Cloud edge-end collaborative virtual power plant scheduling method and system based on dynamic task migration

Abstract

The invention discloses a cloud side-end collaborative virtual power plant scheduling method and system based on dynamic task migration, relates to the technical field of virtual power plant resource scheduling, and solves the problem that a general method is difficult to meet the real-time scheduling requirement of a virtual power plant. According to the invention, task migration is realized through a three-level architecture of a cloud planning long-term strategy, regional collaboration of an edge layer and quick response of a terminal layer in combination with a dynamic load balancing algorithm; the first edge node locally processes regional collaborative tasks, reduces the communication frequency of the cloud server, reduces the end-to-end delay, and then combines federated learning to realize training parameter sharing between the edge nodes and avoid redundant data transmission, thereby effectively improving the response speed of the scheduling system, and improving the scheduling efficiency. And the real-time scheduling requirement of the virtual power plant can be better met.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual power plant resource scheduling, and in particular to a cloud-edge collaborative virtual power plant scheduling method and system based on dynamic task migration. Background Art

[0002] In the field of virtual power plant resource scheduling technology, current mainstream virtual power plant platforms generally rely on a centralized cloud platform architecture to implement global resource scheduling and management. In this architecture, edge nodes are solely responsible for data collection and upload, as well as command reception and execution, forming a hierarchical relationship with the cloud for one-way command transmission. In the early stages of virtual power plant development, this architectural model, leveraging the advantages of centralized management, achieved a certain degree of resource optimization.

[0003] However, with the rapid development of the Energy Internet, the number of terminal devices connected to virtual power plants, such as distributed power sources and controllable loads, has exploded. In this context, the amount of real-time monitoring data and operational status information that the cloud needs to process is growing exponentially. On the one hand, the cloud's centralized processing of massive amounts of terminal data is limited by network bandwidth and computing load, making it unable to meet the real-time scheduling requirements of virtual power plants. On the other hand, the lack of coordination between edge nodes means that redundant data from terminal devices in the same area must be repeatedly uploaded to the cloud, consuming communication resources.

[0004] The limitations of traditional architectures are becoming increasingly apparent in practical applications, making them unable to meet the stringent real-time scheduling and rapid response requirements of virtual power plants. Especially in complex scenarios such as sudden fluctuations in the power system and dramatic changes in renewable energy output, traditional architectures are prone to scheduling delays and imbalanced resource allocation, severely restricting the operational efficiency and reliability of virtual power plants. They are unable to fully leverage their advantages in optimizing energy allocation and improving grid stability, and urgently need to be improved through technological innovation and architectural optimization.

[0005] Based on this, there is an urgent need for a cloud-edge collaborative virtual power plant scheduling method and system based on dynamic task migration. Summary of the Invention

[0006] In response to the problem that existing technologies are difficult to meet the real-time scheduling needs of virtual power plants, the present invention provides a cloud-edge collaborative virtual power plant scheduling method and system based on dynamic task migration, which can better meet the real-time scheduling needs of virtual power plants. The specific technical solution is as follows:

[0007] In a first aspect, an embodiment of the present application provides a cloud-edge collaborative virtual power plant scheduling method based on dynamic task migration, which is applied to a cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration. The system includes a cloud server, an edge node, and a terminal device. The edge node includes a first edge node and a second edge node, and the first edge node and the second edge node are different edge nodes. The method includes:

[0008] The cloud server generates an initial scheduling strategy based on historical load data and environmental data; wherein, the environmental data is environmental data that affects the output of new energy stations aggregated by the virtual power plant; the first edge node obtains the status data of the terminal device managed by the first edge node itself, and generates a task package through the federated learning framework based on the status data and the initial scheduling strategy; the terminal device executes the control instructions in the task package, and sends an abnormality information to the first edge node when an execution abnormality of the control instruction is detected; the first edge node monitors the network status and the computing resource load of the first edge node itself, and when the computing resource load is greater than a preset first threshold, based on the network status and the abnormality information, migrates the task package to be processed to the target migration node, and the target migration node includes the cloud server or the second edge node.

