Energy consumption management method and system based on cloud computing

Abstract

The application provides a cloud computing-based energy consumption management method and system, the cloud computing-based energy consumption management method comprises the following steps: obtaining historical operation data and real-time operation data of a target power grid through a distributed cloud; analyzing the historical operation data by using an energy consumption management model to obtain a first topology structure of the target power grid, wherein the topology structure is a graph data structure; determining an objective function of the first topology structure according to an energy consumption management request, optimizing the first topology structure according to the objective function and the energy consumption management model to obtain a second topology structure; and performing control processing on the target power grid through a power grid dispatching system according to the second topology structure; wherein the energy consumption management model is obtained through training. The application has the beneficial effects of improving the energy consumption management efficiency and reducing the system resource consumption of energy consumption management.

Description

Technical Field

[0001] This invention relates to the field of power energy management technology, and in particular to a cloud computing-based energy consumption management method and system. Background Technology

[0002] The increasingly complex power distribution networks, which include different types of substations, energy storage devices, and load terminals, have a complex topology.

[0003] Graph neural networks are typically used to handle topological relationships; however, traditional methods cannot handle the complex graph structures of distribution networks, making energy management and scheduling efficiency unable to meet the needs of existing distribution networks. Summary of the Invention

[0004] The main objective of this invention is to propose a cloud computing-based energy management method and system, which improves energy management efficiency and reduces system resource consumption in energy management.

[0005] One aspect of the present invention provides a cloud computing-based energy management method, comprising: Historical and real-time operational data of the target power grid are obtained through a distributed cloud platform. The historical operating data is analyzed using an energy management model to obtain the first topology of the target power grid, wherein the topology is a graph data structure; The objective function of the first topology is determined based on the energy management request, and the first topology is optimized based on the objective function and the energy management model to obtain the second topology. According to the second topology, control processing of the target power grid is performed through the power grid dispatching system; The training steps for the energy management model include: The historical operation data, historical power grid diagram structure, and real-time operation data are acquired. After preprocessing the historical operation data, distributed training is performed using a distributed computing framework and graph neural network to obtain the training results. Based on the training results, the training parameters of the graph neural network are adjusted and the training is performed again a preset number of times to obtain the energy management model; Real-time operating data is dynamically input into the energy management model and inferred to obtain an updated power grid diagram structure; The energy management model is optimized based on the updated power grid diagram structure.

[0006] According to the cloud computing-based energy management method, the distributed cloud collects the operating data of the power grid equipment of the target power grid, cleans, normalizes and extracts features from the operating data of the power grid equipment to obtain a power grid representation; the power grid representation is converted into a graph data structure, wherein the graph data structure includes node attributes and edge attributes, the node attributes are the attributes of the power grid equipment, and the edge attributes are the association attributes of the power grid equipment.

[0007] According to the aforementioned cloud computing-based energy management method, which employs a distributed computing framework and graph neural networks for distributed training, the method includes: The distributed computing framework is deployed in the cloud, and the graph data structure is used as the input to the graph neural network. The distributed computing framework is used to perform graph sampling and block processing on the graph data structure to obtain sampling and block results. Based on the sampling and block division results, the graph data structure is batch-partitioned and gradient descent is applied; Cross-validation and selection of optimal parameters are performed using either the Adam or SGD optimizer.

[0008] According to the aforementioned cloud computing-based energy management method, the method further includes: The distributed computing framework is either TensorFlow or PyTorch, and the graph data structure is sampled and partitioned using either GraphSAGE or Cluster-GCN.

[0009] According to the cloud computing-based energy management method, wherein the training results are used to adjust the training parameters of the graph neural network and retrain it a preset number of times to obtain the energy management model, the method further includes: When training the energy management model, either a GPU or a TPU is used for inference acceleration, and TensorRT is used to compress the energy management model. The energy management model is trained according to the objective function and evaluation index, and the hyperparameters of the model are adjusted according to the training results each time.

[0010] According to the cloud computing-based energy management method, real-time operating data is dynamically input into the energy management model and inferred to obtain an updated power grid diagram structure, including: The energy management model is deployed in the cloud. The real-time running data is converted into streaming data using Flink, and the historical power grid structure is updated based on the streaming data to obtain the updated power grid structure. The updated power grid diagram structure is dynamically input into the energy management model for inference.

