Regional energy distributed control method based on edge calculation

Abstract

The invention discloses a regional energy distributed control method based on edge computing, which relates to the technical field of distributed control, and comprises the following steps: collecting regional energy data of edge nodes, and establishing an energy distribution model of the edge nodes; discrete multi-target cascade optimization scheduling is carried out on the edge node energy distribution model, and hierarchical optimization is decoupled; and generating dual-order optimization scheduling based on the column sum constraint, wherein the dual-order optimization scheduling comprises iterative operation of a first-order scheduling algorithm and a second-order scheduling algorithm. Coordinating with other sub-models through electric energy interaction to ensure a global optimization target; according to the method, the challenges of numerous devices and users and large calculation amount in the regional energy internet are effectively handled, scheduling calculation is simplified, and local independence of each sub-model can be kept while global optimization is guaranteed. And a double-order optimization scheduling strategy is adopted, so that the balance of global optimization and local optimization is realized, uncertain factors in the system can be dealt with, and the stability and optimization effect are further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed control, and particularly to a regional energy distributed control method based on edge computing. Background Art

[0002] With the continuous growth of global energy demand and the gradual emergence of renewable energy sources (such as solar energy, wind energy, etc.) as the mainstream energy sources, the traditional centralized energy management model is facing more and more challenges. Especially in the context of the increasing number of distributed energy systems (such as household photovoltaic power generation, wind power communities, energy storage devices, etc.), how to efficiently and intelligently manage and dispatch these distributed energy resources has become the core problem faced by modern energy management systems. Against this background, edge computing, as a new computing architecture, has begun to be widely used in the energy field. Edge computing can not only shorten the data transmission path, reduce latency, but also perform real-time data processing and decision-making locally, thereby improving the response speed and computing efficiency of the system. Combining edge computing, the regional energy distributed control method has gradually become an effective solution, which can intelligently dispatch and manage distributed energy locally, reduce dependence on the central server, and enhance the flexibility and scalability of the system.

[0003] Currently, the Chinese invention patent with the application number CN202411711099.6 discloses a distributed energy regional autonomous regulation and control system and method with multi-level collaborative control, which improves the real-time performance and response speed of the system; enhances the power grid's consumption capacity for clean energy and maximizes the utilization of new energy; reduces the data processing burden, and protects user privacy through edge computing; has the ability of rapid emergency response, and can quickly adjust the load distribution in case of power grid failures or emergencies to ensure the safety and stability of power grid operation. However, in the prior art, due to the large number of devices and users, the scheduling calculation amount is huge, and it is difficult to consider local autonomy while efficiently solving globally, and it is unable to efficiently cope with uncertain factors, so the optimization effect is not obvious. Summary of the Invention

[0004] The technical problem solved by the present invention is that in the prior art, due to the large number of devices and users, the scheduling calculation amount is huge, and it is difficult to consider local autonomy while efficiently solving globally, and it is unable to efficiently cope with uncertain factors, so the optimization effect is not obvious.

[0005] To solve the above technical problem, the present invention provides the following technical solution: A regional energy distributed control method based on edge computing, comprising the following steps: Step S1: Collect regional energy data of edge nodes and establish an energy distribution model of edge nodes; Step S2: Perform discrete multi-objective cascaded optimal scheduling on the edge node energy distribution model and decouple hierarchical optimization; Step S3: Generate a two-stage optimal scheduling based on column sum constraints, and the two-stage optimal scheduling includes iterative operation of a first-stage scheduling algorithm and a second-stage scheduling algorithm.

[0006] Preferably, collecting edge node area energy data includes: Real-time collecting energy data of distributed energy devices through edge computing nodes deployed in the area, where the energy data includes operating status, energy consumption data, and environmental parameters, to form a multi-dimensional energy data stream; The edge computing nodes include intelligent gateways and sensor networks, the distributed energy devices include photovoltaic, energy storage, and gas turbines, the operating status includes start, shutdown, and stagnation, the energy consumption data includes temperature, humidity, pressure, current, voltage, power, energy output, and energy consumption, and the environmental parameters include temperature, humidity, air pressure, and light intensity; The multi-dimensional energy data stream is aggregated to the edge computing nodes through the intelligent gateway, and is marked, synchronized, and preprocessed in the edge computing nodes according to time series, and the preprocessing includes denoising and normalization.