[0009] Preferably, the environmental data includes wind speed data and irradiance data; the cloud server generates an initial scheduling strategy based on historical load data and environmental data, including: the cloud server inputs the historical power load data into a long short-term memory (LSTM) model to obtain predicted load data output by the LSTM model; the cloud server generates an output prediction curve of the new energy station based on the wind speed data and the irradiance data; and generates the initial scheduling strategy based on the predicted load data, the output prediction curve and the output parameters of the traditional power generation units aggregated by the virtual power plant.

[0010] Preferably, the first edge node obtains the status data of the terminal device managed by the edge node itself, and generates a task package through the federated learning framework based on the status data and the initial scheduling strategy, including: the first edge node inputs the status data and the initial scheduling strategy into the regional optimization model, and obtains the task package output by the regional optimization model; wherein each edge node is respectively deployed with the regional optimization model, and the regional optimization model and the federated learning model are federated through gradient aggregation, the weight of the federated learning model is obtained based on the initial scheduling strategy, and the regional optimization model constructs a loss function based on the weight of the federated learning model.

[0011] Preferably, the task package to be processed is migrated to the target migration node based on the network status and the abnormal information, including: calculating a migration score based on the network status, the computing resource load of the second edge node and the priority tag in the task package to be processed; wherein the priority tag is used to indicate the processing priority of the task package to be processed; and selecting the target migration node of the task package to be processed based on the migration score.

[0012] Preferably, migrating the pending task package to the target migration node based on the network status and the abnormal information includes: obtaining the computational complexity of the pending task package; and migrating the pending task package to the cloud server when the computational complexity is higher than a preset second threshold.

[0013] Preferably, after the first edge node obtains the status data of the terminal device managed by the edge node itself, and generates a task package through the federated learning framework based on the status data and the initial scheduling strategy, the method also includes: the first edge node generates a global task mapping table in the blockchain based on the task package; after the task package to be processed is migrated to the target migration node based on the network status and the abnormal information, the method also includes: the first edge node updates the global task mapping table based on the task package to be processed and the target migration node; the target migration node obtains the context information of the task package to be processed based on the global task mapping table, and processes the task package to be processed based on the context information.

[0014] Preferably, the task package is a binary task package, which further includes a device control instruction set and its corresponding execution time window, and a priority tag.

[0015] In a second aspect, an embodiment of the present application provides a cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration, which is applied to the method described in the first aspect. The system includes a cloud server, an edge node, and a terminal device. The edge node includes a first edge node and a second edge node. The first edge node and the second edge node are different edge nodes.

[0016] The cloud server is used to generate an initial dispatch strategy based on historical load data and environmental data; wherein the environmental data is environmental data that affects the output of new energy stations aggregated by the virtual power plant;

[0017] The first edge node is used to obtain status data of a terminal device managed by the first edge node itself, and generate a task package through a federated learning framework based on the status data and the initial scheduling policy;

[0018] The terminal device is used to execute the control instruction in the task package, and send abnormality information to the first edge node when detecting that the execution of the control instruction is abnormal;

[0019] The first edge node is also used to monitor the network status and the computing resource load of the first edge node itself, and when the computing resource load is greater than a preset first threshold, based on the network status and the abnormal information, migrate the task package to be processed to the target migration node, which includes the cloud server or the second edge node.

[0020] Preferably, the first edge node is further used to generate a global task mapping table in the blockchain based on the task package; the first edge node is further used to update the global task mapping table based on the task package to be processed and the target migration node; the target migration node is used to obtain context information of the task package to be processed based on the global task mapping table, and process the task package to be processed based on the context information.

[0021] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the method as described in the first aspect.

