Electric power material demand response method and device

Abstract

The invention discloses an electric power material demand response method and apparatus. The method comprises the steps of obtaining a demand influence file of electric power materials; according to the file type of the demand influence file, matching different file processing strategies, and performing structured processing on the demand influence file to obtain material demand data; according to a pre-constructed material demand generation model, fusing an electric power multivariate knowledge graph, and performing auditing processing on the material demand data to obtain a first material demand file; and performing difference detection on the first material demand file in combination with a material demand standard, establishing association between the first material demand file and a standard term corresponding to the material demand standard, and giving a result of responding to the electric power material demand. The first material demand file is detected in combination with the material demand standard, the result responding to the electric power material demand is obtained, scientificity and reasonability of the electric power material demand are guaranteed, the accuracy of the result responding to the electric power material demand is improved, and reference and help are provided for the electric power material demand.

Description

Technical Field

[0001] The present invention belongs to the technical field of power material management, and in particular relates to a method and device for responding to power material demand. Background Art

[0002] Power supplies refer to the various materials required for the construction, operation, and maintenance of power systems. A power supply plan, responding to power supply demand, pre-arranges the types, quantities, procurement schedules, and supply channels of required power supplies based on the power system's development, operational requirements, and maintenance. A sound power supply plan ensures the smooth construction, operation, and maintenance of power systems, avoids material shortages or backlogs, reduces costs, and improves the reliability and economic efficiency of power systems.

[0003] Currently, the demand for power supplies during power grid construction is large in volume and diverse in variety. Grid material planning primarily relies on approval, aggregation, and on-site surveys by higher-level organizations based on research, statistics, estimates, and reports from lower-level organizations. This ultimately generates an overall material demand plan. This process is not only enormously labor-intensive and resource-intensive, but also requires a significant review workload and a lengthy demand generation cycle. Furthermore, irregular reporting and system input errors during this process can lead to inaccurate material demand plans, hindering procurement planning, project construction, and production scheduling. This also hinders the improvement of lean management capabilities for power supplies.

[0004] Patent application CN120409766A provides a method, system, device, and medium for forecasting power material demand, including: collecting historical power material demand data and influencing factor data, and preprocessing it through cleaning, transformation, and feature extraction; building a deep learning model based on the TensorFlow deep learning framework; using the preprocessed historical data to train the deep learning model, adjust model parameters, and evaluate performance to obtain a final forecasting model; collecting real-time power material demand data and influencing factor data, preprocessing it, and then inputting it into the final forecasting model to forecast power material demand. The deep learning model is constructed using the TensorFlow deep learning framework and combined with the processed historical power material demand data to predict power material demand data, improving forecast accuracy.

[0005] In the above-mentioned related technologies, deep learning models are used to predict material demand. However, responding to the demand for power materials involves many aspects and needs to be modified and adjusted in a timely manner according to actual development conditions. How to ensure the scientificity and rationality of the demand for power materials, improve the accuracy of the demand for power materials, and provide reference and help for the demand for power materials are the problems that need to be solved at present. Summary of the Invention

[0006] In response to the defects in the above-mentioned prior art, the present invention provides a method and device for responding to the demand for electric power materials, the method comprising: obtaining a demand-influencing file for electric power materials; matching different file processing strategies according to the file type of the demand-influencing file, performing structured processing on the demand-influencing file, and obtaining material demand data; based on a pre-built material demand generation model, integrating the electric power multivariate knowledge graph, reviewing and processing the material demand data, and obtaining a first material demand file; combining the material demand standard, performing difference detection on the first material demand file, and establishing an association between the first material demand file and the standard clauses corresponding to the material demand standard, and providing a result of responding to the demand for electric power materials. By processing the demand-influencing file, combining the material demand generation model, obtaining the first material demand file and testing it, and obtaining the result of responding to the demand for electric power materials, the first material demand file is modified and adjusted in a timely manner according to the actual material demand standard, ensuring the scientificity and rationality of the demand for electric power materials, improving the accuracy of the demand for electric power materials, and providing reference and help for the demand for electric power materials.

[0007] In a first aspect, the present invention provides a method for responding to power material demand, specifically comprising the following steps: Obtain demand impact documents for power supplies; According to the file type of demand-affecting documents, different file processing strategies are matched, and demand-affecting documents are structured to obtain material demand data; Based on the pre-built material demand generation model and the integration of the power multi-knowledge graph, the material demand data is reviewed and processed to obtain the first material demand document; In combination with the material demand standard, a difference detection is performed on the first material demand document, and the first material demand document is associated with the standard clauses corresponding to the material demand standard to provide a result that responds to the power material demand.

[0008] Further, the file type includes at least one of text, table, and image; According to the file type of the demand-affecting file, different file processing strategies are matched, and the demand-affecting file is structured to obtain material demand data, including: If the file type of the demand impact document is text, the demand impact document is segmented and tagged with parts of speech, and converted into structured data to obtain the corresponding material demand data; If the file type of the demand impact file is a table, perform structural analysis on the demand impact file, analyze the relationship between the content and position of each cell in the demand impact file, extract the content of the demand impact file and obtain the corresponding material demand data; If the file type of the demand impact document is an image, the text in the demand impact document is enhanced and recognized, and converted into structured data to provide the corresponding material demand data.

[0009] Furthermore, the relationship between the content and position of each cell in the demand impact file is expressed by the cell association degree, specifically:

[0010] Among them, R(c i ,c j ) is the cell association degree between the i-th cell and the j-th cell, α is the association weight coefficient, PosSim(c i ,c j ) is the position similarity between the i-th cell and the j-th cell, ContSim(c i ,c j ) is the content similarity between the i-th cell and the j-th cell.

[0011] Furthermore, based on the pre-built material demand generation model, the power multivariate knowledge graph is integrated to review and process the material demand data to obtain the first material demand file, which is obtained through the following steps: According to the sequence length of material demand data, the rotation position coding extrapolation technology is used to expand the processing window of the material demand generation model; Analyze the correlation similarity between each demand vector of the material demand data and each material vector in the power multivariate knowledge graph, and adjust the attention weight in the attention mechanism of the material demand generation model. The power multivariate knowledge graph includes the material vectors of each power material. Each material vector of each power material includes the technical parameters, coding rules, and procurement specifications of the corresponding power material in the power industry standard. Substitute the material demand data into the adjusted material demand generation model to obtain a first material demand file.

