Edge interaction control and edge node optimization method for regulatory authority transactions

Abstract

The present invention discloses an edge interaction control and edge node optimization method for regulating authority transactions, which obtains the interactive business between each demand response resource supplier and each demander in the area to be optimized; establishes an edge interaction model regarding the economic benefits of demand response resource suppliers and demanders; solves the edge interaction model with the maximum value as the goal, and determines the interaction volume between all demand response resource suppliers and all demanders in the area to be optimized; and allocates the interaction volume between all demand response resource suppliers and all demanders in the area to be optimized to the edge nodes according to a pre-set edge node edge interaction business allocation principle. Advantages: Establishing an edge interaction model, solving the edge interaction model, and obtaining the optimal edge interaction behavior from a data-driven perspective, avoiding the deficiency of a complete parameter set required for model driving; realizing edge node edge interaction business management and control optimization, and optimizing business computing speed and equipment energy consumption.

Description

Technical Field

[0001] The present invention relates to an edge interaction control and edge node optimization method for regulation authority trading, and belongs to the technical field of power grid regulation. Background Art

[0002] As the proportion of distributed resources connected to the grid increases, the trend toward miniaturization, decentralization, and wide-area deployment of demand-side resources becomes increasingly evident. Consumers are gradually shifting toward integrated production and consumption, blurring the boundaries between supply and demand in the system. The contradiction between strong randomness, weak observability, and high requirements becomes prominent, making global information difficult to obtain. Most physical mechanism models suffer from oversimplification, resulting in reduced accuracy. They also assume that elements such as the entire network topology, node loads, and distributed power output are known. However, actual engineering demonstrations and applications still lack complete and accurate secondary equipment for data collection and monitoring. State data synchronization and integration remain difficult to achieve full coverage, and metering layout and accuracy are still far from being fully observable. These practical issues restrict the applicability of model-driven systems. Traditional, fully centralized operation and control technology approaches, due to transaction latency, network congestion, and user privacy issues, are difficult to achieve low-cost, large-scale demand-side resource participation and interaction, nor are they able to meet the plug-and-play requirements of various smart terminals.

[0003] Artificial intelligence, free from the constraints of physical mechanisms, is an innovative attempt to break through model barriers by mining data correlations. Edge computing offers a solution for moving centralized operations and control from the cloud to the user end, providing technical support for interactive edge control and optimization of demand-side resources. Leveraging AI and edge computing to overcome these technical barriers, enhance system flexibility and the ability to integrate large-scale distributed resources, and provide guidance for the flexible interaction of power generation, grid, load, and storage. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide an edge interaction control and edge node optimization method for regulating authority transactions.

[0005] To solve the above technical problems, the present invention provides an edge interaction control and edge node optimization method for regulating authority transactions, comprising:

[0006] Obtain the interactive business between each demand response resource supplier and each demander in the area to be optimized;

[0007] Establishing an edge interaction model regarding the economic benefits of demand response resource suppliers and demanders based on the interactive business between each demand response resource supplier and each demander;

[0008] The maximum value of the edge interaction model is used as the goal to solve the problem, determine the interaction amount between each demand response resource supplier and each demand party, and obtain the interaction amount between all demand response resource suppliers and all demand parties in the area to be optimized based on the interaction amount between each demand response resource supplier and each demand party.

[0009] Obtain edge nodes that can be used for interactive business processing, and allocate the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized to the edge nodes according to the pre-set edge node edge interactive business allocation principle.

[0010] Furthermore, the edge interaction model is:

[0011]

[0012] Among them, ρ∈[0,1], ρ is the economic benefit trade-off factor between the demand response resource supplier and the demander, p t is the real-time electricity price in the spot market p w,t and the medium- and long-term electricity price in the spot market d,t The difference, c t,i is the purchase price of the demand response resource demander in period t, Δd t,i is the interaction amount of demand response resource supplier i in period t, η∈[0,1] is the trade-off factor between the economic benefits and comfort needs of the demand response resource supplier, p r,t is the retail electricity price, The comfort costs paid by the demand response resource providers;

[0013]

[0014] Among them, ε t,i is the self-elasticity coefficient of demand response resource supplier i in period t, c min is the lower limit of the purchase price, Δd t-1,i is the interaction volume of demand response resource supplier i in period t-1, ξ t,t-1,i is the time-price elasticity coefficient of demand response resource supplier i in period t relative to period t-1, c t-1,i is the purchase price of the demand response resource demander in period t-1, d t,i is the electricity demand of demand response resource supplier i in period t, D max 、D min are the upper and lower limits of the interaction volume, and c is the average acquisition price in each period.

[0015] Furthermore, the method of solving the edge interaction model with the maximum value as the goal to determine the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized includes:

[0016] The edge interaction model is solved using the multi-agent deep deterministic strategy MADDPG. The state space, action space, and reward function of each agent are the same. The MADDPG state space S includes the electricity demand d of the demand response resource supplier i in period t. t,i , self-elastic coefficient ε t,i , the time-price elasticity coefficient ξ of period t relative to period t-1 t,t-1,i , the purchase price c of the demand response resource demander in period t-1 t-1,i , expressed as:

[0017] S=(d t,i ,ε t,i ,ξ t,t-1,i ,c t-1,i )

[0018] The MADDPG action space A includes the interaction between the demand response resource supplier and the demander, and the purchase price of the demand response resource demander in period t, which can be expressed as:

[0019] A=(Δd t,i ,c t,i )

[0020] The MADDPG reward function R comprehensively considers the economic benefits of both the demand response resource supplier and the demander, and is expressed as:

[0021]

[0022] Determine the interaction amount between the demand response resource supplier i and each demander according to the reward function R;

[0023] The multi-agent deep deterministic strategy MADDPG is repeatedly used to solve another edge interaction model until the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized is determined.

