Power distribution network resource scheduling management method and system based on large model calculation

Abstract

The invention discloses a power distribution network resource scheduling management method and system based on large model calculation, and relates to the technical field of power grids, and the method comprises the steps: completing the inheritance of a self-neural network base class through a class, processing the relation in a power distribution network, carrying out the feature extraction and mapping through three full connection layers, and providing a nonlinear expression capability through a ReLU activation function; and a hierarchical progressive processing mechanism is adopted to complete the intelligent scheduling decision of the power distribution network resources. According to the method, the operation reliability and safety of the power distribution network are remarkably improved, the intelligent level and the management efficiency of power distribution network resource scheduling are greatly improved, dynamic topology self-adaption and multi-objective optimization are achieved, the performance advantage scheduling response time is shortened by 50%, the economic benefit is improved by 30%, the system reliability is improved by 25%, and the distributed energy consumption rate is improved by 40%; the system is high in practical value adaptability, can be popularized and applied, is low in maintenance cost, is good in expandability, has commercial potential, and provides powerful support for the efficient operation of a modern power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power grids, and in particular to a method and system for dispatching and managing distribution network resources based on large model calculation. Background Art

[0002] With the development of digital information technology, power grid companies continue to promote the construction, operation, and production technology upgrades of intelligent measurement and control universal terminals, with the overall goal of "improving quality and efficiency", and carry out intelligent and integrated construction of intelligent measurement and control universal terminals, giving full play to the role of data as a production factor in the substation field, and strongly supporting the upgrading of digital grids, digital enterprises, digital services, and digital industries. Among them, intelligent measurement and control universal terminals can include various types of power grid sensors.

[0003] At present, the existing technologies for resource dispatching and management of distribution networks mainly have the following problems: traditional dispatching methods are difficult to effectively handle high-dimensional and nonlinear distribution network resource optimization problems; existing deep learning methods are not adaptable enough to dynamic topological changes in distribution networks, and lack the ability to coordinate and dispatch new resources such as distributed energy and energy storage in a unified manner; existing methods are difficult to achieve a balance between multi-objective trade-offs and real-time response; dispatching decisions are not interpretable enough, and it is difficult to gain the trust of operators. Summary of the invention

[0004] In view of the problems existing in the existing distribution network resource dispatching management method and system based on large model calculation, the present invention is proposed.

[0005] Therefore, the present invention solves the problem that the prior art is difficult to achieve multi-objective trade-offs and balanced scheduling decisions in real-time response, and it is difficult to gain the trust of operators. The present invention adopts a distribution network resource scheduling management method and system based on large model calculation to solve this problem.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a distribution network resource dispatching and management method based on large model calculation, which includes: adopting an improved heterogeneous graph neural network structure, inheriting from a neural network base class through a class, and processing the relationship in the distribution network; using a language writing module to construct a three-layer intelligent agent collaborative decision-making mechanism, performing feature extraction and mapping through three fully connected layers, and using a ReLU activation function to provide nonlinear expression capabilities; for the dispatching decision generation under the normal operating state of the distribution network, a hierarchical and progressive processing mechanism is adopted to complete the intelligent dispatching decision of the distribution network resources.

[0008] As a preferred solution of the distribution network resource scheduling management method based on large model calculation of the present invention, the improved heterogeneous graph neural network structure includes an organic combination of node feature encoding, edge feature encoding and multi-head graph attention mechanism, feature encoding includes designing a three-layer deep encoder, setting a dropout rate of 0.1 to prevent overfitting, and using different types of distribution network equipment node features to complete the unified characterization of heterogeneous nodes. The node feature encoder adopts a residual connection structure to retain the original feature information. The edge feature encoding includes a two-layer EdgeEncoder structure to process the line parameter features in the distribution network. The encoder extracts key information of line impedance, flow and capacity, and converts it into a representation vector consistent with the node feature dimension. The edge feature encoder also integrates an attention mechanism so that the model can adaptively pay attention to different types of edge features.

