Grid Fragmented Resource Cloud-Edge Collaborative Control System and Method

Abstract

This application relates to power grid resource control, and discloses a cloud-edge collaborative control system and method for fragmented power grid resources, which includes: an edge terminal uploads edge data (including node status data and a preliminary discriminant scene type) to a cloud server; the cloud server receives topological structure data of the power grid GIS system, and performs clustering analysis on the edge data to obtain an edge node status clustering center and a preliminary discriminant clustering center for the scene type; optimizes the edge node status clustering center based on the topological structure data; fuses the optimized node status clustering center and the scene type discriminant center to form a multi-source fusion feature representation of the power grid operation scene; determines the power grid operation scene type label according to this representation; matches the corresponding policy from the control strategy library and issues it to the edge terminal for execution. In this way, the operation efficiency and stability of the power grid are improved, the adaptability to dynamic changes is enhanced, and it is applicable to efficient resource management and control in a complex power system environment.

Description

Technical Field

[0001] This application relates to power grid resource control, and more specifically, to a cloud-edge collaborative control system and method for fragmented power grid resources. Background Art

[0002] In modern power systems, with the wide application of distributed energy resources (DERs) such as solar energy and wind energy, the power grid structure has become increasingly complex. Traditional centralized control systems are unable to cope with these changes because they are difficult to process a large amount of scattered data in real time and make rapid and accurate responses. In addition, with the development of smart meter and sensor technologies, the amount of data generated by edge devices has increased exponentially, which further exacerbates the burden on existing systems. Against this background, the cloud-edge collaborative control method for fragmented power grid resources has emerged, aiming to solve the limitations of traditional centralized control systems.

[0003] Traditionally, the operation of power systems relies on a central control center to collect and process data from each node and formulate corresponding control strategies based on this information. However, this method has several obvious drawbacks. First, due to data transmission delays and processing capacity limitations, the central control center may not be able to timely obtain and process the status information of all edge devices, thus affecting the timeliness and accuracy of decision-making. Second, when facing a complex power grid environment, a single central controller is difficult to simultaneously consider global optimization and local adaptability, resulting in unsatisfactory control effects. For example, in the case of large load fluctuations or unstable new energy output, traditional control methods may cause problems such as voltage over-limit, affecting the safe and stable operation of the power grid.

[0004] Therefore, an optimized cloud-edge collaborative control solution for fragmented power grid resources is expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide a cloud-edge collaborative control system and method for fragmented power grid resources, which can improve the operation efficiency and stability of the power grid, enhance the adaptability to dynamic changes, and are applicable to efficient resource management and control in complex power system environments.

[0006] According to one aspect of the present application, a cloud-edge collaborative control method for fragmented power grid resources is provided, including: an edge terminal uploads edge data to a cloud server, where the edge data includes edge node status data and a preliminary discrimination scenario type on the edge side; the cloud server receives current power grid topology structure data from the power grid GIS system; the cloud server performs clustering analysis on the edge node status data and the preliminary discrimination scenario type on the edge side to obtain an edge node status clustering center and a preliminary discrimination clustering center for the scenario type; the cloud server optimizes the spatial topology expression of the edge node status clustering center based on the current power grid topology structure data to obtain an optimized edge node status clustering center; the cloud server fuses the preliminary discrimination clustering center for the scenario type and the optimized edge node status clustering center to obtain a multi-source fusion feature representation of the power grid operation scenario; the cloud server determines a power grid operation scenario type label based on the multi-source fusion feature representation of the power grid operation scenario; the cloud server matches a preset control strategy from a control strategy library based on the power grid operation scenario type label and sends the preset control strategy to the corresponding edge terminal.

[0007] In the above cloud-edge collaborative control method for fragmented power grid resources, the edge node status data includes voltage, current, active power, reactive power, and meteorological data, and the preliminary discrimination scenario type on the edge side includes a load level type label, a new energy output level type label, and whether voltage crossing occurs.

[0008] In the above cloud-edge collaborative control method for fragmented power grid resources, when the cloud server performs clustering analysis on the edge node status data and the preliminary discrimination scenario type on the edge side to obtain an edge node status clustering center and a preliminary discrimination clustering center for the scenario type, it includes: fully connecting and encoding each of the edge node status data to obtain a set of edge node status fully connected embedding encoding vectors; performing DBSCAN clustering on the set of edge node status fully connected embedding encoding vectors to obtain the edge node status clustering center.

[0009] In the above cloud-edge collaborative control method for fragmented power grid resources, when the cloud server performs clustering analysis on the edge node status data and the preliminary discrimination scenario type on the edge side to obtain an edge node status clustering center and a preliminary discrimination clustering center for the scenario type, it further includes: performing one-hot encoding on each of the preliminary discrimination scenario types on the edge side to obtain a set of one-hot encoding vectors for the preliminary discrimination scenario types on the edge side; performing K-Means clustering on the set of one-hot encoding vectors for the preliminary discrimination scenario types on the edge side to obtain K local discrimination clustering centers for the scenario type; calculating the weighted sum of the K local discrimination clustering centers for the scenario type to obtain the preliminary discrimination clustering center for the scenario type.