[0024] Furthermore, obtaining an edge node that can be used for interactive service processing includes:

[0025] The node d where the demand-side mobilization center is located, the edge interaction service source node and edge node are s, s∈S′, a(a∈A), S′ is the source node set, A is the edge node set, and the working path set W of (s, d), (s, a), (d, a) is s,d 、W s,a 、W d,a Complete link separation is represented as follows:

[0026]

[0027]

[0028] Among them, ∩ is the intersection, is an empty set;

[0029] The relaxation process of equations (12) and (13) is as follows:

[0030]

[0031]

[0032] Among them, ω s,d 、ω s,a 、ω d,a are any working path set links from source node s to scheduling center node d, source node s to edge node a, and scheduling center node d to edge node a;

[0033] When the working path set is not an empty set, the following constraints are met:

[0034]

[0035]

[0036] Among them, D1 is the limit on the number of links in the working path set;

[0037] The selection criteria for edge nodes are as follows:

[0038]

[0039] Among them, |S a | is the number of source nodes accessing edge interaction services, Indicates that the edge node can cover the intersection of the business source nodes under its jurisdiction. is the set of edge nodes selected for deploying edge servers, Con(·) is the number of edge interaction service source nodes covered by the selected edge nodes, u a is the identifier symbol, when a is selected as an edge node u a is 1, otherwise it is 0, S a The set of edge interaction service source nodes managed by edge nodes, where N is the number of selected edge nodes;

[0040] An edge node that can be used for interactive service processing is acquired based on the edge node selection target.

[0041] Furthermore, the method further comprises:

[0042] Deploy edge servers at the selected edge nodes and offload edge interaction services to the edge servers.

[0043] Furthermore, allocating the interaction amounts of all demand response resource suppliers and all demanders in the area to be optimized to the edge nodes according to a preset edge node edge interaction service allocation principle includes:

[0044] Establish a relationship between the amount of edge interaction expected by the grid and the amount of edge interaction that can be performed by the demand response resources:

[0045]

[0046] Among them, E l,t is the expected edge interaction volume of the edge interaction service l at time t, O t is the demand response resource of the entire network during period t, Δd l,t is the executable edge interaction volume of the demand response resource side edge interaction service l in period t, Δd t is the executable edge interaction amount of all edge interaction services in period t, R t Demand response resources that do not interact;

[0047] Based on the preset edge node edge interaction service allocation principle, the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized is allocated to the edge nodes. The edge node edge interaction service allocation principle is as follows:

[0048]

[0049] Among them, E l is the expected edge interaction volume of the power grid edge interaction service l, ζ l,t ,ξ l,t are the start and end time of the establishment of edge interaction service l, E t is the expected edge interaction volume of all edge interaction services in the power grid during period t, E avg is the control mean value of the edge interaction volume of all edge interaction services expected by the power grid on N edge nodes during period t.

[0050] An edge interaction control and edge node optimization system for regulating authority transactions.

[0051] An acquisition module is used to obtain the interactive business between each demand response resource supplier and each demander in the area to be optimized;

[0052] A construction module is used to establish an edge interaction model regarding the economic benefits of demand response resource suppliers and demand parties based on the interactive business between each demand response resource supplier and each demand party;

[0053] A solution module is used to solve the edge interaction model with the maximum value as the goal, determine the interaction amount between each demand response resource supplier and each demand party, and obtain the interaction amount between all demand response resource suppliers and all demand parties in the area to be optimized based on the interaction amount between each demand response resource supplier and each demand party;

[0054] The allocation module is used to obtain edge nodes that can be used for interactive business processing, and allocate the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized to the edge nodes according to the pre-set edge node edge interactive business allocation principle.

[0055] Furthermore, the edge interaction model is:

[0056]

[0057] Among them, ρ∈[0,1], ρ is the economic benefit trade-off factor between the demand response resource supplier and the demander, p t is the real-time electricity price in the spot market p w,t and the medium- and long-term electricity price in the spot market d,t The difference, c t,i is the purchase price of the demand response resource demander in period t, Δd t,i is the interaction amount of demand response resource supplier i in period t, η∈[0,1] is the trade-off factor between the economic benefits and comfort needs of the demand response resource supplier, p r,t is the retail electricity price, The comfort costs paid by the demand response resource providers;

[0058]

[0059] Among them, ε t,i is the self-elasticity coefficient of demand response resource supplier i in period t, c min is the lower limit of the purchase price, Δd t-1,i is the interaction volume of demand response resource supplier i in period t-1, ξ t,t-1,i is the time-price elasticity coefficient of demand response resource supplier i in period t relative to period t-1, c t-1,i is the purchase price of the demand response resource demander in period t-1, d t,i is the electricity demand of demand response resource supplier i in period t, D max 、D min are the upper and lower limits of the interaction amount, The average purchase price for each period.

[0060] Furthermore, the solution module is used to

[0061] The edge interaction model is solved using the multi-agent deep deterministic strategy MADDPG. The state space, action space, and reward function of each agent are the same. The MADDPG state space S includes the electricity demand d of the demand response resource supplier i in period t. t,i , self-elastic coefficient ε t,i , the time-price elasticity coefficient ξ of period t relative to period t-1 t,t-1,i , the purchase price c of the demand response resource demander in period t-1 t-1,i , expressed as:

[0062] S=(d t,i ,ε t,i ,ξ t,t-1,i ,c t-1,i )

[0063] The MADDPG action space A includes the interaction between the demand response resource supplier and the demander, and the purchase price of the demand response resource demander in period t, which can be expressed as:

[0064] A=(Δd t,i ,c t,i )

[0065] The MADDPG reward function R comprehensively considers the economic benefits of both the demand response resource supplier and the demander, and is expressed as:

[0066]

[0067] Determine the interaction amount between the demand response resource supplier i and each demander according to the reward function R;

[0068] The multi-agent deep deterministic strategy MADDPG is repeatedly used to solve another edge interaction model until the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized is determined.

