Power supply station material management optimization analysis method

Abstract

The invention discloses a power supply station material management optimization analysis method, comprising the following analysis steps: S1, collecting warehouse stock data to obtain warehouse management data and material image data information; s2, carrying out geometric classification on the material image data based on Gaussian distribution; s3, fusing the classified material image data, and outputting image data information; s4, constructing an electric power data information association model, and analyzing the relationship among various storage data; s5, inputting the data information after correlation analysis into an improved YOLOv8 algorithm, analyzing the warehouse stock data, and outputting different electric power warehouse data management and control information; according to the invention, material management of the power supply station can be optimized, manual intervention is greatly reduced, the management precision and efficiency are improved, and intelligent warehouse management is realized.

Description

Technical Field

[0001] This invention belongs to the field of power supply station material management technology, and specifically relates to an optimization analysis method for power supply station material management. Background Technology

[0002] A power supply station is a branch office set up by a power company. It is the end-point agency that directly contacts power users. Its functions include implementing power policies and regulations, carrying out power grid inspection and maintenance, managing electricity metering devices, providing guidance on electricity safety, and developing the electricity market. In practice, the power supply station receives power materials purchased from higher authorities and uses these materials to carry out inspection, maintenance, and emergency repair work based on the operating status and fault conditions of the lines and equipment, thereby improving the reliability of power supply.

[0003] Because power storage and inventory management involves a variety of data and information, with diverse data volumes and types, data collection and analysis are quite difficult. Traditional power supply station storage and warehouse management methods have many drawbacks and cannot adapt to this situation, easily leading to problems such as inaccurate data and low operational efficiency.

[0004] Therefore, in order to solve the above problems, it is necessary to develop an optimization analysis method for the material management of power supply stations. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for optimizing the management of materials in power supply stations. This method optimizes the management of materials in power supply stations, significantly reduces manual intervention, improves management accuracy and efficiency, and realizes intelligent warehouse management.

[0006] The objective of this invention is achieved as follows: a method for optimizing the management of materials in power supply stations, comprising the following analytical steps:

[0007] S1. Based on the power supply station's material management system and the image acquisition equipment installed in the power supply station's warehouse, the warehouse inventory data is collected to obtain warehouse management data and material image data information;

[0008] S2. Geometric classification of material image data based on Gaussian distribution;

[0009] S3. Merge the classified material image data and output the image data information;

[0010] S4. Construct a power data information association model to analyze the relationships between various storage data;

[0011] S5. Input the data information after correlation analysis into the improved YOLOv8 algorithm to analyze the warehouse inventory data and output different power warehouse data management information.

[0012] Furthermore, in step S2, the geometric classification of the material image data includes smooth patches and non-smooth patches, and a two-dimensional Gaussian blur function is used for smooth classification, specifically as follows: In the formula: This represents the distance from the origin on the horizontal axis. This represents the distance from the origin on the vertical axis. It is the standard deviation of the Gaussian distribution.

[0013] Furthermore, the non-smooth surface includes random surface and main direction surface, and a threshold is calculated. Distinguish between random patches and principal orientation patches in an image. The smaller the value, the more random the image vector. In the formula: It is the diagonal of the principal direction vector in the gradient value decomposition of an image patch. The gradient value calculation for the image patch matrix includes: calculating the gradient of each image pixel. , Depend on and Coordinate gradient and Composition, then image vector ( Each pixel in the source image (numbered as the source image number) The gradient value is expressed as: The gradient values ​​of each image patch are decomposed as follows: In the formula: for Gradient value decomposition.

[0014] Furthermore, step S3 specifically includes obtaining high-frequency components based on Max-L1 fusion rules. Low-frequency components are obtained based on weighted average fusion rules. The final fused image .

[0015] Furthermore, the step S3, which obtains the high-frequency components based on the Max-L1 fusion rule, specifically includes: assuming the input image is segmented into... For each image patch, the OMP algorithm is used to process it. Perform sparse coding, dividing each image patch Encoding as sparse coefficients Then, the coefficients are fused using the Max-L1 fusion rule, specifically as follows: In the formula: This represents the fusion vector of the corresponding source image patch. This represents the sparsity coefficient corresponding to each image patch. This represents the dictionary trained using the OMP algorithm. This represents the function value of the OMP algorithm.

