Power distribution network demand response method and device

Abstract

The invention discloses a power distribution network demand response method and device. According to the method, the demand response request of the distribution network operator, the model of each facility and the quotation information are acquired, and the demand response model is constructed based on the data. And then, decomposing the model into a plurality of sub-problems which are processed in parallel, and allocating the sub-problems to edge computing or cloud computing resources for efficient solving by utilizing preset resource performance and a demand response request. The solving result is used for generating scheduling instructions, and the instructions are sent to an energy management system of each facility to guide the energy management system to execute demand response operation and optimize resource allocation. The process not only improves the efficiency and accuracy of demand response, but also enhances the flexibility and economy of the power distribution network, and realizes reasonable allocation of resources and stable operation of the power grid.

Description

Technical Field

[0001] The present invention relates to the field of demand response in distribution networks, and particularly to a method and device for demand response in distribution networks. Background Art

[0002] The intermittent and uncertain characteristics of distributed energy make the power flow control and load balancing of distribution networks more complex. For example, solar power generation will significantly decrease at night and on cloudy days, and wind power generation is significantly affected by wind speed changes, which all increase the uncertainty of grid operation. In addition, the traditional centralized control mode adopted by distribution networks faces problems of insufficient computing resources and communication delays when dealing with a large number of distributed energy sources. Centralized control requires all data to be transmitted to the central control center for processing, which not only increases the burden of data transmission but also may lead to decision-making delays, affecting the real-time response ability of the grid.

[0003] As an effective solution, demand response technology optimizes grid operation by adjusting the load on the user side and improves the stability and economy of the grid. However, most of the existing demand response methods adopt a centralized processing mode, which exposes problems of uneven resource allocation and low market clearing efficiency in the face of high penetration of distributed energy sources. Under the centralized processing mode, all computing tasks are concentrated on the central server, resulting in tight computing resources and being unable to flexibly respond to demand changes in different regions and at different times. At the same time, the market clearing process is complex and involves multiple stakeholders, and centralized processing is difficult to achieve fast and efficient clearing, affecting the activity of the demand response market and the enthusiasm of participants. These effects lead to the inability of the existing technology to accurately and efficiently respond to the demand of the distribution network. Summary of the Invention

[0004] The present invention provides a method and device for demand response in distribution networks to solve the problem that the existing technology cannot accurately and efficiently respond to the demand of the distribution network.

[0005] In a first aspect, the present application provides a method for demand response in distribution networks, including:

[0006] Obtaining a demand response request sent by a distribution network operator, models of each facility, and quotation information of each facility;

[0007] Constructing a demand response model according to the demand response request, models of each facility, and quotation information of each facility;

[0008] Decomposing the demand response model into multiple sub-problems for parallel processing according to the demand response model;

[0009] According to the preset resource performance and the demand response request, send the sub-problems to edge computing resources or cloud computing resources for solution to obtain respective solution results;

[0010] Generate respective scheduling instructions according to the respective solution results;

[0011] Send the respective scheduling instructions to the energy management systems of each facility, so that the energy management systems perform demand response according to the respective scheduling instructions and control the allocation of resources.

[0012] In this application, by obtaining the demand response request of the distribution network operator, the models and quotation information of each facility, a demand response model is constructed, which can accurately reflect the operation state of the distribution network and the accuracy of each facility. Further, the model is decomposed into multiple sub-problems for parallel processing, and these sub-problems are allocated to edge computing resources or cloud computing resources for efficient solution according to the preset resource performance and demand response request. This decomposition and allocation strategy not only improves the computing efficiency, but also reduces the communication delay, ensuring a quick response. The solution results are used to generate scheduling instructions, which are sent to the energy management systems of each facility to guide them to perform demand response operations and optimize resource allocation. This process significantly improves the operation efficiency and stability of the distribution network, while reducing the operation cost, realizing efficient, accurate and cost-effective demand response. This application solves the problem in the prior art that the demand of the distribution network cannot be accurately and efficiently responded to.

[0013] As a preferred embodiment of the first aspect, the constructing a demand response model according to the demand response request, the models of each facility and the quotation information of each facility is specifically:

[0014] Determine a demand response market clearing model according to the demand response request;

[0015] Calculate the allocation situation of each facility and each preset distributed new energy according to the demand response market clearing model;

[0016] Calculate the cost of power interaction between the distribution network operator and the main grid and the cost of reducing the power generation of renewable power sources according to the models of each facility;

[0017] Determine the demand response overhead of each facility according to the quotation information of each facility;

[0018] Set an objective function according to the allocation situation of each facility and each preset distributed new energy, the cost of power interaction between the distribution network operator and the main grid, the cost of reducing the power generation of renewable power sources and the demand response overhead of each facility;

[0019] Solve the objective function according to the principle of minimizing the total cost preset to obtain a demand response model.

