An artificial intelligence-based power grid service planning method and device, and a storage medium

Abstract

The application provides a power grid business planning method and device based on artificial intelligence and a storage medium, the method comprising the following steps: constructing a power grid heterogeneous structure topology graph based on electrical connection, power transmission or business correlation of all physical devices in the power grid, and mapping the power grid physical topology structure and business logic; training a power grid business planning intelligent model based on historical data of power grid operation; processing the power grid heterogeneous structure topology graph by using the trained power grid business planning intelligent model to obtain a planning feature matrix; taking power flow, voltage, line capacity and new energy output as constraints, taking power supply reliability and new energy consumption as collaborative optimization targets, and solving the planning feature matrix to obtain an optimal solution, and outputting the optimal solution as a decision scheme. The application improves the intelligent level of power grid planning and guarantees the safe, stable and economic operation of a new power system.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence and power grid planning integration technology, specifically to an artificial intelligence-based power grid business planning method, device, and storage medium. Background Technology

[0002] Traditional power grid business planning methods are mostly based on optimization models such as linear programming and genetic algorithms. They rely on manual extraction of power grid characteristics, making it difficult to accurately depict the nonlinear relationships of the power grid topology. Furthermore, they have poor adaptability to dynamically changing data such as load and renewable energy output, resulting in planning schemes with insufficient timeliness, weak coordination, and low practicality. They cannot meet the intelligent, refined, and dynamic planning requirements of new power systems.

[0003] In existing technologies, the weights in graph neural networks cannot reflect the physical characteristics of the power grid, and the attention weights cannot reflect the planning business type of the power grid. Therefore, to apply graph neural networks to power grid planning, it is necessary to design the corresponding parameters in the graph neural network according to the physical characteristics of the power grid. Summary of the Invention

[0004] In view of one or more technical defects in the prior art, the present invention proposes the following technical solution.

[0005] An artificial intelligence-based power grid business planning method, characterized in that the method includes: The construction steps involve building a heterogeneous topology diagram of the power grid based on electrical connections, power transmission, or business relationships among all physical devices in the power grid, and mapping the physical topology of the power grid to business logic. Training steps: Train a smart model for power grid business planning based on historical power grid operation data, and use the trained smart model for power grid business planning to process the heterogeneous topology map of the power grid to obtain the planning feature matrix; The decision-making process involves taking power flow, voltage, line capacity, and renewable energy output as constraints, and power supply reliability and renewable energy absorption as synergistic optimization objectives. The optimal solution is obtained from the planning feature matrix, and the optimal solution is output as the decision scheme.

[0006] Furthermore, the constructed heterogeneous power grid topology is as follows: G (t) ={V,E,A (t) H (t)}, where V={v1,v2,...,v N} represents a set of nodes, v i A node represents a busbar, grid equipment, load node, renewable energy plant, or business unit. E represents an edge set, where an edge indicates an electrical connection, power transmission, or business relationship between two nodes. (t)The characteristic matrix represents the edge characteristics, which are used to represent the impedance, current carrying capacity, loss, and transmission efficiency between nodes. H (t) The feature vectors representing nodes include voltage, capacity, load, renewable energy output, and planning constraints. The power grid is represented as a spatiotemporally heterogeneous topological graph sequence G={G (1) G (2) ,……,G (T)}, where t represents any time between 1 and T, N is the total number of nodes, N is an integer of at least 2, 1≤i≤N, and T is the current time.

[0007] Furthermore, the intelligent model for power grid business planning is a graph convolutional neural network constructed based on the heterogeneous topology of the power grid, and the graph convolutional neural network uses adaptive weights for forward propagation during training.

[0008] The adaptive weights are calculated as follows: ; The topological similarity between the i-th and j-th nodes is calculated using the feature vectors of the i-th and j-th nodes. This represents the electrical difference between the i-th and j-th nodes. ; This represents the degree of business correlation between the i-th and j-th nodes, which is dynamically determined based on the power grid planning tasks. Where 1≤j≤N, i≠j, i≠k, This represents a combination of nodes that have a connecting edge to the i-th node out of N nodes. This represents the impedance of the i-th node. This represents the impedance of the j-th node. It is a constant.

