Power resource deployment prediction method and system based on industrial chain power consumption data analysis

Abstract

The present application relates to a power resource deployment prediction method based on industrial chain power consumption data analysis, and the specific steps include: obtaining the power consumption information and production capacity information of each enterprise in each cycle on the industrial chain, and the transaction information between each enterprise; according to the transaction information between each enterprise, a one-way graph structure for representing the transmission relationship between each enterprise in the industrial chain is constructed, and the weight information of each one-way edge in the one-way graph structure is determined; the power consumption of the upstream enterprise in the current cycle and the last cycle is collected, and the power consumption change rate of the upstream enterprise in the current cycle is calculated and input as the attribute information of the initial node in the one-way graph structure; the one-way graph structure, the attribute information of the initial node, the weight information of the one-way edge and the historical information of the middle and downstream enterprises are input into the pre-trained graph convolution network, and the graph features of each enterprise node in the one-way graph structure are extracted.

Description

Technical Field

[0001] This invention relates to the field of power dispatching technology, specifically to a method and system for predicting power resource allocation based on power consumption data analysis of the industrial chain. Background Technology

[0002] Artificial intelligence models such as graph convolutional networks can be used to comprehensively analyze the electricity consumption of upstream and downstream enterprises in the industrial chain, thereby accurately predicting the electricity consumption trend of the industrial chain. In existing technologies, the method of predicting the electricity consumption trend of the industrial chain by comprehensively analyzing the electricity consumption information of all enterprises in the current period assumes that the electricity consumption changes of all enterprises in the industrial chain are synchronized.

[0003] In reality, changes in electricity consumption have a transmission process along the industrial chain. For example, when the electricity consumption of upstream enterprises changes in the current cycle, this change is often not reflected in the electricity consumption of midstream and downstream enterprises in the same cycle. It is usually only in subsequent cycles that the electricity consumption of midstream and downstream enterprises changes accordingly. Existing technologies using artificial intelligence for electricity trend forecasting cannot predict the transmission of changes in electricity consumption from upstream enterprises along the industrial chain, and therefore cannot accurately predict the electricity consumption of midstream and downstream enterprises in subsequent cycles.

[0004] The existing Chinese invention patent with patent number "CN119150181A" discloses a graph neural network method for extracting electricity anomalies in the industrial chain. It first uses an improved LSTM to extract high-dimensional time-series features of each enterprise, and then inputs these features as nodes into a graph neural network for spatial relationship aggregation. The nodes in the graph structure are enterprise feature vectors, the edges are the relationship strength between enterprises, and the input is multi-dimensional historical time-series data of a single enterprise. All of these are static data and cannot dynamically reflect the transmission process of electricity consumption changes in the industrial chain.

[0005] To address the problems existing in current technologies, this paper uses artificial intelligence models to simulate transmission relationships, predicts changes in electricity consumption of downstream enterprises in the industrial chain, and further predicts the electricity consumption trend of the industrial chain, thereby enabling more accurate and timely allocation and scheduling of power resources. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes a method and system for predicting power resource allocation based on power consumption data analysis of the industrial chain.

[0007] The technical solution of the present invention is as follows: On the one hand, this invention proposes a method for predicting power resource allocation based on electricity consumption data analysis of the industrial chain, the specific steps of which include: Obtain electricity consumption and production capacity information of various enterprises in the industrial chain at various stages, as well as transaction information between various enterprises; Based on the transaction information between various enterprises, a one-way graph structure is constructed to represent the transmission relationship between enterprises in the industrial chain, and the weight information of each one-way edge in the one-way graph structure is determined. Collect the electricity consumption of upstream enterprises in the current cycle and the previous cycle, calculate the rate of change of electricity consumption of upstream enterprises in the current cycle, and input it into the attribute information of the initial node in the one-way graph structure. The unidirectional graph structure, the attribute information of the initial node, the weight information of the unidirectional edge, and the historical information of the midstream and downstream enterprises are input into the pre-trained graph convolutional network to extract the graph features of each enterprise node in the unidirectional graph structure. The graph features of each enterprise node are input into a neural network to predict the rate of change in electricity consumption of each enterprise in subsequent periods, and to allocate power resources accordingly.