[0069] Furthermore, the allocation module is used to

[0070] The node d where the demand-side mobilization center is located, the edge interaction service source node and edge node are s, s∈S′, a(a∈A), S′ is the source node set, A is the edge node set, and the working path set W of (s, d), (s, a), (d, a) is s,d 、W s,a 、W d,a Complete link separation is represented as follows:

[0071]

[0072]

[0073] Among them, ∩ is the intersection, is an empty set;

[0074] The relaxation process of equations (12) and (13) is as follows:

[0075]

[0076]

[0077] Among them, ω s,d 、ω s,a 、ω d,a are any working path set links from source node s to scheduling center node d, source node s to edge node a, and scheduling center node d to edge node a;

[0078] When the working path set is not an empty set, the following constraints are met:

[0079]

[0080]

[0081] Among them, D1 is the limit on the number of links in the working path set;

[0082] The selection criteria for edge nodes are as follows:

[0083]

[0084] Among them, |S a | is the number of source nodes accessing edge interaction services, Indicates that the edge node can cover the intersection of the business source nodes under its jurisdiction. is the set of edge nodes selected for deploying edge servers, Con(·) is the number of edge interaction service source nodes covered by the selected edge nodes, u a is the identifier symbol, when a is selected as an edge node u a is 1, otherwise it is 0, S a The set of edge interaction service source nodes managed by edge nodes, where N is the number of selected edge nodes;

[0085] An edge node that can be used for interactive service processing is acquired based on the edge node selection target.

[0086] Furthermore, the allocation module is also used to

[0087] Deploy edge servers at the selected edge nodes and offload edge interaction services to the edge servers.

[0088] Furthermore, the allocation module is used to

[0089] Establish a relationship between the amount of edge interaction expected by the grid and the amount of edge interaction that can be performed by the demand response resources:

[0090]

[0091] Among them, E l,t is the expected edge interaction volume of the edge interaction service l at time t, O t is the demand response resource of the entire network during period t, Δd l,t is the executable edge interaction volume of the demand response resource side edge interaction service l in period t, Δd t is the executable edge interaction amount of all edge interaction services in period t, R t Demand response resources that do not interact;

[0092] Based on the preset edge node edge interaction service allocation principle, the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized is allocated to the edge nodes. The edge node edge interaction service allocation principle is as follows:

[0093]

[0094] Among them, E l is the expected edge interaction volume of the power grid edge interaction service l, ζ l,t ,ξ l,t are the start and end time of the establishment of edge interaction service l, E t is the expected edge interaction volume of all edge interaction services in the power grid during period t, E avg is the control mean value of the edge interaction volume of all edge interaction services expected by the power grid on N edge nodes during period t.

[0095] Furthermore, the deployment of the edge interaction model relies on the NVIDIA Jetson Xavier NX artificial intelligence chip to embed the edge interaction model into the response subject energy management system for packaging.

[0096] A computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods described above.

[0097] A computing device comprising:

[0098] One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the methods described.

[0099] The beneficial effects achieved by the present invention are:

[0100] An edge interaction model is established to characterize the edge interaction mechanism between demand response resource suppliers and demanders, and the edge interaction model is solved to obtain the optimal edge interaction behavior from a data-driven perspective. This avoids the deficiency of the complete parameter set required for model driving, deeply explores the correlation relationship between demand resource edge interaction data, and reveals the physical characteristics of edge interaction from a data perspective.

[0101] Select edge server deployment nodes, offload edge interaction services to edge servers, coordinate edge-end business processing, allocate edge interaction services taking into account business reliability, optimize edge node edge interaction business management and control, optimize business computing speed and equipment energy consumption, and reduce the impact of business under imperfect communication conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 It is a schematic flow diagram of the present invention. DETAILED DESCRIPTION

[0103] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0104] Example:

[0105] like Figure 1 As shown in the figure, a method of edge interaction control and edge node optimization for regulating authority transactions is proposed, including edge interaction model, edge interaction model solution, edge interaction model deployment, edge node selection, and edge interaction business allocation;

[0106] The edge interaction model describes the edge interaction mechanism between the demand response resource supplier and the demand side based on the relevant information of the interactive business between the demand response resource supplier and the demand side in the area to be optimized (response subject characteristics);

[0107] The edge interaction model is solved by using multi-agent deep reinforcement learning to obtain the optimal edge interaction behavior from a data-driven perspective;

[0108] The edge interaction model deployment relies on the NVIDIA Jetson Xavier NX artificial intelligence chip to embed the edge interaction model into the response subject energy management system for packaging;

[0109] The selection of edge nodes takes into account the limited computing power and communication resources of the terminal packaged chips, and selects edge server deployment nodes;

[0110] Edge interaction service allocation and service reliability are taken into account to optimize the management and control of edge interaction services at edge nodes.

[0111] Furthermore, the edge interaction model describes the edge interaction mechanism between the demand response resource supplier and the demander based on the characteristics of the response subject. The specific model is constructed as follows:

[0112] (1) Based on the hierarchical and partitioned control authority characteristics, the demand response resource supplier comprehensively considers economic benefits and comfort needs according to the current price signal, breaks through the traditional centralized operation and control technical barriers, sinks the cloud master station service to the edge of the network, interacts with the demand response resource demander at the edge of the network, and adjusts its own response behavior with the goal of maximizing its own benefits. It is modeled as the following optimization problem:

[0113]

[0114] Among them, H is the set of all time periods in a day, η∈[0,1] is the trade-off factor between the economic benefits and comfort needs of the demand response resource supplier, c t,i is the purchase price of demand response resource demand side, c max 、c min are the upper and lower limits of the purchase price, p r,t is the retail electricity price, Δd t,i is the interaction volume of demand response resource supplier i in period t, is the dissatisfaction function, which represents the comfort cost paid by the demand response resource supplier;