[0007] Preferably, establishing an edge node energy distribution model based on the energy data includes: Integrating the geographical location partition and functional hierarchy of edge nodes through Geographic Information System (GIS), and hierarchically describing the multi-energy flow coupling characteristics in combination with the geographical location partition and functional hierarchy of edge nodes. The geographical location partition includes the city center, suburbs, and edge areas, the functional hierarchy includes core nodes and terminal nodes, and the multi-energy flow coupling characteristics include the first temperature, the second temperature, electricity, and gas; The hierarchical description includes: Dividing the current edge node energy distribution model into several levels; Using virtual power plant technology to aggregate distributed energy resources into a virtual entity to form a schedulable virtual power plant; Based on the integrated function of the virtual power plant, perform intelligent scheduling on the edge node energy distribution model through an optimization algorithm. The goals of intelligent scheduling include maximizing energy utilization efficiency, minimizing scheduling cost, meeting the needs of each node, and optimizing energy flow direction. Deploy the optimized edge node energy distribution model to the actual regional energy Internet, and the regional energy Internet includes all edge nodes, core nodes, energy stations, and energy storage devices.

[0008] Preferably, the step S2 includes: The optimization objectives for setting the edge node energy distribution model include reducing the energy loss of the edge node energy distribution model, improving the energy utilization rate of the edge node energy distribution model, and reducing the cost of the edge node energy distribution model; The discretization process of the scheduling decision includes: discretizing all energy data for each time period by category. For each category of energy data, continuous data is divided into several discrete intervals according to a predetermined interval. The categories include operating status, energy consumption data, and environmental parameters.

[0009] Preferably, the decoupled hierarchical optimization of the edge node energy distribution model includes: Decoupling the edge node energy distribution model of the region into several virtual energy distribution sub-models, and uniformly managing each virtual energy distribution sub-model. The unified management includes managing the power interaction between the virtual energy distribution sub-models through a schedulable virtual power plant. The internal scheduling planning of each virtual energy distribution sub-model is autonomously optimized by the local controller, and the target cascading method is used to coordinate the edge node energy distribution model and the virtual energy distribution sub-model through penalty functions and consistency parameters.

[0010] Preferably, the coordination process specifically includes: Regarding the edge node energy distribution model as the top-level management node to overall balance the global energy supply and demand and interactively optimize the multi-dimensional energy data stream; Regarding the virtual energy distribution sub-model as the lower-level execution node, adjusting the scheduling strategy according to the local device constraints and feeding back the interactive power to the upper-level management node. The local device constraints include the preset charge and discharge limits of energy storage devices and the dynamic characteristics of the regional energy Internet. The dynamic characteristics include the dynamic feature quantities obtained by neural network feature extraction of the preprocessed multi-dimensional energy data stream.

[0011] Preferably, the step S3 includes: The first-order scheduling algorithm includes initializing the parameters of the optimized edge node energy distribution model. The parameters include the first penalty function factor, the second penalty function factor, the first convergence criterion, the second convergence criterion, the iteration counter k = 1, the first penalty function iteration multiplier, and the second penalty function iteration multiplier; The first penalty function factor is the penalty function factor for controlling the error of power interaction information; The second penalty function factor is the penalty function factor for controlling the scheduling error of the virtual energy distribution sub-model; The first convergence criterion is the convergence error threshold for power interaction between virtual energy distribution sub-models; The second convergence criterion is the convergence error threshold for power interaction between the virtual energy distribution sub-model and the edge node energy distribution model; The first penalty function iteration multiplier is used to update the multiplier factor of the first penalty function factor; The second penalty function iteration multiplier is used to update the multiplier factor of the second penalty function factor.

[0012] By iteratively updating the first interaction power variable and the second interaction power variable until the convergence condition is satisfied, the convergence condition is iterating to a preset iteration threshold.