[0009] As a preferred solution of the distribution network resource dispatching management method based on large model calculation of the present invention, wherein: the processing of the relationship in the distribution network includes encoding the input node feature x to obtain the initial node embedding vector h, encoding the edge feature edge_attr to obtain the edge representation vector edge_h, and through a four-layer graph attention network, comprehensively considering the local and global topological relationships of the nodes, iteratively updating the feature representation of the nodes, introducing the modulation effect of the edge feature, considering the similarity of the node features when calculating the attention weights between nodes, and taking the features of the connecting edges as modulation factors;

[0010] The modulation factor includes the system adopting a dual detection strategy to accurately compare node features and edge features. During the node feature detection process, the differences between nodes in the new and old topological maps are determined, and whether there are new or deleted nodes. For co-existing nodes, the relative error of the feature vector is calculated to determine whether the node features have changed, and the detection sensitivity is flexibly adjusted through configurable threshold parameters.

[0011] As a preferred solution of the distribution network resource dispatching management method based on large model calculation of the present invention, the attention mechanism includes adopting a lightweight neural network structure, inputting local state information and decision information of the upper-layer coordination agent, performing feature fusion through a 128-dimensional hidden layer, setting the action space dimension to one-fourth of the global action space, and introducing a residual connection mechanism;

[0012] The introduction of the residual connection mechanism includes generating a scheduling strategy based on the overall state of the system, combining the regional coordination agent with the global strategy and the regional state, extracting regional features through the _get_region_state function, generating regional coordination instructions, executing the agent to fuse local state information and coordination instructions, completing control actions, and establishing a two-way information flow mechanism. The two-way information flow mechanism includes a downward decision flow to ensure the effective transmission of the strategy, and the upward state feedback provides a basis for upper-level decision-making. Through the two-way information flow, the system completes dynamic optimization and adjustment.

[0013] As a preferred solution of the distribution network resource dispatching management method based on large model calculation of the present invention, the three-layer intelligent agent collaborative decision-making mechanism is constructed, including setting a global dispatching intelligent agent at the top layer, responsible for the formulation of the overall dispatching strategy, configuring multiple regional coordination intelligent agents in the middle layer, completing the coordinated optimization of resources in the region, and deploying local execution intelligent agents at the bottom layer to execute control instructions;

[0014] Feature extraction and mapping are performed through three fully connected layers. The hidden layer dimension is 512. The ReLU activation function is used to provide nonlinear expression capabilities. The output layer generates the action space probability distribution through the Softmax function to complete the probability modeling of the global scheduling strategy.

[0015] The probability modeling includes adopting the PPO algorithm to design the architecture, introducing the adaptive clipping ratio and multiple loss function optimization mechanism; in the algorithm initialization stage, the network model, optimizer and clipping ratio framework are constructed, the network model includes adopting the deep neural network structure to approximate the policy function and the value function, the optimizer selects the Adam optimizer, and the clipping ratio is set to 0.2 to limit the policy update step size;

[0016] The PPO algorithm design architecture includes calculating the ratio of action probabilities under the new and old strategies, introducing a clipping mechanism to constrain the strategy ratio by taking the exponential operation of the logarithmic difference; when the ratio exceeds the range of [0.8, 1.2], clipping is performed.

[0017] As a preferred solution of the distribution network resource dispatching management method based on large model calculation of the present invention, wherein: the dispatching decision generation for the normal operation of the distribution network includes processing using an emergency processing mechanism of an emergency processing mode, and when an abnormal working condition is detected, switching to the emergency control mode in a timely manner;

[0018] The emergency control mode includes fault handling by integrating deep learning and the system's hybrid decision-making mechanism. This mode accurately locates system anomalies through an intelligent fault location module, extracts millisecond-level fault features based on an improved convolutional neural network, and quickly locates fault points for distribution network resource scheduling through in-depth analysis of transient waveforms and multi-dimensional data of phase characteristics.