[0115] Δd t,i This reflects the willingness of demand response resource suppliers to respond to the purchase price signal. Previous edge interaction models believed that it was only related to the purchase price in period t, ignoring the degree to which demand response resource suppliers responded to the purchase price difference between adjacent periods. That is, if the current purchase price is higher than the previous purchase price, it will stimulate additional response enthusiasm, Δd t,i Increase, on the contrary, Δd t,i Reduce; Here we introduce the time-price elasticity mechanism to modify the model described in formula 1;

[0116]

[0117] Where: t,i is the self-elasticity coefficient of the demand response resource supplier i in period t, reflecting the percentage of interaction adjustment caused by a 1% change in the purchase price in period t, ξ t,t-1,i is the time-price elasticity coefficient of demand response resource supplier i in period t relative to period t-1, d t,i is the electricity demand of demand response resource supplier i in period t, D max 、D min are the upper and lower limits of the interaction amount, is the average purchase price for each period;

[0118] The dissatisfaction function is usually considered to be a convex function whose value increases as the amount of interaction increases. Here, a quadratic function is used to represent the dissatisfaction function Δd t,i :

[0119]

[0120] Where: α i >0 is a constant, representing the comfort demand sensitivity of the demand response resource supplier i. The larger the value, the higher the dissatisfaction and the lower the response willingness, and vice versa; β i is the dissatisfaction auxiliary coefficient, which has the same effect as α i similar.

[0121] (2) Demand response resource demanders interact with suppliers at the edge to obtain a certain amount of interaction for their own use or participate in the electricity wholesale market through aggregators, thereby improving their own economic benefits and maximizing their own profits by seeking the optimal purchase price. This can be modeled as the following optimization problem:

[0122]

[0123] p t =p w,t -p d,t (5)

[0124] Among them, p w,t 、p d,t are the real-time electricity price and medium- and long-term electricity price in the spot market, respectively, and I is the set of demand response resource suppliers;

[0125] (3) Taking into account the economic benefits of both the demand response resource supplier and the demander, the objectives of constructing an edge interactive control model for regulation authority trading are as follows:

[0126]

[0127]

[0128] Among them, ρ∈[0,1], ρ is the economic benefit trade-off factor between the demand response resource supplier and the demander.

[0129] Furthermore, the edge interaction model solution utilizes multi-agent deep reinforcement learning to obtain the optimal edge interaction behavior from a data-driven perspective, breaking through the model-driven barriers. The specific solution steps are as follows:

[0130] (1) Here, the multi-agent deep deterministic strategy (MADDPG) is used to solve the edge interaction model. The state space, action space, and reward function of each agent are the same. The MADDPG state space S includes the electricity demand d of the demand response resource supplier i in time period t.t,i , self-elastic coefficient ε t,i , the time-price elasticity coefficient ξ of period t relative to period t-1 t,t-1,i , the purchase price c of the demand response resource demander in period t-1 t-1,i , which can be expressed as:

[0131] S=(d t,i ,ε t,i ,ξ t,t-1,i ,c t-1,i ) (8)

[0132] (2) The MADDPG action space A includes the interaction between the demand response resource supplier and the demand response resource demander, and the purchase price of the demand response resource demander in period t, which can be expressed as:

[0133] A=(Δd t,i ,c t,i ) (9)

[0134] (3) The MADDPG reward function R comprehensively considers the economic benefits of both the demand response resource supplier and the demander, and can be expressed as:

[0135]

[0136] The agent continuously interacts with the environment to arrive at the optimal edge interaction behavior strategy, and uses the following formula to measure the performance of the strategy:

[0137]

[0138] Among them, π θ is the interaction strategy, θ is the interaction strategy learning parameter, P π is the state transfer function, Q(s,π θ (s)) is the action value function, is the expected function. The larger the function value is, the better the performance is.

[0139] Furthermore, the edge interaction model deployment relies on the NVIDIA Jetson Xavier NX artificial intelligence chip to embed the edge interaction model into the response subject energy management system for packaging. The traditional operation and control system places the model on the cloud host, which has problems such as high communication latency, large bandwidth usage, and difficulty in privacy protection. The artificial intelligence chip provides technical support for the sinking of cloud services to achieve edge autonomy and provides a hardware acceleration carrier for artificial intelligence data-driven algorithm models. The specific deployment steps are as follows:

[0140] (1) Use a USB data cable to connect the NVIDIA Jetson Xavier NX chip (lower computer) to the PC, enter the NVIDIA Jetson Xavier NX chip Ubuntu system through the PC (upper computer), and download Python 3.X, Jupyter, and Visdom environment software;

[0141] (2) Use the pip3installxxx command to install the dependency package, copy the PC-side edge interaction model to the Jupyter notebook of the NVIDIA JetsonXavier NX chip system through the copy.deepcopy() deep copy command, and click Run;

[0142] (3) Deploy the TCP communication protocol in the upper and lower seats, as follows:

[0143] i) Import the socket module

[0144] import socket

[0145] ii) Create a TCP socket

[0146] tcp=socket.socket(socket.AF_INET,socket.SOCK_STREAM)

[0147] iii) Connection and Binding

[0148] Connection: tcp.connect("ip",port) #ip is the IP address, port is the port

[0149] Bind: tcp.bind(("ip",port))

[0150] (4) After the communication protocol is deployed, the data line is disconnected and the NVIDIA Jetson Xavier NX chip equipped with the edge interaction model is embedded in the response subject energy management system for packaging;

[0151] (5) Remotely access the control chip through the host computer, use the following commands to communicate and interact, send control instructions through the browser, remotely develop control, receive corresponding data, etc.:

[0152] i) Remote login, username is the name of the login seat, ip is the IP address;

[0153] sshusername@ip

[0154] ii) Data is encrypted and sent between boarding and disembarking stations using a hash algorithm;

[0155] Encryption: tcp.encode()

[0156] Send: tcp.send()

[0157] ii) Data is decrypted and received between boarding and disembarking stations using a hash algorithm;

[0158] Decryption: tcp.decode()

[0159] Receive: tcp.recv()

[0160] (6) When the communication interaction between the upper and lower cameras ends, the socket is closed. The upper camera can also be a mobile phone, tablet or other device.