[0010] In the above grid fragmentation resource cloud-edge collaborative control method, the cloud server optimizes the spatial topological expression of the edge node state clustering center based on the current grid topological structure data to obtain an optimized edge node state clustering center, including: extracting grid topological feature encoder for the current grid topological structure data to obtain a grid topological structure feature encoding matrix; mapping the edge node state clustering center to the topological space of the grid topological structure feature encoding matrix through vector multiplication to obtain the optimized edge node state clustering center.

[0011] In the above grid fragmentation resource cloud-edge collaborative control method, the cloud server fuses the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain a multi-source fusion feature representation of the grid operation scenario, including: cascading the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain the multi-source fusion feature representation of the grid operation scenario.

[0012] In the above grid fragmentation resource cloud-edge collaborative control method, the cloud server determines the grid operation scenario type label based on the multi-source fusion feature representation of the grid operation scenario, including: inputting the multi-source fusion feature representation of the grid operation scenario into a scenario classifier based on a support vector machine model to obtain the grid operation scenario type label.

[0013] In the above grid fragmentation resource cloud-edge collaborative control method, cascading the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain the multi-source fusion feature representation of the grid operation scenario, including: using the equivalence of the pseudo-inverse norm under the homogeneous space structure to perform cross-dimensional isomorphism on the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain an optimized preliminary discriminant clustering center of the scenario type; cascading the optimized preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain the multi-source fusion feature representation of the grid operation scenario.

[0014] In the above grid fragmented resource cloud-edge collaborative control method, by using the equivalence of the pseudo-inverse norm under the homogeneous space structure, the initial discriminant clustering center of the scenario type and the optimized edge node state clustering center are cross-dimensionally isomorphic to obtain the optimized initial discriminant clustering center of the scenario type, including: performing joint space trivialization on the initial discriminant clustering center of the scenario type based on the optimized edge node state clustering center to obtain a joint space trivialization matrix; performing a full-dimensional homogeneous mapping on the initial discriminant clustering center of the scenario type and the optimized edge node state clustering center in the complex manifold space to obtain a full-dimensional homogeneous mapping vector; based on the joint space trivialization matrix and the full-dimensional homogeneous mapping vector, using the pseudo-inverse matrix to perform group action homogenization on the initial discriminant clustering center of the scenario type based on the homogeneous space structure to obtain the optimized initial discriminant clustering center of the scenario type.

[0015] According to another aspect of the present application, there is also provided a grid fragmented resource cloud-edge collaborative control system, including: an edge data acquisition module for uploading edge data including edge node state data and a preliminarily discriminated scenario type on the edge side from an edge terminal to a cloud server; a grid topology structure data receiving module for the cloud server to receive current grid topology structure data from a grid GIS system; a clustering analysis module for the cloud server to perform clustering analysis on the edge node state data and the preliminarily discriminated scenario type on the edge side to obtain an edge node state clustering center and an initial discriminant clustering center of the scenario type; a spatial topology expression optimization module for the cloud server to optimize the spatial topology expression of the edge node state clustering center based on the current grid topology structure data to obtain an optimized edge node state clustering center; a grid operation scenario multi-source fusion module for the cloud server to fuse the initial discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain a multi-source fusion feature representation of the grid operation scenario; a scenario type label determination module for the cloud server to determine a grid operation scenario type label based on the multi-source fusion feature representation of the grid operation scenario; a preset control strategy sending module for the cloud server to match a preset control strategy from a control strategy library based on the grid operation scenario type label and send the preset control strategy to the corresponding edge terminal.

[0016] Compared with the prior art, the grid fragmented resource cloud-edge collaborative control system and method provided by this application include: the edge terminal uploads edge data (including node status data and preliminary discriminant scenario types) to the cloud server; the cloud server receives the topological structure data of the grid GIS system, and performs clustering analysis on the edge data to obtain the edge node status clustering center and the preliminary discriminant clustering center of the scenario type; optimizes the edge node status clustering center based on the topological structure data; fuses the optimized node status clustering center and the scenario type discriminant center to form a multi-source fusion feature representation of the grid operation scenario; determines the grid operation scenario type label according to this representation; matches the corresponding policy from the control policy library and sends it to the edge terminal for execution. In this way, the grid operation efficiency and stability are improved, the adaptability to dynamic changes is enhanced, and it is applicable to efficient resource management and control in complex power system environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 It is a schematic flowchart of the grid fragmented resource cloud-edge collaborative control method according to an embodiment of the present application.

[0019] Figure 2 It is a schematic flowchart of S3 in the grid fragmented resource cloud-edge collaborative control method according to an embodiment of the present application.

[0020] Figure 3 It is a schematic flowchart of another embodiment of S3 in the grid fragmented resource cloud-edge collaborative control method according to an embodiment of the present application.

[0021] Figure 4 It is a schematic flowchart of S4 in the grid fragmented resource cloud-edge collaborative control method according to an embodiment of the present application.

