understand the structure and semantics of the input sequence. By combining position information with the word vector of the word, Positional Encoding can help the model better understand the order and contextual relationship of words in the input sequence, and improve the performance of the model when processing natural language tasks.

[0139] Softmax is a commonly used activation function, usually used in the last layer in multi-classification problems. It converts the raw scores output by the neural network into a probability distribution, so that the probability value of each category is between 0 and 1, and the sum of the probabilities of all categories is 1. This makes it easy to interpret and compare the output results.

[0140] Optionally, the Softmax function is usually used as the activation function of the output layer to convert the output of the last layer of the neural network into a probability distribution, and then make classification predictions based on the probability values. The derivative of the Softmax function is also easy to calculate and can be conveniently used in the back-propagation algorithm for parameter updates. The Softmax function plays an important role in deep learning and is a key component in achieving multi-classification tasks.

[0141] Linear usually refers to linear functions in neural networks, also called fully connected layers or affine transformations. In neural networks, linear functions are usually used to connect hidden layers and output layers. In the hidden layer, linear functions are used to weight and sum input features to generate new feature representations; in the output layer, linear functions are used to map the feature representations of the hidden layer to the output categories. By stacking multiple linear functions and nonlinear activation functions, neural networks can learn complex nonlinear relationships and achieve more powerful fitting capabilities.

[0142] In this embodiment, by combining AI automatic image recognition technology and communication technology, the cumbersome problem of ledger entry in the current low-voltage distribution network is reduced, the entry steps are reduced, the low-voltage ledger information is quickly identified and uploaded, and the current low-voltage transparency and digitalization of the power grid are accelerated.

[0143] Optionally, the present invention uses image acquisition and DERT neural networks to perform AI training on a large number of images, add attention to key positions on the images, and identify and classify image parameters. The classified and identified parameters are transmitted back to the background in the form of command feedback, and data forms are automatically generated to solve the difficulties in the entry of low-voltage equipment ledgers in the prior art. At the same time, AI training can replace the shortcomings of slow manual entry and lack of accuracy in information transmission, and can realize rapid and accurate identification of equipment ledgers, and can summarize ledger information, so as to quickly establish a basic ledger database.

[0144] In this embodiment, a large number of historical low-voltage distribution network equipment images are trained through image acquisition and neural networks, an attention mechanism is added to the part of the image related to equipment parameters, the equipment images are classified, the image parameters of the equipment images are identified, and the image parameters are returned to the database background in the form of command feedback, and a data form is automatically generated. Since the present invention can automatically realize the entry of the ledger of the low-voltage distribution network, the low efficiency of manual entry is avoided, thereby solving the technical problem of low efficiency of low-voltage distribution network ledger entry, and further achieving the technical effect of improving the efficiency of low-voltage distribution network ledger entry.

[0145] Example 3

[0146] According to an embodiment of the present invention, a data storage device for a distribution network is also provided. It should be noted that the data storage device for a distribution network can be used to execute the data storage method for a distribution network in Embodiment 1.

[0147] Figure 6 Schematic diagram of a data storage device for a distribution network according to an embodiment of the present invention. Figure 6As shown, the data storage device 600 of the power distribution network may include: an acquisition unit 601 , an analysis unit 602 , a classification unit 603 and a storage unit 604 .

[0148] The acquisition unit 601 is used to acquire the equipment image of the power distribution network.

[0149] The analysis unit 602 is used to input the device image into the target image recognition model for analysis to obtain the device parameters of the device image, wherein the target image recognition model is trained by the device image samples and corresponding device parameter samples of the distribution network.

[0150] The classification unit 603 is used to classify the device parameters and obtain classification results.

[0151] The storage unit 604 is used to transmit the device image and the corresponding classification result back to the server for storage.

[0152] Optionally, the data storage device 600 of the distribution network may further include: a first acquisition unit, used to acquire historical equipment images of the distribution network; an extraction unit, used to extract equipment image samples and corresponding equipment parameter samples from the historical equipment images; and a training unit, used to train an initial image recognition model using equipment image samples and corresponding equipment parameter samples to obtain a target image recognition model.