[0009] Furthermore, based on the type of power grid business plan, nodes that are important to the business plan are identified, and attention weights are applied to those important nodes.

[0010] Furthermore, if the type is capacity expansion, then the important nodes are power plants, substations, and large electricity consumers; if the type is renewable energy consumption, then the important nodes are renewable energy power plants and large electricity consumers; if the type is improving power supply reliability, then the important nodes are power plants, renewable energy power plants, substations, and large electricity consumers.

[0011] The attention weight The calculation method is as follows: ; in, Calculate the attention vector. These are learnable weight row vectors used to... Convert to a scalar Let i be the feature vector of node i. Bias vector, This represents the hyperbolic tangent activation function.

[0012] Furthermore, the graph convolutional neural network is trained using a joint loss function consisting of topological feature reconstruction loss, topological smoothness loss, and planning task supervision loss.

[0013] The joint loss function is:

[0014] in, This represents the topological information of the i-th node. This represents the topological information of the i-th node after graph embedding. , Let i and j represent the feature vectors of the i-th and j-th nodes, respectively. This represents the label value of the planned task at time t. This represents the predicted value of the planned task at time t. It is a constant between 0.1 and 0.3. It is a constant between 0.75 and 1.0.

[0015] Furthermore, the constraints of power flow, voltage, line capacity, and renewable energy output specifically mean: power flow constraint means the sum of active and reactive current inputs and outputs at all nodes is 0; voltage constraint means voltage fluctuates within a certain range; voltage constraint means it must not exceed the maximum design capacity of each line; and renewable energy output constraint means it must not exceed the sum of renewable energy outputs. The power supply reliability calculation method is as follows: ,in, The proportion of normal power supply throughout the year, Average duration of power outages throughout the year The number of hours in a year; the optimization goal of renewable energy consumption is to maximize the consumption of electricity generated by renewable energy. The optimal solution is obtained by using the Pareto method, and then the optimal power grid business planning scheme is determined by comprehensively scoring the Pareto optimal solution based on the analytic hierarchy process.

[0016] The present invention also proposes an artificial intelligence-based power grid business planning device, characterized in that the device comprises: The construction unit constructs a heterogeneous topology diagram of the power grid based on electrical connections, power transmission, or business relationships among all physical devices in the power grid, mapping the physical topology of the power grid to business logic; Training Unit: Train a smart model for power grid business planning based on historical power grid operation data, and use the trained smart model for power grid business planning to process the heterogeneous topology map of the power grid to obtain the planning feature matrix; The decision-making unit, constrained by power flow, voltage, line capacity, and renewable energy output, and with power supply reliability and renewable energy absorption as the synergistic optimization objectives, seeks the optimal solution for the planning feature matrix and outputs the optimal solution as the decision scheme.

[0017] Furthermore, the constructed heterogeneous power grid topology is as follows: G (t) ={V,E,A (t) H (t)}, where V={v1,v2,...,v N} represents a set of nodes, v i A node represents a busbar, grid equipment, load node, renewable energy plant, or business unit. E represents an edge set, where an edge indicates an electrical connection, power transmission, or business relationship between two nodes. (t) The characteristic matrix represents the edge characteristics, which are used to represent the impedance, current carrying capacity, loss, and transmission efficiency between nodes. H (t) The feature vectors representing nodes include voltage, capacity, load, renewable energy output, and planning constraints. The power grid is represented as a spatiotemporally heterogeneous topological graph sequence G={G (1) G (2) ,……,G (T)}, where t represents any time between 1 and T, N is the total number of nodes, N is an integer of at least 2, 1≤i≤N, and T is the current time.

[0018] Furthermore, the intelligent model for power grid business planning is a graph convolutional neural network constructed based on the heterogeneous topology of the power grid, and the graph convolutional neural network uses adaptive weights for forward propagation during training.

[0019] The attention weight The calculation method is as follows: ; in, Calculate the attention vector. These are learnable weight row vectors used to... Convert to a scalar Let i be the feature vector of node i. Bias vector, This represents the hyperbolic tangent activation function.