[0011] According to the aforementioned cloud-based energy management method, the energy management model performs reasoning, including: Obtain the node attributes and edge attributes of the historical power grid graph structure, and update the node attributes of the historical power grid graph structure according to the updated power grid graph structure; wherein the historical power grid graph structure is represented as follows: V is the set of nodes in the historical power grid graph, representing the devices in the power grid. The matrix of node attributes is... d is the dimension of the node attributes, and the matrix of edge attributes is... 'm' is the dimension of the edge attribute; the node attributes of the historical power grid graph structure are updated as follows: After the update, the node attributes are: , where f is the process function for updating node attributes in the power grid diagram structure; Graph convolution and attention are used to compute node embeddings based on updated node attributes and update edge attributes. The graph convolution operation is as follows: in, H ( l ) is the first l The node feature matrix of the layer, A ~= A + I It involves adding a self-connected adjacency matrix. yes The degree matrix, W ( l ) is the first l The learnable weight matrix of the layer, σ is the activation function, and the attention mechanism calculates the attention weights between nodes using the following formula: in, It is a node i and nodes j Attention score between, a a It is a learnable weight vector, where || represents the vector concatenation operation; The node embedding Z can be obtained by weighted summation. Normalized attention weights; Edge attributes based on node embedding are , A function to update edge attributes; An optimization scheme is generated based on the updated node attributes, edge attributes, and objective function. The objective function includes one of the following: minimizing loss, maximizing power supply reliability, and minimizing energy consumption. This is the objective function. These are the constraints for power balance constraints; The optimization scheme is used to characterize the starting and stopping of power grid equipment and transmission lines between power grid equipment in the target power grid, specifically: .

[0012] According to the cloud computing-based energy management method described above, the objective function further includes: The objective function for minimizing the loss is: in Let (i,j) be the resistance of edge (i,j). Let be the squared power flow value of edge (i,j); The objective function for achieving the highest reliability of the power supply is: in It is a node i Reliability; The objective function for minimizing energy consumption is: in It is a node i Unit energy consumption cost It is a node i Power output.

[0013] Another aspect of the present invention discloses a cloud computing-based energy management system, characterized in that it includes: The first module is used to acquire historical and real-time operating data of the target power grid through a distributed cloud. The second module is used to analyze the historical operating data using an energy consumption management model to obtain the first topology of the target power grid, wherein the topology is a graph data structure. The third module is used to determine the objective function of the first topology based on the energy management request, and optimize the first topology based on the objective function and the management model to obtain the second topology. The fourth module is used to perform control processing on the target power grid according to the second topology; The fifth module is used to acquire the historical operating data, historical power grid diagram structure and real-time operating data. After preprocessing the historical operating data, it uses a distributed computing framework and graph neural network for distributed training to obtain the training results. The sixth module is used to adjust the training parameters of the graph neural network and retrain it a preset number of times based on the training results to obtain the energy management model. The seventh module is used to dynamically input real-time operating data into the energy management model and perform reasoning to obtain an updated power grid diagram structure; The eighth module is used to optimize the energy management model based on the updated power grid diagram structure.

[0014] The beneficial effects of this invention are as follows: by using cloud computing (cloud deployment), the training and inference of graph neural networks in the cloud computing environment can efficiently handle the power grid topology optimization problem, thereby improving the efficiency of energy consumption management; at the same time, by collecting power grid data in real time in the cloud and using GNN to generate the optimal topology reconstruction scheme, it can adapt to large power grids (such as the State Grid) containing millions of nodes and edges, and has low complexity compared to existing technologies. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a schematic diagram of the system composition according to an embodiment of the present invention.

[0016] Figure 2 This is a schematic diagram of the cloud computing energy management method according to an embodiment of the present invention.

[0017] Figure 3 This is a schematic diagram of the distributed training process according to an embodiment of the present invention.

[0018] Figure 4 This is a schematic diagram of the energy management model training process according to an embodiment of the present invention.

[0019] Figure 5 This is a schematic diagram of the reasoning process of an energy management model according to an embodiment of the present invention.

[0020] Figure 6 This is a schematic diagram of the reasoning process of another energy management model in an embodiment of the present invention.