[0012] Furthermore, the attention weight is specifically expressed as:

[0013] Among them, A k is the attention weight corresponding to the k-th material vector in the electric power multivariate knowledge graph, S k is the semantic similarity between the kth material vector and the material demand data in the electric power multivariate knowledge graph, λ k is the important coefficient of the kth material vector in the electric power multivariate knowledge graph, and M is the total number of material vectors in the electric power multivariate knowledge graph.

[0014] Furthermore, the required reward function is specifically expressed as:

[0015] Among them, β1 is the weight coefficient corresponding to the demand basic reward, β2 is the weight coefficient corresponding to the demand incremental reward, β3 is the weight coefficient corresponding to the demand penalty term, R base is the demand-based reward, R inc is the demand incremental reward, P i is the demand penalty item for the i-th violation.

[0016] Furthermore, the material demand generation model includes a demand input layer, an attention mechanism, a demand reward layer, and a demand output layer; The attention mechanism is connected to the demand input layer. Material demand data enters the attention mechanism through the demand input layer to extract the core demand features in the material demand data. The demand output layer is connected to the attention mechanism to generate intermediate demand documents based on the core demand features; The demand reward layer is connected to the demand output layer, analyzes the compliance status of the intermediate demand documents, obtains the total demand reward, and feeds it back to the attention mechanism and demand output layer; Among them, the demand reward layer includes the basic reward layer, the incremental reward layer, the violation penalty layer and the reward fusion layer; The basic reward layer is connected to the demand output layer to analyze the matching degree between the intermediate demand documents and the corresponding procurement specifications to obtain the basic reward for the intermediate demand; The incremental reward layer is connected to the demand output layer to analyze the fluctuation difference between each value in the intermediate demand file and the corresponding standard value to obtain the incremental reward of the intermediate demand; The violation penalty layer is connected to the demand output layer to analyze the violations between the data in the intermediate demand file and give the intermediate demand penalty items; The reward fusion layer is connected to the basic reward layer, incremental reward layer, and violation penalty layer respectively. It is used to fuse the intermediate demand basic reward, intermediate demand incremental reward, and intermediate demand penalty items to obtain the total demand reward and provide feedback to the attention mechanism and demand output layer.

[0017] Furthermore, the construction of the material demand generation model is obtained through the following steps: Obtain historical material demand data; Analyze the matching degree between historical material demand data and corresponding procurement specifications to obtain demand-based rewards; Analyze the fluctuation difference between each value in the historical material demand data and the corresponding standard value to obtain the demand increment reward; Analyze the violations between various data in the historical material demand data and give demand penalty items; Integrate the demand base reward, demand incremental reward, and demand penalty term, train the initial neural network model until convergence, and obtain the first demand generation model; Insert a low-rank matrix into the attention mechanism of the first demand generation model, freeze the model parameters of the first demand generation model, adjust and optimize the parameters corresponding to the low-rank matrix, and obtain the material demand generation model.

[0018] Furthermore, it also includes: A two-stage distillation framework is constructed, where the first stage is used to output the raw score distribution of the demand teacher model, and the second stage is used to minimize the output difference between the demand student model and the demand teacher model on the power material demand generation task; Analyze the influence of the model matrix in the demand teacher model on the output results of the demand teacher model, and give the Hessian matrix of the demand teacher model; Perform Cholesky decomposition on the Hessian matrix of the demand teacher model to obtain a lower triangular matrix; The model matrix in the demand teacher model is divided into blocks to obtain multiple parameter sub-matrices; Quantize the model parameters in each parameter submatrix in turn, and combine with the lower triangular matrix to give the corresponding quantization error; Repeat the process of quantizing the model parameters until all model parameters are quantized and the quantization error converges, and obtain the compressed model matrix; Integrate the compression model matrix into the demand student model to complete the construction of the demand student model; The demand teacher model or demand student model is used as the material demand generation model.

[0019] Furthermore, it also includes the deployment and optimization of the demand teacher model and demand student model: Deploy the required teacher model on cloud devices and the required student model on edge devices; The demand student model obtains the demand impact file on the edge device, gives the corresponding response results of the power material demand, and synchronizes the response results of the power material demand to the cloud device; The cloud device responds to the edge device, and the demand teacher model reviews the results of the response to the power material demand and gives the review results; Based on the audit results, the demand student model is optimized.

[0020] Furthermore, the results of responding to power material demands are synchronized to cloud devices, including: Compress the results of responding to the power material demand to obtain a demand binary file; Encrypt the required binary file based on the pre-configured encryption algorithm to obtain the required encrypted file; Based on the encrypted channel between cloud devices and edge devices, the encrypted files are transmitted to the cloud devices; The cloud device responds to the edge device, and the demand teacher model reviews the results of the response to the power material demand and provides the review results, including: The cloud device obtains the required encrypted file from the corresponding encrypted channel; According to the device model of the edge device, the corresponding decryption key is searched from the key library, and the required encrypted file is decrypted to obtain the compressed file; Decompress the compressed file, and the teacher model is required to review the decompressed compressed file and give the review result.

[0021] Furthermore, material demand standards include basic material standards, professional material standards, and dynamic material standards; In combination with the material demand standards, a difference detection is performed on the first material demand document, and the first material demand document is associated with the standard clauses corresponding to the material demand standards, and the results of responding to the power material demand are given, specifically including: Based on the basic material standards, verify the uniqueness of various power materials in the first material demand document and select corresponding colors for marking; Based on professional material standards, match the technical parameters of various power materials in the first material demand document with the corresponding equipment requirements, and select corresponding colors for marking; Based on the dynamic material standard, the quantity of various power materials in the first material demand file is associated with the dynamic inventory and marked with corresponding colors; Traverse the first material demand file and give the result of responding to the power material demand.

[0022] In a second aspect, the present invention further provides a device for responding to power material demand, which adopts any of the above-mentioned methods for responding to power material demand, including: Data acquisition module, used to obtain demand impact documents for power materials; The data processing module is used to match different file processing strategies according to the file type of the demand-affecting file, perform structured processing on the demand-affecting file, and obtain material demand data; A demand generation module is used to generate a material demand model based on a pre-built model, integrate the multi-dimensional knowledge graph of electric power, review and process the material demand data, and obtain a first material demand document; The demand review module performs difference detection on the first material demand document in combination with the material demand standard, associates the first material demand document with the standard clauses corresponding to the material demand standard, and gives a result that responds to the power material demand.