[0153] Optionally, the data storage device 600 of the power distribution network may further include: a correction unit, used to perform correction processing on the device image to obtain a corrected device image; and a cutting unit, used to cut out image characters from the corrected device image.

[0154] Optionally, the analysis unit 602 may include: an extraction module for extracting character features of the device image, wherein the character features are used to indicate basic information of the device image; and an input module for inputting the character features into a target image recognition model to obtain device parameters of the device image.

[0155] Optionally, the input module may include: a recognition subunit, which is used to recognize character features using a target image recognition model to obtain device parameters of the device image.

[0156] Optionally, the data storage device 600 of the distribution network may further include: a generation unit for generating a data form from at least one device image and at least one device parameter; and an integration unit for integrating the data form to obtain a device database of the distribution network.

[0157] In this embodiment, the device image of the distribution network is obtained; the device image is input into the target image recognition model for analysis to obtain the device parameters of the device image, wherein the target image recognition model is obtained by training the device image samples and corresponding device parameter samples of the distribution network; the device parameters are classified to obtain the classification results; the device image and the corresponding classification results are transmitted back to the server for storage. In other words, the device image of the distribution network is input into the target image recognition model for analysis to obtain the device parameters of the device image, and then the device image and its corresponding device parameters are transmitted back to the server for storage. Since the present invention can automatically process the device image, it avoids the shortcomings of information loss caused by manual entry of device information, thereby solving the technical problem of low data storage efficiency of the distribution network, and further achieving the technical effect of improving the data storage efficiency of the distribution network.

[0158] Example 4

[0159] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0160] Example 5

[0161] An embodiment of the present application provides a processor, and the processor is used to run a program, wherein the program executes and implements the methods in various embodiments of the present invention when it is run.

[0162] Example 6

[0163] The embodiments of the present application further provide a computer program product, including computer instructions, which implement the methods in various embodiments of the present invention when executed by a processor.

[0164] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0165] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0166] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0167] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0168] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0169] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent functional component, it 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 all or part of the technical solution can be embodied in the form of a software functional component, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc., which can store program codes.

[0170] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A data storage method for a distribution network, characterized in that: include: Obtain equipment images of the distribution network; Inputting the device image into a target image recognition model for analysis to obtain device parameters of the device image, wherein the target image recognition model is trained by device image samples and corresponding device parameter samples of the power distribution network; Classifying the device parameters to obtain classification results; The device image and the corresponding classification result are transmitted back to the server for storage.

2. The method according to claim 1, characterized in that The method further comprises: Acquiring historical equipment images of the power distribution network; Extracting the device image sample and the corresponding device parameter sample from the historical device image; The initial image recognition model is trained using the device image samples and the corresponding device parameter samples to obtain the target image recognition model.

3. The method according to claim 1, characterized in that Before inputting the device image into the target image recognition model for analysis to obtain the device parameters of the device image, the method further includes: Performing correction processing on the device image to obtain a corrected device image; Cut out the image characters from the correction device image.

4. The method according to claim 1, characterized in that: The device image is input into the target image recognition model for analysis to obtain device parameters of the device image, including: Extracting character features of the device image, wherein the character features are used to indicate basic information of the device image; The character features are input into the target image recognition model to obtain the device parameters of the device image.

5. The method according to claim 4, characterized in that Inputting the character features into the target image recognition model to obtain the device parameters of the device image includes: The target image recognition model is used to recognize the character features to obtain the device parameters of the device image.

6. The method according to claim 1, characterized in that The method further comprises: Generate a data form using at least one of the device images and at least one of the device parameters; The data tables are integrated to obtain a device database of the distribution network.

7. A data storage device for a distribution network, characterized in that: include: An acquisition unit, used for acquiring equipment images of the distribution network; An analysis unit, used for inputting the device image into a target image recognition model for analysis to obtain device parameters of the device image, wherein the target image recognition model is trained by device image samples and corresponding device parameter samples of the power distribution network; A classification unit, used to classify the device parameters and obtain a classification result; The storage unit is used to transmit the device image and the corresponding classification result back to the server for storage.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed by a processor, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

9. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 6 when running.

10. A computer program product, characterized in that The computer program product comprises computer instructions, which implement the method according to any one of claims 1 to 6 when executed by a processor.

Images