[0020] Furthermore, the adaptive weights are calculated as follows: ; The topological similarity between the i-th and j-th nodes is calculated using the feature vectors of the i-th and j-th nodes. This represents the electrical difference between the i-th and j-th nodes. ; This represents the degree of business correlation between the i-th and j-th nodes, which is dynamically determined based on the power grid planning tasks. Where 1≤j≤N, i≠j, i≠k, This represents a combination of nodes that have a connecting edge to the i-th node out of N nodes. This represents the impedance of the i-th node. This represents the impedance of the j-th node. It is a constant.

[0021] Furthermore, based on the type of power grid business plan, nodes that are important to the business plan are identified, and attention weights are applied to those important nodes.

[0022] Furthermore, the graph convolutional neural network is trained using a joint loss function consisting of topological feature reconstruction loss, topological smoothness loss, and planning task supervision loss.

[0023] The joint loss function is:

[0024] in, This represents the topological information of the i-th node. This represents the topological information of the i-th node after graph embedding. , Let i and j represent the feature vectors of the i-th and j-th nodes, respectively. This represents the label value of the planned task at time t. This represents the predicted value of the planned task at time t. It is a constant between 0.1 and 0.3. It is a constant between 0.75 and 1.0.

[0025] Furthermore, the constraints of power flow, voltage, line capacity, and renewable energy output specifically mean: power flow constraint means the sum of active and reactive current inputs and outputs at all nodes is 0; voltage constraint means voltage fluctuates within a certain range; voltage constraint means it must not exceed the maximum design capacity of each line; and renewable energy output constraint means it must not exceed the sum of renewable energy outputs. The power supply reliability calculation method is as follows: ,in, The proportion of normal power supply throughout the year, Average duration of power outages throughout the year The number of hours in a year; the optimization goal of renewable energy consumption is to maximize the consumption of electricity generated by renewable energy. The optimal solution is obtained by using the Pareto method, and then the optimal power grid business planning scheme is determined by comprehensively scoring the Pareto optimal solution based on the analytic hierarchy process.

[0026] The present invention also proposes a computer-readable storage medium storing computer program code, which, when executed by a computer, performs any of the methods described above.

[0027] The technical advantages of this invention are as follows: This invention provides an artificial intelligence-based power grid business planning method, device, and storage medium. The method includes: a construction step S101, constructing a heterogeneous power grid topology map based on electrical connections, power transmission, or business relationships among all physical devices in the power grid, and mapping the physical topology to business logic; a training step S102, training a power grid business planning intelligent model based on historical power grid operation data, and processing the heterogeneous power grid topology map using the trained intelligent model to obtain a planning feature matrix; and a decision-making step S103, using power flow, voltage, line capacity, and renewable energy output as constraints, and power supply reliability and renewable energy absorption as collaborative optimization objectives, finding the optimal solution for the planning feature matrix, and outputting the optimal solution as a decision scheme. This invention applies graph neural networks to power grid business planning, enabling autonomous extraction of power grid topology features, deep integration of multiple business needs, and rapid response to dynamic planning decisions, thereby improving the intelligence level of power grid planning and ensuring the safe, stable, and economical operation of the new power system. Attached Figure Description

[0028] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0029] Figure 1 This is a flowchart of an artificial intelligence-based power grid business planning method according to an embodiment of the present invention.

[0030] Figure 2 This is a structural diagram of an artificial intelligence-based power grid business planning device according to an embodiment of the present invention. Detailed Implementation

[0031] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0032] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0033] Figure 1 This invention illustrates an artificial intelligence-based power grid business planning method, which includes: In step S101, a heterogeneous topology diagram of the power grid is constructed based on electrical connections, power transmission, or business relationships among all physical devices in the power grid, and the physical topology of the power grid is mapped to business logic. Training step S102: Train the intelligent model for power grid business planning based on historical data of power grid operation, and use the trained intelligent model for power grid business planning to process the heterogeneous topology map of the power grid to obtain the planning feature matrix; In decision-making step S103, the optimal solution is obtained for the planning feature matrix, with power flow, voltage, line capacity and new energy output as constraints and power supply reliability and new energy consumption as the synergistic optimization objectives. The optimal solution is then output as the decision scheme.