[0016] Furthermore, in step S3, the low-frequency component is obtained based on the weighted average fusion rule. Specifically, the steps include: First, calculating the L2 norm of the corresponding high-frequency components. Let the constraint matrix A have full rank. The formula for minimizing the L2 norm is expressed as: ,in, Indicates the L2 norm value. To provide parameters, we introduce Lagrange multipliers. : , Let Lagrange multipliers be the introduced factors. Then the fused low-frequency components are represented as follows: .

[0017] Furthermore, step S4 includes the following steps: constructing an input data model. Extract warehouse inventory management data information , exist The correlation matrix is ​​used to determine the relationships between different entities and to formulate corresponding warehousing and inventory management plans. Represented as: Warehouse inventory data information relationship matrix Represented as: In the formula: This represents multiple data sets appearing in warehouse inventory. Indicates the number of data types. This represents a collection of data information on warehouse inventory types, quantities, inbound / outbound frequencies, orders, product names, storage locations, warehouses, and batch numbers. This indicates information that affects data collection when collecting warehouse inventory data. This indicates the number of measurements taken to obtain updated warehouse management data samples or image data samples. , The relationship between them is represented as: , Represent matrices respectively and A collection of individual elements.

[0018] Furthermore, the improved YOLOv8 algorithm model structure in step S5 includes a backbone network, a neck network, and a detection head. The backbone network adopts an architecture combining convolutional neural networks and Transformers, and the neck network adopts a feature pyramid network that introduces channel attention and spatial attention mechanisms.

[0019] Furthermore, the backbone network specifically includes Stem, a downsampling module, and RepVitSEBlock; given an input... Then Stem is represented as follows: In the formula: This represents a convolution with a kernel size of 3 and a stride of 2. This represents the GELU activation function; the downsampling module consists of RepVitBlock and FFN, specifically represented as follows: In the formula: This represents a regular convolution with a kernel size of 1. This represents a depthwise convolution with a kernel size of 3. In the formula: RepVitSEBlock represents a depthwise convolution with a kernel size of 1; In the formula: This represents the channel attention mechanism; where, , This represents the activation function.

[0020] Furthermore, the neck network includes multiple fusion modules, which are represented as follows: In the formula: This indicates a splicing operation. This indicates that the concatenated features are separated into two features. , express Reparameterized convolution.

[0021] Due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0022] (1) By performing geometric classification of material image data based on Gaussian distribution and fusing the classified material image data, the collected images are processed by hierarchical image fusion, which effectively improves the accuracy of subsequent model recognition.

[0023] (2) By constructing a power data information association model, based on the extracted various warehouse inventory data information, the correlation is calculated, and the relationship between various warehouse data is analyzed, which facilitates the formulation of appropriate warehouse inventory management and control schemes.

[0024] (3) By using the improved YOLOv8 algorithm to analyze the warehouse inventory data, output different power warehouse data management information, improve management accuracy and efficiency, and realize intelligent warehouse management. Attached Figure Description

[0025] Figure 1 This is a flowchart of the present invention.

[0026] Figure 2 This is a structural diagram of the improved YOLOv8 algorithm model in step S5 of this invention.

[0027] Figure 3 This is the backbone network structure diagram in the improved YOLOv8 algorithm of step S5 of the present invention.

[0028] Figure 4 This is a structural diagram of the neck network fusion module in the improved YOLOv8 algorithm of step S5 of the present invention. Detailed Implementation

[0029] The technical solution of the present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0030] like Figure 1 , Figure 2 , Figure 3 , Figure 4 As shown, a method for optimizing the management of materials in power supply stations includes the following analytical steps:

[0031] S1. Based on the power supply station's material management system and the image acquisition equipment installed in the power supply station's warehouse, the warehouse inventory data is collected to obtain warehouse management data and material image data information.

[0032] S2. Perform geometric classification of material image data based on Gaussian distribution.

[0033] Preferably, in step S2, the geometric classification of the material image data includes smooth patches and non-smooth patches, and a two-dimensional Gaussian blur function is used for smooth classification, specifically as follows: In the formula: This represents the distance from the origin on the horizontal axis. This represents the distance from the origin on the vertical axis. It is the standard deviation of the Gaussian distribution.