[0023] The present invention provides a method and device for responding to power material demand, which has at least the following beneficial effects: (1) By processing the demand impact file and combining it with the material demand generation model, the first material demand file is obtained and tested to obtain the power material demand results. According to the actual material demand standards, the first material demand file is modified and adjusted in a timely manner to ensure the scientificity and rationality of the power material demand, improve the accuracy of the power material demand, and provide reference and help for the power material demand.

[0024] (2) By building a demand reward layer including a basic reward layer, an incremental reward layer, a violation penalty layer and a reward fusion layer, the demand reward function is integrated into the material demand generation model. Based on the synergy between the demand reward function and the low-rank matrix, the performance of the first demand generation model in the power material demand task is effectively optimized. The demand reward function provides feedback signals through a three-level reward mechanism to fine-tune the first demand generation model to ensure that the model can generate high-quality power material demand.

[0025] (3) In the process of training the material demand generation model, by freezing the model parameters of the first demand generation model and adjusting and optimizing the parameters corresponding to the low-rank matrix, the training efficiency of the material demand generation model can be improved and the demand for computing resources can be reduced; at the same time, only adjusting the low-rank matrix can reduce the complexity of the material demand generation model, reduce the risk of overfitting, and enhance the generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A flowchart of a method for responding to power material demand provided by an embodiment of the present invention; Figure 2 A flowchart of data processing provided by an embodiment of the present invention; Figure 3 A flowchart of obtaining a first material demand document provided in an embodiment of the present invention; Figure 4 This is an architecture diagram of a material demand generation model provided in an embodiment of the present invention; Figure 5 A training flow chart of a material demand generation model provided in an embodiment of the present invention; Figure 6 This is a structural block diagram of a device for responding to power material demand provided by an embodiment of the present invention.

[0027] Among them, 201 is a data acquisition module; 202 is a data processing module; 203 is a demand generation module; and 204 is a demand review module. DETAILED DESCRIPTION

[0028] To better understand the above technical solution, the following will be described in detail with reference to the accompanying drawings and specific implementation methods. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0029] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "an," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0030] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or device. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or device comprising the element.

[0031] Electricity is a system of production and consumption consisting of power generation, transmission, transformation, distribution, and consumption. As a clean and efficient energy source, electricity is an essential foundation for modern production and life, and is widely used in various fields, including industrial manufacturing, residential life, and transportation.

[0032] With the continuous development of the power industry, the scale of power systems is expanding, and the types and quantities of power materials are increasing. This places higher demands on the accuracy, timeliness, and scientific nature of power material plans that respond to power material demand. How can we manage and support the generation, review, and optimization of power material plans through information and intelligent means to improve the efficiency and level of plan management?

[0033] Power material planning is a crucial component of power system operation and management, directly impacting key aspects of power companies, including material procurement, inventory management, and operational cost control. Power material demand is influenced by a variety of factors, including seasonal demand fluctuations, cyclical fluctuations, economic conditions, meteorological conditions, and equipment status. The complex relationships and nonlinear correlations between these factors make traditional methods unable to meet the refined management needs of power companies. In response to this situation, a demand generation model is needed. This model can analyze the demand for various power materials and serve as the basis for developing procurement batch scheduling plans, material inventory procurement plans, and special material batch procurement plans, thus supporting the implementation of various procurement models.

[0034] At the same time, power material planning involves multi-source data, including text, tables, images and other forms. Existing technologies mainly rely on manual or simple automated tools to clean and analyze these data, with slow processing speed and low accuracy, and it is difficult to meet the needs of large-scale data processing. The power field is highly professional and has a complex knowledge system. The existing system is difficult to fully integrate professional knowledge in the power field in the process of plan generation and review in response to material needs, resulting in deviations between plans and actual needs. The existing large-scale models have large parameter scales and high computing resource consumption, making them difficult to deploy on edge devices. They cannot meet the needs of real-time processing in edge scenarios of the power system, resulting in insufficient adaptability and practicality of the system, affecting the use effect of power material planning.

[0035] To this end, the present invention provides a method for responding to power material demand, including obtaining a demand impact file for power materials; matching different file processing strategies according to the file type of the demand impact file, performing structured processing on the demand impact file, and obtaining material demand data; based on a pre-constructed material demand generation model, integrating a power multivariate knowledge graph, and reviewing and processing the material demand data to obtain a first material demand file; combining the material demand standard, performing difference detection on the first material demand file, and establishing an association between the first material demand file and the standard clauses corresponding to the material demand standard to provide a result of responding to power material demand. By processing the demand impact file, combining the material demand generation model, obtaining the first material demand file and testing it, and obtaining the result of responding to power material demand, the first material demand file is modified and adjusted in a timely manner according to the actual material demand standard, ensuring the scientificity and rationality of power material demand, improving the accuracy of power material demand, and providing reference and help for power material demand.

[0036] The present invention adopts OCR / LayoutLMv3 technology to achieve efficient cleaning and analysis of multimodal data, solves the problem of low data processing efficiency, improves the speed and accuracy of data processing, and solves the problem of insufficient domain knowledge integration through RoPE extrapolation and injection of power multi-knowledge graph into the material demand generation model, making the power material demand more in line with actual needs. Through model compression technology, the problem of limited model deployment is solved, and while meeting the real-time processing requirements of edge scenarios, the scientific nature of plan optimization is ensured. The overall automation of the entire link of power material demand from data to decision-making is realized, and the efficiency and quality of power material demand generation are improved.

[0037] like Figure 1 As shown, an embodiment of the present invention provides a method for responding to power material demand, which specifically includes the following steps: S101: Obtain the demand impact document for power materials.

[0038] Specifically, the power material demand impact file is a file that affects the power material demand within the planning period, for example, information or files such as maintenance records and replacement records of each power material and the use environment of the power material.

[0039] In a specific example, a power company needs to plan its annual power material procurement, which involves multiple types of power materials such as power generation equipment, transmission lines, and substation facilities. The demand-influencing documents include unstructured data such as handwritten inspection records, paper report images, and equipment nameplate photos.

[0040] S102: According to the file type of the demand-affecting file, different file processing strategies are matched, and the demand-affecting file is structured to obtain material demand data.