[0034] This invention addresses the shortcomings of the background technology, namely, that traditional power grid business planning methods, which are mostly based on optimization models such as linear programming and genetic algorithms, rely on manual extraction of power grid features, making it difficult to accurately characterize the nonlinear relationships of the power grid topology, and have poor adaptability to dynamically changing data such as load and renewable energy output. This results in planning schemes with insufficient timeliness, weak coordination, and low practicality, failing to meet the intelligent, refined, and dynamic planning requirements of new power systems. The invention creatively proposes constructing a heterogeneous power grid topology map based on electrical connections, power transmission, or business relationships among all physical devices in the power grid. This maps the physical topology of the power grid to business logic. Then, a trained intelligent power grid business planning model processes the heterogeneous topology map to obtain a planning feature matrix. Finally, with power flow, voltage, line capacity, and renewable energy output as constraints, and power supply reliability and renewable energy absorption as collaborative optimization objectives, the optimal solution is obtained from the planning feature matrix, and this optimal solution is output as the decision scheme. This enables the effective capture of topological associations and feature dependencies between nodes, which is highly compatible with the natural topological structure of the power grid. Applying graph neural networks to power grid business planning can achieve autonomous extraction of power grid topological features, deep integration of multiple business needs, and rapid response of dynamic planning decisions, thereby improving the intelligence level of power grid planning and ensuring the safe, stable, and economical operation of the new power system. This is one of the important inventive concepts of this invention.

[0035] In one embodiment, to accurately extract topological associations and feature dependencies between nodes, this invention proposes a method for constructing a heterogeneous power grid topology graph. The constructed heterogeneous power grid topology graph is: G (t) ={V,E,A(t) H (t)}, where V={v1,v2,...,v N} represents a set of nodes, v i A node represents a busbar, grid equipment, load node, renewable energy plant, or business unit. E represents an edge set, where an edge indicates an electrical connection, power transmission, or business relationship between two nodes. (t) The characteristic matrix represents the edge characteristics, which are used to represent the impedance, current carrying capacity, loss, and transmission efficiency between nodes. H (t) The feature vectors representing nodes include voltage, capacity, load, renewable energy output, and planning constraints. The power grid is represented as a spatiotemporally heterogeneous topological graph sequence G={G (1) G (2) ,……,G (T)}, where t represents any time between 1 and T, N is the total number of nodes, N is an integer of at least 2, 1≤i≤N, and T is the current time.

[0036] The proposed method for modeling heterogeneous topology of power grids for multi-service coupling breaks through the limitations of traditional single topology modeling. It unifies the modeling of power grid equipment, load nodes, new energy power plants, and service units as heterogeneous nodes, and constructs a multi-dimensional feature graph model containing electrical attributes, service attributes, and topological associations. This achieves a deep mapping between the physical topology of the power grid and the business logic, and can dynamically reflect the operating status of the power grid at any time, thereby making the power grid task planning more accurate. This is another important inventive concept of the present invention.

[0037] In one embodiment, the intelligent model for power grid business planning is a graph convolutional neural network constructed based on the heterogeneous topology of the power grid, and the graph convolutional neural network uses adaptive weights for forward propagation during training.

[0038] The adaptive weights are calculated as follows: ; The topological similarity between the i-th and j-th nodes is calculated using the feature vectors of the i-th and j-th nodes. This represents the electrical difference between the i-th and j-th nodes. ; This represents the degree of business correlation between the i-th and j-th nodes, which is dynamically determined based on the power grid planning tasks. Where 1≤j≤N, i≠j, i≠k, This represents a combination of nodes that have a connecting edge to the i-th node out of N nodes. This represents the impedance of the i-th node. This represents the impedance of the j-th node. It is a constant.

[0039] In this invention, the weights of traditional graph convolutional neural networks (GCNNs) are improved. The weights of traditional GCNNs are determined solely by topological connections, which fails to reflect the electrical characteristics and business importance of devices in the power grid. Consequently, the inference cannot be correlated with the physical characteristics of the power grid. This invention improves the forward propagation function by adding adaptive weights to the traditional forward propagation function. Furthermore, it constructs an adaptive weight calculation method based on the topological similarity between nodes, the electrical differences between nodes, and the business relevance between nodes. This ensures that the trained GCNN is more consistent with the physical characteristics of the power grid, which is another important inventive concept of this invention.

[0040] In one embodiment, if nodes that are important to the impact of the power grid business plan are determined based on the type of the business plan, then attention weights are applied to those important nodes.