[0019] As a preferred solution of the distribution network resource dispatching management method based on large model calculation of the present invention, the hierarchical and progressive processing mechanism includes adopting a distributed collaborative control strategy, the control system sorts the dispatching instructions according to time priority, quickly locates the fault point of the distribution network resource dispatching, establishes a real-time feedback mechanism, and completes the intelligent dispatching decision of the distribution network resources.

[0020] In a second aspect, an embodiment of the present invention provides a distribution network resource dispatching and management system based on large model calculation, which includes: a processing module, which adopts an improved heterogeneous graph neural network structure, inherits from a neural network base class through a class, and processes the relationship in the distribution network;

[0021] The building module uses a language writing module to build a three-layer agent collaborative decision-making mechanism, extracts and maps features through three fully connected layers, and uses the ReLU activation function to provide nonlinear expression capabilities;

[0022] The dispatching module generates dispatching decisions under normal operating conditions of the distribution network and adopts a hierarchical and progressive processing mechanism to complete intelligent dispatching decisions on distribution network resources.

[0023] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, it implements any step of the above-mentioned distribution network resource scheduling management method based on large model calculation.

[0024] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned distribution network resource scheduling and management method based on large model calculation is implemented.

[0025] The beneficial effects of the present invention are as follows: the present invention realizes the refined and intelligent management of distribution network resources by adopting an improved heterogeneous graph neural network structure, a three-layer intelligent agent collaborative decision-making mechanism and a hierarchical and progressive processing mechanism. It not only improves the accuracy of resource characterization through node feature coding and edge feature coding, but also enhances the dynamic optimization capability and decision-making efficiency of the model through a multi-head graph attention mechanism and a bidirectional information flow mechanism. The emergency processing mode and the rapid fault point location technology ensure the rapid response and recovery of the system under abnormal conditions, and significantly improve the operation reliability and safety of the distribution network. Overall, the present invention greatly improves the intelligence level and management efficiency of distribution network resource scheduling, and provides strong support for the efficient operation of modern power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0027] Figure 1 A specific flow chart of a distribution network resource dispatching management method and system based on large model calculation is provided for one embodiment of the present invention.

[0028] Figure 2 A schematic diagram of a knowledge migration mechanism of a distribution network resource scheduling management method and system based on large model calculation is provided for one embodiment of the present invention.

[0029] Figure 3 A schematic diagram of an improved PPO algorithm of a distribution network resource scheduling management method and system based on large model calculation is provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0033] The present invention is described in detail with reference to the schematic diagram. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0034] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0035] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0036] Example 1

[0037] Reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, and provides a distribution network resource scheduling management method based on large model calculation, comprising:

[0038] S1: An improved heterogeneous graph neural network structure is adopted, including the organic combination of node feature encoding, edge feature encoding and multi-head graph attention mechanism. Feature encoding includes designing a three-layer deep encoder, setting a dropout rate of 0.1 to prevent overfitting, and integrating the node features of different types of distribution network equipment to complete the unified representation of heterogeneous nodes. The node feature encoder adopts a residual connection structure to retain the original feature information. The edge feature encoding includes a two-layer EdgeEncoder structure to process the line parameter characteristics in the distribution network. The encoder extracts key information of line impedance, flow and capacity, and converts it into a representation vector consistent with the node feature dimension. The attention mechanism is also integrated in the edge feature encoder, so that the model can adaptively pay attention to different types of edge features.

[0039] In terms of node feature encoding, the present invention designs a three-layer deep NodeEncoder encoder, and prevents overfitting by setting a dropout rate of 0.1. The encoder can map different types of distribution network equipment node features, such as operating parameters of transformers, circuit breakers, busbars, etc., to a unified hidden_dim-dimensional feature space, thereby achieving unified characterization of heterogeneous nodes. The node feature encoder adopts a residual connection structure to ensure the stability of deep network training while retaining the original feature information.