[0041] Specifically, the file type includes at least one of text, table, and image.

[0042] Further, refer to Figure 2 , according to the file type of the demand-affecting file, match different file processing strategies, perform structured processing on the demand-affecting file, and obtain material demand data, including: If the file type of the demand impact document is text, the demand impact document is segmented and tagged with parts of speech, and converted into structured data to obtain the corresponding material demand data; If the file type of the demand impact file is a table, perform structural analysis on the demand impact file, analyze the relationship between the content and position of each cell in the demand impact file, extract the content of the demand impact file and obtain the corresponding material demand data; If the file type of the demand impact document is an image, the text in the demand impact document is enhanced and recognized, and converted into structured data to provide the corresponding material demand data.

[0043] The relationship between the content and position of each cell in the demand impact file is expressed by the cell correlation degree, specifically:

[0044] Among them, R(c i ,c j ) is the cell association degree between the i-th cell and the j-th cell, α is the association weight coefficient, PosSim(c i ,c j ) is the position similarity between the i-th cell and the j-th cell, ContSim(c i ,c j ) is the content similarity between the i-th cell and the j-th cell.

[0045] In a specific implementation, when the file type of the demand-impacting document is text, a BERT model variant that integrates BiLSTM (bidirectional long short-term memory network) + CRF (conditional random field) is used to perform word segmentation, part-of-speech tagging, and entity recognition on the demand-impacting document, extract key information such as "transformer model specifications" and "cable quantity", and convert it into structured data to obtain the corresponding material demand data.

[0046] In the BERT model variant that integrates BiLSTM and CRF, BERT is fed into the demand impact document to obtain the contextual representation of each word and output a fixed-dimensional vector representing the semantic information of each word in the demand impact document. The BERT output is then fed into the BiLSTM layer to further capture the dependencies between the previous and next steps in the sequence. The BiLSTM layer then outputs a vector containing both forward and backward information. Finally, the BiLSTM output is fed into the CRF layer to model the dependencies between label sequences. The CRF layer then outputs the final label sequence, completing the processing of the demand impact document.

[0047] When the demand impact file is a table, the Layout Language Model (LayoutLMv3) is used to perform structural analysis on the demand impact file, analyzing the relationship between the content and position of each cell in the demand impact file. The content of the demand impact file is extracted to obtain the corresponding material demand data. In a specific example, an image of an electric power material inventory report containing a slashed header is presented. Optical Character Recognition (OCR) technology is used to perform text recognition on the image of the electric power material inventory report, extracting the text content and its location information. The text content and its corresponding bounding box are extracted from the OCR results. The extracted text and bounding box information are converted into a format that can be processed by the LayoutLMv3 model and input into the LayoutLMv3 model. This allows the model to understand the table structure and the cell content and location information. The association between cells is calculated to establish the row and column relationships of the table. Based on the cell association, the row and column structure of the table is determined, and the content of each cell is extracted. The content is then organized according to the row and column structure to obtain the corresponding material demand data.

[0048] When the demand impact document is an image, a multi-scale OCR engine is used to enhance and recognize the text within it, converting it into structured data to produce the corresponding material demand data. In a specific example, the demand impact document is an image of a device nameplate. The blurred text within the device nameplate image is enhanced and recognized, and the result is converted into structured data in JSON format to produce the corresponding material demand data. For example, the parameter information for a "10kV circuit breaker" is stored in a standardized manner.

[0049] S103: Based on the pre-built material demand generation model, the power multi-knowledge graph is integrated to review and process the material demand data to obtain a first material demand file.

[0050] Reference Figure 3 , specifically including: According to the sequence length of material demand data, the rotation position coding extrapolation technology is used to expand the processing window of the material demand generation model; Analyze the correlation similarity between each demand vector of the material demand data and each material vector in the power multivariate knowledge graph, and adjust the attention weight in the attention mechanism of the material demand generation model. The power multivariate knowledge graph includes the material vectors of each power material. Each material vector of each power material includes the technical parameters, coding rules, and procurement specifications of the corresponding power material in the power industry standard. Substitute the material demand data into the adjusted material demand generation model to obtain a first material demand file.

[0051] Among them, the attention weight is specifically expressed as:

[0052] Among them, A k is the attention weight corresponding to the k-th material vector in the electric power multivariate knowledge graph, S k is the semantic similarity between the kth material vector and the material demand data in the electric power multivariate knowledge graph, λ k is the important coefficient of the kth material vector in the electric power multivariate knowledge graph, and M is the total number of material vectors in the electric power multivariate knowledge graph.

[0053] It's understandable that RoPE extrapolation technology is directly applicable to models based on the Transformer architecture. By adjusting the model's RoPE parameters (such as rotation angle and scaling factor), the model can process longer sequences at runtime without retraining. RoPE extrapolation can be implemented in various ways. For example, Position Interpolation (PI) reduces the rotation radius to map positions outside the training range back to the training range. NTK-Aware interpolation differentiates between high-frequency (low-dimensional) and low-frequency (high-dimensional) components, extrapolating the high-frequency components (with a smaller scaling margin) and interpolating the low-frequency components (with a larger scaling margin). Dynamic NTK dynamically adjusts the scaling factor to gradually reduce the rotation radius based on the current sequence length, avoiding sudden performance changes. In other examples, other methods can be used to expand the processing window of the material demand generation model, which are not limited here.

[0054] In a specific implementation, NTK-Aware interpolation is adopted, which is specifically expressed as follows:

[0055] Among them, β' is the adjusted RoPE parameter, β is the original RoPE parameter, L extra is the sequence length of material demand data, L train is the maximum sequence length during material demand generation model training (i.e., the processing window of the material demand generation model), d is the dimension of the material demand generation model, and i is the current dimension index.

[0056] The adjusted RoPE parameters are substituted into the material demand generation model to expand the processing window of the material demand generation model.

[0057] Analyze the correlation similarity between each demand vector of material demand data and each material vector in the electric power multivariate knowledge graph, and adjust the attention weight in the attention mechanism in the material demand generation model to achieve the effect of strengthening the correlation weight between the knowledge in the electric power material field and the general semantics.