[0022] Figure 5 It is a schematic block diagram of the grid fragmented resource cloud-edge collaborative control system according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described here.

[0024] Figure 1 It is a schematic flowchart of the cloud-edge collaborative control method for fragmented power grid resources according to an embodiment of the present application. As Figure 1 shown, the cloud-edge collaborative control method for fragmented power grid resources includes: S1, the edge terminal uploads edge data to the cloud server, and the edge data includes edge node status data and a preliminary discriminant scenario type on the edge side; S2, the cloud server receives the current power grid topology structure data from the power grid GIS system; S3, the cloud server performs clustering analysis on the edge node status data and the preliminary discriminant scenario type on the edge side to obtain an edge node status clustering center and a preliminary discriminant clustering center for the scenario type; S4, the cloud server optimizes the spatial topology expression of the edge node status clustering center based on the current power grid topology structure data to obtain an optimized edge node status clustering center; S5, the cloud server fuses the preliminary discriminant clustering center for the scenario type and the optimized edge node status clustering center to obtain a multi-source fusion feature representation of the power grid operation scenario; S6, the cloud server determines a power grid operation scenario type label based on the multi-source fusion feature representation of the power grid operation scenario; S7, the cloud server matches a preset control strategy from the control strategy library based on the power grid operation scenario type label and sends the preset control strategy to the corresponding edge terminal.

[0025] Specifically, in step S1, the edge terminal uploads edge data to the cloud server, and the edge data includes edge node status data and a preliminary discriminant scenario type on the edge side. It should be understood that traditional centralized control systems are difficult to process a large amount of scattered data in real time and make rapid and accurate responses. Therefore, it is necessary for the edge terminal to upload edge data to the cloud server. In this way, a large amount of real-time monitoring data can be transmitted to the cloud for comprehensive analysis, thereby improving the timeliness and accuracy of decision-making. In one embodiment, the edge node status data includes voltage, current, active power, reactive power, and meteorological data, and the preliminary discriminant scenario type on the edge side includes a load level type label, a new energy output level type label, and whether a voltage over-limit occurs. Specifically, the edge node status data (such as voltage, current, active power, reactive power, etc.) provides key information on the current operating conditions of each node in the power grid, while the meteorological data helps predict the future power grid load change trend for a period of time. In addition, the preliminary discriminant scenario type on the edge side (such as a load level type label, a new energy output level type label, and whether a voltage over-limit occurs) provides preliminary scenario classification information for the cloud server, making the subsequent clustering analysis and strategy matching more efficient.

[0026] In a specific embodiment, it is first necessary to deploy corresponding sensors and data acquisition devices at each edge terminal. For example, in the application scenario of the smart grid, smart meters and sensors are installed in each substation and distribution station to monitor parameters such as voltage, current, active power, and reactive power in real time, and record weather conditions (such as temperature, humidity, wind speed, etc.). After these data are collected by the edge terminal, they are uploaded to the cloud server through a secure and reliable communication network (such as 4G / 5G or fiber optic network). During the upload process, the data is usually compressed and encrypted to ensure the security and integrity of the data.

[0027] Specifically, in step S2, the cloud server receives the current power grid topology structure data from the power grid GIS system. It should be understood that the topology structure data of the power grid is crucial for realizing efficient and accurate monitoring and control of the power grid operation status. The power grid Geographic Information System (GIS) provides detailed power grid topology structure information, including substations, transmission lines, distribution lines, and their connection relationships, etc. The cloud server receives the current power grid topology structure data from the power grid GIS system in order to comprehensively understand the physical architecture of the power grid and perform more accurate data analysis and strategy formulation on this basis.

[0028] Specifically, the power grid topology structure data provides a basic framework for understanding the power grid operation status. Through these data, the electrical connection relationships between each node and the overall layout of the network can be clarified, which is of great significance for identifying potential fault points and optimizing the power distribution path. For example, in the event of a local fault, based on the power grid topology structure data, the affected area can be quickly located and corresponding isolation measures can be taken to prevent the spread of the fault. In addition, the power grid topology structure data also supports the simulation and prediction of the power grid behavior under different scenarios, which helps to formulate a more scientific and reasonable scheduling plan.

[0029] In one embodiment, the power grid GIS system is deployed in the local data center or the cloud, and is responsible for maintaining and updating the detailed topology structure information of the power grid. These information are stored in a standardized format, such as the CIM (Common Information Model) model, to facilitate interoperability with other systems. In a specific embodiment, triggered regularly every day or according to demand, the power grid GIS system will push the latest power grid topology structure data to the central cloud server. These data include but are not limited to substation locations, transmission line lengths and capacities, switch states, etc. After the cloud server receives these data, it first parses and validates them to ensure the consistency and integrity of the data. Subsequently, the cloud server combines these data with the real-time operation data obtained from the edge terminal for in-depth analysis.