[0022] Compared with the existing technology, the beneficial effects of the present invention are: through the three-level architecture of long-term strategy planning in the cloud, regional collaboration at the edge layer, and rapid response at the terminal layer, task migration is realized in combination with a dynamic load balancing algorithm; the first edge node locally processes regional collaboration tasks, reduces the frequency of cloud communication, reduces end-to-end latency, and then combines federated learning to realize model parameter sharing between edge nodes, avoids redundant data transmission, thereby effectively improving the response speed of the scheduling system, and can better meet the implementation and scheduling needs of virtual power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0024] Figure 1 A flowchart of a cloud-edge-device collaborative virtual power plant scheduling method based on dynamic task migration provided in an embodiment of the present application;

[0025] Figure 2 A schematic diagram of the structure of a cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0028] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0029] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0030] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0031] In order to solve the problem that traditional methods are difficult to meet the real-time scheduling needs of virtual power plants, the present invention provides a cloud-edge collaborative virtual power plant scheduling method and system based on dynamic task migration, which can better meet the real-time scheduling needs of virtual power plants.

[0032] See also Figure 1 , Figure 1 A flow chart of a cloud-edge collaborative virtual power plant scheduling method based on dynamic task migration is provided for an embodiment of the present application. The method is applied to a cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration, the system including a cloud server, an edge node and a terminal device, the edge node including a first edge node and a second edge node, the first edge node and the second edge node being different edge nodes; Figure 1 As shown, the method includes:

[0033] Step 101: The cloud server generates an initial scheduling strategy based on historical load data and environmental data.

[0034] Among them, the cloud server can be a blade server, a high-density server, a rack server, a cabinet server, a general server, a graphics processing unit (GPU) server, a data processing unit (DPU) server or an artificial intelligence (AI) server.

[0035] Preferably, a long-term prediction model and a blockchain evidence storage module are run on the cloud server. The long-term prediction model is used to predict the load on the electricity demand side and the output of new energy stations in the power system; the blockchain evidence storage module is used to store relevant data on scheduling plans and scheduling task execution, and uses the tamper-proof characteristics of the blockchain to perform data encryption protection, data sharing and coordination, auditing and tracing, and improve the reliability and fault tolerance of scheduling actions.

[0036] Exemplarily, the cloud server is deployed in a data center and equipped with a GPU cluster, and runs a long-term prediction model and a blockchain evidence storage module based on the GPU cluster.

[0037] The historical load data is the load data generated by the electricity demand side in the past. For example, the historical load data can be a load curve generated by the load data of a province over the past 30 days. The daily load data is discretized into 96 data points, and the time interval between two adjacent data points is 15 minutes.

[0038] The environmental data is environmental data that affects the output of the new energy sites aggregated by the virtual power plant. Preferably, the environmental data includes 24-hour irradiance and wind speed forecasts provided by the Meteorological Bureau; specifically, the environmental data includes 24-hour wind speed forecasts for the area where the wind power station is located, and 24-hour irradiance data for the area where the photovoltaic power station is located.

[0039] A virtual power plant is an energy management system that coordinates and controls distributed generation, demand-side response, and energy storage resources, allowing it to participate in power market transactions and grid operations in a virtual manner. This virtual power plant can be deployed as software on the cloud server.

[0040] Among them, the power grid is a power grid connected to new energy, which includes new energy power generation units such as wind turbines and photovoltaic power generation units, and can also include traditional thermal power generation units; the power generation units aggregated by the virtual power plant refer to the virtual power plant integrating many scattered power generation units of different types and sizes in the power grid for unified management and coordinated operation.

[0041] The virtual power plant can store the above historical load data during operation, and the cloud server can obtain the historical load data from the virtual power plant.

[0042] Preferably, the cloud server inputs the historical power load data into a long short-term memory (LSTM) model to obtain the predicted load data output by the LSTM model; the cloud server generates an output prediction curve of the new energy station based on the wind speed data and the irradiance data; and generates the initial scheduling strategy based on the predicted load data, the output prediction curve and the output parameters of the traditional power generation units aggregated by the virtual power plant.

[0043] The cloud server generates an output prediction curve for the wind power station in the new energy site based on the wind speed data, and generates an output prediction curve for the photovoltaic power station in the new energy site based on the irradiance data.

[0044] Preferably, the cloud server can construct a mixed integer programming model with predicted load data as a constraint and minimizing the operating cost of the virtual power plant as the objective function; then the predicted load data, the output prediction curve and the output parameters of the traditional generator sets aggregated by the virtual power plant are input into the mixed integer programming model to solve the optimal scheduling strategy.