[0058] The multivariate knowledge graph for electric power includes the material vectors for each electric power material. Each material vector for an electric power material includes the technical parameters, coding rules, and procurement specifications for the corresponding electric power material in the electric power industry standards. Specifically, the electric power industry standards are obtained, and the technical parameters, coding rules, and procurement specifications for various electric power materials, including power generation, transmission, and substation equipment, are collected in the electric power industry standards to construct a triple knowledge graph containing entities, attributes, and relationships, such as "lightning arrester - applicable voltage - 110kV." In this example, the multivariate knowledge graph for electric power is a triple knowledge graph. In other examples, the content of the knowledge graph can be increased or decreased based on actual conditions, and this is not limited to this.

[0059] Finally, the material demand data is substituted into the adjusted material demand generation model to obtain the first material demand file.

[0060] Further, refer to Figure 4 ,The material demand generation model includes a demand input layer, an attention mechanism, a demand reward layer, and a demand output layer; The attention mechanism is connected to the demand input layer. Material demand data enters the attention mechanism through the demand input layer to extract the core demand features in the material demand data. The demand output layer is connected to the attention mechanism to generate intermediate demand documents based on the core demand features; The demand reward layer is connected to the demand output layer, analyzes the compliance status of the intermediate demand documents, obtains the total demand reward, and feeds it back to the attention mechanism and demand output layer; Among them, the demand reward layer includes the basic reward layer, the incremental reward layer, the violation penalty layer and the reward fusion layer; The basic reward layer is connected to the demand output layer to analyze the matching degree between the intermediate demand documents and the corresponding procurement specifications to obtain the basic reward for the intermediate demand; The incremental reward layer is connected to the demand output layer to analyze the fluctuation difference between each value in the intermediate demand file and the corresponding standard value to obtain the incremental reward of the intermediate demand; The violation penalty layer is connected to the demand output layer to analyze the violations between the data in the intermediate demand file and give the intermediate demand penalty items; The reward fusion layer is connected to the basic reward layer, incremental reward layer, and violation penalty layer respectively. It is used to fuse the intermediate demand basic reward, intermediate demand incremental reward, and intermediate demand penalty items to obtain the total demand reward and provide feedback to the attention mechanism and demand output layer.

[0061] Further, refer to Figure 5 ,The construction of the material demand generation model is ,obtained through the following steps: Obtain historical material demand data; Analyze the matching degree between historical material demand data and corresponding procurement specifications to obtain demand-based rewards; Analyze the fluctuation difference between each value in the historical material demand data and the corresponding standard value to obtain the demand increment reward; Analyze the violations between various data in the historical material demand data and give demand penalty items; Integrate the demand base reward, demand incremental reward, and demand penalty term, train the initial neural network model until convergence, and obtain the first demand generation model; Insert a low-rank matrix into the attention mechanism of the first demand generation model, freeze the model parameters of the first demand generation model, adjust and optimize the parameters corresponding to the low-rank matrix, and obtain the material demand generation model.

[0062] Among them, the demand reward function is obtained by integrating the demand basic reward, demand incremental reward and demand penalty term, which is specifically expressed as:

[0063] Among them, β1 is the weight coefficient corresponding to the demand basic reward, β2 is the weight coefficient corresponding to the demand incremental reward, β3 is the weight coefficient corresponding to the demand penalty term, R base is the demand-based reward, R inc is the demand incremental reward, P i is the demand penalty item for the i-th violation.

[0064] Based on the demand reward function, after the initial neural network model is trained to obtain the first demand generation model, a low-rank matrix is ​​inserted into the attention mechanism of the first demand generation model to adapt to the power material demand task.

[0065] In a specific example, the weight matrix of the attention mechanism of the first demand generation model is W, and the inserted low-rank matrices are A and B, then the new weight matrix W' is:

[0066] Among them, the dimension of the low-rank matrix A is d×r, the dimension of the low-rank matrix B is r×d, and r is the rank of the low-rank matrix.

[0067] Next, the model parameters of the first demand generation model are frozen, and only the parameters corresponding to the newly added low-rank matrix are fine-tuned. This significantly reduces computing resource consumption while maintaining model performance. In this example, the AdamW optimizer is used for fine-tuning. Iterative training is performed using historical material demand data. Based on the demand reward function, the low-rank matrix parameters are adjusted to optimize model performance, resulting in a material demand generation model.

[0068] In a specific example, consider a power material demand task requiring the generation of a procurement plan. Based on the completed first demand generation model, low-rank matrices A and B are inserted into the attention mechanism of the first demand generation model. The model parameters of the first demand generation model are frozen, and only the parameters corresponding to the low-rank matrices are fine-tuned. Historical material demand data is prepared, including historical plans, historical procurement specifications, and historical material quantities.

[0069] Input the historical material demand data into the first demand generation model to generate a plan P. Calculate the cosine similarity between the generated plan P and the standard procurement specification S to obtain the demand-based reward:

[0070] The preset reasonable quantity range is [Q min ,Q max ], the amount of materials generated is Q, and the demand increment reward is calculated:

[0071] Setting: The deduction corresponding to the model specification conflict is P spec The deduction points corresponding to the delivery date contradiction are P delivery , calculate the demand penalty term P penalty :

[0072] The total reward is obtained by summing the demand base reward and the demand incremental reward, minus the demand penalty. The loss function is then calculated, and the parameters corresponding to the low-rank matrix are updated using the AdamW optimizer. Through multiple iterative training, the low-rank matrix parameters are optimized to improve model performance.

[0073] By building a demand reward layer consisting of a basic reward layer, an incremental reward layer, a violation penalty layer, and a reward fusion layer, the demand reward function is integrated into the material demand generation model. The synergy between the demand reward function and the low-rank matrix effectively optimizes the performance of the first demand generation model on the power material demand task. The demand reward function provides feedback signals through a three-level reward mechanism, fine-tuning the first demand generation model to ensure that the model can generate high-quality power material demand.

[0074] Furthermore, it also includes: A two-stage distillation framework is constructed, where the first stage is used to output the raw score distribution of the demand teacher model, and the second stage is used to minimize the output difference between the demand student model and the demand teacher model on the power material demand generation task; Analyze the influence of the model matrix in the demand teacher model on the output results of the demand teacher model, and give the Hessian matrix of the demand teacher model; Perform Cholesky decomposition on the Hessian matrix of the demand teacher model to obtain a lower triangular matrix; The model matrix in the demand teacher model is divided into blocks to obtain multiple parameter sub-matrices; Quantize the model parameters in each parameter submatrix in turn, and combine with the lower triangular matrix to give the corresponding quantization error; Repeat the process of quantizing the model parameters until all model parameters are quantized and the quantization error converges, and obtain the compressed model matrix; Integrate the compression model matrix into the demand student model to complete the construction of the demand student model; The demand teacher model or demand student model is used as the material demand generation model.