[0161] #Close the socket

[0162] tcp.close()

[0163] Furthermore, the selection of edge nodes takes into account the limited computing power and communication resources of the terminal packaged chips. When the user end accesses a large number of smart terminal devices, it causes a large amount of terminal chip resources to be consumed, the edge interactive services are blocked, and the cost of re-awakening the chip is high. It is necessary to rely on higher-performance edge servers to offload the edge interactive services to the edge servers. The edge terminals collaborate to complete the business processing. The specific selection steps for edge server deployment nodes are as follows:

[0164] (1) The node d where the demand-side mobilization center is located, the edge interaction service source node and edge node are s (s∈S) and a (a∈A), S′ is the source node set, A is the edge node set, and the working path set W of (s, d), (s, a), and (d, a) is s,d 、W s,a 、W d,a Complete link separation is represented as follows:

[0165]

[0166]

[0167] Among them, ∩ is the intersection, is an empty set;

[0168] (2) Equations 12 and 13 have a wide set of feasible solutions, which makes it difficult for edge node selection to converge quickly. Therefore, they are relaxed as follows:

[0169]

[0170]

[0171] Among them, ω s,d 、ω s,a 、ωd,a are any working path set links from source node s to scheduling center node d, source node s to edge node a, and scheduling center node d to edge node a;

[0172] (3) When the working path set is not an empty set, the following constraints are met:

[0173]

[0174]

[0175] Among them, D1 is the limit on the number of links in the working path set;

[0176] (4) The selection criteria for edge nodes are as follows:

[0177]

[0178] Among them, |S a | is the number of source nodes accessing edge interaction services, is the set of edge nodes selected for deploying edge servers, Con(·) is the number of edge interaction service source nodes covered by the selected edge nodes, u a is the identifier symbol, when a is selected as an edge node u a is 1, otherwise it is 0, S a The set of edge interaction service source nodes managed by edge nodes, where N is the number of selected edge nodes;

[0179] (5) Deploy edge servers at the selected edge nodes so that the terminal chip can offload edge interaction services to the edge servers.

[0180] Furthermore, the edge interaction service allocation takes service reliability into consideration to optimize the management and control of edge interaction services at edge nodes. The specific allocation and control optimization steps are as follows:

[0181] (1) Establish the relationship between the amount of edge interaction expected by the grid and the amount of edge interaction that can be performed by the demand response resources:

[0182]

[0183] Among them, E l,t is the expected edge interaction volume of the edge interaction service l at time t, O t is the demand response resource of the entire network during period t, Δd l,t is the executable edge interaction volume of the demand response resource side edge interaction service l in period t, Δd t is the executable edge interaction amount of all edge interaction services in period t, R t Demand response resources that do not interact;

[0184] (2) Principles for allocating edge node-edge interaction services:

[0185]

[0186] Among them, E l is the expected edge interaction volume of the power grid edge interaction service l, ζ l,t ,ξ l,t are the start and end time of the establishment of edge interaction service l, E t is the expected edge interaction volume of all edge interaction services in the power grid during period t, E avg The edge interaction volume of all edge interaction services expected by the power grid in period t on N edge nodes should be controlled at a mean value. The edge interaction volume controlled by each edge node in period t should be close to the control mean value.

[0187] The present invention also provides an edge interaction control and edge node optimization system for regulating authority transactions.

[0188] An acquisition module is used to obtain the interactive business between each demand response resource supplier and each demander in the area to be optimized;

[0189] A construction module is used to establish an edge interaction model regarding the economic benefits of demand response resource suppliers and demand parties based on the interactive business between each demand response resource supplier and each demand party;

[0190] A solution module is used to solve the edge interaction model with the maximum value as the goal, determine the interaction amount between each demand response resource supplier and each demand party, and obtain the interaction amount between all demand response resource suppliers and all demand parties in the area to be optimized based on the interaction amount between each demand response resource supplier and each demand party;

[0191] The allocation module is used to obtain edge nodes that can be used for interactive business processing, and allocate the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized to the edge nodes according to the pre-set edge node edge interactive business allocation principle.

[0192] Furthermore, the edge interaction model is:

[0193]

[0194] Among them, ρ∈[0,1], ρ is the economic benefit trade-off factor between the demand response resource supplier and the demander, p t is the real-time electricity price in the spot market p w,t and the medium- and long-term electricity price in the spot market d,t The difference, c t,i is the purchase price of the demand response resource demander in period t, Δd t,iis the interaction amount of demand response resource supplier i in period t, η∈[0,1] is the trade-off factor between the economic benefits and comfort needs of the demand response resource supplier, p r,t is the retail electricity price, The comfort costs paid by the demand response resource providers;

[0195]

[0196] Among them, ε t,i is the self-elasticity coefficient of demand response resource supplier i in period t, c min is the lower limit of the purchase price, Δd t-1,i is the interaction volume of demand response resource supplier i in period t-1, ξ t,t-1,i is the time-price elasticity coefficient of demand response resource supplier i in period t relative to period t-1, c t-1,i is the purchase price of the demand response resource demander in period t-1, d t,i is the electricity demand of demand response resource supplier i in period t, D max 、D min are the upper and lower limits of the interaction amount, The average purchase price for each period.