[0020] Calculate the cost of power interaction between the distribution network operator and the main grid and the cost of reducing renewable power generation according to the models of each facility, specifically:

[0021]

[0022] In the formula, π t represents the price of input or output power, θ a,t is the reduction amount of renewable power generation determined by the distribution network operator, is the cost per 1MW reduction in power generation, x 0,t represents the power exchange between the distribution network and the main grid, is the cost of power interaction between the distribution network operator and the main grid and reducing renewable power generation, T represents the set of continuous time slots, t represents the t-th time slot, y0 = {P ia,t , P aj,t , Q ia,t , Q aj,t , V a,t , x 0,t , θ a,t}, θ a,t is the reduction amount of renewable power generation determined by the distribution network operator, P ia,t is the active power flowing from node i ∈ A to node a ∈ A, Q ia,t is the reactive power flowing from node i ∈ A to node a ∈ A, x 0,t represents the power exchange between the distribution network and the main grid, V a,t represents the voltage of node a.

[0023] Calculate the allocation of each facility and each preset distributed new energy according to the demand response market clearing model, specifically:

[0024] Calculate the allocation of each facility and each preset distributed new energy according to minimizing the total system cost of the demand response market clearing model under preset constraint conditions;

[0025] Among them, the calculation formula for minimizing the total system cost of the demand response market clearing model under preset constraint conditions is:

[0026]

[0027] In the formula, d n (y n ) is the demand response cost of each facility, N is a set representing a group of flexibility facilities, and n represents a certain flexibility facility among them.

[0028] In this preferred embodiment, the present application constructs an accurate demand response model by comprehensively considering demand response requests, facility models, and bid information. First, a market clearing model is determined based on the demand response request, and then the allocation status of each facility and distributed new energy is calculated. Next, the power interaction between the distribution network operator and the main grid and the cost of reducing renewable energy generation are calculated using the facility models. Then, in combination with the facility bid information, the demand response expenses of each facility are determined. Finally, an objective function is set based on the above allocation, cost, and expenses, and the solution is obtained by following the principle of minimizing the total cost, thereby obtaining an optimized demand response model. This process not only improves the accuracy of resource allocation but also effectively reduces the operating cost, enhances the economy and stability of the distribution network, and realizes the efficiency and optimization of demand response.

[0029] As a preferred embodiment of the first aspect, according to the demand response model, the demand response model is decomposed into multiple sub-problems for parallel processing, specifically as follows:

[0030] According to the demand response model, the power consumption of the assets of each facility, the reduction amount of renewable energy generation, the power exchange between the distribution network operator and the main grid, and the distribution network power flow security constraints are obtained;

[0031] According to the power consumption of the assets of each facility, the reduction amount of renewable energy generation, the power exchange between the distribution network operator and the main grid, and the distribution network power flow security constraints, the demand response model is decomposed into sub-problems for parallel processing;

[0032] Among them, each sub-problem corresponds to the demand response operation of one or more facilities.

[0033] In this preferred embodiment, the present application accurately extracts key information by analyzing the demand response model, including the power consumption of the assets of each facility, the reduction amount of renewable energy generation, the power exchange between the distribution network operator and the main grid, and the distribution network power flow security constraints. Based on this detailed data, the model is cleverly decomposed into multiple sub-problems for parallel processing, and each sub-problem is specifically associated with the demand response operation of one or more facilities. This decomposition strategy greatly improves the computational efficiency because it allows multiple sub-problems to be processed simultaneously, making full use of the advantages of parallel computing. At the same time, by optimizing each sub-problem separately, the solution accuracy and response speed of the overall model are significantly improved, thereby enhancing the rapid adaptability of the distribution network to demand changes, ensuring the stability and economy of the grid operation, and realizing the efficient execution of demand response operations.

[0034] In the second aspect, the present application provides a device for responding to the demand of a distribution network. The device for responding to the demand of a distribution network includes an acquisition module, a construction module, a decomposition module, a solution module, and a response module;

[0035] The acquisition module is used to acquire the demand response request sent by the distribution network operator, the models of each facility, and the quotation information of each facility;

[0036] The construction module is used to construct a demand response model according to the demand response request, the models of each facility, and the quotation information of each facility;

[0037] The decomposition module is used to decompose the demand response model into multiple sub-problems for parallel processing according to the demand response model;

[0038] The solution module is used to send the sub-problems to the edge computing resources or cloud computing resources for solution according to the preset resource performance and the demand response request, and obtain each solution result;

[0039] The response module is used to generate each scheduling instruction according to each solution result;

[0040] Send each of the scheduling instructions to the energy management system of each facility, so that the energy management system executes demand response according to each of the scheduling instructions and controls the allocation of resources.