[0075] In a specific implementation, it is necessary to prepare a demand teacher model and a demand student model, wherein the demand teacher model is a large-scale language model that has been trained, has high performance and a large number of parameters, and the demand student model is a lightweight model. By constructing a two-stage distillation framework, through the first stage and the second stage, the output performance of the demand student model is guaranteed while reducing the number of parameters and computational costs.

[0076] By combining the original score distribution of the demand teacher model obtained in the first stage with the model matrix of the demand teacher model, we analyze the impact of the model matrix in the demand teacher model on the output results of the demand teacher model. For each weight parameter in the model matrix, we calculate its second-order derivative and give the Hessian matrix of the demand teacher model. Then, we perform Cholesky decomposition on the Hessian matrix to obtain the lower triangular matrix L, where H = L·L T , H is the Hessian matrix, L T is the transposed matrix of the lower triangular matrix L.

[0077] The weight matrix is ​​divided into multiple small blocks, namely parameter sub-matrices, and the model parameters in each block are quantized one by one. The quantization formula is:

[0078] Among them, W quant is the quantized model parameter, W float It is the floating-point parameter of the model parameter, and scale is the quantization scale factor used to map the floating-point parameter to the quantization range.

[0079] Analyze and calculate the error between the output of the demand teacher model before and after quantization, and calculate the corresponding quantization error. Based on the quantization error, adjust the quantized model parameters to improve model accuracy. Repeat the quantization and adjustment process until all model parameters are quantized. Integrate the compressed model matrix into the demand student model to complete the construction of the demand student model. Through quantization and adjustment, the demand student model can reduce the number of parameters while maintaining high performance.

[0080] Depending on the specific situation, either the demand teacher model or the demand student model can be used as the material demand generation model. For example, when the equipment can meet the needs of the demand teacher model, the demand teacher model can be used as the material demand generation model. When rapid results are required, the demand student model can be used as the material demand generation model, with no restrictions. In this example, the trained material demand generation model is used as the demand teacher model, and model compression is performed to obtain the demand student model.

[0081] Furthermore, it also includes the deployment and optimization of the demand teacher model and demand student model: Deploy the required teacher model on cloud devices and the required student model on edge devices; The demand student model obtains the demand impact file on the edge device, gives the corresponding response results of the power material demand, and synchronizes the response results of the power material demand to the cloud device; The cloud device responds to the edge device, and the demand teacher model reviews the results of the response to the power material demand and gives the review results; Based on the audit results, the demand student model is optimized.

[0082] In one specific implementation, the demand student model is deployed on an edge device, running locally, to generate the corresponding first material demand file. To ensure consistency between the edge device's output and the cloud device's, the edge device periodically uploads the output to the cloud device. Upon receiving the output uploaded by the edge device, the cloud device reviews the output.

[0083] The cloud device compares the output results uploaded by the edge device with the expected results from the cloud device. The cloud device checks for consistency. If any discrepancies are found, the cloud device further analyzes the cause. If an error is found, indicating a failure, the cloud device sends feedback to the edge device, requesting that the task be re-executed or the corresponding student model be updated. Conversely, if no error is found, indicating a passing result, the cloud device does not send any feedback to the edge device.

[0084] In order to ensure the security of information transmission between cloud devices and edge devices and improve data transmission efficiency, the results of responding to power material needs are compressed and encrypted for transmission.

[0085] Furthermore, the results of responding to power material demands are synchronized to cloud devices, including: Compress the results of responding to the power material demand to obtain a demand binary file; Encrypt the required binary file based on the pre-configured encryption algorithm to obtain the required encrypted file; Based on the encrypted channel between cloud devices and edge devices, the encrypted files are transmitted to the cloud devices; The cloud device responds to the edge device, and the demand teacher model reviews the results of the response to the power material demand and provides the review results, including: The cloud device obtains the required encrypted file from the corresponding encrypted channel; According to the device model of the edge device, the corresponding decryption key is searched from the key library, and the required encrypted file is decrypted to obtain the compressed file; Decompress the compressed file, and the teacher model is required to review the decompressed compressed file and give the review result.

[0086] In one specific embodiment, after the edge device obtains the result of responding to the power material demand, it compresses the result to obtain a binary file of the demand. In this example, the result of responding to the power material demand can be compressed using Huffman coding, LZW coding, Gzip, Bzip2, or LZMA. Other file compression methods can be used in other embodiments, and this is not limited to this. After the file compression is completed, the demand binary file is encrypted based on an encryption algorithm pre-configured in the edge device to obtain an encrypted demand file. In the present invention, each edge device is pre-configured with a corresponding encryption algorithm. Different encryption algorithms are configured based on the actual use scenario of the edge device to meet the encryption requirements of different edge devices. Encryption algorithms can include AES, DES, 3DES, and RC4. Based on an encrypted channel between the cloud device and the edge device, the encrypted demand file is transmitted to the cloud device. For example, TLS (Transport Layer Security) is used to establish a secure communication channel.

[0087] The cloud device obtains the edge device's requested encrypted file through the corresponding encrypted channel. Based on the edge device's device model, the cloud device searches the key library for the corresponding decryption key, decrypts the requested encrypted file, and obtains a compressed file. The edge device's device model is a unique identifier for each edge device, and the corresponding decryption key is retrieved from the key library based on the edge device's device model. The key library is a database used by the cloud device to store encryption and decryption keys for each edge device and is updated based on communication with the edge device. The compressed file is then decompressed, and the demand teacher model reviews the decompressed compressed file and provides the review results.

[0088] During the interaction between cloud devices and edge devices, files are compressed before being encrypted. This improves transmission efficiency while reducing the amount of data encrypted, thereby increasing encryption efficiency. The integration of file encryption and encrypted channels further ensures secure interactions between cloud devices and edge devices.

[0089] S104: performing a difference detection on the first material demand document in combination with the material demand standard, and establishing an association between the first material demand document and the standard clauses corresponding to the material demand standard, and providing a result of responding to the power material demand.