[0040] Edge feature encoding adopts a two-layer EdgeEncoder structure, which is specially used to process line parameter features in distribution networks. The encoder can effectively extract key information such as line impedance, flow, capacity, etc., and convert it into a representation vector consistent with the node feature dimension. The edge feature encoder also integrates the attention mechanism, which enables the model to adaptively pay attention to different types of edge features, thereby improving the targetedness of feature extraction.

[0041] At the graph attention propagation level, the present invention innovatively designs a four-layer superimposed multi-head graph attention network. Each layer of MultiHeadGATLayer contains 8 attention heads. The generalization ability of the model is improved through batch normalization and dropout mechanisms. The multi-head attention mechanism enables the model to learn the relationships between nodes from multiple subspaces at the same time, significantly improving the comprehensiveness and robustness of feature extraction.

[0042] During the forward propagation process, the model first encodes the input node feature x to obtain the initial node embedding vector h, then encodes the edge feature edge_attr to obtain the edge representation vector edge_h. Finally, through the four-layer graph attention network, the local and global topological relationships of the nodes are comprehensively considered and the feature representation of the nodes is iteratively updated. This hierarchical feature extraction method ensures that the model can fully capture the multi-scale topological information in the distribution network.

[0043] An important innovation of this module is the introduction of the modulation effect of edge features in attention calculation. Specifically, when calculating the attention weights between nodes, not only the similarity of node features is considered, but also the characteristics of the connecting edges are used as modulation factors, making the attention allocation more consistent with the actual physical meaning. This design significantly improves the model's ability to perceive the power flow characteristics of the distribution network.

[0044] Ultimately, the improved heterogeneous graph neural network can output node representation vectors with rich semantic information, providing a reliable feature basis for subsequent scheduling decisions. Experiments show that this module has a significant technical advantage over traditional methods, with an improvement of 35% in feature extraction effect and 40% in dynamic scene adaptability.

[0045] S1.1: Processing the relationship in the distribution network includes encoding the input node feature x to obtain the initial node embedding vector h, encoding the edge feature edge_attr, and obtaining the edge representation vector edge_h. Through the four-layer graph attention network, the local and global topological relationships of the nodes are comprehensively considered, the feature representation of the nodes is iteratively updated, and the modulation effect of the edge feature is introduced. When calculating the attention weight between nodes, the similarity of the node features is considered, and the features of the connecting edges are used as modulation factors, so that the attention allocation is more in line with the actual physical meaning. The algorithm of the dynamic topological reconstruction mechanism of the attention mechanism is:

[0046]

[0047]

[0048] Furthermore, the knowledge transfer mechanism of the present invention shows significant advantages in practical applications. The reasoning speed of the lightweight model trained by this mechanism is more than 3 times faster than that of the original teacher model, while the accuracy rate only drops by less than 5 percentage points. In addition, the mechanism has good generalization and can adapt to distribution network systems of different scales and topologies, providing a flexible solution for actual deployment.

[0049] In terms of model updating, the knowledge transfer mechanism of the present invention supports incremental learning. When a new operating mode or fault type appears in the system, targeted training can be performed on the teacher model first, and then the new knowledge can be transferred to the deployed lightweight model through knowledge distillation, thereby achieving dynamic improvement of model capabilities and avoiding the computational overhead caused by complete retraining.

[0050] The specific algorithm of the knowledge transfer mechanism of the attention mechanism is:

[0051]

[0052] S2: A three-layer intelligent collaborative decision-making mechanism is constructed using a language writing module. Feature extraction and mapping are performed through three fully connected layers, and the ReLU activation function is used to provide nonlinear expression capabilities.

[0053] Among them, a three-layer intelligent agent collaborative decision-making mechanism was constructed, including setting up a global scheduling intelligent agent at the top level, responsible for the formulation of the overall scheduling strategy, configuring multiple regional coordination intelligent agents in the middle level to complete the coordinated optimization of resources in the region, and deploying local execution intelligent agents at the bottom level to execute control instructions;

[0054] Feature extraction and mapping are performed through three fully connected layers. The hidden layer dimension is 512. The ReLU activation function is used to provide nonlinear expression capabilities. The output layer generates the action space probability distribution through the Softmax function to complete the probability modeling of the global scheduling strategy.