[0021] Figure 7 This is a system block diagram of an embodiment of the present invention. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings. Throughout the description, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" can be used interchangeably. Terms such as "first," "second," etc., are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features. In the following description, the consecutive reference numerals for method steps are for ease of review and understanding. Adjusting the implementation order of steps, in conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0023] refer to Figure 1 ,in Figure 1 This is a schematic diagram of the system composition according to an embodiment of the present invention. It includes a target power grid 100, a cloud server 200, and a power grid dispatching system 300. The cloud server 200 is the cloud platform, configured according to the complexity of the target power grid. The cloud server 200 is used to acquire historical operating data, historical power grid diagram structures, and real-time operating data. After preprocessing the historical operating data, it performs distributed training using a distributed computing framework and graph neural network to obtain training results. Based on the training results, it adjusts the training parameters of the graph neural network and retrains it a preset number of times to obtain an energy consumption management model. Real-time operating data is dynamically input into the energy consumption management model and used for inference to obtain an updated power grid diagram structure. The energy consumption management model is then optimized based on the updated power grid diagram structure.

[0024] The power grid dispatching system 300 is used to receive optimization schemes and perform topology optimization on the target power grid 100.

[0025] refer to Figure 2 ,in Figure 2 This is a schematic flowchart of a cloud computing energy management method according to an embodiment of the present invention, which includes, but is not limited to, steps S100 to S800, wherein S100 to S400 is the energy management method flowchart, including: S100 acquires historical and real-time operating data of the target power grid through a distributed cloud.

[0026] In some embodiments, the distributed cloud collects power grid equipment operation data of the target power grid, cleans, normalizes, and extracts features from the power grid equipment operation data to obtain a power grid representation; the power grid representation is then converted into a graph data structure, wherein the graph data structure includes node attributes and edge attributes, where node attributes are the attributes of power grid equipment and edge attributes are the association attributes of power grid equipment.

[0027] S200 analyzes historical operating data using an energy management model to obtain the first topology of the target power grid, where the topology is a graph data structure.

[0028] S300: Determine the objective function of the first topology based on the energy management request, and optimize the first topology based on the objective function and the energy management model to obtain the second topology.

[0029] In some embodiments, it can be understood that the first topology is the historical topology of the target power grid, and the second topology is the updated topology based on real-time operating data. For example, the topology is updated when any distribution device or load device fails or is actively shut down.

[0030] S400 performs control processing on the target power grid through the power grid dispatching system according to the second topology.

[0031] In some embodiments, it is understood that when the power grid dispatching system changes the power grid from a first topology to a second topology, any nodes and edges in the power grid that have changed will be modified accordingly. Dynamic graph neural networks (Dynamic GNNs) can be used to handle the dynamic changes in the power grid topology.

[0032] Among them, S500~S800 is the training process for the energy management model, including: The S500 acquires historical operating data, historical power grid diagram structure, and real-time operating data. After preprocessing the historical operating data, it uses a distributed computing framework and graph neural network for distributed training to obtain the training results.

[0033] refer to Figure 3 The distributed training process diagram shown includes, but is not limited to, steps S510-S540: The S510 uses a distributed computing framework deployed in the cloud, with graph data structures as input to graph neural networks.

[0034] S520 uses a distributed computing framework to perform graph sampling and block processing on graph data structures, obtaining sampling and block results.

[0035] S530 performs batch partitioning and gradient descent on the graph data structure based on the sampling and block results.

[0036] The S540 uses either the Adam or SGD optimizer for cross-validation and to select the optimal parameters.

[0037] In some embodiments, one of the distributed computing frameworks TensorFlow and PyTorch, and one of the graph data structures, GraphSAGE and Cluster-GCN, are used for graph sampling and block processing of graph data structures.

[0038] Understandably, graph sampling is used with GraphSAGE or Cluster-GCN to reduce computational costs.

[0039] In some embodiments, reference Figure 4 The diagram shown illustrates the training process for the energy management model, including but not limited to steps S550-S560: S600 adjusts the training parameters of the graph neural network based on the training results and retrains it a preset number of times to obtain the energy management model.

[0040] The S650 uses either a GPU or a TPU for inference acceleration when training the power management model, and uses TensorRT for model compression. The S660 trains the energy management model based on the objective function and evaluation indicators, and adjusts the model hyperparameters based on the training results each time it is trained.

[0041] In some embodiments, the evaluation metrics include accuracy, recall, F1 score, and may also be optimized according to the objective function of the energy management task, such as energy minimization.

[0042] The S700 dynamically inputs real-time operating data into the energy management model and performs inference to obtain an updated power grid diagram structure.

[0043] In some embodiments, reference Figure 5 The diagram showing the reasoning process of the energy management model includes, but is not limited to, steps S710 to S730: S710 deploys an energy management model via the cloud; The S720 converts real-time running data into streaming data using Flink, updates the historical power grid structure based on the streaming data, and obtains the updated power grid structure. The S730 dynamically inputs the updated power grid diagram structure into the energy management model for inference.