[0034] Preferably, the non-smooth surface patches include random surface patches and main direction surface patches, determined by calculating a threshold. Distinguish between random patches and principal orientation patches in an image. The smaller the value, the more random the image vector. In the formula: It is the diagonal of the principal direction vector in the gradient value decomposition of an image patch. The gradient value calculation for the image patch matrix includes: calculating the gradient of each image pixel. , Depend on and Coordinate gradient and Composition, then image vector ( Each pixel in the source image (numbered as the source image number) The gradient value is expressed as: The gradient values ​​of each image patch are decomposed as follows: In the formula: for Gradient value decomposition.

[0035] S3. Merge the classified material image data and output the image data information.

[0036] Preferably, step S3 specifically includes obtaining high-frequency components based on Max-L1 fusion rules. Low-frequency components are obtained based on weighted average fusion rules. The final fused image .

[0037] Preferably, obtaining high-frequency components based on the Max-L1 fusion rule specifically includes: assuming the input image is segmented into... For each image patch, the OMP algorithm is used to process it. Perform sparse coding, dividing each image patch Encoding as sparse coefficients Then, the coefficients are fused using the Max-L1 fusion rule, specifically as follows: In the formula: This represents the fusion vector of the corresponding source image patch. This represents the sparsity coefficient corresponding to each image patch. This represents the dictionary trained using the OMP algorithm. This represents the function value of the OMP algorithm.

[0038] Preferably, the low-frequency component is obtained based on the weighted average fusion rule. Specifically, the steps include: First, calculating the L2 norm of the corresponding high-frequency components. Let the constraint matrix A have full rank. The formula for minimizing the L2 norm is expressed as: ,in, Indicates the L2 norm value. To provide parameters, we introduce Lagrange multipliers. : , Let Lagrange multipliers be the introduced factors. Then the fused low-frequency components are represented as follows: .

[0039] S4. Construct a power data information association model to analyze the relationships between various storage data.

[0040] Preferably, step S4 includes the following steps: constructing an input data model. Extract warehouse inventory management data information , exist The correlation matrix is ​​used to determine the relationships between different entities and to formulate corresponding warehousing and inventory management plans. Represented as: Warehouse inventory data information relationship matrix Represented as: In the formula: This represents multiple data sets appearing in warehouse inventory. Indicates the number of data types. This represents a collection of data information on warehouse inventory types, quantities, inbound / outbound frequencies, orders, product names, storage locations, warehouses, and batch numbers. This indicates information that affects data collection when collecting warehouse inventory data. This indicates the number of measurements taken to obtain updated warehouse management data samples or image data samples. , The relationship between them is represented as: , Represent matrices respectively and A collection of individual elements.

[0041] S5. Input the data information after correlation analysis into the improved YOLOv8 algorithm to analyze the warehouse inventory data and output different power warehouse data management information.

[0042] Preferably, the improved YOLOv8 algorithm model structure in step S5 includes a backbone network, a neck network, and a detection head. The backbone network adopts an architecture combining convolutional neural networks and Transformers, and the neck network adopts a feature pyramid network that incorporates channel attention and spatial attention mechanisms.

[0043] Preferably, the backbone network specifically includes Stem, a downsampling module, and RepVitSEBlock; given an input Then Stem is represented as follows: In the formula: This represents a convolution with a kernel size of 3 and a stride of 2. This represents the GELU activation function; the downsampling module consists of RepVitBlock and FFN, specifically represented as follows: In the formula: This represents a regular convolution with a kernel size of 1. This represents a depthwise convolution with a kernel size of 3. In the formula: RepVitSEBlock represents a depthwise convolution with a kernel size of 1; In the formula: This represents the channel attention mechanism; where, , This represents the activation function.

[0044] Preferably, the neck network includes multiple fusion modules, which are represented as follows: In the formula: This indicates a splicing operation. This indicates that the concatenated features are separated into two features. , express Reparameterized convolution.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for optimizing the management of materials in power supply stations, characterized in that: The analysis includes the following steps: S1. Based on the power supply station's material management system and the image acquisition equipment installed in the power supply station's warehouse, the warehouse inventory data is collected to obtain warehouse management data and material image data information; S2. Geometric classification of material image data based on Gaussian distribution; S3. Merge the classified material image data and output the image data information; S4. Construct a power data information association model to analyze the relationships between various storage data; S5. Input the data information after correlation analysis into the improved YOLOv8 algorithm to analyze the warehouse inventory data and output different power warehouse data management information.