[0030] Specifically, in step S3, the cloud server performs clustering analysis on the edge node status data and the initially discriminated scenario types on the edge side to obtain the edge node status clustering center and the initially discriminated scenario type clustering center. It should be understood that clustering analysis helps to extract representative features from a large amount of edge node status data. For example, in a smart grid containing multiple substations and distribution stations, each station uploads a large amount of real-time monitoring data, including parameters such as voltage, current, active power, and reactive power. If these raw data are directly used for analysis, not only is the computational complexity high, but it is also easily affected by noise and outliers. Through clustering analysis, similar data points can be grouped into one category to form several clustering centers, which represent different types of operating states or scenarios, making subsequent analysis and decision-making more efficient and accurate. At the same time, clustering analysis can help identify potential problems in the power grid. For example, when the voltage levels in certain areas are generally high, clustering analysis can quickly locate these abnormal areas and take corresponding measures for adjustment. In addition, clustering analysis can also be used to predict possible future problems, such as voltage over-limit phenomena during peak load periods. By performing clustering analysis on historical data, specific operating trends under certain patterns can be discovered, so as to formulate countermeasures in advance.

[0031] In one embodiment, as Figure 2 shown, in step S3, the cloud server performs clustering analysis on the edge node status data and the initially discriminated scenario types on the edge side to obtain the edge node status clustering center and the initially discriminated scenario type clustering center, including: S31, performing fully connected encoding on each of the edge node status data to obtain a set of fully connected embedded encoding vectors for the edge node status; S32, performing DBSCAN clustering on the set of fully connected embedded encoding vectors for the edge node status to obtain the edge node status clustering center.

[0032] Specifically, first, the cloud server performs fully connected encoding on the edge node status data to generate a set of fully connected embedded encoding vectors. Next, the cloud server uses the DBSCAN algorithm to perform clustering analysis on the set of fully connected embedded encoding vectors. DBSCAN is a density-based clustering algorithm that does not require pre-specifying the number of clusters and can effectively handle noisy data. Specifically, the DBSCAN algorithm determines which points belong to the same cluster by defining two key parameters: Eps (neighborhood radius) and MinPts (minimum number of points). If the number of points within the Eps neighborhood of a certain point is greater than or equal to MinPts, then this point is regarded as a core point, and all points within its neighborhood are grouped into the same cluster. In this way, similar node status data can be grouped into one category to form the edge node status clustering center.

[0033] In another embodiment, asFigure 3 As shown, in step S3, the cloud server performs clustering analysis on the edge node status data and the edge-side preliminary discriminant scenario types to obtain the edge node status clustering center and the scenario type preliminary discriminant clustering center, further including: S33, performing one-hot encoding on each of the edge-side preliminary discriminant scenario types to obtain a set of edge-side preliminary discriminant scenario type one-hot encoding vectors; S34, performing K-Means clustering on the set of edge-side preliminary discriminant scenario type one-hot encoding vectors to obtain K local discriminant clustering centers for scenario types; S35, calculating the weighted sum of the K local discriminant clustering centers for scenario types to obtain the scenario type preliminary discriminant clustering center.

[0034] That is, in addition to the edge node status data, the cloud server also needs to perform clustering analysis on the edge-side preliminary discriminant scenario types. For the edge-side preliminary discriminant scenario types, the cloud server uses the K-Means algorithm for clustering analysis. Specifically, the cloud server first performs one-hot encoding on the edge-side preliminary discriminant scenario types to generate a set of one-hot encoding vectors. Then, the K-Means algorithm is used to cluster this set to obtain K local discriminant clustering centers for scenario types. The K-Means algorithm is a commonly used clustering method that divides data into K clusters through iterative optimization, with each cluster represented by a center point (centroid). The goal of the K-Means algorithm is to minimize the distance between points within the cluster, thus forming a compact clustering result. Finally, to further improve the accuracy of the clustering result, the cloud server also needs to calculate the weighted sum of these K local discriminant clustering centers for scenario types to finally obtain the scenario type preliminary discriminant clustering center.

[0035] Specifically, in step S4, the cloud server optimizes the spatial topological expression of the edge node state clustering center based on the current power grid topological structure data to obtain an optimized edge node state clustering center. It should be understood that the optimization of the spatial topological expression helps to combine the edge node state clustering center with the actual power grid topological structure. For example, in a complex smart grid, a large amount of real-time monitoring data is uploaded by each substation and distribution station. After clustering analysis, these data form multiple clustering centers. However, relying solely on these clustering centers cannot fully reflect the actual operating conditions of the power grid because they ignore the electrical connection relationships and network topological structures between nodes. Through the optimization of the spatial topological expression, the clustering centers can be mapped into the topological space of the power grid topological structure feature coding matrix, so that each clustering center has clear physical meaning and location information. This not only improves the accuracy of data analysis but also provides strong support for subsequent control strategy matching. At the same time, the optimization of the spatial topological expression can help identify potential power grid problems and bottlenecks. For example, when the voltage levels in certain areas are generally high, through the optimization of the spatial topological expression, these abnormal areas can be quickly located and corresponding measures can be taken for adjustment. In addition, the optimization of the spatial topological expression can also be used to predict potential future problems, such as voltage over-limit phenomena during peak load periods. By optimizing the spatial topological expression of historical data, the operating trends under specific patterns can be discovered, and corresponding countermeasures can be formulated in advance.