[0044] The objective function includes one of minimizing losses, maximizing power supply reliability, and minimizing energy consumption. The optimization scheme is used to characterize the starting and stopping of grid equipment and transmission lines between grid equipment in the target grid.

[0045] In some embodiments, the energy management model performs reasoning, including: Obtain the node attributes and edge attributes of the historical power grid graph structure, and update the node attributes of the historical power grid graph structure according to the updated power grid graph structure; wherein the historical power grid graph structure is represented as follows: V is the set of nodes in the historical power grid graph, representing the devices in the power grid. The matrix of node attributes is... d is the dimension of the node attributes, and the matrix of edge attributes is... 'm' is the dimension of the edge attribute; the node attributes of the historical power grid graph structure are updated as follows: After the update, the node attributes are: , where f is the process function for updating node attributes in the power grid diagram structure; Graph convolution and attention are used to compute node embeddings based on updated node attributes and update edge attributes. The graph convolution operation is as follows: in, H ( l ) is the first l The node feature matrix of the layer, A ~= A + I It involves adding a self-connected adjacency matrix. yes The degree matrix, W ( l ) is the first l The learnable weight matrix of the layer, σ is the activation function, and the attention mechanism calculates the attention weights between nodes using the following formula: in, It is a node i and nodes j Attention score between, a a It is a learnable weight vector, where || represents the vector concatenation operation; The node embedding Z can be obtained by weighted summation. Normalized attention weights; Edge attributes based on node embedding are , A function to update edge attributes; An optimization scheme is generated based on the updated node attributes, edge attributes, and objective function. The objective function includes one of the following: minimizing loss, maximizing power supply reliability, and minimizing energy consumption. This is the objective function. These are the constraints for power balance constraints; The optimization scheme is used to characterize the starting and stopping of power grid equipment and transmission lines between power grid equipment in the target power grid, specifically: .

[0046] In some embodiments, the objective function is as follows: The objective function for minimizing the loss is: in Let (i,j) be the resistance of edge (i,j). Let be the squared power flow value of edge (i,j); The objective function for achieving the highest reliability of the power supply is: in It is a node i Reliability; The objective function for minimizing energy consumption is: in It is a node i Unit energy consumption cost It is a node i Power output.

[0047] Understandably, the objective function can be selected in other ways depending on the power grid energy consumption management.

[0048] In some embodiments, referring to another energy management model inference flowchart, which includes, but is not limited to, steps S731-S733: S731, Obtain the node attributes and edge attributes of the historical power grid graph structure, and update the node attributes of the historical power grid graph structure according to the updated power grid graph structure; S732 uses graph convolution and attention to compute node embeddings based on updated node attributes and update edge attributes; S733 generates an optimization scheme based on the updated node attributes, edge attributes, and objective function.

[0049] In some embodiments, the cloud deployment uses the cloud service AWS SageMaker to deploy the GNN model (energy management model, i.e., GNN model), wherein the cloud service AWS SageMaker is deployed via a RESTful API interface.

[0050] In some embodiments, hardware such as GPUs / TPUs is used to accelerate inference, and ONNX or TensorRT is employed for model optimization and compression.

[0051] In some embodiments, real-time inference includes streaming data processing: including processing real-time data streams using Kafka or Flink, updating the grid graph structure, and dynamically inputting GNN models for inference.

[0052] In some embodiments, node / edge feature update: update the features of nodes and edges based on real-time data, perform graph convolution or attention calculation, generate node embeddings, and generate power grid optimization schemes (such as switching operation suggestions).

[0053] The S800 optimizes the energy management model based on the updated power grid diagram structure to reduce data transmission.

[0054] In some embodiments, the inference results obtained above can also be fed back to the power grid dispatching system to adjust the model parameters according to the actual operating effect.

[0055] In some embodiments, edge computing is used for data transmission to the cloud, power grid dispatching system, and power grid equipment to improve transmission efficiency.

[0056] In some embodiments, the specific process of model training is as follows: (1) Preparation of test dataset After training, we divided the data from the IEEE 118-bus system into training and testing sets: Training set: 80% of the node data, used to train the model.

[0057] Test set: 20% of the node data, used to evaluate model performance.

[0058] Assuming we have completed model training and made predictions on the test set, the following are the detailed results for the test set.