[0008] As a preferred embodiment, the architecture of the graph convolutional network includes three message passing layers: a power consumption change rate passing layer, a node feature enhancement layer, and a power consumption change rate combined with edge feature passing layer. The electricity consumption change rate transmission layer transmits messages about electricity consumption change rates based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain. The node feature enhancement layer transmits messages about the rate of change in electricity consumption based on the node information of the target node. The electricity consumption change rate combined with edge feature transmission layer not only transmits the electricity consumption change rate message based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain, but also transmits edge features based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain combined with the weight information of each unidirectional edge.

[0009] In a preferred embodiment, the first perceptron MLP_1 is used to extract graph features at each enterprise node in the electricity consumption rate transfer layer; the second perceptron MLP_2 is used to extract features at each node in the node feature enhancement layer; the electricity consumption rate combined with edge feature transfer layer uses the third perceptron MLP_3 to extract graph features at each node and the fourth perceptron MLP_4 to extract edge features.

[0010] In a preferred embodiment, when the first perceptron MLP_1 performs graph feature extraction, if the target node is an upstream node, only the information of the target node itself is input; if the target node is a mid-to-downstream node, the source node information of the target node and the directed edge information between the source node and the target node are input. When the second perceptron MLP_2 and the third perceptron MLP_3 extract graph features, if the target node is an upstream node, only the information of the target node itself is input. If the target node is a mid-to-downstream node, the information of the source node of the target node, the directed edge information between the source node and the target node, and the information of the target node itself are input. When the fourth perceptron MLP_4 performs edge feature extraction, it inputs the source node information of the target edge, the target edge's own information, and the target node's information.

[0011] In a preferred embodiment, the weight information of each unidirectional edge in the unidirectional graph structure is determined according to the proportion of the transaction volume between the upstream enterprise and different downstream enterprises.

[0012] As a preferred implementation, it also includes calculating the correlation index between the production capacity and electricity consumption of each enterprise based on the electricity consumption information and production capacity information of each enterprise in each period. Then, calculate the change in electricity consumption based on the electricity consumption change rate of each enterprise in the subsequent cycle, and further calculate the change in production capacity.

[0013] As a preferred embodiment, the specific method for calculating the correlation index between the production capacity and electricity consumption of each enterprise is as follows:

[0014] In the formula, This is a correlation index between production capacity and electricity consumption; To count the number of months; For the target number of months; The electricity consumption over n months; The production capacity is for n months.

[0015] On the other hand, this invention proposes a power resource allocation and prediction system based on the analysis of electricity consumption data in the industrial chain, comprising: The data acquisition module obtains electricity consumption and production capacity information of various enterprises in the industrial chain at various cycles, as well as transaction information between various enterprises; The feature graph construction module constructs a one-way graph structure to represent the transmission relationship between enterprises in the industrial chain based on the transaction information between various enterprises, and determines the weight information of each one-way edge in the one-way graph structure. The feature graph initial parameter setting module collects the electricity consumption of upstream enterprises in the current cycle and the previous cycle, calculates the rate of change of electricity consumption of upstream enterprises in the current cycle, and inputs it into the unidirectional graph structure as the attribute information of the initial node. The graph feature extraction module inputs the one-way graph structure, the attribute information of the initial node, the weight information of the one-way edge, and the historical information of the midstream and downstream enterprises into the pre-trained graph convolutional network to extract the graph features of each enterprise node in the one-way graph structure. The power resource adjustment module inputs the graph features of each enterprise node into a neural network to predict the rate of change in electricity consumption of each enterprise in subsequent periods and allocates power resources accordingly.

[0016] On the other hand, the present invention proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the power resource allocation and prediction method based on power consumption data analysis of the industrial chain as described in any embodiment of the present invention.

[0017] On the other hand, the present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the power resource allocation and prediction method based on industrial chain power consumption data analysis as described in any embodiment of the present invention.

[0018] The present invention has the following beneficial effects: 1. This invention constructs a directed graph structure and uses a graph convolutional network to simulate the step-by-step transmission of messages along the industry chain, thereby enabling the prediction of changes in the downstream of the subsequent cycle based on changes in the upstream of the current cycle.