[0013] Preferably, the first-order scheduling algorithm includes: Step S301: Obtain the first electrical energy interaction information and the second electrical energy interaction information between each virtual energy distribution sub-model respectively; The first electrical energy interaction information is the unit quantity of electrical energy interaction from the edge node energy distribution model to the virtual energy distribution sub-model; The second electrical energy interaction information is the unit quantity of electrical energy interaction from the virtual energy distribution sub-model to the edge node energy distribution model; Step S302: Solve the optimization problems of all virtual energy distribution sub-models based on the optimization algorithm to obtain the first electrical energy scheduling and the second electrical energy scheduling between each virtual energy distribution sub-model; The first electrical energy scheduling is the electrical energy output from the virtual energy distribution sub-model to the edge node energy distribution model; The second electrical energy scheduling is the electrical energy output from the virtual energy distribution sub-model to other virtual energy distribution sub-models; Step S303: Update the penalty function factor according to the optimization result, and the update process includes: The first penalty function factor + 1 = the first penalty function iteration multiplier × the first penalty function factor; The second penalty function factor + 1 = the second penalty function iteration multiplier × the second penalty function factor; Step S304: Update the iteration counter k = k + 1, and return to Step S302 until the convergence criterion is satisfied; The convergence criterion is that when the electrical energy exchange error between the virtual energy distribution sub-models satisfies the convergence condition, the convergence condition includes: |The first electrical energy interaction information - the first electrical energy scheduling| < the first convergence criterion; |The second electrical energy interaction information - the second electrical energy scheduling| < the second convergence criterion; Step S305: When the convergence criterion is satisfied simultaneously, end the iteration process and output the optimization result. The optimization result includes the electrical energy scheduling plan and the electrical energy exchange strategy of each virtual energy distribution sub-model.

[0014] Preferably, the second-order scheduling algorithm includes: Adopt the column constraint generation algorithm CCG, and correct the decision deviation in the first-order scheduling algorithm through interactive iteration with the first-order scheduling algorithm.

[0015] Preferably, the second-order scheduling algorithm specifically includes: Step S311: Initialize the lower bound of the sub-problem to negative infinity, the upper bound of the sub-problem to positive infinity, set the iteration counter r = 1 of the sub-problem, and set the third convergence criterion; Step S312: Set the convergence condition as: the upper bound of the sub-problem - the lower bound of the sub-problem ≥ the third convergence criterion; Step S313: Iteratively optimize the optimization result of the first-order scheduling algorithm, update the dual variables and objective constraints in the calculation process of the first-order scheduling algorithm, and update the upper and lower bounds of the sub-problem; Step S314: Update r = r + 1; Step S315: Take the currently updated sub-problem as the optimization object for optimization, recursively optimize the upper and lower bounds of the sub-problem of the sub-problem, and solve to obtain the upper and lower bounds that meet the third convergence criterion; Step S316: Stop the iteration when the convergence condition is met; All data volumes of the regional energy distributed control method based on edge computing are within a unit time.

[0016] Advantages of the present invention: This method analyzes and optimizes by decoupling the edge node energy distribution model into multiple small integrated energy sub-models, and coordinates with other sub-models through power interaction to ensure the global optimization goal; through the regional energy distributed control method based on edge computing, it can effectively cope with the challenges of numerous devices and users and large computational volume in the regional energy Internet. By decoupling the edge node energy distribution model from the virtual energy distribution sub-model and using discrete multi-objective cascaded optimization scheduling, it not only simplifies the scheduling calculation but also maintains the local autonomy of each sub-model while ensuring global optimization. Adopting a two-stage optimization scheduling strategy, the balance between global optimization and local optimization is achieved through the collaborative work of the first-order scheduling algorithm and the second-order scheduling algorithm, which can cope with uncertain factors in the system and further improve the stability and optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the basic process of a regional energy distributed control method based on edge computing provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention is provided in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments.