[0059] (2) Test results data. The following is a comparison of the model's predictions and actual values ​​on the test set. Please refer to Table 1 for the data: Table 1 Test Result Data Table In some implementations, the mean absolute error (MAE) is also used for evaluation, as shown in Table 2 for the error distribution.

[0060] It can be determined that the power grid optimization scheme obtained by adopting the technical solution of the present invention can efficiently handle the power grid topology optimization problem in a short time and improve the efficiency of energy consumption management.

[0061] refer to Figure 7 ,in, Figure 7 This is a system block diagram of an embodiment of the present invention, comprising: a first module 710 for acquiring historical and real-time operating data of a target power grid through a distributed cloud; a second module 720 for analyzing the historical operating data using an energy management model to obtain a first topology of the target power grid, wherein the topology is a graph data structure; a third module 730 for determining the objective function of the first topology based on an energy management request, and optimizing the first topology based on the objective function and the management model to obtain a second topology; a fourth module 740 for performing control processing on the target power grid based on the second topology; a fifth module 750 for acquiring historical operating data, historical power grid graph structures, and real-time operating data, preprocessing the historical operating data, and then performing distributed training using a distributed computing framework and a graph neural network to obtain training results; a sixth module 760 for adjusting the training parameters of the graph neural network based on the training results and retraining it a preset number of times to obtain an energy management model; a seventh module 770 for dynamically inputting real-time operating data into the energy management model and performing inference to obtain an updated power grid graph structure; and an eighth module 780 for optimizing the energy management model based on the updated power grid graph structure.

[0062] This invention also provides an electronic device, which includes a processor and a memory; The memory stores the program; The processor executes a program to perform the aforementioned cloud-based energy management method; the electronic device has the function of carrying and running the cloud-based energy management software system provided in the embodiments of the present invention, such as a personal computer, minicomputer, mainframe, workstation, network or distributed computing environment, standalone or integrated computer platform, or communicating with charged particle tools or other imaging devices, etc.

[0063] This invention also provides a computer-readable storage medium storing a program that is executed by a processor to implement the cloud computing-based energy management method described above.

[0064] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0065] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned cloud-based energy management method.

[0066] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0067] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0069] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0070] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0071] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0072] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0073] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A cloud computing-based energy management method, characterized in that, include: Historical and real-time operational data of the target power grid are obtained through a distributed cloud platform. The historical operating data is analyzed using an energy management model to obtain the first topology of the target power grid, wherein the topology is a graph data structure; The objective function of the first topology is determined according to the energy management request. The first topology is optimized according to the objective function and the energy management model to obtain the second topology. The first topology is the topology of the target power grid at a historical moment. The second topology is the topology updated according to real-time operating data. The objective function is one of loss minimization, maximum power supply reliability and minimum energy consumption. According to the second topology, control processing of the target power grid is performed through the power grid dispatching system; The training steps for the energy management model include: The historical operation data, historical power grid diagram structure, and real-time operation data are acquired. After preprocessing the historical operation data, distributed training is performed using a distributed computing framework and graph neural network to obtain the training results. Based on the training results, the training parameters of the graph neural network are adjusted and the training is performed again a preset number of times to obtain the energy management model; Real-time operating data is dynamically input into the energy management model and inferred to obtain an updated power grid diagram structure; The energy management model is optimized based on the updated power grid diagram structure; The distributed cloud collects the power grid equipment operation data of the target power grid, cleans, normalizes, and extracts features from the power grid equipment operation data to obtain a power grid representation; the power grid representation is then converted into a graph data structure, which includes node attributes and edge attributes. The node attributes are the attributes of the power grid equipment, and the edge attributes are the association attributes of the power grid equipment. The use of a distributed computing framework and graph neural networks for distributed training includes: The distributed computing framework is deployed in the cloud, and the graph data structure is used as the input to the graph neural network. The distributed computing framework is used to perform graph sampling and block processing on the graph data structure to obtain sampling and block results. Based on the sampling and block division results, the graph data structure is batch-partitioned and gradient descent is applied; Cross-validation and selection of optimal parameters are performed using either the Adam or SGD optimizer.

2. The energy management method based on cloud computing according to claim 1, characterized in that, The method further includes: The distributed computing framework is either TensorFlow or PyTorch, and the graph data structure is sampled and partitioned using either GraphSAGE or Cluster-GCN.