[0036] In a specific embodiment, as Figure 4 shown, in step S4, the cloud server optimizes the spatial topological expression of the edge node state clustering center based on the current power grid topological structure data to obtain an optimized edge node state clustering center, including: S41, performing a power grid topological feature extractor based on convolutional coding on the current power grid topological structure data to obtain a power grid topological structure feature coding matrix; S42, mapping the edge node state clustering center into the topological space of the power grid topological structure feature coding matrix through vector multiplication to obtain the optimized edge node state clustering center.

[0037] That is to say, first, the cloud server processes the current power grid topological structure data with a power grid topological feature extractor based on convolutional coding to obtain a power grid topological structure feature coding matrix. Convolutional coding is a commonly used feature extraction method. Through the combination of convolutional layers and pooling layers, it can effectively capture the local features and global dependencies in the power grid topological structure. Specifically, the convolutional encoder performs multi-layer convolutional operations on the power grid topological structure data. Each layer of convolution extracts feature maps of different scales, and finally forms a high-dimensional feature coding matrix.

[0038] Next, after obtaining the power grid topology structure feature coding matrix, the cloud server maps the edge node state clustering center into the topology space of the power grid topology structure feature coding matrix through vector multiplication, so as to obtain the optimized edge node state clustering center. This mapping process can be realized through matrix operations, that is, multiplying the edge node state clustering center vector by the power grid topology structure feature coding matrix to obtain a new vector representation. This new vector representation not only contains the information of the original clustering center, but also combines the power grid topology structure features.

[0039] Specifically, in step S5, the cloud server fuses the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain a multi-source fusion feature representation of the power grid operation scenario. It should be understood that multi-source fusion helps to comprehensively understand the operation state of the power grid from multiple dimensions. For example, in a complex smart grid, a large amount of real-time monitoring data is uploaded by each substation and distribution station, and these data form the edge node state clustering center after clustering analysis. However, relying solely on these clustering centers cannot fully reflect the actual operation status of the power grid because they ignore the preliminary discriminant results of the scenario type. By combining the preliminary discriminant clustering center of the scenario type with the optimized edge node state clustering center, a multi-dimensional feature representation can be formed, so as to more comprehensively understand the actual operation of the power grid. At the same time, multi-source fusion can help identify potential power grid problems and bottlenecks. For example, when the voltage levels in certain areas are generally high, multi-source fusion can quickly locate these abnormal areas and take corresponding measures for adjustment. In addition, multi-source fusion can also be used to predict possible future problems, such as voltage over-limit phenomena during peak load periods. By performing multi-source fusion on historical data, the operation trends under specific patterns can be discovered, so as to formulate countermeasures in advance.

[0040] In a specific embodiment, the cloud server fuses the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain a multi-source fusion feature representation of the power grid operation scenario, including: cascading the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain the multi-source fusion feature representation of the power grid operation scenario. This multi-source fusion feature representation not only contains the key information of the node state, but also combines the preliminary discriminant results of the scenario type, laying a foundation for more accurate scenario classification.

[0041] In particular, after mapping the edge node state clustering center to the topological space of the power grid topological structure feature encoding matrix through vector multiplication to obtain the optimized edge node state clustering center, the vector heterogeneity between the optimized edge node state clustering center and the preliminary discriminant clustering center of the scenario type will become more obvious. Therefore, it is expected to improve the classification accuracy of the scenario classifier by vector homogenization of the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center belonging to different representation dimension spaces.

[0042] In a preferred embodiment, cascading the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain the multi-source fusion feature representation of the power grid operation scenario includes: using the equivalence of the pseudo-inverse norm under the homogeneous space structure to perform cross-dimensional isomorphism on the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain an optimized preliminary discriminant clustering center of the scenario type; cascading the optimized preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain the multi-source fusion feature representation of the power grid operation scenario.

[0043] Specifically, using the equivalence of the pseudo-inverse norm under the homogeneous space structure to perform cross-dimensional isomorphism on the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain an optimized preliminary discriminant clustering center of the scenario type includes: First, let the edge node state clustering center be represented as , and the power grid topological structure feature encoding matrix be represented as , then the optimized edge node state clustering center is , and the preliminary discriminant clustering center of the scenario type is , then in , where is the corresponding base space, it is expected to achieve , where is the co-space.