[0045] Preferably, the cloud server can generate a strategy hash value based on the solved optimal scheduling strategy, and write the measurement hash value into the blockchain.

[0046] Step 102: The first edge node obtains status data of the terminal device managed by the first edge node itself, and generates a task package through a federated learning framework based on the status data and the initial scheduling strategy.

[0047] Among them, edge nodes can be terminal devices, storage devices, switches or servers. For example, edge nodes can be distributed in substations where virtual power plants are aggregated, specifically edge computing gateways with built-in federated learning engines and migration decision modules.

[0048] The terminal device may include a photovoltaic inverter, an energy storage converter, and a load controller. Preferably, the terminal device supports the Modbus protocol (modbus over TCP, Modbus / TCP) based on the transmission control protocol, and the MQTT protocol (message queuing telemetry transport).

[0049] Specifically, edge devices refer to computing devices located at the edge of the network, close to data sources or users; terminal devices are devices that directly interact with users or the physical world. Edge devices are responsible for collecting data generated by terminal devices, forwarding instructions from cloud servers to the corresponding terminal devices, and monitoring the terminal devices' execution of these instructions. Edge devices play a certain role in managing and coordinating terminal devices. They can centrally manage and schedule multiple terminal devices based on local operating conditions and policies.

[0050] Therefore, it is understandable that in a distributed scenario, a virtual power plant aggregates multiple edge devices to manage terminal devices in different locations and areas.

[0051] Preferably, the first edge node inputs the state data and the initial scheduling strategy into the regional optimization model, and the regional optimization model outputs the task package; wherein each edge node is respectively deployed with the regional optimization model, and the regional optimization model and the federated learning model are federated learned through gradient aggregation, the weight of the federated learning model is obtained based on the initial scheduling strategy, and the regional optimization model constructs a loss function based on the weight of the federated learning model.

[0052] The federated learning framework achieves a balance between data privacy protection and collaborative optimization by distributing training models across edge nodes. Its core principle is to keep the data static while the model is dynamic. Edge nodes only share model parameters, not raw data, and encryption technology is used to ensure data security.

[0053] Specifically, the initial scheduling policy serves as the global constraints and characteristic prior knowledge of the federated learning model. This initial scheduling measurement is parsed into a feature vector, where each dimension corresponds to a scheduling parameter (such as predicted load, renewable energy output ratio, etc.). This vector serves as the initialization weight for the federated learning model, ensuring that the local optimization direction of the edge node is consistent with the global cloud-based policy. Furthermore, during the local training phase of federated learning, the initial policy is introduced as a factor in the deviation penalty term when defining the loss function of the regional optimization model.

[0054] Among them, in the data preprocessing stage, the first edge node adds noise to the real-time status data of the terminal device (such as power output, battery SOC), and performs differential privacy desensitization on the status data, so that the data of a single device cannot be reversely inferred. Then, the first edge node locally trains the regional optimization model based on the desensitized status data, and then extracts the model gradient parameters of the regional optimization model after local training, performs homomorphic encryption, and uploads it to the cloud server; the cloud server aggregates the gradient parameters of each edge node in the ciphertext state to avoid information leakage in the intermediate links. Exemplarily, after the cloud server securely weights the gradients of each edge node, it updates the global weight of the federated learning model based on the calculation results.

[0055] The first edge node then obtains the global gradient parameters sent by the cloud server and updates the local regional optimization model based on the newly sent global gradient parameters.

[0056] After multiple training processes as described above, the first edge node obtains the status data and the initial scheduling strategy, and can directly input the status data and the initial scheduling strategy into the trained regional optimization model to obtain a binary task package containing a priority label output by the regional optimization model.

[0057] Specifically, the output of the regional optimization model is encoded into a binary instruction stream (such as device start / stop instructions and power adjustment parameters) and encapsulated into a lightweight task package. The task package uses lightweight binary encapsulation, and the instruction set is decoupled from the device (for example, standardized power adjustment instructions are defined rather than device-specific protocols), ensuring cross-node execution compatibility.