[0019] 2. This invention quantifies the intensity of electricity consumption changes during the transmission process by defining the edge weight as a proportion based on the actual transaction volume between enterprises.

[0020] 3. This invention achieves multi-level, high-fidelity graph feature extraction through a three-layer message passing mechanism of graph convolutional networks. The first layer quickly establishes preliminary information associations between upstream and downstream nodes to ensure the effective injection of changing signals. The second layer incorporates the target node's own features during the transmission process, avoiding information overlay and making the feature expression of each node more distinctive, reflecting both external transmission and its own characteristics. The third layer innovatively updates edge features synchronously, making the relationship (edge) between nodes no longer static, but dynamically evolving with the transmission process.

[0021] 4. This invention not only predicts electricity consumption, but also organically combines electricity consumption forecasting with capacity forecasting through the innovative indicator of "capacity correlation index". After predicting electricity consumption, future capacity can be further calculated. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the electric vehicle industry chain. Figure 2 A single-line diagram used to represent the transmission relationship between enterprises in the industrial chain; Figure 3 This is a diagram of the architecture of a graph convolutional network. Figure 4 This is a schematic diagram of the process of message passing to upstream nodes through the first MLP; Figure 5 This is a flowchart illustrating the process of message passing between midstream or downstream nodes via the first MLP. Figure 6 This is a schematic diagram of the process of message passing to upstream nodes through the second MLP; Figure 7 This is a flowchart illustrating the process of message passing between midstream or downstream nodes via a second MLP. Figure 8 This is a flowchart illustrating the process of message passing to upstream nodes via a third-party MLP. Figure 9 This is a flowchart illustrating the process of message passing between midstream or downstream nodes via a third-party MLP. Figure 10 This is a schematic diagram of the message passing process through the fourth MLP edge. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0025] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0026] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0027] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0028] Example 1: A power resource allocation and forecasting method based on electricity consumption data analysis of the industrial chain includes the following steps: Obtain electricity consumption and production capacity information of various enterprises in the industrial chain at various stages, as well as transaction information between various enterprises; In this embodiment, taking the electric vehicle industry chain as an example, a schematic diagram of the upstream and downstream of the industry chain is shown. (Reference) Figure 1 As shown, upstream companies include: battery cathode material suppliers, battery anode material suppliers, electrolyte suppliers, rare earth permanent magnet suppliers, bearing manufacturers, shaft suppliers, etc.; midstream companies include battery system suppliers and motor suppliers, etc.; downstream companies include vehicle manufacturers, etc.

[0029] Using a monthly statistical period, calculate the electricity consumption W1~W of enterprises at all levels over the past N months, starting from the current date. N and production capacity Q1~Q N .

[0030] Collect transaction information between enterprises at all levels, such as transaction information between battery cathode material suppliers, battery anode material suppliers, electrolyte suppliers and battery system suppliers; transaction information between permanent magnet suppliers, bearing manufacturers and shaft suppliers and motor suppliers; and transaction information between battery system suppliers and motor suppliers and vehicle manufacturers, etc.

[0031] Based on the transaction information between various enterprises, a one-way graph structure is constructed to represent the transmission relationship between enterprises in the industrial chain, and the weight information of each one-way edge in the one-way graph structure is determined. In this embodiment, based on the collected transaction information between various enterprises, a one-way graph structure is constructed to represent the transmission relationship between enterprises in the industrial chain, such as... Figure 2 As shown in the diagram. The directed edges in the one-way graph structure indicate the direction of transmission between firms.

[0032] The weight information of each directed edge in a one-way graph structure is determined based on the proportion of transactions between upstream enterprises and different downstream enterprises. For example, the directed edge E between nodes N1 and N7. 1,7 Directed edge E with nodes N1 and N8 1,8 weight w 1,7 and w 1,8 This provides the ratio of transaction volume between the company corresponding to N1 and the companies corresponding to N7 and N8, respectively, and w 1,7 and w 1,8 The sum is 1. By analogy, we can determine... Figure 3 The weights of each directed edge in the directed graph shown.