[0044] Then, first based on the optimized edge node state clustering center perform joint space trivialization on the preliminary discriminant clustering center of the scenario type to obtain the joint space trivialization matrix, which is expressed as: ; where, represents the edge node state clustering center, represents the transpose of the preliminary discriminant clustering center of the scenario type, and are all column vectors, represents the inverse matrix of the power grid topological structure feature encoding matrix, represents matrix multiplication, The base space representing the optimized edge node state clustering center The base space representing the preliminary discriminant clustering center of the scenario type Represents and The joint tensor product space of Represents the joint space trivialization matrix

[0045] Define As a full-dimensional homogeneous mapping in the complex manifold space, perform a full-dimensional homogeneous mapping in the complex manifold space on the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center to obtain a full-dimensional homogeneous mapping vector, denoted as: ; where Represents the full-dimensional homogeneous mapping vector Represents mapping the vector in the space to the space

[0046] Based on the joint space trivialization matrix and the full-dimensional homogeneous mapping vector, use the pseudo-inverse matrix to perform group action homogenization on the preliminary discriminant clustering center of the scenario type based on the homogeneous space structure to obtain the optimized preliminary discriminant clustering center of the scenario type, denoted as: ; where Represents the inverse matrix of the joint space trivialization matrix Represents the Frobenius norm of the inverse matrix of the joint space trivialization matrix Represents element-wise multiplication Represents element-wise addition Represents the weight hyperparameter Represents the optimized preliminary discriminant clustering center of the scenario type

[0047] That is, utilize the pseudo-inverse norm equivalence under the homogeneous space structure to achieve cross-dimensional isomorphism without dimension scaling of the preliminary discriminant clustering center of the scenario type and the optimized edge node state clustering center, thereby improving the classification accuracy while avoiding the introduction of redundant specifications

[0048] Specifically, in step S6, the cloud server determines the power grid operation scenario type label based on the multi-source fusion feature representation of the power grid operation scenario. It should be understood that determining the power grid operation scenario type label can help identify potential power grid problems and bottlenecks. For example, when the voltage levels in certain areas are generally high, the specific scenario type label can be used to quickly locate these abnormal areas and take corresponding measures for adjustment. In addition, determining the scenario type label can also be used to predict possible future problems, such as voltage over-limit phenomena during peak load periods. By classifying historical data into scenarios, the operation trends under specific patterns can be discovered, enabling the formulation of countermeasures in advance. At the same time, determining the power grid operation scenario type label is of great significance for optimizing control strategies. By identifying different operation scenario types, the most appropriate control strategy can be selected according to the specific situation. For example, at high load levels, increasing power generation capacity or adjusting load distribution can be chosen; while when the output of new energy fluctuates greatly, measures such as charging and discharging energy storage devices can be taken to smooth the output.

[0049] In a specific embodiment, the cloud server determines the power grid operation scenario type label based on the multi-source fusion feature representation of the power grid operation scenario, including: inputting the multi-source fusion feature representation of the power grid operation scenario into a scenario classifier based on a support vector machine model (SVM) to obtain the power grid operation scenario type label. Among them, SVM is a commonly used machine learning classification algorithm that maximizes the margin between different categories by finding the optimal hyperplane, thus achieving an efficient classification effect. Specifically, the training process of the SVM model includes the following steps: First, data preparation, using historical power grid operation data and their corresponding scenario type labels as training samples. After preprocessing and feature extraction, these samples form a feature vector set. Then, model training, training the SVM model with the training samples. During the training process, the SVM model automatically adjusts parameters to find the optimal hyperplane, making the samples of different categories have the largest margin. Next, model verification, verifying the trained SVM model through methods such as cross-validation to ensure its good generalization ability. If the model performs poorly, the model performance can be improved by adjusting parameters or increasing training samples. Finally, classification prediction, in practical applications, the cloud server inputs the multi-source fusion feature representation of the power grid operation scenario into the trained SVM model, and the model outputs the corresponding power grid operation scenario type label. These labels can be used to guide the subsequent matching and execution of control strategies.

[0050] Specifically, in step S7, the cloud server matches a preset control strategy from the control strategy library based on the power grid operation scenario type label and sends the preset control strategy to the corresponding edge terminal. It should be understood that matching the preset control strategy based on the power grid operation scenario type label helps to comprehensively understand the operation status of the power grid from multiple dimensions and select the most appropriate control measures according to the specific situation. For example, in a complex smart grid, a large amount of real-time monitoring data is uploaded by each substation and distribution station. After clustering analysis and multi-source fusion, these data form power grid operation scenario type labels. These labels can clarify the current operation status of the power grid, such as high load status, low load status, voltage over-limit status, etc. By matching these labels with the preset control strategies in the control strategy library, the most suitable control strategy for the current operation status can be selected, thereby improving the response speed and accuracy of the system.

[0051] In a specific embodiment, the control strategy matching and sending can be achieved through the following steps: First, a strategy library containing various preset control strategies is established on the cloud server. These strategies are predefined based on historical data and expert experience and cover control measures under different operation scenarios. For example, for the high load status, the strategy library may include strategies such as increasing the power generation capacity and adjusting the load distribution; for the voltage over-limit status, the strategy library may include strategies such as adjusting the tap position of the transformer and starting the reactive power compensation device. Then, the cloud server inputs the power grid operation scenario type label into the strategy library matching module, and automatically selects the most suitable control strategy according to the label. The matching process can be achieved through a rule engine or a machine learning model. For example, if the current label is the high load status, the system will select the strategy of increasing the power generation capacity or adjusting the load distribution; if the label is the voltage over-limit status, the system will select the strategy of adjusting the tap position of the transformer or starting the reactive power compensation device. Finally, after the matching is completed, the cloud server sends the selected control strategy to the corresponding edge terminal through a secure and reliable communication network. After receiving the strategy, the edge terminal will perform corresponding operations according to the instructions. For example, if the strategy is to adjust the tap position of the transformer, the edge terminal will send a control signal to the transformer controller to adjust its tap position; if the strategy is to start the reactive power compensation device, the edge terminal will send a start signal to the reactive power compensation device to make it start working.