[0058] Exemplarily, the region optimization model may use a binary serialization protocol such as Protocol Buffers to compress the instruction size to less than 30% of the original text.

[0059] According to the "Guidelines for Power System Security and Stability," the first edge node can prioritize task packages based on frequency deviation thresholds. Therefore, the task package can include the deviation between the grid frequency at the time the task package is generated and the rated grid frequency as a priority tag. When the frequency deviation between the grid frequency corresponding to the task package and the rated grid frequency is less than 0.2 Hz, the grid frequency is within the normal range, and the task package has a low priority. When the frequency deviation is between 0.2 Hz and 0.5 Hz, the corresponding grid is in a warning state, and the task package requires priority processing. When the frequency deviation is greater than 0.5 Hz, the corresponding grid is in an emergency state, and the task package requires immediate processing.

[0060] The larger the frequency deviation, the faster the adjustment is required (for example, a level 3 deviation requires a response within 500ms), and high-priority task packets will be assigned to low-latency nodes.

[0061] Preferably, the task package also includes a device control instruction set and its corresponding execution time window.

[0062] Step 103: The terminal device executes the control instruction in the task package, and sends abnormality information to the first edge node when detecting that the execution of the control instruction is abnormal.

[0063] Among them, when an abnormal situation occurs when the terminal device executes the control instruction, resulting in the control instruction being unable to be completed, the terminal device can collect relevant abnormal information and report it to the corresponding first edge node.

[0064] For example, after the photovoltaic inverter receives the task package, if the current irradiance drops suddenly and the control instruction of 15% light abandonment cannot be completed, the photovoltaic inverter first detects that the output deviation is greater than the preset processing 10% and this state lasts for 5 minutes, triggering the abnormal code E103 (output is uncontrollable); then the photovoltaic inverter collects the abnormal information and uploads it.

[0065] Step 104: The first edge node monitors the network status and the computing resource load of the first edge node itself, and when the computing resource load is greater than a preset first threshold, based on the network status and the abnormal information, migrates the pending task package to the target migration node.

[0066] The target migration node includes the cloud server or the second edge node.

[0067] The pending task packages include task packages that have not been sent to the terminal device, and task packages that are not completed due to anomalies that occur during execution by the terminal device.

[0068] Preferably, the first edge node can calculate a migration score based on the network status, the computing resource load of the second edge node and the priority tag of the task package to be processed; and select a target migration node for the task package to be processed based on the migration score.

[0069] The calculation formula of the migration score may include:

[0070]

[0071] Among them, α, β, and γ are weight coefficients, L max is the maximum tolerable delay, L is the network status delay; R is the computing resource load of the second edge node, specifically the proportion of the available CPU / GPU computing power of the second edge node; U is the priority of the task packet to be processed.

[0072] When the grid frequency deviation continues to increase, the γ value can be automatically increased to strengthen the migration priority of emergency tasks.

[0073] Preferably, the first edge node can obtain the computational complexity of the task package to be processed; and when the computational complexity is higher than a preset second threshold, migrate the task package to be processed to the cloud server.

[0074] Among them, the computing power of the edge node is limited. When the computational complexity of the task package to be processed is high, the first edge node can force the task package to be processed to be migrated to the cloud server for processing to save the computing power of the edge node and avoid affecting the overload of the edge node.

[0075] The computational complexity can be obtained by weighted summing of the number of floating-point operations of subtasks such as instruction parsing, optimization calculation, and communication overhead.

[0076] Among them, after the first edge node receives the abnormal information, it can call the locally stored historical optimal strategy (such as the strategy with the highest success rate in the same scenario in the past 24 hours) to replace the current task package; then generate an encrypted event report containing the abnormal time, device ID, and error code, and upload it to the cloud server synchronously; then update the scheduling strategy stored in the blockchain based on the analysis results of the cloud server.