[0033] Collect the electricity consumption of upstream enterprises in the current cycle and the previous cycle, calculate the rate of change of electricity consumption of upstream enterprises in the current cycle, and input it into the attribute information of the initial node in the one-way graph structure. In this embodiment, the formula for calculating the rate of change in electricity consumption of upstream enterprises in the current period is as follows:

[0034] The electricity consumption change rate reflects the degree of change in the electricity consumption of the corresponding enterprise.

[0035] Therefore, the electricity consumption change rate of each enterprise in the Nth period can be used as the attribute information of the upstream enterprise's corresponding nodes N1~N6.

[0036] Thus constructing with Figure 2 The graph structure shown corresponds to the graph data. In this graph data, the attribute of nodes N1~N6 corresponding to upstream enterprises is the electricity consumption change rate of the corresponding enterprise in the current period (i.e., the Nth period), while the data of other nodes is empty. In addition, the attribute information of each edge is the weight information calculated in the above steps.

[0037] The unidirectional graph structure, the attribute information of the initial node, the weight information of the unidirectional edge, and the historical information of the midstream and downstream enterprises are input into the pre-trained graph convolutional network to extract the graph features of each enterprise node in the unidirectional graph structure. In this embodiment, a graph convolutional network is used to extract features from graph data and predict the transmission of changes in electricity consumption rates from upstream enterprises to midstream and downstream enterprises.

[0038] Figure 3 For the architecture of graph convolutional networks, refer to Figure 3 As shown, the graph convolutional network includes multiple message passing layers. These message passing layers are divided into three parts: the first part passes messages about the rate of change in electricity consumption based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industry chain (i.e., transmission of electricity consumption changes); the second part passes messages about the rate of change in electricity consumption by combining the node information of the target node (i.e., enhancement of node features); and the third part passes messages not only to nodes but also to edges (i.e., for feature extraction of the graph structure).

[0039] In addition, each message passing layer in the first part includes a first sensor MLP_1; each message passing layer in the second part includes a second sensor MLP_2; and each message passing layer in the third part includes a third sensor MLP_3 and a fourth sensor MLP_4.

[0040] The following sections will explain the three parts of the message passing layer. 1) Transmission of electrical changes This message passing layer updates the target node's features solely based on the features of the source node and its incoming edges. Specifically, let's take one instance of this message passing layer as an example: refer to Figure 4 As shown, when a message needs to be passed to a node of an upstream enterprise (e.g., node N1), only the node's own information is input into MLP_1 to obtain the node characteristics after the message is passed.

[0041] refer to Figure 5 As shown, when message passing is to be performed on a node (e.g., node N8) of a midstream or downstream enterprise, the message will be passed relative to the source node (e.g., nodes N1, N2, and N3) and the directed edge (e.g., edge E) between the source node and the node. 18 E 28 and E 38 The information is input into MLP_1 to obtain the node characteristics of the nodes of the midstream or downstream enterprise.

[0042] Similarly, for each message passing layer in this section, the above method is used to pass messages to each node. Thus, based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industry chain, the message transmission of electricity consumption change rate (i.e., electricity consumption change transmission) is carried out.

[0043] 2) Enhancement of node features This message passing layer, for midstream or downstream enterprises, performs message passing based on the source node corresponding to the target node, the corresponding incoming edges, and the characteristics of the target node itself, updating the node characteristics of the target node. Specifically, let's take a specific message passing layer as an example: refer to Figure 6 As shown, in the second message passing layer, when a message needs to be passed to the node of the upstream enterprise, only the characteristics of the node itself (N1) are input into MLP_2 to obtain the corresponding node characteristics.

[0044] refer to Figure 7 As shown, in the second part of the message passing layer, when a message needs to be passed to a node (e.g., node N8) of a midstream or downstream enterprise, the information of the source node (e.g., nodes N1, N2, and N3) relative to that node, and the directed edge between the source node and that node (e.g., edge E) are passed. 18 E 28 and E 38 The information of the node (node ​​N8) and its own information are input into MLP_2 to obtain the node characteristics of the node of the midstream or downstream enterprise.