[0041] In one embodiment, if the type is capacity expansion, the important nodes are power plants, substations, and large electricity consumers; if the type is renewable energy consumption, the important nodes are renewable energy power plants and large electricity consumers; if the type is improving power supply reliability, the important nodes are power plants, renewable energy power plants, substations, and large electricity consumers.

[0042] The attention weight The calculation method is as follows: ; in, Calculate the attention vector. These are learnable weight row vectors used to... Convert to a scalar Let i be the feature vector of node i. Bias vector, This represents the hyperbolic tangent activation function.

[0043] In order to deeply associate the planned business types with nodes, this invention creatively proposes to add attention to nodes related to business types, that is, to multiply their feature vectors by a weight to increase their feature vectors, thereby ensuring that different nodes are given corresponding weights for different business types. This also simplifies the way attention weights are calculated in the prior art, making the trained graph neural network more in line with the physical characteristics of the power grid. Furthermore, this invention proposes a specific formula for calculating attention weights, which is another important technical concept of this invention.

[0044] In one embodiment, the graph convolutional neural network is trained using a joint loss function consisting of topological feature reconstruction loss, topological smoothness loss, and planning task supervision loss.

[0045] The joint loss function is:

[0046] in, This represents the topological information of the i-th node. This represents the topological information of the i-th node after graph embedding. , Let i and j represent the feature vectors of the i-th and j-th nodes, respectively. This represents the label value of the planned task at time t. This represents the predicted value of the planned task at time t. It is a constant between 0.1 and 0.3. It is a constant between 0.75 and 1.0.

[0047] To adapt to the complex task of power grid business planning, this invention proposes a multi-task joint loss function. This joint loss function is constructed based on topology feature reconstruction loss, topology smoothness loss, and planning task supervision loss, thereby enabling the trained graph neural network to minimize the joint loss and ensuring that the output embedding simultaneously satisfies topology fidelity, temporal regularity, and planning accuracy. This is another important technical concept of this invention.

[0048] In one embodiment, the constraints of power flow, voltage, line capacity, and renewable energy output specifically mean: power flow constraint means the sum of active and reactive current inputs and outputs at all nodes is 0; voltage constraint means the voltage fluctuates within a certain range; voltage constraint means it must not exceed the maximum design capacity of each line; and renewable energy output constraint means it must not exceed the sum of renewable energy outputs. The power supply reliability calculation method is as follows: ,in, The proportion of normal power supply throughout the year, Average duration of power outages throughout the year The number of hours in a year; the optimization goal of renewable energy consumption is to maximize the consumption of electricity generated by renewable energy. The optimal solution is obtained by using the Pareto method, and then the optimal power grid business planning scheme is determined by comprehensively scoring the Pareto optimal solution based on the analytic hierarchy process.

[0049] In this invention, the optimization solution is based on reliability and absorption rate as objectives. The Pareto method can be used to solve the problem and obtain a set of Pareto optimal planning schemes. Then, the comprehensive optimal scheme is determined by the analytic hierarchy process or the entropy weight method, thereby ensuring that the generated power grid scheme is more in line with actual needs. This is another important inventive concept of this invention.

[0050] Figure 2 An artificial intelligence-based power grid business planning device of the present invention is shown, the device comprising: Construction unit 201 constructs a heterogeneous topology diagram of the power grid based on electrical connections, power transmission, or business relationships among all physical devices in the power grid, and maps the physical topology of the power grid to business logic; Training Unit 202: Train a smart model for power grid business planning based on historical data of power grid operation, and use the trained smart model for power grid business planning to process the heterogeneous topology map of the power grid to obtain the planning feature matrix; Decision unit 203, with power flow, voltage, line capacity and new energy output as constraints, and with power supply reliability and new energy absorption as collaborative optimization objectives, finds the optimal solution for the planning feature matrix and outputs the optimal solution as the decision scheme.