[0077] For example, when executing the "Discharge 300kW" command in a task package, a certain energy storage power station detected that the battery temperature exceeded 50°C (the safety threshold), triggering an abnormal signal. The edge node searched the local policy library and found that when the historical ambient temperature was 48°C, the "step discharge" strategy had a success rate of 97%. It then executed the "Discharge 300kW" command using this "step discharge" strategy. It is understandable that after completing the command execution using the historically optimal strategy, the cloud server does not need to migrate the task package for that command.

[0078] Preferably, the first edge node can generate a global task mapping table in the blockchain based on the task package; after migrating the pending task package to the target migration node based on the network status and the exception information, the first edge node updates the global task mapping table based on the pending task package and the target migration node; the target migration node obtains the context information of the pending task package based on the global task mapping table, and processes the pending task package based on the context information.

[0079] Among them, the update of the global task mapping table includes: establishing a topological relationship map between task packages and physical devices on the cloud server; when task migration occurs, marking the migration path and the location of the new execution node in the map; and dynamically adjusting the connection weights between nodes in the map based on the changes in node load after migration.

[0080] The target migration node can calculate the hash value of the header of the pending task package and match it with the hash value stored in the blockchain to confirm whether the pending task package is a legal task package.

[0081] After determining that the task package to be processed is a legitimate task package, the target migration node can parse the device instruction set in the task package, generate a dependency graph, determine the resources required to execute the instruction, and verify whether the resources of the terminal device it manages can complete the execution of the instruction.

[0082] When the corresponding terminal device has no relevant resources to execute the instruction, the target migration node can initiate an equivalent replacement strategy.

[0083] Then, the target migration node may dynamically adjust the priority of processing the pending task package based on the context information; and then process the pending task package based on the priority.

[0084] Through context-aware intelligent adaptation, task packages can be "plug-and-play" in new environments, reducing the need for manual intervention; by launching alternative solutions through dynamic remapping, the impact of device heterogeneity can be reduced.

[0085] In the embodiment of the present application, a three-level architecture is implemented through long-term strategy planning at the cloud level, regional collaboration at the edge level, and rapid response at the terminal level, combined with a dynamic load balancing algorithm to achieve task migration; wherein the first edge node locally processes regional collaboration tasks, reduces the frequency of cloud communications, and reduces end-to-end latency, and then combines federated learning to achieve data sharing between edge nodes, avoiding redundant data transmission, thereby effectively improving the response speed of the scheduling system and better meeting the implementation scheduling needs of virtual power plants.

[0086] The above describes the method part provided by the embodiment of the present application. The following describes the system part provided by the embodiment of the present application.

[0087] See also Figure 2 , Figure 2 A schematic diagram of the structure of a cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system 1 includes a cloud server 10, an edge node 20 and a terminal device 30. The edge node 20 includes a first edge node 21 and a second edge node 22. The first edge node 21 and the second edge node 22 are different edge nodes.

[0088] It is understandable that in Figure 2 In the specific example, the number of the cloud server 10, the first edge node 21 and the second edge node 22 is 1, and the terminal device 30 includes terminal devices a to f; in actual application, the scheduling system 1 can be any system including Figure 2The present embodiment does not limit the specific structure of the scheduling system 1. The number of cloud servers 10, first edge nodes 21 and second edge nodes 22 in the scheduling system 1 can be greater, and the number of terminal devices 30 can be greater or less, and the present embodiment does not specifically limit this.

[0089] The cloud server 10 is used to generate an initial scheduling strategy based on historical load data and environmental data; wherein the environmental data is environmental data that affects the output of new energy stations aggregated by the virtual power plant;

[0090] The first edge node 21 is used to obtain status data of the terminal device managed by the first edge node 21 itself, and generate a task package through a federated learning framework based on the status data and the initial scheduling policy;

[0091] The terminal device 30 is configured to execute the control instruction in the task package and, when detecting an execution abnormality of the control instruction, send abnormality information to the first edge node 21;

[0092] The first edge node 21 is also used to monitor the network status and the computing resource load of the first edge node 21 itself, and when the computing resource load is greater than a preset first threshold, based on the network status and the abnormal information, migrate the task package to be processed to the target migration node, which includes the cloud server 30 or the second edge node 22.