[0019] Refer to Figure 1, which is an embodiment of the present invention, provides a distributed control method for regional energy based on edge computing, including the following steps: Step S1: Collect the regional energy data of edge nodes and establish an energy distribution model for edge nodes; Step S2: Perform discrete multi-objective cascaded optimization scheduling on the energy distribution model of edge nodes and decouple hierarchical optimization; Step S3: Generate a two-stage optimization scheduling based on column sum constraints, and the two-stage optimization scheduling includes the iterative operation of a first-stage scheduling algorithm and a second-stage scheduling algorithm.

[0020] Collecting the regional energy data of edge nodes includes: Real-time collect the energy data of distributed energy devices through edge computing nodes deployed in the region. The energy data includes operating status, energy consumption data, and environmental parameters, forming a multi-dimensional energy data stream; The edge computing nodes include intelligent gateways and sensor networks. The distributed energy devices include photovoltaic, energy storage, and gas turbines. The operating status includes start-up, shutdown, and stagnation. The energy consumption data includes temperature, humidity, pressure, current, voltage, power, energy output, and energy consumption. The environmental parameters include temperature, humidity, air pressure, and light intensity; The multi-dimensional energy data stream converges to the edge computing nodes through the intelligent gateway and is marked, synchronized, and preprocessed in the edge computing nodes according to the time series. The preprocessing includes denoising and normalization.

[0021] Establishing an energy distribution model for edge nodes based on energy data includes: Integrate the geographical location partition and functional hierarchy of edge nodes through the Geographic Information System (GIS), and describe the multi-energy flow coupling characteristics in layers in combination with the geographical location partition and functional hierarchy of edge nodes. The geographical location partition includes the city center, suburbs, and edge areas. The functional hierarchy includes core nodes and terminal nodes. The core nodes are used for large-scale coordination and control of the overall operation of the main edge computing node network. The terminal nodes are used for local scheduling. The multi-energy flow coupling characteristics include the first temperature, the second temperature, electricity, and gas; The layered description includes: Divide the current energy distribution model of edge nodes into several layers; for example, the first temperature (heating) and the second temperature (cooling) are important parameters of the thermal system. Especially in the city center area, the management and scheduling of the temperature flow require centralized coordination of the core nodes to schedule the coupling between the heat energy and cold energy devices to improve the energy efficiency of the system; The core nodes need to coordinate the coupling between electricity and gas. For example, when the natural gas supply is insufficient, the energy storage device and the power system can be used to provide supplements. The terminal nodes balance the demand for electricity and gas by controlling the start-up or shutdown of local devices; Using virtual power plant technology to aggregate distributed energy resources into a virtual entity, forming a dispatchable virtual power plant. By aggregating multiple distributed resources into a unified control unit, the dispatching task is simplified. The virtual power plant can balance demand and supply by integrating the multi-energy flow coupling characteristics of different types, reduce energy waste, and improve the overall system stability.

[0022] Based on the integrated function of the virtual power plant, the intelligent dispatching of the energy distribution model of edge nodes is carried out through an optimization algorithm. The goals of intelligent dispatching include maximizing energy utilization efficiency, minimizing dispatching costs, meeting the demands of each node, and optimizing the energy flow direction. The optimized energy distribution model of edge nodes is deployed into the actual regional energy Internet, which includes all edge nodes, core nodes, energy stations, and energy storage devices.

[0023] The regional energy Internet dispatching system optimized based on the collected data and virtual power plant technology can effectively improve the utilization efficiency of energy resources, reduce the system operation cost, and ensure the stability and reliability of energy supply.

[0024] Step S2 includes: This method realizes the decoupling of the power coupling relationship between the global energy dispatching platform and each sub-model through virtualization technology. The global energy dispatching platform does not need to obtain the device dispatching information inside each sub-model, but coordinates by regarding each sub-model as a whole. Each virtual energy distribution sub-model only focuses on the power interaction with the global energy dispatching platform and does not pay attention to the internal dispatching details of other sub-models.

[0025] The set optimization goals of the energy distribution model of edge nodes include reducing the energy loss of the energy distribution model of edge nodes, improving the energy utilization rate of the energy distribution model of edge nodes, and reducing the cost of the energy distribution model of edge nodes. Due to the discreteness of equipment and energy forms, the discretization processing of dispatching decisions includes: discretizing all energy data in each time period by category. For each category of energy data, the continuous data is divided into several discrete intervals according to a predetermined interval. The categories include operating status, energy consumption data, and environmental parameters.