[0197] Furthermore, the solution module is used to

[0198] The edge interaction model is solved using the multi-agent deep deterministic strategy MADDPG. The state space, action space, and reward function of each agent are the same. The MADDPG state space S includes the electricity demand d of the demand response resource supplier i in period t. t,i , self-elastic coefficient ε t,i , the time-price elasticity coefficient ξ of period t relative to period t-1 t,t-1,i , the purchase price c of the demand response resource demander in period t-1 t-1,i , expressed as:

[0199] S=(d t,i ,ε t,i ,ξ t,t-1,i ,c t-1,i )

[0200] The MADDPG action space A includes the interaction between the demand response resource supplier and the demander, and the purchase price of the demand response resource demander in period t, which can be expressed as:

[0201] A=(Δd t,i ,c t,i )

[0202] The MADDPG reward function R comprehensively considers the economic benefits of both the demand response resource supplier and the demander, and is expressed as:

[0203]

[0204] Determine the interaction amount between the demand response resource supplier i and each demander according to the reward function R;

[0205] The multi-agent deep deterministic strategy MADDPG is repeatedly used to solve another edge interaction model until the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized is determined.

[0206] Furthermore, the allocation module is used to

[0207] The node d where the demand-side mobilization center is located, the edge interaction service source node and edge node are s, s∈S′, a(a∈A), S′ is the source node set, A is the edge node set, and the working path set W of (s, d), (s, a), (d, a) is s,d 、W s,a 、W d,a Complete link separation is represented as follows:

[0208]

[0209]

[0210] Among them, ∩ is the intersection, is an empty set;

[0211] The relaxation process of equations (12) and (13) is as follows:

[0212]

[0213]

[0214] Among them, ω s,d 、ω s,a 、ω d,a are any working path set links from source node s to scheduling center node d, source node s to edge node a, and scheduling center node d to edge node a;

[0215] When the working path set is not an empty set, the following constraints are met:

[0216]

[0217]

[0218] Among them, D1 is the limit on the number of links in the working path set;

[0219] The selection criteria for edge nodes are as follows:

[0220]

[0221] Among them, |S a | is the number of source nodes accessing edge interaction services, Indicates that the edge node can cover the intersection of the business source nodes under its jurisdiction. is the set of edge nodes selected for deploying edge servers, Con(·) is the number of edge interaction service source nodes covered by the selected edge nodes, u a is the identifier symbol, when a is selected as an edge node u a is 1, otherwise it is 0, S a The set of edge interaction service source nodes managed by edge nodes, where N is the number of selected edge nodes;

[0222] An edge node that can be used for interactive service processing is acquired based on the edge node selection target.

[0223] Furthermore, the allocation module is also used to

[0224] Deploy edge servers at the selected edge nodes and offload edge interaction services to the edge servers.

[0225] Furthermore, the allocation module is used to

[0226] Establish a relationship between the amount of edge interaction expected by the grid and the amount of edge interaction that can be performed by the demand response resources:

[0227]

[0228] Among them, E l,t is the expected edge interaction volume of the edge interaction service l at time t, O t is the demand response resource of the entire network during period t, Δd l,t is the executable edge interaction volume of the demand response resource side edge interaction service l in period t, Δd t is the executable edge interaction amount of all edge interaction services in period t, R t Demand response resources that do not interact;

[0229] Based on the preset edge node edge interaction service allocation principle, the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized is allocated to the edge nodes. The edge node edge interaction service allocation principle is as follows:

[0230]

[0231] Among them, E l is the expected edge interaction volume of the power grid edge interaction service l, ζ l,t ,ξl,t are the start and end time of the establishment of edge interaction service l, E t is the expected edge interaction volume of all edge interaction services in the power grid during period t, E avg is the control mean value of the edge interaction volume of all edge interaction services expected by the power grid on N edge nodes during period t.

[0232] Furthermore, the deployment of the edge interaction model relies on the NVIDIA Jetson Xavier NX artificial intelligence chip to embed the edge interaction model into the response subject energy management system for packaging.

[0233] The corresponding present invention also provides a computer-readable storage medium storing one or more programs, characterized in that the one or more programs include instructions, which, when executed by a computing device, enable the computing device to perform any of the methods described.

[0234] The present invention also provides a computing device, comprising:

[0235] One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the methods described.

[0236] One of the beneficial effects of this solution is that it establishes an edge interaction model based on the characteristics of the response subject, characterizes the edge interaction mechanism between the demand response resource supplier and the demander, and uses multi-agent deep reinforcement learning MADDPG to solve the model, obtaining the optimal edge interaction behavior from a data-driven perspective. This avoids the deficiency of the complete parameter set required for model driving, deeply explores the correlation relationship between the demand resource edge interaction data, and reveals the physical characteristics of the edge interaction from a data perspective.

[0237] One of the beneficial effects of this solution is that it relies on the NVIDIA Jetson Xavier NX artificial intelligence chip to carry the edge interaction model, embeds it into the response subject energy management system for packaging, and breaks through the barriers of traditional centralized operation and control systems such as high communication latency, large bandwidth usage, and difficult privacy protection. It provides technical support for the sinking of cloud services to achieve edge autonomy and provides a hardware acceleration carrier for artificial intelligence data-driven algorithm models.

[0238] One of the beneficial effects of this solution is that, considering the limited computing power and communication resources of the terminal packaged chip, edge server deployment nodes are selected, edge interaction services are offloaded to the edge server, the edge and the end collaborate to complete business processing, and edge interaction services are allocated considering business reliability, so as to achieve edge node edge interaction business management and control optimization, optimize business computing speed and equipment energy consumption, and reduce the impact of business under imperfect communication conditions.