[0055] The improved PPO algorithm is used for processing. The specific algorithm is:

[0056]

[0057]

[0058] Furthermore, the multi-agent network structure adopts a layered architecture design to build an efficient three-layer agent collaborative decision-making mechanism. A global scheduling agent is set up at the top level to be responsible for the formulation of the overall scheduling strategy; multiple regional coordination agents are configured in the middle level to achieve coordinated optimization of resources in the region; and local execution agents are deployed at the bottom level to execute specific control instructions.

[0059] The global scheduling agent adopts a deep neural network structure. The input layer dimension is the product of the state space dimension and the number of agents. Feature extraction and mapping are performed through three fully connected layers. The hidden layer dimension is 512, and the ReLU activation function is used to provide nonlinear expression capabilities. The output layer generates the action space probability distribution through the Softmax function to achieve probabilistic modeling of the global scheduling strategy.

[0060] The regional coordination agent has designed an adaptive network structure based on the characteristics of different regions. The input dimension of each regional coordination agent is the same as the state dimension of a single agent, and feature transformation is performed through two hidden layers with a hidden layer dimension of 256. Considering the refined requirements of regional scheduling, the action space dimension is set to half of the global action space to achieve precise control within the region. At the same time, the attention mechanism is introduced to enable the coordination agent to adaptively focus on important state information within the region.

[0061] As the most basic control unit, the local execution agent adopts a lightweight neural network structure. The input includes local state information and decision information of the upper-level coordination agent, and features are fused through a 128-dimensional hidden layer. Considering the specific requirements of the execution level, the action space dimension is set to one-fourth of the global action space to ensure the accurate execution of control instructions. At the same time, the residual connection mechanism is introduced to effectively prevent the gradient disappearance problem.

[0062] In the process of action selection, the present invention designs a top-down decision transmission mechanism. First, the global scheduling agent generates a macro scheduling strategy based on the overall state of the system. Secondly, the regional coordination agent combines the global strategy and regional state, extracts regional features through the _get_region_state function, and generates regional coordination instructions. Finally, the local execution agent fuses the local state information and coordination instructions to generate specific control actions. This hierarchical decision-making mechanism not only ensures the realization of the global optimization goal, but also takes into account the flexibility of local control.

[0063] In order to improve the reliability of decision-making, the present invention establishes a two-way information flow mechanism between intelligent agents. The downward decision flow ensures the effective transmission of strategies, and the upward state feedback provides a basis for upper-level decisions. Through this two-way information flow, the system can quickly adapt to environmental changes and achieve dynamic optimization and adjustment. At the same time, each layer of intelligent agents adopts an asynchronous update mechanism to avoid decision bottlenecks and improve system response speed.

[0064] In practical applications, the multi-agent network structure of the present invention exhibits good scalability and robustness. By adjusting the number of agents and network parameters, it can flexibly adapt to distribution network systems of different sizes. At the same time, the multi-level decision-making mechanism and redundant design improve the fault tolerance of the system, ensuring that basic dispatching functions can be maintained even when some agents fail.

[0065] For the dispatch decision generation under the normal operation of the distribution network, a hierarchical and progressive processing mechanism is adopted to complete the intelligent dispatch decision of distribution network resources.

[0066] Among them, the dispatch decision generation under the normal operation of the distribution network includes an algorithm processing of the emergency processing mode, and the algorithm includes:

[0067]

[0068] Preferably, a hierarchical and progressive processing mechanism is adopted, including adopting a distributed collaborative control strategy, the control system sorts the scheduling instructions according to time priority, realizes precise control of various types of equipment through an improved feedforward-feedback control algorithm, and establishes a real-time feedback mechanism.