[0090] Specifically, material requirement standards include basic material standards, specialized material standards, and dynamic material standards. Basic material standards are used to verify the uniqueness of power material codes, ensuring that each power material has a unique code and preventing duplicate or incorrect material codes. Specialized material standards include the technical parameters of various power materials and corresponding equipment requirements. For example, insulation level and voltage level matching requirements ensure that the insulation level of the equipment meets the requirements of its voltage level, thereby ensuring the safety and reliability of power materials. Dynamic material standards link real-time inventory and procurement cycle data to ensure the rationality and timeliness of procurement requirements, reducing inventory backlogs and stock-out risks.

[0091] Furthermore, in combination with the material demand standards, a difference detection is performed on the first material demand document, and the first material demand document is associated with the standard clauses corresponding to the material demand standards, and a result of responding to the power material demand is given, specifically including: Based on the basic material standards, verify the uniqueness of various power materials in the first material demand document and select corresponding colors for marking; Based on professional material standards, match the technical parameters of various power materials in the first material demand document with the corresponding equipment requirements, and select corresponding colors for marking; Based on the dynamic material standard, the quantity of various power materials in the first material demand file is associated with the dynamic inventory and marked with corresponding colors; Traverse the first material demand file and give the result of responding to the power material demand.

[0092] In a specific implementation, it is determined whether the material code of each electric power material appearing in the first material requirement file is unique in the existing electric power material code library. If so, no marking is required; if not, the corresponding material code in the first material requirement file is highlighted using a color corresponding to the basic material standard.

[0093] Determine whether the technical parameters of various power materials match the corresponding equipment requirements, as well as whether the voltage levels of various power materials match their insulation levels. If the voltage level meets the corresponding insulation level, it meets the standard and does not need to be marked. Otherwise, the relevant technical parameters and equipment requirements of the corresponding power materials in the first material requirement document are highlighted using the color corresponding to the professional material standard.

[0094] Determine the relationship between the quantities of various power materials in the first material demand document and the quantities in the dynamic inventory. It is understood that the quantities in the dynamic inventory change in real time, and there will be certain discrepancies between the dynamic inventory at the time the first material demand document is issued and the dynamic inventory in the first material demand document. The rationality of the first material demand document needs to be checked based on real-time inventory and procurement cycles. If the discrepancy is within the corresponding range, no marking is made. Conversely, if the discrepancy is outside the corresponding range, the corresponding power material in the first material demand document is highlighted using the color corresponding to the dynamic material standard.

[0095] In a specific example, when labeling the first material demand document, the material code is first determined based on the basic material standard. For example, if the material code is 00001, which is unique in the power material code library, no labeling is required. Then, based on the professional material standard, the power material with material code 00001 has an equipment voltage level of 10kV and an equipment insulation level of 30kV. Since the equipment voltage level is lower than the equipment insulation level, it meets the requirements and does not require labeling. Based on the dynamic material standard, the quantity of material with material code 00001 in real-time inventory is 50, and the planned purchase quantity in the first material demand document is 30. The required quantity is 70, and 50 + 30 > 70, which meets the demand. However, (50 + 30 - 70) / 70 = 1 / 7 ∈ [0, 0.2] is within the unsafe quantity range and is highlighted using the warning color corresponding to the dynamic material standard. For example, the quantity of a material with material code 00001 in real-time inventory is 75. The planned purchase quantity in the first material demand document is 50, and the demand is 100. 75 + 50 > 100, meeting the demand. Furthermore, (75 + 50 - 100) / 100 = 0.25 ∈ [0.2, 0.4], which is within the safe quantity range and does not require marking. If the planned purchase quantity exceeds the backlog range, the corresponding color marking will be applied.

[0096] By comprehensively verifying and marking the first material requirement document, we ensure that the power material requirements comply with basic, professional and dynamic material standards, improve the accuracy and rationality of the power material requirements, and facilitate the rapid identification and processing of key differences.

[0097] This invention ensures the compliance of power material demand by combining power industry standards. At the same time, it also deploys demand teacher models and demand student models in the cloud and edge to meet the usage requirements of edge sites (such as substations) and maximize the efficiency of the material demand generation model.

[0098] Reference Figure 6 The embodiment of the present invention provides a device for responding to power material demand, including: Data acquisition module 201, used to obtain power material demand impact files; The data processing module 202 is used to match different file processing strategies according to the file type of the demand-affecting file, perform structured processing on the demand-affecting file, and obtain material demand data; The demand generation module 203 is used to generate a material demand model based on a pre-built model, integrate the multi-dimensional knowledge graph of electric power, review and process the material demand data, and obtain a first material demand document; The demand review module 204 performs a difference detection on the first material demand document in combination with the material demand standard, associates the first material demand document with the standard clauses corresponding to the material demand standard, and provides a result of responding to the power material demand.

[0099] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0100] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such changes and modifications as fall within the scope of the claims and their equivalents.

Claims

1. A method for responding to power material demand, characterized in that: include: Obtain demand impact documents for power supplies; According to the file type of demand-affecting documents, different file processing strategies are matched, and demand-affecting documents are structured to obtain material demand data; Based on the pre-built material demand generation model and the integration of the power multi-knowledge graph, the material demand data is reviewed and processed to obtain the first material demand document; In combination with the material demand standard, a difference detection is performed on the first material demand document, and the first material demand document is associated with the standard clauses corresponding to the material demand standard to provide a result that responds to the power material demand.

2. The method for responding to power material demand according to claim 1, wherein: The file type includes at least one of text, table, and image; According to the file type of the demand-affecting file, different file processing strategies are matched, and the demand-affecting file is structured to obtain material demand data, including: If the file type of the demand impact document is text, the demand impact document is segmented and tagged with parts of speech, and converted into structured data to obtain the corresponding material demand data; If the file type of the demand impact file is a table, perform structural analysis on the demand impact file, analyze the relationship between the content and position of each cell in the demand impact file, extract the content of the demand impact file and obtain the corresponding material demand data; If the file type of the demand impact document is an image, the text in the demand impact document is enhanced and recognized, and converted into structured data to provide the corresponding material demand data.