[0026] The decoupling hierarchical optimization of the energy distribution model of edge nodes includes: Decouple the edge node energy distribution model of the region into several virtual energy distribution sub-models, and uniformly manage each virtual energy distribution sub-model. The unified management includes managing the electrical energy interaction between the virtual energy distribution sub-models through a dispatchable virtual power plant. The internal scheduling planning of each virtual energy distribution sub-model is autonomously optimized by the local controller. The target cascading method is used to coordinate the edge node energy distribution model and the virtual energy distribution sub-models through penalty functions and consistency parameters to achieve the balance between global optimization and local autonomy.

[0027] The coordination process specifically includes: Take the edge node energy distribution model as the uppermost management node to overall balance the global energy supply and demand, and interactively optimize the multi-dimensional energy data stream; Take the virtual energy distribution sub-models as the lower-layer execution nodes, adjust the scheduling strategy according to the local device constraints, and feedback the interactive power to the upper-layer management node. The local device constraints include the preset charge and discharge limits of the energy storage devices and the dynamic characteristics of the regional energy Internet. The dynamic characteristics include the dynamic feature quantities obtained by neural network feature extraction of the preprocessed multi-dimensional energy data stream.

[0028] Step S3 includes: The first-order scheduling algorithm includes initializing the parameters of the optimized edge node energy distribution model. The parameters include the first penalty function factor, the second penalty function factor, the first convergence criterion, the second convergence criterion, the iteration counter k = 1, the first penalty function iteration multiplier, and the second penalty function iteration multiplier; The first penalty function factor is the penalty function factor for controlling the error of the electrical energy interaction information; The second penalty function factor is the penalty function factor for controlling the scheduling error of the virtual energy distribution sub-model; The first convergence criterion is the convergence error threshold for the electrical energy interaction between the virtual energy distribution sub-models; The second convergence criterion is the convergence error threshold for the electrical energy interaction between the virtual energy distribution sub-model and the edge node energy distribution model; The first penalty function iteration multiplier is the multiplier factor for updating the first penalty function factor; The second penalty function iteration multiplier is the multiplier factor for updating the second penalty function factor.

[0029] Iteratively update the first interactive power variable and the second interactive power variable until the convergence condition is met. The convergence condition is to iterate to the preset iteration threshold.

[0030] The first-order scheduling algorithm ensures the robustness of the system under uncertain scenarios.

[0031] The first-order scheduling algorithm includes: Step S301: Obtain the first electric energy interaction information and the second electric energy interaction information between each virtual energy distribution sub-model respectively; The first electric energy interaction information is the unit quantity of electric energy interaction from the edge node energy distribution model to the virtual energy distribution sub-model; The second electric energy interaction information is the unit quantity of electric energy interaction from the virtual energy distribution sub-model to the edge node energy distribution model; Step S302: Solve the optimization problems of all virtual energy distribution sub-models based on the optimization algorithm to obtain the first electric energy scheduling and the second electric energy scheduling between each virtual energy distribution sub-model; The first electric energy scheduling is the electric energy output by the virtual energy distribution sub-model to the edge node energy distribution model; The second electric energy scheduling is the electric energy output by the virtual energy distribution sub-model to other virtual energy distribution sub-models; Step S303: Update the penalty function factor according to the optimization result, and the update process includes: The first penalty function factor + 1 = the first penalty function iteration multiplier × the first penalty function factor; The second penalty function factor + 1 = the second penalty function iteration multiplier × the second penalty function factor; The update process of these two factors ensures that the system controls the error more and more strictly in each round of iteration, helping the optimization process converge faster; Step S304: Update the iteration counter k = k + 1, and return to Step S302 until the convergence criterion is met; The convergence criterion is that when the electric energy exchange error between the virtual energy distribution sub-models meets the convergence condition, the convergence condition includes: |The first electric energy interaction information - the first electric energy scheduling| < the first convergence criterion; |The second electric energy interaction information - the second electric energy scheduling| < the second convergence criterion; Step S305: When the convergence criterion is met simultaneously, end the iteration process and output the optimization result. The optimization result includes the electric energy scheduling plan and the electric energy exchange strategy of each virtual energy distribution sub-model.