[0093] Preferably, the environmental data includes wind speed data and irradiance data; the cloud server 30 is specifically used to input the historical power load data into the long short-term memory (LSTM) model to obtain the predicted load data output by the LSTM model; based on the wind speed data and the irradiance data, the output prediction curve of the new energy station is generated; based on the predicted load data, the output prediction curve and the output parameters of the traditional power generation units aggregated by the virtual power plant, the initial scheduling strategy is generated.

[0094] Preferably, the first edge node 21 is specifically used to input the state data and the initial scheduling strategy into the regional optimization model, and obtain the regional optimization model to output the task package; wherein each edge node is respectively deployed with the regional optimization model, and the regional optimization model and the federated learning model are federated learned through gradient aggregation, the weight of the federated learning model is obtained based on the initial scheduling strategy, and the regional optimization model constructs a loss function based on the weight of the federated learning model.

[0095] Preferably, the first edge node 21 is specifically configured to calculate a migration score based on the network status, the computing resource load of the second edge node, and the priority tag of the task package to be processed; and select a target migration node for the task package to be processed based on the migration score.

[0096] Preferably, the first edge node 21 is specifically configured to obtain the computational complexity of the task package to be processed; and when the computational complexity is higher than a preset second threshold, migrate the task package to be processed to the cloud server.

[0097] Preferably, after the first edge node 21 obtains the status data of the terminal device managed by the edge node itself, and generates a task package through the federated learning framework based on the status data and the initial scheduling strategy, the first edge node 21 is also used to generate a global task mapping table in the blockchain based on the task package; after the first edge node 21 migrates the task package to be processed to the target migration node based on the network status and the abnormal information, the first edge node 21 is also used to update the global task mapping table based on the task package to be processed and the target migration node; the target migration node is used to obtain the context information of the task package to be processed based on the global task mapping table, and process the task package to be processed based on the context information.

[0098] Preferably, the task package is a binary task package, which also includes a device control instruction set and its corresponding execution time window.

[0099] The cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration provided in the embodiment of the present application can be understood by referring to the corresponding content of the aforementioned method embodiment part, and will not be repeated here.

[0100] like Figure 3 As shown, Figure 3 A possible logical structure diagram of a computing device provided in an embodiment of the present application. The computing device 300 includes: a processor 301, a communication interface 302, a memory 303, and a bus 304. The processor 301, the communication interface 302, and the memory 303 are interconnected via the bus 304. In the embodiment of the present application, the processor 301 is used to control and manage the actions of the computing device 300. For example, the processor 301 is used to execute Figure 1 The steps in the embodiments and / or other processes for the technology described herein. The communication interface 302 is used to support the computing device 300 to communicate. The memory 303 is used to store program codes and data of the computing device 300.

[0101] Among them, the processor 301 can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It can implement or execute the various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and so on. The bus 304 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0102] In another embodiment of the present application, a computer-readable storage medium is further provided, wherein the computer-readable storage medium includes instructions, which, when executed on a computer, causes the computer to execute the above-mentioned Figure 1 The method described in the embodiment.

[0103] Those skilled in the art will appreciate that the units of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0104] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0105] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0106] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0107] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0108] If the functions are implemented in the form of 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 the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk or optical disk, and other media that can store program codes.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A cloud-edge-device collaborative virtual power plant scheduling method based on dynamic task migration, characterized in that: A cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration is applied, the system including a cloud server, an edge node, and a terminal device, the edge node including a first edge node and a second edge node, the first edge node and the second edge node being different edge nodes; the method including: The cloud server generates an initial scheduling strategy based on historical load data and environmental data; wherein the environmental data is environmental data that affects the output of new energy stations aggregated by the virtual power plant; The first edge node obtains status data of a terminal device managed by the first edge node itself, and generates a task package through a federated learning framework based on the status data and the initial scheduling policy; The terminal device executes the control instruction in the task package, and sends abnormal information to the first edge node when detecting an execution abnormality of the control instruction; The first edge node monitors the network status and the computing resource load of the first edge node itself, and when the computing resource load is greater than a preset first threshold, based on the network status and the abnormal information, migrates the task package to be processed to the target migration node, and the target migration node includes the cloud server or the second edge node.