[0045] Similarly, for each message passing layer in this section, the above method is used to pass messages to each node. Thus, the message passing layer combines the node information of the target node to pass messages about the electricity consumption change rate (i.e., the enhancement of node characteristics).

[0046] 3) Feature extraction of graph structures This message passing layer not only passes messages to nodes but also to edges. Specifically, let's take one example from this message passing layer: refer to Figure 8 As shown, in the third message passing layer, when a message needs to be passed to the upstream enterprise's node, only the node's own (N1) features are input into MLP_3 to obtain the corresponding node features.

[0047] refer to Figure 9 As shown, in the third part of the message passing layer, when a message needs to be passed to a node (e.g., node N8) of a midstream or downstream enterprise, the information of the source node (e.g., nodes N1, N2, and N3) relative to that node, and the directed edge between the source node and that node (e.g., edge E) are passed. 18 E 28 and E 38 The information of the node (node ​​N8) and its own information are input into MLP_3 to obtain the node characteristics of the node of the midstream or downstream enterprise.

[0048] In other words, the message passing operation for nodes in the third message passing layer is similar to that in the second message passing layer, except that a different perceptron is used.

[0049] refer to Figure 10 As shown, in the third part, the message passing layer, message passing also needs to be performed on edges. For example, for edge E... 18 The source node (N1) of the edge and the edge itself (E) 18 The information of the target node (N8) and the edge features of the edge are input into MLP_4 to obtain the edge features of the edge.

[0050] Similarly, for each message passing layer in this section, the above method is used to pass messages to each node and each edge.

[0051] Thus, through the message passing process of the three message passing layers described above, the graph features corresponding to the graph data are obtained.

[0052] The graph features of each enterprise node are input into a neural network to predict the rate of change in electricity consumption of each enterprise in subsequent periods, and to allocate power resources accordingly.

[0053] In this embodiment, the features of each node in the graph are input into a neural network model to predict the rate of change in electricity consumption of the enterprise corresponding to that node in the subsequent period (i.e., period N+1).

[0054] For example: Reference Figure 9As shown, the node features of node N7 corresponding to the midstream enterprise are input into the neural network, thereby outputting the rate of change in electricity consumption of the enterprise corresponding to that node in subsequent cycles (e.g., cycle N+1).

[0055] Similarly, this method can be used to determine the electricity consumption change rate of various enterprises in the midstream and downstream sectors.

[0056] After forecasting the electricity consumption change rate of various midstream and downstream enterprises, it is possible to further forecast the electricity consumption of midstream and downstream enterprises in subsequent cycles.

[0057] Therefore, based on the predicted electricity consumption of each enterprise within each regional power grid, the electricity consumption of each region in subsequent cycles can be determined. This allows for the allocation and scheduling of power resources in each region.

[0058] As a preferred embodiment of this example, the architecture of the graph convolutional network includes three message passing layers: a power consumption change rate passing layer, a node feature enhancement layer, and a power consumption change rate combined with edge feature passing layer. The electricity consumption change rate transmission layer transmits messages about electricity consumption change rates based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain. The node feature enhancement layer transmits messages about the rate of change in electricity consumption based on the node information of the target node. The electricity consumption change rate combined with edge feature transmission layer not only transmits the electricity consumption change rate message based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain, but also transmits edge features based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain combined with the weight information of each unidirectional edge.

[0059] In a preferred embodiment of this example, each enterprise node in the electricity consumption change rate transmission layer uses a first sensor MLP_1 for graph feature extraction; each node in the node feature enhancement layer uses a second sensor MLP_2 for feature extraction; and the electricity consumption change rate combined with edge feature transmission layer uses a third sensor MLP_3 for graph feature extraction at each node and a fourth sensor MLP_4 for edge feature extraction.