[0032] By introducing the robust optimization method, deal with the uncertainty sources in the system and ensure that each sub-model can still operate efficiently under the most unfavorable conditions.

[0033] Through the way of column and constraint generation, gradually optimize the constraint conditions in the scheduling problem to ensure the coordination between the global and local of the system.

[0034] The optimization scheduling process gradually optimizes the electric energy distribution of the energy system through an iterative process. This process ensures the effective distribution of electric energy, optimizes the energy utilization efficiency and meets the constraint conditions of the system.

[0035] The second-order scheduling algorithm includes: Adopt the column constraint generation algorithm CCG, and correct the decision deviation in the first-order scheduling algorithm through interactive iteration with the first-order scheduling algorithm.

[0036] For the worst-case scenarios monitored in real time, such as extreme weather and sudden load changes, dynamically adjust the energy distribution strategy, optimize the real-time scheduling plan under the multi-energy flow coupling, and ensure economy and security.

[0037] Specifically, the second-order scheduling algorithm includes: Step S311: Initialize the lower bound of the sub-problem to negative infinity, the upper bound of the sub-problem to positive infinity, set the iteration counter r of the sub-problem to 1, and set the third convergence criterion to judge whether the algorithm converges; Step S312: Set the convergence condition as: the upper bound of the sub-problem - the lower bound of the sub-problem ≥ the third convergence criterion; Step S313: Iteratively optimize the optimization result of the first-order scheduling algorithm, update the dual variables and objective constraints in the calculation process of the first-order scheduling algorithm, and update the upper and lower bounds of the sub-problem; Step S314: Update r = r + 1, and the new solution will be used as the input value for the next round; Step S315: Optimize the currently updated sub-problem as the optimization object, recursively optimize the upper and lower bounds of the sub-problems of the sub-problem, and solve to obtain the upper and lower bounds that meet the third convergence criterion; This process ensures the dynamic tracking and adjustment of all constraints, guarantees that the quality of the solution is improved as much as possible in each iteration, and the update of the constraint conditions ensures that the sub-problem can be improved within the feasible solution space after each optimization.

[0038] Step S316: Stop the iteration when the convergence condition is met; This algorithm continuously updates the upper and lower bounds and constraint conditions of the sub-problem through an alternating process of solving a master problem and a sub-problem. In each iteration, the algorithm updates the current solution and introduces new dual variables and constraints, and finally searches for the global optimal solution. The core idea of the algorithm is to optimize each sub-problem through iterative solution and update, and gradually converge to the optimal solution.

[0039] All data volumes of the regional energy distributed control method based on edge computing are within a unit time.

[0040] This method analyzes and optimizes the edge node energy distribution model by decoupling it into multiple small-scale integrated energy sub-models, and coordinates with other sub-models through electrical energy interaction to ensure the global optimization goal. Through the regional energy distributed control method based on edge computing, it can effectively address the challenges of numerous devices and users and large computational volume in the regional energy Internet. By decoupling the edge node energy distribution model from the virtual energy distribution sub-model and using discrete multi-objective cascaded optimization scheduling, it not only simplifies the scheduling calculation but also maintains the local autonomy of each sub-model while ensuring global optimization. The dual-stage optimization scheduling strategy is adopted to achieve the balance between global optimization and local optimization through the collaborative work of the first-stage scheduling algorithm and the second-stage scheduling algorithm, which can cope with the uncertain factors in the system and further improve the stability and optimization effect.

[0041] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk. These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the functions specified in one process Figure 1 a process or multiple processes and / or boxes Figure 1 specified in a box or multiple boxes.

[0042] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A regional energy distributed control method based on edge computing, characterized in that Including the following steps: Step S1: Collect energy data in the edge node area and establish an edge node energy distribution model; Step S2: Conduct discrete multi-objective cascaded optimal scheduling on the edge node energy distribution model and decouple hierarchical optimization; Step S3: Generate a two-stage optimal scheduling based on column sum constraints, and the two-stage optimal scheduling includes iterative operation of a first-stage scheduling algorithm and a second-stage scheduling algorithm.