2. The method according to claim 1, characterized in that The environmental data includes wind speed data and irradiance data; the cloud server generates an initial scheduling strategy based on the historical load data and environmental data, including: The cloud server inputs the historical power load data into a long short-term memory (LSTM) model to obtain predicted load data output by the LSTM model; The cloud server generates an output prediction curve of the new energy station based on the wind speed data and the irradiance data; The initial scheduling strategy is generated based on the predicted load data, the output prediction curve and the output parameters of the traditional power generation units aggregated by the virtual power plant.

3. The method according to claim 1 or 2, characterized in that The first edge node obtains status data of a terminal device managed by the first edge node itself, and generates a task package through a federated learning framework based on the status data and the initial scheduling policy, including: The first edge node inputs the state data and the initial scheduling strategy into a regional optimization model to obtain the task package output by the regional optimization model; Among them, each edge node is respectively deployed with the regional optimization model, the regional optimization model and the preset federated learning model perform federated learning through gradient aggregation, the weight of the federated learning model is obtained based on the initial scheduling strategy, and the regional optimization model constructs a loss function based on the weight of the federated learning model.

4. The method according to claim 1 or 2, characterized in that The migrating the pending task package to a target migration node based on the network status and the abnormal information includes: Calculating a migration score based on the network status, the computing resource load of the second edge node, and the priority tag in the task package to be processed; wherein the priority tag is used to indicate the processing priority of the task package to be processed; Based on the migration score, a target migration node of the task package to be processed is selected.

5. The method according to claim 1 or 2, characterized in that The migrating the pending task package to a target migration node based on the network status and the abnormal information includes: Obtaining the computational complexity of the task package to be processed; When the computational complexity is higher than a preset second threshold, the to-be-processed task package is migrated to the cloud server.

6. The method according to claim 1 or 2, characterized in that After the first edge node obtains status data of a terminal device managed by the first edge node itself and generates a task package through a federated learning framework based on the status data and the initial scheduling policy, the method further includes: The first edge node generates a global task mapping table in the blockchain based on the task package; After migrating the pending task package to the target migration node based on the network status and the abnormal information, the method further includes: The first edge node updates the global task mapping table based on the task package to be processed and the target migration node; The target migration node obtains context information of the to-be-processed task package based on the global task mapping table, and processes the to-be-processed task package based on the context information.

7. The method according to claim 1 or 2, characterized in that The task package is a binary task package, and the task package also includes a device control instruction set and its corresponding execution time window, as well as a priority tag.

8. A cloud-edge collaborative virtual power plant scheduling system based on dynamic task migration, characterized in that: The method applied to any one of claims 1 to 7, wherein the system includes a cloud server, an edge node, and a terminal device, the edge node includes a first edge node and a second edge node, and the first edge node and the second edge node are different edge nodes; The cloud server generates an initial scheduling strategy based on historical load data and environmental data; wherein the environmental data is environmental data that affects the output of new energy stations aggregated by the virtual power plant; The first edge node obtains status data of a terminal device managed by the first edge node itself, and generates a task package through a federated learning framework based on the status data and the initial scheduling policy; The terminal device executes the control instruction in the task package, and sends abnormal information to the first edge node when detecting an execution abnormality of the control instruction; The first edge node monitors the network status and the computing resource load of the first edge node itself, and when the computing resource load is greater than a preset first threshold, based on the network status and the abnormal information, migrates the task package to be processed to the target migration node, and the target migration node includes the cloud server or the second edge node.

9. The system according to claim 8, characterized in that The first edge node is further configured to generate a global task mapping table in the blockchain based on the task package; The first edge node is further configured to update the global task mapping table based on the task package to be processed and the target migration node; The target migration node is configured to obtain context information of the to-be-processed task package based on the global task mapping table, and process the to-be-processed task package based on the context information.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

Images