3. The method for responding to power material demand according to claim 1, wherein: According to the pre-built material demand generation model, the power multi-knowledge graph is integrated to review and process the material demand data to obtain the first material demand file, which is obtained through the following steps: According to the sequence length of material demand data, the rotation position coding extrapolation technology is used to expand the processing window of the material demand generation model; Analyze the correlation similarity between each demand vector of the material demand data and each material vector in the power multivariate knowledge graph, and adjust the attention weight in the attention mechanism of the material demand generation model. The power multivariate knowledge graph includes the material vectors of each power material. Each material vector of each power material includes the technical parameters, coding rules, and procurement specifications of the corresponding power material in the power industry standard. Substitute the material demand data into the adjusted material demand generation model to obtain a first material demand file.

4. The method for responding to power material demand according to claim 3, wherein: The material demand generation model includes a demand input layer, an attention mechanism, a demand reward layer, and a demand output layer; The attention mechanism is connected to the demand input layer. Material demand data enters the attention mechanism through the demand input layer to extract the core demand features in the material demand data. The demand output layer is connected to the attention mechanism to generate intermediate demand documents based on the core demand features; The demand reward layer is connected to the demand output layer, analyzes the compliance status of the intermediate demand documents, obtains the total demand reward, and feeds it back to the attention mechanism and demand output layer; Among them, the demand reward layer includes the basic reward layer, the incremental reward layer, the violation penalty layer and the reward fusion layer; The basic reward layer is connected to the demand output layer to analyze the matching degree between the intermediate demand documents and the corresponding procurement specifications to obtain the basic reward for the intermediate demand; The incremental reward layer is connected to the demand output layer to analyze the fluctuation difference between each value in the intermediate demand file and the corresponding standard value to obtain the incremental reward of the intermediate demand; The violation penalty layer is connected to the demand output layer to analyze the violations between the data in the intermediate demand file and give the intermediate demand penalty items; The reward fusion layer is connected to the basic reward layer, incremental reward layer, and violation penalty layer respectively. It is used to fuse the intermediate demand basic reward, intermediate demand incremental reward, and intermediate demand penalty items to obtain the total demand reward and provide feedback to the attention mechanism and demand output layer.

5. The method for responding to power material demand according to claim 4, wherein: The construction of the material demand generation model is achieved through the following steps: Obtain historical material demand data; Analyze the matching degree between historical material demand data and corresponding procurement specifications to obtain demand-based rewards; Analyze the fluctuation difference between each value in the historical material demand data and the corresponding standard value to obtain the demand increment reward; Analyze the violations between various data in the historical material demand data and give demand penalty items; Integrate the demand base reward, demand incremental reward, and demand penalty term, train the initial neural network model until convergence, and obtain the first demand generation model; Insert a low-rank matrix into the attention mechanism of the first demand generation model, freeze the model parameters of the first demand generation model, adjust and optimize the parameters corresponding to the low-rank matrix, and obtain the material demand generation model.

6. The method for responding to power material demand according to claim 3, wherein: Also includes: A two-stage distillation framework is constructed, where the first stage is used to output the raw score distribution of the demand teacher model, and the second stage is used to minimize the output difference between the demand student model and the demand teacher model on the power material demand generation task; Analyze the influence of the model matrix in the demand teacher model on the output results of the demand teacher model, and give the Hessian matrix of the demand teacher model; Perform Cholesky decomposition on the Hessian matrix of the demand teacher model to obtain a lower triangular matrix; The model matrix in the demand teacher model is divided into blocks to obtain multiple parameter sub-matrices; Quantize the model parameters in each parameter submatrix in turn, and combine with the lower triangular matrix to give the corresponding quantization error; Repeat the process of quantizing the model parameters until all model parameters are quantized and the quantization error converges, and obtain the compressed model matrix; Integrate the compression model matrix into the demand student model to complete the construction of the demand student model; The demand teacher model or demand student model is used as the material demand generation model.

7. The method for responding to power material demand according to claim 6, wherein: It also includes the deployment and optimization of the demand teacher model and demand student model: Deploy the required teacher model on cloud devices and the required student model on edge devices; The demand student model obtains the demand impact file on the edge device, gives the corresponding response results of the power material demand, and synchronizes the response results of the power material demand to the cloud device; The cloud device responds to the edge device, and the demand teacher model reviews the results of the response to the power material demand and gives the review results; Based on the audit results, the demand student model is optimized.

8. The method for responding to power material demand according to claim 7, wherein: Synchronize the results of responding to power material needs to cloud devices, including: Compress the results of responding to the power material demand to obtain a demand binary file; Encrypt the required binary file based on the pre-configured encryption algorithm to obtain the required encrypted file; Based on the encrypted channel between cloud devices and edge devices, the encrypted files are transmitted to the cloud devices; The cloud device responds to the edge device, and the demand teacher model reviews the results of the response to the power material demand and provides the review results, including: The cloud device obtains the required encrypted file from the corresponding encrypted channel; According to the device model of the edge device, the corresponding decryption key is searched from the key library, and the required encrypted file is decrypted to obtain the compressed file; Decompress the compressed file, and the teacher model is required to review the decompressed compressed file and give the review result.

9. The method for responding to power material demand according to claim 1, wherein: Material demand standards include basic material standards, professional material standards and dynamic material standards; In combination with the material demand standards, a difference detection is performed on the first material demand document, and the first material demand document is associated with the standard clauses corresponding to the material demand standards, and the results of responding to the power material demand are given, specifically including: Based on the basic material standards, verify the uniqueness of various power materials in the first material demand document and select corresponding colors for marking; Based on professional material standards, match the technical parameters of various power materials in the first material demand document with the corresponding equipment requirements, and select corresponding colors for marking; Based on the dynamic material standard, the quantity of various power materials in the first material demand file is associated with the dynamic inventory and marked with corresponding colors; Traverse the first material demand file and give the result of responding to the power material demand.

10. A device for responding to power material demand, characterized in that: The method for responding to power material demand according to any one of claims 1 to 9 comprises: Data acquisition module, used to obtain demand impact documents for power materials; The data processing module is used to match different file processing strategies according to the file type of the demand-affecting file, perform structured processing on the demand-affecting file, and obtain material demand data; A demand generation module is used to generate a material demand model based on a pre-built model, integrate the multi-dimensional knowledge graph of electric power, review and process the material demand data, and obtain a first material demand document; The demand review module performs difference detection on the first material demand document in combination with the material demand standard, associates the first material demand document with the standard clauses corresponding to the material demand standard, and gives a result that responds to the power material demand.

Images