[0060] In a preferred embodiment of this example, when the first perceptron MLP_1 performs graph feature extraction, if the target node is an upstream node, only the information of the target node itself is input; if the target node is a mid-to-downstream node, the source node information of the target node and the directed edge information between the source node and the target node are input. When the second perceptron MLP_2 and the third perceptron MLP_3 extract graph features, if the target node is an upstream node, only the information of the target node itself is input. If the target node is a mid-to-downstream node, the information of the source node of the target node, the directed edge information between the source node and the target node, and the information of the target node itself are input. When the fourth perceptron MLP_4 performs edge feature extraction, it inputs the source node information of the target edge, the target edge's own information, and the target node's information.

[0061] In a preferred embodiment of this invention, the weight information of each unidirectional edge in the unidirectional graph structure is determined based on the proportion of transactions between the upstream enterprise and different downstream enterprises.

[0062] As a preferred embodiment of this example, it further includes calculating the correlation index between the production capacity and electricity consumption of each enterprise based on the electricity consumption information and production capacity information of each enterprise in each period. Then, calculate the change in electricity consumption based on the electricity consumption change rate of each enterprise in the subsequent cycle, and further calculate the change in production capacity.

[0063] As a preferred embodiment of this example, the specific method for calculating the correlation index between the production capacity and electricity consumption of each enterprise is as follows:

[0064] In the formula, This is a correlation index between production capacity and electricity consumption; To count the number of months; For the target number of months; The electricity consumption over n months; The production capacity is for n months.

[0065] In this embodiment, the production capacity of each enterprise is predicted based on its capacity correlation index in subsequent periods (e.g., the N+1th period): Specifically, in predicting the electricity consumption W of each enterprise in the subsequent period N+1 Then, the production capacity of each enterprise in subsequent cycles can be further predicted based on the following formula.

[0066] Since this formula only contains Q N+1 This is an unknown, so it can be solved.

[0067] Therefore, this plan can not only predict the transmission of changes in electricity consumption in the industrial chain, but also the transmission of production capacity in the industrial chain.

[0068] Example 2: A power resource allocation and forecasting system based on electricity consumption data analysis of the industrial chain includes: The data acquisition module obtains electricity consumption and production capacity information of various enterprises in the industrial chain at various cycles, as well as transaction information between various enterprises; The feature graph construction module constructs a one-way graph structure to represent the transmission relationship between enterprises in the industrial chain based on the transaction information between various enterprises, and determines the weight information of each one-way edge in the one-way graph structure. The feature graph initial parameter setting module collects the electricity consumption of upstream enterprises in the current cycle and the previous cycle, calculates the rate of change of electricity consumption of upstream enterprises in the current cycle, and inputs it into the unidirectional graph structure as the attribute information of the initial node. The graph feature extraction module inputs the one-way graph structure, the attribute information of the initial node, the weight information of the one-way edge, and the historical information of the midstream and downstream enterprises into the pre-trained graph convolutional network to extract the graph features of each enterprise node in the one-way graph structure. The power resource adjustment module inputs the graph features of each enterprise node into a neural network to predict the rate of change in electricity consumption of each enterprise in subsequent periods and allocates power resources accordingly.

[0069] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for predicting power resource allocation based on electricity consumption data analysis of the industrial chain, characterized in that, The specific steps include: Obtain electricity consumption and production capacity information of various enterprises in the industrial chain at various stages, as well as transaction information between various enterprises; Based on the transaction information between various enterprises, a one-way graph structure is constructed to represent the transmission relationship between enterprises in the industrial chain, and the weight information of each one-way edge in the one-way graph structure is determined. Collect the electricity consumption of upstream enterprises in the current cycle and the previous cycle, calculate the rate of change of electricity consumption of upstream enterprises in the current cycle, and input it into the attribute information of the initial node in the one-way graph structure. The unidirectional graph structure, the attribute information of the initial node, the weight information of the unidirectional edge, and the historical information of the midstream and downstream enterprises are input into the pre-trained graph convolutional network to extract the graph features of each enterprise node in the unidirectional graph structure. The graph features of each enterprise node are input into a neural network to predict the rate of change in electricity consumption of each enterprise in subsequent periods, and to allocate power resources accordingly.