2. The regional energy distributed control method based on edge computing according to claim 1, characterized in that Collecting energy data in the edge node area includes: Real-time collect energy data of distributed energy devices through edge computing nodes deployed in the area, and the energy data includes operating status, energy consumption data and environmental parameters, forming a multi-dimensional energy data stream; The edge computing nodes include intelligent gateways and sensor networks, the distributed energy devices include photovoltaic, energy storage and gas turbines, the operating status includes start, shutdown and stagnation, the energy consumption data includes temperature, humidity, pressure, current, voltage, power, energy output and energy consumption, and the environmental parameters include temperature, humidity, air pressure and light intensity; The multi-dimensional energy data stream converges to the edge computing node through the intelligent gateway, and is marked, synchronized and pre-processed in the edge computing node according to the time series, and the pre-processing includes denoising and normalization.

3. The regional energy distributed control method based on edge computing according to claim 2, wherein, Establishing an edge node energy distribution model based on the energy data includes: Integrate the geographical location partition and functional hierarchy of the edge node through the Geographic Information System (GIS), and describe the multi-energy flow coupling characteristics in layers in combination with the geographical location partition and functional hierarchy of the edge node. The geographical location partition includes the city center, suburbs and edge areas, the functional hierarchy includes core nodes and terminal nodes, and the multi-energy flow coupling characteristics include the first temperature, the second temperature, electricity and gas; The layer description includes: Divide the current edge node energy distribution model into several levels; Use virtual power plant technology to aggregate distributed energy resources into a virtual entity to form a schedulable virtual power plant; Based on the integrated function of the virtual power plant, conduct intelligent scheduling on the edge node energy distribution model through an optimization algorithm. The goals of intelligent scheduling include maximizing energy utilization efficiency, minimizing scheduling cost, meeting the needs of each node and optimizing the energy flow direction, and deploy the optimized edge node energy distribution model to the actual regional energy Internet, which includes all edge nodes, core nodes, energy stations and energy storage devices.

4. The edge computing-based regional energy distributed control method according to claim 3, wherein The said Step S2 includes: Set the optimization goals of the edge node energy distribution model, including reducing the energy loss of the edge node energy distribution model, improving the energy utilization rate of the edge node energy distribution model and reducing the cost of the edge node energy distribution model; The discretization process of the scheduling decision includes: discretize all energy data in each period according to categories respectively. For each category of energy data, divide the continuous data into several discrete intervals according to a predetermined interval, and the categories include operating status, energy consumption data and environmental parameters.

5. The edge-computing-based regional energy distributed control method according to claim 4, wherein The decoupled hierarchical optimization of the edge node energy distribution model includes: Decouple the edge node energy distribution model of the area into several virtual energy distribution sub-models, and uniformly manage each virtual energy distribution sub-model. The unified management includes managing the electrical energy interaction between the virtual energy distribution sub-models through a schedulable virtual power plant. The internal scheduling planning of each virtual energy distribution sub-model is autonomously optimized by a local controller. The objective cascading method is used to coordinate the edge node energy distribution model and the virtual energy distribution sub-models through penalty functions and consistency parameters.

6. The regional energy distributed control method based on edge computing according to claim 5, characterized in that The coordination process specifically includes: Regarding the edge node energy distribution model as the top-level management node, overall planning the global energy supply and demand balance, and interactively optimizing the multi-dimensional energy data stream; Regarding the virtual energy distribution sub-model as the lower-level execution node, adjusting the scheduling strategy according to local device constraints, and feeding back the interactive power to the upper-level management node. The local device constraints include the preset charge and discharge limits of energy storage devices and the dynamic characteristics of the regional energy Internet. The dynamic characteristics include the dynamic feature quantities obtained by performing neural network feature extraction on the preprocessed multi-dimensional energy data stream.