2. The method for optimizing the management of materials in a power supply station according to claim 1, characterized in that: In step S2, the geometric classification of the material image data includes smooth patches and non-smooth patches. A two-dimensional Gaussian blur function is used for smooth classification, specifically as follows: In the formula: This represents the distance from the origin on the horizontal axis. This represents the distance from the origin on the vertical axis. It is the standard deviation of the Gaussian distribution.

3. The method for optimizing the management of materials in a power supply station according to claim 2, characterized in that: The non-smooth surface patches include random surface patches and main direction surface patches, and a threshold is calculated. Distinguish between random patches and principal orientation patches in an image. The smaller the value, the more random the image vector. In the formula: It is the diagonal of the principal direction vector in the gradient value decomposition of an image patch. The gradient value calculation for the image patch matrix includes: calculating the gradient of each image pixel. , Depend on and Coordinate gradient and Composition, then image vector ( Each pixel in the source image (numbered as the source image number) The gradient value is expressed as: The gradient values ​​of each image patch are decomposed as follows: In the formula: for Gradient value decomposition.

4. The method for optimizing the management of materials in a power supply station according to claim 1, characterized in that: Step S3 specifically includes obtaining high-frequency components based on Max-L1 fusion rules. Low-frequency components are obtained based on weighted average fusion rules. The final fused image .

5. The method for optimizing the management of materials in a power supply station according to claim 4, characterized in that: The step S3, which obtains high-frequency components based on the Max-L1 fusion rule, specifically includes: assuming the input image is segmented into... For each image patch, the OMP algorithm is used to process it. Perform sparse coding, dividing each image patch Encoding as sparse coefficients Then, the coefficients are fused using the Max-L1 fusion rule, specifically as follows: In the formula: This represents the fusion vector of the corresponding source image patch. This represents the sparsity coefficient corresponding to each image patch. This represents the dictionary trained using the OMP algorithm. This represents the function value of the OMP algorithm.

6. The method for optimizing the management of materials in a power supply station according to claim 4, characterized in that: In step S3, the low-frequency component is obtained based on the weighted average fusion rule. Specifically, the steps include: First, calculating the L2 norm of the corresponding high-frequency components. Let the constraint matrix A have full rank. The formula for minimizing the L2 norm is expressed as: ,in, Indicates the L2 norm value. To provide parameters, we introduce Lagrange multipliers. : , Let Lagrange multipliers be the introduced factors. Then the fused low-frequency components are represented as follows: .

7. The method for optimizing the management of materials in a power supply station according to claim 1, characterized in that: Step S4 includes the following steps: constructing an input data model. Extract warehouse inventory management data information , exist The correlation matrix is ​​used to determine the relationships between different entities and to formulate corresponding warehousing and inventory management plans. Represented as: Warehouse inventory data information relationship matrix Represented as: In the formula: This represents multiple data sets appearing in warehouse inventory. Indicates the number of data types. This represents a collection of data information on warehouse inventory types, quantities, inbound / outbound frequencies, orders, product names, storage locations, warehouses, and batch numbers. This indicates information that affects data collection when collecting warehouse inventory data. This indicates the number of measurements taken to obtain updated warehouse management data samples or image data samples. , The relationship between them is represented as: , Represent matrices respectively and A collection of individual elements.

8. The method for optimizing the management of materials in a power supply station according to claim 1, characterized in that: The improved YOLOv8 algorithm model structure in step S5 includes a backbone network, a neck network, and a detection head. The backbone network adopts an architecture that combines convolutional neural networks and Transformers, and the neck network adopts a feature pyramid network that introduces channel attention and spatial attention mechanisms.

9. The method for optimizing the management of materials in a power supply station according to claim 8, characterized in that: The backbone network specifically includes Stem, a downsampling module, and RepVitSEBlock; given an input... Then Stem is represented as follows: In the formula: This represents a convolution with a kernel size of 3 and a stride of 2. This represents the GELU activation function; the downsampling module consists of RepVitBlock and FFN, specifically represented as follows: In the formula: This represents a regular convolution with a kernel size of 1. This represents a depthwise convolution with a kernel size of 3. In the formula: RepVitSEBlock represents a depthwise convolution with a kernel size of 1; In the formula: This represents the channel attention mechanism; where, , This represents the activation function.

10. The method for optimizing the management of materials in a power supply station according to claim 8, characterized in that: The neck network includes multiple fusion modules, which are represented as follows: In the formula: This indicates a splicing operation. This indicates that the concatenated features are separated into two features. , express Reparameterized convolution.

Images