[0052] In summary, the cloud-edge collaborative control method for fragmented power grid resources provided in this application has been clarified. This method uploads edge data including status data such as voltage and current and initially discriminates the scenario type from the edge terminal to the cloud server, and processes it in combination with the topological structure data of the power grid GIS system. The cloud server first performs clustering analysis on the edge node status data and scenario type to obtain the edge node status clustering center and the initial discrimination clustering center of the scenario type; then optimizes the edge node status clustering center based on the power grid topological structure data. By fusing the optimized node status clustering center and the scenario type discrimination center, a multi-source fusion feature representation of the power grid operation scenario is formed, and the power grid operation scenario type label is determined. Finally, the corresponding preset control strategy is matched from the control strategy library according to the label and sent to the edge terminal for execution. This method improves the operation efficiency and stability of the power grid, enhances the adaptability of the system to dynamic changes, and is suitable for efficient resource management and control in complex power system environments.

[0053] This application also provides a cloud-edge collaborative control system for fragmented power grid resources, as Figure 5 shown. The cloud-edge collaborative control system 500 for fragmented power grid resources includes: an edge data acquisition module 510, configured to upload edge data from the edge terminal to the cloud server, where the edge data includes edge node status data and the initially discriminated scenario type on the edge side; a power grid topological structure data receiving module 520, configured to receive the current power grid topological structure data from the power grid GIS system by the cloud server; a clustering analysis module 530, configured to perform clustering analysis on the edge node status data and the initially discriminated scenario type on the edge side by the cloud server to obtain the edge node status clustering center and the initial discrimination clustering center of the scenario type; a spatial topological expression optimization module 540, configured to optimize the spatial topological expression of the edge node status clustering center based on the current power grid topological structure data by the cloud server to obtain the optimized edge node status clustering center; a multi-source fusion module 550 for power grid operation scenarios, configured to fuse the initial discrimination clustering center of the scenario type and the optimized edge node status clustering center by the cloud server to obtain a multi-source fusion feature representation of the power grid operation scenario; a scenario type label determination module 560, configured to determine the power grid operation scenario type label based on the multi-source fusion feature representation of the power grid operation scenario by the cloud server; and a preset control strategy sending module 570, configured to match a preset control strategy from the control strategy library based on the power grid operation scenario type label by the cloud server and send the preset control strategy to the corresponding edge terminal.

[0054] This application embodiment also provides a computer-readable storage medium, in which computer program code is stored. When the computer program code runs on a computer, the computer is caused to execute the above-related method steps to implement a cloud-edge collaborative control method for fragmented power grid resources provided in the above embodiment.

[0055] The embodiment of the present application also provides a computer program product. When the computer program product runs on a computer, it causes the computer to execute the above-related steps to implement a grid fragmented resource cloud-edge collaborative control method provided by the above embodiment.

[0056] Among them, the system, computer-readable storage medium or computer program product provided by the embodiment of the present application are all used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method provided above, and will not be elaborated here.

[0057] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments.

[0058] The processes depicted in the accompanying drawings do not necessarily require the particular order shown or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous. Each embodiment in this specification is described in a progressive manner, and the same or similar parts among the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A cloud-edge collaborative control method for fragmented power grid resources, characterized in that: include: The edge terminal uploads edge data to the cloud server. The edge data includes edge node status data and the initial scene type identified by the edge side. The cloud server receives the current power grid topology data from the power grid GIS system; The cloud server performs cluster analysis on the edge node status data and the edge side preliminary identification scene type to obtain the edge node status cluster center and the scene type preliminary identification cluster center; The cloud server optimizes the spatial topological expression of the edge node state cluster center based on the current power grid topology data to obtain an optimized edge node state cluster center; The cloud server fuses the scene type preliminary identification cluster center and the optimized edge node state cluster center to obtain a multi-source fusion feature representation of the power grid operation scene; The cloud server determines a type label of the power grid operation scenario based on the multi-source fusion feature representation of the power grid operation scenario; The cloud server matches a preset control strategy from a control strategy library based on the power grid operation scenario type tag, and sends the preset control strategy to the corresponding edge terminal; The edge node status data includes voltage, current, active power, reactive power and meteorological data, and the edge side initially determines the scene type including load level type label, new energy output level type label and whether voltage exceeds the limit; The cloud server performs cluster analysis on the edge node status data and the edge side preliminary identification scene type to obtain the edge node status cluster center and the scene type preliminary identification cluster center, including: Fully connect encoding each of the edge node state data to obtain a set of edge node state fully connected embedded coding vectors; Performing DBSCAN clustering on the set of edge node state fully connected embedded coding vectors to obtain the edge node state clustering center; The cloud server performs cluster analysis on the edge node status data and the edge side preliminary identification scene type to obtain the edge node status cluster center and the scene type preliminary identification cluster center, and further includes: Performing one-hot encoding on each of the edge-side initially discriminated scene types to obtain a set of one-hot encoding vectors of the edge-side initially discriminated scene types; Performing K-Means clustering on the set of one-hot encoding vectors of the edge-side preliminary scene type discrimination to obtain K scene type local discrimination cluster centers; A weighted sum of the K scene type local discriminant cluster centers is calculated to obtain the scene type preliminary discriminant cluster center.