[0069] In a preferred embodiment, a distribution network resource dispatching and management system based on large model calculation includes a processing module, which adopts an improved heterogeneous graph neural network structure, inherits from a neural network base class through a class, and processes relationships in the distribution network; a construction module, which uses a language writing module to construct a three-layer intelligent agent collaborative decision-making mechanism, performs feature extraction and mapping through three fully connected layers, and uses a ReLU activation function to provide nonlinear expression capabilities; a dispatching module, which generates dispatching decisions under normal operating conditions of the distribution network, and adopts a hierarchical and progressive processing mechanism to complete intelligent dispatching decisions on distribution network resources.

[0070] This embodiment also provides a computer device, which is suitable for the multi-source power grid information fusion method based on the Internet of Things, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the multi-source power grid information fusion method based on the Internet of Things proposed in the above embodiment.

[0071] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0072] In summary, the present invention realizes the refined and intelligent management of distribution network resources by adopting an improved heterogeneous graph neural network structure, a three-layer intelligent agent collaborative decision-making mechanism and a hierarchical and progressive processing mechanism. It not only improves the accuracy of resource characterization through node feature coding and edge feature coding, but also enhances the dynamic optimization capability and decision-making efficiency of the model through multi-head graph attention mechanism and bidirectional information flow mechanism. The emergency processing mode and fault point rapid location technology ensure the rapid response and recovery of the system under abnormal conditions, and significantly improve the operation reliability and safety of the distribution network. On the whole, the present invention greatly improves the intelligence level and management efficiency of distribution network resource scheduling, and provides strong support for the efficient operation of modern power systems.

[0073] Example 2

[0074] Reference Figure 1 to Figure 3 , which is the second embodiment of the present invention, and this embodiment provides a distribution network resource scheduling management method based on large model calculation. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0075] In the experimental process of the technology of the present invention, a distribution network resource dispatching and management system based on an improved heterogeneous graph neural network structure was first constructed. The system is written in Python language, inherited from the neural network base class, and is used to process complex relationships in the distribution network. The experimental object is a medium-sized distribution network containing 100 nodes and 150 edges. The node types include substations, distribution lines, user loads, etc., and the edge features include key information such as line impedance, flow and capacity.

[0076] The first step of the experiment is to encode the node and edge features. The node feature encoder uses a three-layer deep encoder structure, sets a dropout rate of 0.1 to prevent overfitting, and retains the original feature information through a residual connection structure. The edge feature encoder uses a two-layer EdgeEncoder structure to extract line parameter features and convert them into representation vectors consistent with the node feature dimension.

[0077] The second step is to iteratively update the feature representation of the nodes by comprehensively considering the local and global topological relationships of the nodes through a four-layer graph attention network (GAT). When calculating the attention weights between nodes, the modulation effect of edge features is introduced to make the attention allocation more consistent with the actual physical meaning. In the experiment, the initial embedding vector dimension of each node is 64, and the edge representation vector dimension is also 64. After four layers of GAT iterations, the final node embedding vector dimension remains 64.

[0078] The third step is to build a three-layer agent collaborative decision-making mechanism. The top-level global scheduling agent is responsible for the formulation of the overall scheduling strategy, the middle-level regional coordination agent completes the coordinated optimization of resources in the region, and the bottom-level local execution agent executes control instructions. The agent extracts and maps features through three fully connected layers. The hidden layer dimension is 512. The ReLU activation function is used to provide nonlinear expression capabilities. The output layer generates the action space probability distribution through the Softmax function. In the experiment, the global scheduling agent generates a scheduling strategy every 10 seconds, the regional coordination agent updates the optimization plan every 5 seconds, and the local execution agent executes control instructions every 1 second.