3. The energy management method based on cloud computing according to claim 1, characterized in that, The step of adjusting the training parameters of the graph neural network and retraining it a preset number of times based on the training results to obtain the energy management model further includes: When training the energy management model, either a GPU or a TPU is used for inference acceleration, and TensorRT is used to compress the energy management model. The energy management model is trained according to the objective function and evaluation index, and the hyperparameters of the model are adjusted according to the training results each time.

4. The energy management method based on cloud computing according to claim 1, characterized in that, The step of dynamically inputting real-time operating data into the energy management model and performing inference to obtain an updated power grid diagram structure includes: The energy management model is deployed in the cloud. The real-time running data is converted into streaming data using Flink, and the historical power grid structure is updated based on the streaming data to obtain the updated power grid structure. The updated power grid diagram structure is dynamically input into the energy management model for inference.

5. The cloud computing-based energy management method according to claim 4, characterized in that, The energy management model performs reasoning, including: Obtain the node attributes and edge attributes of the historical power grid graph structure, and update the node attributes of the historical power grid graph structure according to the updated power grid graph structure; wherein the historical power grid graph structure is represented as follows: V represents the set of nodes in the historical power grid diagram, E represents the devices in the power grid, and the matrix of node attributes is... , Let be a real number, d be the dimension of the node attributes, and the matrix of edge attributes be... 'm' is the dimension of the edge attribute; the node attributes of the historical power grid graph structure are updated as follows: After the update, the node attributes are: Where f is the process function for updating node attributes in the power grid graph structure. This represents the updated set of nodes in the power grid diagram structure. This indicates the updated equipment in the power grid diagram structure; Graph convolution and attention are used to compute node embeddings based on updated node attributes and update edge attributes. The graph convolution operation is as follows: ， in, It is the first The node feature matrix of the layer, It is the first The node feature matrix of the layer, This indicates that the original adjacency matrix is ​​used. With the identity matrix The adjacency matrix obtained by self-connection yes The degree matrix, It is the first The learnable weight matrix of the layer, σ is the activation function, and the attention mechanism calculates the attention weights between nodes using the following formula: ， in, It is a node and nodes Attention scores between It is a learnable weight vector. This represents the vector concatenation operation. For nodes eigenvectors, For nodes eigenvectors, Indicates transpose. For activation functions; The node embedding Z can be obtained by weighted summation. ， ， For normalized attention weights, For nodes Embedded, For nodes The set of neighboring nodes, For the index of neighboring nodes, Indicates a node of Sum of the exponential scores of the neighboring nodes; The updated edge attributes are obtained based on node embedding Z. , A function to update edge attributes; An optimization scheme is generated based on the updated node attributes, edge attributes, and objective function. The objective function includes one of the following: minimizing loss, maximizing power supply reliability, and minimizing energy consumption. ， This is the objective function. These are the constraints for power balance constraints; The optimization scheme is used to characterize the starting and stopping of power grid equipment and transmission lines between power grid equipment in the target power grid, specifically: 。 6. The energy management method based on cloud computing according to claim 5, characterized in that, The objective function also includes: The objective function for minimizing the loss is: ， in For the edge The resistance, For the edge The square value of the power flow, Represents the set of loss lines in the target power grid; The objective function for achieving the highest reliability of the power supply is: ， in It is the reliability of node i; The objective function for minimizing energy consumption is: ， in It is a node Unit energy consumption cost It is a node Power output.

7. A cloud computing-based energy management system according to any one of claims 1-6, characterized in that, include: The first module is used to acquire historical and real-time operating data of the target power grid through a distributed cloud. The second module is used to analyze the historical operating data using an energy consumption management model to obtain the first topology of the target power grid, wherein the topology is a graph data structure. The third module is used to determine the objective function of the first topology based on the energy management request, optimize the first topology based on the objective function and the energy management model to obtain a second topology. The first topology is the topology of the target power grid at a historical moment, and the second topology is the updated topology based on real-time operating data. The objective function is one of minimizing losses, achieving the highest power supply reliability, and minimizing energy consumption. The fourth module is used to perform control processing on the target power grid according to the second topology; The fifth module is used to acquire the historical operating data, historical power grid diagram structure and real-time operating data. After preprocessing the historical operating data, it uses a distributed computing framework and graph neural network for distributed training to obtain the training results. The sixth module is used to adjust the training parameters of the graph neural network and retrain it a preset number of times based on the training results to obtain the energy management model. The seventh module is used to dynamically input real-time operating data into the energy management model and perform reasoning to obtain an updated power grid diagram structure; The eighth module is used to optimize the energy management model based on the updated power grid diagram structure.

Images