[0239] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0240] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0241] These computer program instructions may also be stored in a computer readable memory that can direct 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 an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0242] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0243] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for edge interaction control and edge node optimization for regulating authority transactions, characterized in that: include: Obtain the interactive business between each demand response resource supplier and each demander in the area to be optimized; Establishing an edge interaction model regarding the economic benefits of demand response resource suppliers and demanders based on the interactive business between each demand response resource supplier and each demander; The maximum value of the edge interaction model is used as the goal to solve the problem, determine the interaction amount between each demand response resource supplier and each demand party, and obtain the interaction amount between all demand response resource suppliers and all demand parties in the area to be optimized based on the interaction amount between each demand response resource supplier and each demand party. Obtain edge nodes that can be used for interactive business processing, and allocate the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized to the edge nodes according to the pre-set edge node edge interactive business allocation principle; The edge interaction model is: Among them, ρ∈[0,1], ρ is the economic benefit trade-off factor between the demand response resource supplier and the demander, p t is the real-time electricity price in the spot market p w,t and the medium- and long-term electricity price in the spot market d,t The difference, c t,i is the purchase price of the demand response resource demander in period t, Δd t,i is the interaction amount of demand response resource supplier i in period t, η∈[0,1] is the trade-off factor between the economic benefits and comfort needs of the demand response resource supplier, p r,t is the retail electricity price, The comfort costs paid by the demand response resource providers; Among them, ε t,i is the self-elasticity coefficient of demand response resource supplier i in period t, c min is the lower limit of the purchase price, Δd t-1,i is the interaction volume of demand response resource supplier i in period t-1, ξ t,t-1,i is the time-price elasticity coefficient of demand response resource supplier i in period t relative to period t-1, c t-1,i is the purchase price of the demand response resource demander in period t-1, d t,i is the electricity demand of demand response resource supplier i in period t, D max 、D min are the upper and lower limits of the interaction amount, The average purchase price for each period.

2. The edge interaction control and edge node optimization method for regulating authority transactions according to claim 1 is characterized in that: The method of solving the edge interaction model with the maximum value as the goal to determine the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized includes: The edge interaction model is solved using the multi-agent deep deterministic strategy MADDPG. The state space, action space, and reward function of each agent are the same. The MADDPG state space S includes the electricity demand d of the demand response resource supplier i in period t. t,i , self-elastic coefficient ε t,i , the time-price elasticity coefficient ξ of period t relative to period t-1 t,t-1,i , the purchase price c of the demand response resource demander in period t-1 t-1,i , expressed as: S=(d t,i ,he t,i ,x t,t-1,i ,c t-1,i ) The MADDPG action space A includes the interaction between the demand response resource supplier and the demander, and the purchase price of the demand response resource demander in period t, which can be expressed as: A=(Δd t,i ,c t,i ) The MADDPG reward function R comprehensively considers the economic benefits of both the demand response resource supplier and the demander, and is expressed as: Where, I is the set of demand response resource suppliers; H is the set of all time periods in a day; Determine the interaction amount between the demand response resource supplier i and each demander according to the reward function R; The multi-agent deep deterministic strategy MADDPG is repeatedly used to solve another edge interaction model until the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized is determined.

3. The edge interaction control and edge node optimization method for regulating authority transactions according to claim 1 is characterized in that: The acquiring of an edge node that can be used for interactive service processing includes: The mobilization center node on the demand side is d, the edge interaction service source node is s, and the edge node is a, s∈S′, a∈A, S′ is the source node set, A is the edge node set, and the working path set W of (s, d), (s, a), (d, a) is s,d 、W s,a 、W d,a Complete link separation is represented as follows: Among them, ∩ is the intersection, is an empty set; The relaxation process of equations (12) and (13) is as follows: Among them, ω s,d 、ω s,a 、ω d,a are any working path set links from source node s to scheduling center node d, source node s to edge node a, and scheduling center node d to edge node a; When the working path set is not an empty set, the following constraints are met: Among them, D1 is the limit on the number of links in the working path set; The selection criteria for edge nodes are as follows: Among them, |S a | is the number of source nodes accessing edge interaction services, Indicates that the edge node can cover the intersection of the business source nodes under its jurisdiction. is the set of edge nodes selected for deploying edge servers, Con(·) is the number of edge interaction service source nodes covered by the selected edge nodes, u a is the identifier symbol, when a is selected as an edge node u a is 1, otherwise it is 0, S a The set of edge interaction service source nodes managed by edge nodes, where N is the number of selected edge nodes; An edge node that can be used for interactive service processing is acquired based on the edge node selection target.

4. The edge interaction control and edge node optimization method for regulating authority transactions according to claim 3 is characterized in that: The method further comprises: Deploy edge servers at the selected edge nodes and offload edge interaction services to the edge servers.

5. The edge interaction control and edge node optimization method for regulating authority transactions according to claim 1 is characterized in that: The method of allocating the interaction amounts of all demand response resource suppliers and all demanders in the area to be optimized to the edge nodes according to the preset edge node edge interaction service allocation principle includes: Establish a relationship between the amount of edge interaction expected by the grid and the amount of edge interaction that can be performed by the demand response resources: Among them, E l,t is the expected edge interaction volume of the edge interaction service l at time t, O t is the demand response resource of the entire network during period t, Δd l,t is the executable edge interaction volume of the demand response resource side edge interaction service l in period t, Δd t is the executable edge interaction amount of all edge interaction services in period t, R t Demand response resources that do not interact; Based on the preset edge node edge interaction service allocation principle, the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized is allocated to the edge nodes. The edge node edge interaction service allocation principle is as follows: Among them, E l is the expected edge interaction volume of the power grid edge interaction service l, ζ l,t ,ξ l,t are the start and end time of the establishment of edge interaction service l, E t is the expected edge interaction volume of all edge interaction services in the power grid during period t, E avg is the control mean value of the edge interaction volume of all edge interaction services expected by the power grid on N edge nodes during period t.