[0079] The fourth step is to generate a dispatch decision under normal operating conditions of the distribution network. The experiment adopts a hierarchical and progressive processing mechanism. Through a distributed collaborative control strategy, the control system sorts dispatch instructions according to time priority, and achieves precise control of various types of equipment through an improved feedforward-feedback control algorithm, while establishing a real-time feedback mechanism. The experimental results show that the system can efficiently dispatch resources under normal operating conditions, the average dispatch response time is shortened to less than 2 seconds, and the dispatch accuracy rate reaches more than 95%. The experimental data of the present invention are shown in Table 1 below:

[0080] Table 1 Experimental data table of the present invention

[0081] Data Types Numeric Number of nodes 100 Number of edges 150 Node embedding vector dimensions 64 Edge representation vector dimension 64 Dropout rate 0.1 Number of graph attention network layers 4 Hidden layer dimensions 512 Global scheduling agent scheduling frequency Every 10 seconds Regional coordination agent update frequency Every 5 seconds Local execution agent execution frequency Every 1 second Average dispatch response time < 2 seconds Scheduling accuracy >95％ Emergency mode start time Within 1 second Initial emergency dispatch completion time Within 3 seconds Restoration time of grid stability Within 5 seconds

[0082] Table 1 clearly shows the key data points in the experiment, which helps to understand the experimental process and its results. The comparison between the present invention and the prior art is shown in Table 2 below:

[0083] Table 2 Comparison between the present invention and the prior art

[0084]

[0085] Table 2 clearly demonstrates the advantages of the technical solution of the present invention in many aspects, highlighting its significant improvements in node and edge feature encoding, attention mechanism, agent collaborative decision-making, scheduling response time, scheduling accuracy, emergency handling capability, model complexity, computational accuracy, economy and scalability.

[0086] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distribution network resource dispatching management method based on large model calculation, characterized in that: include, An improved heterogeneous graph neural network structure is adopted, which inherits from the neural network base class through classes to handle the relationship in the distribution network; A three-layer intelligent agent collaborative decision-making mechanism was constructed using a language writing module. Feature extraction and mapping were performed through three fully connected layers, and the ReLU activation function was used to provide nonlinear expression capabilities. For the dispatch decision generation under the normal operation of the distribution network, a hierarchical and progressive processing mechanism is adopted to complete the intelligent dispatch decision of distribution network resources.

2. The distribution network resource dispatching management method based on large model calculation as claimed in claim 1, characterized in that: The improved heterogeneous graph neural network structure includes an organic combination of node feature encoding, edge feature encoding and a multi-head graph attention mechanism. The feature encoding includes designing a three-layer deep encoder, setting a dropout rate of 0.1 to prevent overfitting, and integrating the node features of different types of distribution network equipment to complete the unified representation of heterogeneous nodes. The node feature encoder adopts a residual connection structure to retain the original feature information. The edge feature encoding includes adopting a double-layer EdgeEncoder structure to process the line parameter features in the distribution network. The encoder extracts key information of line impedance, flow and capacity, and converts it into a representation vector consistent with the node feature dimension. The edge feature encoder also integrates an attention mechanism so that the model can adaptively pay attention to different types of edge features.

3. The distribution network resource dispatching management method based on large model calculation as claimed in claim 2 is characterized in that: The processing of the relationship in the distribution network includes encoding the input node feature x to obtain the initial node embedding vector h, encoding the edge feature edge_attr to obtain the edge representation vector edge_h, and using a four-layer graph attention network to comprehensively consider the local and global topological relationships of the nodes, iteratively update the feature representation of the nodes, introduce the modulation effect of the edge feature, consider the similarity of the node features when calculating the attention weights between nodes, and use the features of the connecting edges as modulation factors; The modulation factor includes the system adopting a dual detection strategy to accurately compare node features and edge features. During the node feature detection process, the differences between nodes in the new and old topological maps are determined, and whether there are new or deleted nodes. For co-existing nodes, the relative error of the feature vector is calculated to determine whether the node features have changed, and the detection sensitivity is flexibly adjusted through configurable threshold parameters.