[0051] This invention addresses the shortcomings of the background technology, namely, that traditional power grid business planning methods, which are mostly based on optimization models such as linear programming and genetic algorithms, rely on manual extraction of power grid features, making it difficult to accurately characterize the nonlinear relationships of the power grid topology, and have poor adaptability to dynamically changing data such as load and renewable energy output. This results in planning schemes with insufficient timeliness, weak coordination, and low practicality, failing to meet the intelligent, refined, and dynamic planning requirements of new power systems. The invention creatively proposes constructing a heterogeneous power grid topology map based on electrical connections, power transmission, or business relationships among all physical devices in the power grid. This maps the physical topology of the power grid to business logic. Then, a trained intelligent power grid business planning model processes the heterogeneous topology map to obtain a planning feature matrix. Finally, with power flow, voltage, line capacity, and renewable energy output as constraints, and power supply reliability and renewable energy absorption as collaborative optimization objectives, the optimal solution is obtained from the planning feature matrix, and this optimal solution is output as the decision scheme. This enables the effective capture of topological associations and feature dependencies between nodes, which is highly compatible with the natural topological structure of the power grid. Applying graph neural networks to power grid business planning can achieve autonomous extraction of power grid topological features, deep integration of multiple business needs, and rapid response of dynamic planning decisions, thereby improving the intelligence level of power grid planning and ensuring the safe, stable, and economical operation of the new power system. This is one of the important inventive concepts of this invention.

[0052] In one embodiment, to accurately extract topological associations and feature dependencies between nodes, this invention proposes a method for constructing a heterogeneous power grid topology graph. The constructed heterogeneous power grid topology graph is: G (t) ={V,E,A (t) H (t)}, where V={v1,v2,...,v N} represents a set of nodes, vi A node represents a busbar, grid equipment, load node, renewable energy plant, or business unit. E represents an edge set, where an edge indicates an electrical connection, power transmission, or business relationship between two nodes. (t) The characteristic matrix represents the edge characteristics, which are used to represent the impedance, current carrying capacity, loss, and transmission efficiency between nodes. H (t) The feature vectors representing nodes include voltage, capacity, load, renewable energy output, and planning constraints. The power grid is represented as a spatiotemporally heterogeneous topological graph sequence G={G (1) G (2) ,……,G (T)}, where t represents any time between 1 and T, N is the total number of nodes, N is an integer of at least 2, 1≤i≤N, and T is the current time.

[0053] The proposed method for modeling heterogeneous topology of power grids for multi-service coupling breaks through the limitations of traditional single topology modeling. It unifies the modeling of power grid equipment, load nodes, new energy power plants, and service units as heterogeneous nodes, and constructs a multi-dimensional feature graph model containing electrical attributes, service attributes, and topological associations. This achieves a deep mapping between the physical topology of the power grid and the business logic, and can dynamically reflect the operating status of the power grid at any time, thereby making the power grid task planning more accurate. This is another important inventive concept of the present invention.

[0054] In one embodiment, the intelligent model for power grid business planning is a graph convolutional neural network constructed based on the heterogeneous topology of the power grid, and the graph convolutional neural network uses adaptive weights for forward propagation during training.

[0055] The adaptive weights are calculated as follows: ; The topological similarity between the i-th and j-th nodes is calculated using the feature vectors of the i-th and j-th nodes. This represents the electrical difference between the i-th and j-th nodes. ; This represents the degree of business correlation between the i-th and j-th nodes, which is dynamically determined based on the power grid planning tasks. Where 1≤j≤N, i≠j, i≠k, This represents a combination of nodes that have a connecting edge to the i-th node out of N nodes. This represents the impedance of the i-th node. This represents the impedance of the j-th node. It is a constant.

[0056] In this invention, the weights of traditional graph convolutional neural networks (GCNNs) are improved. The weights of traditional GCNNs are determined solely by topological connections, which fails to reflect the electrical characteristics and business importance of devices in the power grid. Consequently, the inference cannot be correlated with the physical characteristics of the power grid. This invention improves the forward propagation function by adding adaptive weights to the traditional forward propagation function. Furthermore, it constructs an adaptive weight calculation method based on the topological similarity between nodes, the electrical differences between nodes, and the business relevance between nodes. This ensures that the trained GCNN is more consistent with the physical characteristics of the power grid, which is another important inventive concept of this invention.

[0057] In one embodiment, if nodes that are important to the impact of the power grid business plan are determined based on the type of the business plan, then attention weights are applied to those important nodes.