2. The power resource allocation and prediction method based on industrial chain electricity consumption data analysis according to claim 1, characterized in that, The architecture of the graph convolutional network includes three message passing layers: a power consumption change rate passing layer, a node feature enhancement layer, and a power consumption change rate combined with edge feature passing layer. The electricity consumption change rate transmission layer transmits messages about electricity consumption change rates based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain. The node feature enhancement layer transmits messages about the rate of change in electricity consumption based on the node information of the target node. The electricity consumption change rate combined with edge feature transmission layer not only transmits the electricity consumption change rate message based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain, but also transmits edge features based on the transmission characteristics between nodes corresponding to enterprises at different levels of the industrial chain combined with the weight information of each unidirectional edge.

3. The power resource allocation and prediction method based on industrial chain electricity consumption data analysis according to claim 2, characterized in that, In the electricity consumption change rate transmission layer, the first perceptron MLP_1 is used to extract graph features at each enterprise node; in the node feature enhancement layer, the second perceptron MLP_2 is used to extract features at each node; in the electricity consumption change rate combined with edge feature transmission layer, the third perceptron MLP_3 is used to extract graph features at each node, and the fourth perceptron MLP_4 is used to extract edge features.

4. The power resource allocation and prediction method based on industrial chain electricity consumption data analysis according to claim 3, characterized in that, When the first perceptron MLP_1 performs graph feature extraction, if the target node is an upstream node, it only inputs the information of the target node itself; if the target node is a mid-to-downstream node, it inputs the source node information of the target node and the directed edge information between the source node and the target node. When the second perceptron MLP_2 and the third perceptron MLP_3 extract graph features, if the target node is an upstream node, only the information of the target node itself is input. If the target node is a mid-to-downstream node, the information of the source node of the target node, the directed edge information between the source node and the target node, and the information of the target node itself are input. When the fourth perceptron MLP_4 performs edge feature extraction, it inputs the source node information of the target edge, the target edge's own information, and the target node's information.

5. The power resource allocation and prediction method based on industrial chain electricity consumption data analysis according to claim 1, characterized in that, The weight information of each unidirectional edge in the unidirectional graph structure is determined according to the proportion of the transaction volume between the upstream enterprise and different downstream enterprises.

6. The power resource allocation and prediction method based on industrial chain electricity consumption data analysis according to claim 1, characterized in that, It also includes calculating the correlation index between the production capacity and electricity consumption of each enterprise based on the electricity consumption and production capacity information of each enterprise in each cycle; Then, calculate the change in electricity consumption based on the electricity consumption change rate of each enterprise in the subsequent cycle, and further calculate the change in production capacity.

7. The power resource allocation and prediction method based on industrial chain electricity consumption data analysis according to claim 6, characterized in that, The specific method for calculating the correlation index between the production capacity and electricity consumption of each enterprise is as follows: In the formula, This is a correlation index between production capacity and electricity consumption; To count the number of months; For the target number of months; The electricity consumption over n months; The production capacity is for n months.

8. A power resource allocation and forecasting system based on electricity consumption data analysis of the industrial chain, characterized in that, include: The data acquisition module obtains electricity consumption and production capacity information of various enterprises in the industrial chain at various cycles, as well as transaction information between various enterprises; The feature graph construction module constructs a one-way graph structure to represent the transmission relationship between enterprises in the industrial chain based on the transaction information between various enterprises, and determines the weight information of each one-way edge in the one-way graph structure. The feature graph initial parameter setting module collects the electricity consumption of upstream enterprises in the current cycle and the previous cycle, calculates the rate of change of electricity consumption of upstream enterprises in the current cycle, and inputs it into the unidirectional graph structure as the attribute information of the initial node. The graph feature extraction module inputs the one-way graph structure, the attribute information of the initial node, the weight information of the one-way edge, and the historical information of the midstream and downstream enterprises into the pre-trained graph convolutional network to extract the graph features of each enterprise node in the one-way graph structure. The power resource adjustment module inputs the graph features of each enterprise node into a neural network to predict the rate of change in electricity consumption of each enterprise in subsequent periods and allocates power resources accordingly.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the power resource allocation and prediction method based on power consumption data analysis of the industrial chain as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the power resource allocation and prediction method based on power consumption data analysis of the industrial chain as described in any one of claims 1 to 7.

Images