2. The cloud-edge collaborative control method for fragmented power grid resources according to claim 1 is characterized in that: The cloud server optimizes the spatial topological expression of the edge node state cluster center based on the current power grid topology data to obtain an optimized edge node state cluster center, including: Performing a convolutional coding-based power grid topology feature extractor on the current power grid topology data to obtain a power grid topology feature coding matrix; The edge node state cluster center is mapped to the topological space of the power grid topology structure feature coding matrix through vector multiplication to obtain the optimized edge node state cluster center.

3. The cloud-edge collaborative control method for fragmented power grid resources according to claim 2 is characterized in that: The cloud server fuses the scene type preliminary identification cluster center and the optimized edge node state cluster center to obtain a multi-source fusion feature representation of the power grid operation scene, including: The scene type preliminary identification cluster center and the optimized edge node state cluster center are cascaded to obtain a multi-source fusion feature representation of the power grid operation scene.

4. The cloud-edge collaborative control method for fragmented power grid resources according to claim 3 is characterized in that: The cloud server determines the type label of the power grid operation scenario based on the multi-source fusion feature representation of the power grid operation scenario, including: The multi-source fusion feature representation of the power grid operation scenario is input into a scenario classifier based on a support vector machine model to obtain a type label of the power grid operation scenario.

5. The cloud-edge collaborative control method for fragmented power grid resources according to claim 4 is characterized in that: The scene type preliminary identification cluster center and the optimized edge node state cluster center are cascaded to obtain a multi-source fusion feature representation of the power grid operation scene, including: By using the pseudo-inverse norm equivalence under the homogeneous space structure, the scene type preliminary discrimination cluster center and the optimized edge node state cluster center are subjected to cross-dimensional isomorphism to obtain the optimized scene type preliminary discrimination cluster center; The optimized scene type preliminary identification cluster center and the optimized edge node state cluster center are cascaded to obtain a multi-source fusion feature representation of the power grid operation scenario.

6. The cloud-edge collaborative control method for fragmented power grid resources according to claim 5 is characterized in that: By utilizing the pseudo-inverse norm equivalence under the homogeneous space structure, the scene type preliminary discrimination cluster center and the optimized edge node state cluster center are cross-dimensionally isomorphic to obtain the optimized scene type preliminary discrimination cluster center, including: Based on the optimized edge node state cluster center, the scene type preliminary discrimination cluster center is subjected to joint spatial trivialization to obtain a joint spatial trivialization matrix; Performing full-dimensional homogeneous mapping on the scene type preliminary identification cluster center and the optimized edge node state cluster center in a complex manifold space to obtain a full-dimensional homogeneous mapping vector; Based on the joint space trivialization matrix and the full-dimensional homogeneous mapping vector, the scene type preliminary discrimination cluster center is homogenized based on the group action of the homogeneous space structure using a pseudo-inverse matrix to obtain an optimized scene type preliminary discrimination cluster center.

7. A power grid fragmentation resource cloud-edge collaborative control system, used to implement the power grid fragmentation resource cloud-edge collaborative control method according to any one of claims 1-6, characterized in that: include: Edge data acquisition module, used for edge terminals to upload edge data to the cloud server. The edge data includes edge node status data and the initial scene type identified by the edge side; A power grid topology data receiving module is used for the cloud server to receive the current power grid topology data from the power grid GIS system; A cluster analysis module, used for the cloud server to perform cluster analysis on the edge node status data and the edge side preliminary identification scene type to obtain the edge node status cluster center and the scene type preliminary identification cluster center; A spatial topology expression optimization module, used for the cloud server to optimize the spatial topology expression of the edge node state cluster center based on the current power grid topology structure data to obtain an optimized edge node state cluster center; A multi-source fusion module for power grid operation scenarios, used for the cloud server to fuse the scene type preliminary identification cluster center and the optimized edge node state cluster center to obtain a multi-source fusion feature representation of the power grid operation scenario; A scenario type label determination module is used for the cloud server to determine the type label of the power grid operation scenario based on the multi-source fusion feature representation of the power grid operation scenario; The preset control strategy sending module is used for the cloud server to match the preset control strategy from the control strategy library based on the power grid operation scenario type label, and send the preset control strategy to the corresponding edge terminal.

Images