[0058] In one embodiment, if the type is capacity expansion, the important nodes are power plants, substations, and large electricity consumers; if the type is renewable energy consumption, the important nodes are renewable energy power plants and large electricity consumers; if the type is improving power supply reliability, the important nodes are power plants, renewable energy power plants, substations, and large electricity consumers.

[0059] The attention weight The calculation method is as follows: ; in, Calculate the attention vector. These are learnable weight row vectors used to... Convert to a scalar Let i be the feature vector of node i. Bias vector, This represents the hyperbolic tangent activation function.

[0060] In order to deeply associate the planned business types with nodes, this invention creatively proposes to add attention to nodes related to business types, that is, to multiply their feature vectors by a weight to increase their feature vectors, thereby ensuring that different nodes are given corresponding weights for different business types. This also simplifies the way attention weights are calculated in the prior art, making the trained graph neural network more in line with the physical characteristics of the power grid. Furthermore, this invention proposes a specific formula for calculating attention weights, which is another important technical concept of this invention.

[0061] In one embodiment, the graph convolutional neural network is trained using a joint loss function consisting of topological feature reconstruction loss, topological smoothness loss, and planning task supervision loss.

[0062] The joint loss function is:

[0063] in, This represents the topological information of the i-th node. This represents the topological information of the i-th node after graph embedding. , Let i and j represent the feature vectors of the i-th and j-th nodes, respectively. This represents the label value of the planned task at time t. This represents the predicted value of the planned task at time t. It is a constant between 0.1 and 0.3. It is a constant between 0.75 and 1.0.

[0064] To adapt to the complex task of power grid business planning, this invention proposes a multi-task joint loss function. This joint loss function is constructed based on topology feature reconstruction loss, topology smoothness loss, and planning task supervision loss, thereby enabling the trained graph neural network to minimize the joint loss and ensuring that the output embedding simultaneously satisfies topology fidelity, temporal regularity, and planning accuracy. This is another important technical concept of this invention.

[0065] In one embodiment, the constraints of power flow, voltage, line capacity, and renewable energy output specifically mean: power flow constraint means the sum of active and reactive current inputs and outputs at all nodes is 0; voltage constraint means the voltage fluctuates within a certain range; voltage constraint means it must not exceed the maximum design capacity of each line; and renewable energy output constraint means it must not exceed the sum of renewable energy outputs. The power supply reliability calculation method is as follows: ,in, The proportion of normal power supply throughout the year, Average duration of power outages throughout the year The number of hours in a year; the optimization goal of renewable energy consumption is to maximize the consumption of electricity generated by renewable energy. The optimal solution is obtained by using the Pareto method, and then the optimal power grid business planning scheme is determined by comprehensively scoring the Pareto optimal solution based on the analytic hierarchy process.

[0066] In this invention, the optimization solution is based on reliability and absorption rate as objectives. The Pareto method can be used to solve the problem and obtain a set of Pareto optimal planning schemes. Then, the comprehensive optimal scheme is determined by the analytic hierarchy process or the entropy weight method, thereby ensuring that the generated power grid scheme is more in line with actual needs. This is another important inventive concept of this invention.

[0067] One embodiment of the present invention provides a computer storage medium storing a computer program. When the computer program on the computer storage medium is executed by a processor, the above-described method is implemented. The computer storage medium may be a hard disk, DVD, CD, flash memory, or other storage device.

[0068] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0069] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the apparatus described in various embodiments or some parts of the embodiments of this application.