4. The distribution network resource dispatching management method based on large model calculation as claimed in claim 3 is characterized by: The attention mechanism includes adopting a lightweight neural network structure, inputting local state information and decision information of the upper-layer coordination agent, performing feature fusion through a 128-dimensional hidden layer, setting the action space dimension to one-fourth of the global action space, and introducing a residual connection mechanism; The introduction of the residual connection mechanism includes generating a scheduling strategy based on the overall state of the system, combining the regional coordination agent with the global strategy and the regional state, extracting regional features through the _get_region_state function, generating regional coordination instructions, executing the agent to fuse local state information and coordination instructions, completing control actions, and establishing a two-way information flow mechanism. The two-way information flow mechanism includes a downward decision flow to ensure the effective transmission of the strategy, and the upward state feedback provides a basis for upper-level decision-making. Through the two-way information flow, the system completes dynamic optimization and adjustment.

5. The distribution network resource dispatching management method based on large model calculation as claimed in claim 4 is characterized in that: The three-layer intelligent agent collaborative decision-making mechanism is constructed, including setting a global scheduling intelligent agent at the top layer, responsible for the formulation of the overall scheduling strategy, configuring multiple regional coordination intelligent agents in the middle layer to complete the coordinated optimization of resources in the region, and deploying local execution intelligent agents at the bottom layer to execute control instructions; Feature extraction and mapping are performed through three fully connected layers. The hidden layer dimension is 512. The ReLU activation function is used to provide nonlinear expression capabilities. The output layer generates the action space probability distribution through the Softmax function to complete the probability modeling of the global scheduling strategy. The probabilistic modeling includes adopting the PPO algorithm to design the architecture, introducing adaptive clipping ratio and multiple loss function optimization mechanisms; In the algorithm initialization stage, a network model, an optimizer and a clipping ratio framework are constructed. The network model includes a deep neural network structure for approximating the policy function and the value function. The optimizer uses the Adam optimizer, and the clipping ratio is set to 0.2 to limit the policy update step size. The PPO algorithm design architecture includes calculating the ratio of action probabilities under the new and old strategies, introducing a clipping mechanism to constrain the strategy ratio by taking the exponential operation of the logarithmic difference; when the ratio exceeds the range of [0.8, 1.2], clipping is performed.

6. The distribution network resource dispatching management method based on large model calculation as claimed in claim 5, characterized in that: The generation of dispatching decisions for the normal operation of the distribution network includes adopting an emergency processing mechanism of an emergency processing mode for processing, and when an abnormal operating condition is detected, switching to an emergency control mode in a timely manner; The emergency control mode includes fault handling by integrating deep learning and the system's hybrid decision-making mechanism. This mode accurately locates system anomalies through an intelligent fault location module, extracts millisecond-level fault features based on an improved convolutional neural network, and quickly locates fault points for distribution network resource scheduling through in-depth analysis of transient waveforms and multi-dimensional data of phase characteristics.

7. The distribution network resource dispatching management method based on large model calculation as claimed in claim 6, characterized in that: The hierarchical and progressive processing mechanism includes adopting a distributed collaborative control strategy, the control system sorts the dispatch instructions according to time priority, quickly locates the fault point through the distribution network resource dispatch, establishes a real-time feedback mechanism, and completes the intelligent dispatch decision of the distribution network resources.

8. A distribution network resource dispatching and management system based on large model calculation, based on the distribution network resource dispatching and management method based on large model calculation according to any one of claims 1 to 7, characterized in that: include, The processing module uses an improved heterogeneous graph neural network structure and inherits from the neural network base class through classes to process the relationships in the distribution network; The building module uses a language writing module to build a three-layer agent collaborative decision-making mechanism, extracts and maps features through three fully connected layers, and uses the ReLU activation function to provide nonlinear expression capabilities; The dispatching module generates dispatching decisions under normal operating conditions of the distribution network and adopts a hierarchical and progressive processing mechanism to complete intelligent dispatching decisions on distribution network resources.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distribution network resource dispatching management method based on large model calculation according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distribution network resource dispatching management method based on large model calculation according to any one of claims 1 to 7 are implemented.

Images