6. An edge interaction control and edge node optimization system for regulating authority transactions, characterized in that: An acquisition module is used to obtain the interactive business between each demand response resource supplier and each demander in the area to be optimized; A construction module is used to establish an edge interaction model regarding the economic benefits of demand response resource suppliers and demand parties based on the interactive business between each demand response resource supplier and each demand party; A solution module is used to solve the edge interaction model with the maximum value as the goal, determine the interaction amount between each demand response resource supplier and each demand party, and obtain the interaction amount between all demand response resource suppliers and all demand parties in the area to be optimized based on the interaction amount between each demand response resource supplier and each demand party; An allocation module is used to obtain edge nodes that can be used for interactive business processing, and allocate the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized to the edge nodes according to a pre-set edge node edge interactive business allocation principle; The edge interaction model is: Among them, ρ∈[0,1], ρ is the economic benefit trade-off factor between the demand response resource supplier and the demander, p t is the real-time electricity price in the spot market p w,t and the medium- and long-term electricity price in the spot market d,t The difference, c t,i is the purchase price of the demand response resource demander in period t, Δd t,i is the interaction amount of demand response resource supplier i in period t, η∈[0,1] is the trade-off factor between the economic benefits and comfort needs of the demand response resource supplier, p r,t is the retail electricity price, The comfort costs paid by the demand response resource providers; Among them, ε t,i is the self-elasticity coefficient of demand response resource supplier i in period t, c min is the lower limit of the purchase price, Δd t-1,i is the interaction volume of demand response resource supplier i in period t-1, ξ t,t-1,i is the time-price elasticity coefficient of demand response resource supplier i in period t relative to period t-1, c t-1,i is the purchase price of the demand response resource demander in period t-1, d t,i is the electricity demand of demand response resource supplier i in period t, D max 、D min are the upper and lower limits of the interaction amount, The average purchase price for each period.

7. The edge interaction control and edge node optimization system for regulating authority transactions according to claim 6 is characterized in that: The solution module is used to The edge interaction model is solved using the multi-agent deep deterministic strategy MADDPG. The state space, action space, and reward function of each agent are the same. The MADDPG state space S includes the electricity demand d of the demand response resource supplier i in period t. t,i , self-elastic coefficient ε t,i , the time-price elasticity coefficient ξ of period t relative to period t-1 t,t-1,i , the purchase price c of the demand response resource demander in period t-1 t-1,i , expressed as: S=(d t,i ,he t,i ,x t,t-1,i ,c t-1,i ) The MADDPG action space A includes the interaction between the demand response resource supplier and the demander, and the purchase price of the demand response resource demander in period t, which can be expressed as: A=(Δd t,i ,c t,i ) The MADDPG reward function R comprehensively considers the economic benefits of both the demand response resource supplier and the demander, and is expressed as: Where, I is the set of demand response resource suppliers; H is the set of all time periods in a day; Determine the interaction amount between the demand response resource supplier i and each demander according to the reward function R; The multi-agent deep deterministic strategy MADDPG is repeatedly used to solve another edge interaction model until the interaction amount between all demand response resource suppliers and all demanders in the area to be optimized is determined.

8. The edge interaction control and edge node optimization system for regulating authority transactions according to claim 6 is characterized in that: The distribution module is used to The mobilization center node on the demand side is d, the edge interaction service source node is s, and the edge node is a, s∈S′, a∈A, S′ is the source node set, A is the edge node set, and the working path set W of (s, d), (s, a), (d, a) is s,d 、W s,a 、W d,a Complete link separation is represented as follows: Among them, ∩ is the intersection, is an empty set; The relaxation process of equations (12) and (13) is as follows: Among them, ω s,d 、ω s,a 、ω d,a are any working path set links from source node s to scheduling center node d, source node s to edge node a, and scheduling center node d to edge node a; When the working path set is not an empty set, the following constraints are met: Among them, D1 is the limit on the number of links in the working path set; The selection criteria for edge nodes are as follows: Among them, |S a | is the number of source nodes accessing edge interaction services, Indicates that the edge node can cover the intersection of the business source nodes under its jurisdiction. is the set of edge nodes selected for deploying edge servers, Con(·) is the number of edge interaction service source nodes covered by the selected edge nodes, u a is the identifier symbol, when a is selected as an edge node u a is 1, otherwise it is 0, S a The set of edge interaction service source nodes managed by edge nodes, where N is the number of selected edge nodes; An edge node that can be used for interactive service processing is acquired based on the edge node selection target.

9. The edge interaction control and edge node optimization system for regulating authority transactions according to claim 8 is characterized in that: The distribution module is also used to Deploy edge servers at the selected edge nodes and offload edge interaction services to the edge servers.

10. The edge interaction control and edge node optimization system for regulating authority transactions according to claim 6 is characterized in that: The distribution module is used to Establish a relationship between the amount of edge interaction expected by the grid and the amount of edge interaction that can be performed by the demand response resources: Among them, E l,t is the expected edge interaction volume of the edge interaction service l at time t, O t is the demand response resource of the entire network during period t, Δd l,t is the executable edge interaction volume of the demand response resource side edge interaction service l in period t, Δd t is the executable edge interaction amount of all edge interaction services in period t, R t Demand response resources that do not interact; Based on the preset edge node edge interaction service allocation principle, the interaction volume of all demand response resource suppliers and all demand parties in the area to be optimized is allocated to the edge nodes. The edge node edge interaction service allocation principle is as follows: Among them, E l is the expected edge interaction volume of the power grid edge interaction service l, ζ l,t ,ξ l,t are the start and end time of the establishment of edge interaction service l, E t is the expected edge interaction volume of all edge interaction services in the power grid during period t, E avg is the control mean value of the edge interaction volume of all edge interaction services expected by the power grid on N edge nodes during period t.

11. The edge interaction control and edge node optimization system for regulating authority transactions according to claim 6 is characterized in that: The deployment of the edge interaction model relies on the NVIDIA Jetson Xavier NX artificial intelligence chip to embed the edge interaction model into the response subject energy management system for packaging.

12. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any one of the methods according to claims 1 to 5 .

13. A computing device, characterized in that include, One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for executing any one of the methods according to claims 1 to 5.

Images