[0070] Finally, it should be noted that the above embodiments are for illustration only and not for limiting the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention without departing from the spirit and scope of the present invention. Any modifications or partial substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A power grid business planning method based on artificial intelligence, characterized in that, The method includes: The construction steps involve building a heterogeneous topology diagram of the power grid based on electrical connections, power transmission, or business relationships among all physical devices in the power grid. Training steps: Train a smart model for power grid business planning based on historical power grid operation data, and use the trained smart model for power grid business planning to process the heterogeneous topology map of the power grid to obtain the planning feature matrix; The decision-making process involves taking power flow, voltage, line capacity, and renewable energy output as constraints, and power supply reliability and renewable energy consumption as synergistic optimization objectives. The optimal solution is obtained from the planning feature matrix, and the optimal solution is output as the decision scheme. The constructed heterogeneous power grid topology is as follows: G (t) ={V,E,A (t) H (t) }, where V={v1,v2,...,v N } represents a set of nodes, v i A node represents a busbar, grid equipment, load node, renewable energy plant, or business unit. E represents an edge set, where an edge indicates an electrical connection, power transmission, or business relationship between two nodes. (t) The characteristic matrix represents the edge characteristics, which are used to represent the impedance, current carrying capacity, loss, and transmission efficiency between nodes. H (t) The feature vectors representing nodes include voltage, capacity, load, renewable energy output, and planning constraints. The power grid is represented as a spatiotemporally heterogeneous topological graph sequence G={G (1) G (2) ,……,G (T) }, where t represents any time between 1 and T, N is the total number of nodes, N is an integer of at least 2, 1≤i≤N, and T is the current time; The intelligent model for power grid business planning is a graph convolutional neural network constructed based on the heterogeneous topology of the power grid. The graph convolutional neural network uses adaptive weights for forward propagation during training. The adaptive weights are calculated as follows: ; The topological similarity between the i-th and j-th nodes is calculated using the feature vectors of the i-th and j-th nodes. This represents the electrical difference between the i-th and j-th nodes. ; This represents the degree of business correlation between the i-th and j-th nodes, which is dynamically determined based on the power grid planning tasks. Where 1≤j≤N, i≠j, i≠k, This represents a combination of nodes that have a connecting edge to the i-th node out of N nodes. This represents the impedance of the i-th node. This represents the impedance of the j-th node. It is a constant.

2. The method according to claim 1, characterized in that, If the type of power grid business plan determines the nodes that are important to the business plan, then attention weights are applied to those nodes.

3. The method according to claim 2, characterized in that, The graph convolutional neural network is trained using a joint loss function consisting of topological feature reconstruction loss, topological smoothness loss, and planning task supervision loss.

4. A power grid business planning device based on artificial intelligence, characterized in that, The device includes: The construction unit builds a heterogeneous topology diagram of the power grid based on electrical connections, power transmission, or business relationships among all physical devices in the power grid. Training Unit: Train a smart model for power grid business planning based on historical power grid operation data, and use the trained smart model for power grid business planning to process the heterogeneous topology map of the power grid to obtain the planning feature matrix; The decision-making unit, with power flow, voltage, line capacity and new energy output as constraints, and with power supply reliability and new energy consumption as the synergistic optimization objectives, seeks the optimal solution of the planning feature matrix and outputs the optimal solution as the decision scheme. The constructed heterogeneous power grid topology is as follows: G (t) ={V,E,A (t) H (t) }, where V={v1,v2,...,v N } represents a set of nodes, v i A node represents a busbar, grid equipment, load node, renewable energy plant, or business unit. E represents an edge set, where an edge indicates an electrical connection, power transmission, or business relationship between two nodes. (t) The characteristic matrix represents the edge characteristics, which are used to represent the impedance, current carrying capacity, loss, and transmission efficiency between nodes. H (t) The feature vectors representing nodes include voltage, capacity, load, renewable energy output, and planning constraints. The power grid is represented as a spatiotemporally heterogeneous topological graph sequence G={G (1) G (2) ,……,G (T) }, where t represents any time between 1 and T, N is the total number of nodes, N is an integer of at least 2, 1≤i≤N, and T is the current time; The intelligent model for power grid business planning is a graph convolutional neural network constructed based on the heterogeneous topology of the power grid. The graph convolutional neural network uses adaptive weights for forward propagation during training. The adaptive weights are calculated as follows: ; The topological similarity between the i-th and j-th nodes is calculated using the feature vectors of the i-th and j-th nodes. This represents the electrical difference between the i-th and j-th nodes. ; This represents the degree of business correlation between the i-th and j-th nodes, which is dynamically determined based on the power grid planning tasks. Where 1≤j≤N, i≠j, i≠k, This represents a combination of nodes that have a connecting edge to the i-th node out of N nodes. This represents the impedance of the i-th node. This represents the impedance of the j-th node. It is a constant.

5. The apparatus according to claim 4, characterized in that, If the type of power grid business plan determines the nodes that are important to the business plan, then attention weights are applied to those nodes.

6. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1-3.

Images