7. The regional energy distributed control method based on edge computing according to claim 6, wherein The step S3 includes: The first-order scheduling algorithm includes initializing the parameters of the optimized edge node energy distribution model. The parameters include the first penalty function factor, the second penalty function factor, the first convergence criterion, the second convergence criterion, the iteration counter k = 1, the first penalty function iteration multiplier, and the second penalty function iteration multiplier; The first penalty function factor is the penalty function factor for controlling the error of electrical energy interaction information; The second penalty function factor is the penalty function factor for controlling the scheduling error of the virtual energy distribution sub-model; The first convergence criterion is the convergence error threshold for the electrical energy interaction between virtual energy distribution sub-models; The second convergence criterion is the convergence error threshold for the electrical energy interaction between the virtual energy distribution sub-model and the edge node energy distribution model; The first penalty function iteration multiplier is the multiplier factor for updating the first penalty function factor; The second penalty function iteration multiplier is the multiplier factor for updating the second penalty function factor; Iteratively update the first interaction power variable and the second interaction power variable until the convergence condition is met. The convergence condition is to iterate to the preset iteration threshold.

8. The edge-computing-based regional energy distributed control method according to claim 7, wherein The first-order scheduling algorithm includes: Step S301: Obtain the first electrical energy interaction information and the second electrical energy interaction information between each virtual energy distribution sub-model respectively; The first electrical energy interaction information is the unit quantity of electrical energy interaction from the edge node energy distribution model to the virtual energy distribution sub-model; The second electrical energy interaction information is the unit quantity of electrical energy interaction from the virtual energy distribution sub-model to the edge node energy distribution model; Step S302: Solve the optimization problems of all virtual energy distribution sub-models based on the optimization algorithm to obtain the first electrical energy scheduling and the second electrical energy scheduling between each virtual energy distribution sub-model; The first electrical energy scheduling is the electrical energy output by the virtual energy distribution sub-model to the edge node energy distribution model; The second electrical energy scheduling is the electrical energy output by the virtual energy distribution sub-model to other virtual energy distribution sub-models; Step S303: Update the penalty function factor according to the optimization result. The update process includes: The first penalty function factor + 1 = the first penalty function iteration multiplier × the first penalty function factor; The second penalty function factor + 1 = the second penalty function iteration multiplier × the second penalty function factor; Step S304: Update the iteration counter k = k + 1, and return to Step S302 until the convergence criterion is satisfied; The convergence criterion is that when the power exchange error between the virtual energy distribution sub-models satisfies the convergence condition, the convergence condition includes: |The first power interaction information - the first power dispatch| < the first convergence criterion; |The second power interaction information - the second power dispatch| < the second convergence criterion; Step S305: When the convergence criterion is satisfied simultaneously, end the iteration process and output the optimization result. The optimization result includes the power dispatch plan and the power exchange strategy of each virtual energy distribution sub-model.

9. The edge-computing-based regional energy distributed control method according to claim 8, wherein The second-order scheduling algorithm includes: Adopt the column constraint generation algorithm CCG, and correct the decision deviation in the first-order scheduling algorithm through interactive iteration with the first-order scheduling algorithm.

10. The edge-computing-based regional energy distributed control method according to claim 9, wherein, The second-order scheduling algorithm specifically includes: Step S311: Initialize the lower bound of the sub-problem to negative infinity, the upper bound of the sub-problem to positive infinity, set the iteration counter r of the sub-problem to 1, and set the third convergence criterion; Step S312: Set the convergence condition as: the upper bound of the sub-problem - the lower bound of the sub-problem ≥ the third convergence criterion; Step S313: Iteratively optimize the optimization result of the first-order scheduling algorithm, update the dual variable and the objective constraint in the calculation process of the first-order scheduling algorithm, and update the upper bound and the lower bound of the sub-problem; Step S314: Update r = r + 1; Step S315: Optimize the currently updated sub-problem as the optimization object, recursively optimize the upper bound and the lower bound of the sub-problem of the sub-problem, and solve to obtain the upper bound and the lower bound that meet the third convergence criterion; Step S316: Stop the iteration when the convergence condition is met; All data volumes of the regional energy distributed control method based on edge computing are within a unit time.

Images