[0041] This device uses five modules to work in division of labor and coordination, which can respond to the needs of the distribution network more efficiently. This application constructs a demand response model by acquiring the demand response request of the distribution network operator, the models of each facility, and the quotation information. This model can accurately reflect the operation status of the distribution network and the accuracy of each facility. Further, the model is decomposed into multiple sub-problems for parallel processing. These sub-problems are allocated to the edge computing resources or cloud computing resources for efficient solution according to the preset resource performance and the demand response request. This decomposition and allocation strategy not only improves the computing efficiency, but also reduces the communication delay, ensuring a quick response. The solution results are used to generate scheduling instructions, which are sent to the energy management system of each facility to guide it to execute the demand response operation and optimize the resource allocation. This process significantly improves the operation efficiency and stability of the distribution network, while reducing the operation cost, realizing efficient, accurate and cost-effective demand response. This application solves the problem in the prior art that the demand of the distribution network cannot be accurately and efficiently responded to.

[0042] As a preferred embodiment of the second aspect, constructing a demand response model according to the demand response request, the models of each facility, and the quotation information of each facility specifically includes:

[0043] Determine a demand response market clearing model according to the demand response request;

[0044] Calculate the allocation situation of each facility and each preset distributed new energy according to the demand response market clearing model;

[0045] Based on the models of each facility, calculate the cost of power interaction between the distribution network operator and the main grid and the cost of reducing renewable power generation;

[0046] Based on the quotation information of each facility, determine the demand response overhead of each facility;

[0047] Based on the distribution of each facility and each preset distributed new energy, the cost of power interaction between the distribution network operator and the main grid, the cost of reducing renewable power generation, and the demand response overhead of each facility, set up the objective function;

[0048] According to the principle of minimizing the total cost preset, solve the objective function to obtain the demand response model.

[0049] In this preferred embodiment, the present application constructs an accurate demand response model by comprehensively considering the demand response request, the models of each facility and the quotation information. First, determine the market clearing model according to the demand response request, and then calculate the distribution status of each facility and distributed new energy. Then, use the models of each facility to calculate the cost of power interaction between the distribution network operator and the main grid and the cost of reducing renewable energy generation. Combine the facility quotation information to clarify the demand response overhead of each facility. Finally, set up the objective function based on the above distribution, cost and overhead, and solve it according to the principle of minimizing the total cost, so as to obtain the optimized demand response model. This process not only improves the accuracy of resource allocation, but also effectively reduces the operating cost, enhances the economy and stability of the distribution network, and realizes the efficiency and optimization of demand response.

[0050] As a preferred embodiment of the second aspect, according to the demand response model, decompose the demand response model into multiple sub-problems for parallel processing, specifically:

[0051] According to the demand response model, obtain the power consumption of the assets of each facility, the reduction amount of renewable energy generation, the power exchange amount between the distribution network operator and the main grid, and the distribution network power flow security constraint;

[0052] According to the power consumption of the assets of each facility, the reduction amount of renewable energy generation, the power exchange amount between the distribution network operator and the main grid, and the distribution network power flow security constraint, decompose the demand response model into each sub-problem for parallel processing;

[0053] Among them, each sub-problem corresponds to the demand response operation of one or more facilities.

[0054] In this preferred embodiment, the present application accurately extracts key information by analyzing the demand response model, including the power consumption of each facility asset, the reduction in renewable energy generation, the power exchange volume between the distribution network operator and the main grid, and the distribution network power flow security constraints. Based on this detailed data, the model is ingeniously decomposed into multiple sub-problems for parallel processing, and each sub-problem is specifically associated with the demand response operations of one or more facilities. This decomposition strategy greatly improves the computational efficiency because it allows multiple sub-problems to be processed simultaneously, fully leveraging the advantages of parallel computing. At the same time, by optimizing each sub-problem separately, the solution accuracy and response speed of the overall model are significantly improved, thereby enhancing the rapid adaptation ability of the distribution network to demand changes, ensuring the stability and economy of the power grid operation, and achieving the efficient execution of demand response operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 : A flowchart of an embodiment of the method for responding to the demand of the distribution network provided by the present application;

[0056] Figure 2 : A structural diagram of an embodiment of the demand response architecture based on cloud-edge collaboration provided by the present application;

[0057] Figure 3 : A structural diagram of an embodiment of the device for responding to the demand of the distribution network provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0059] Embodiment 1

[0060] Please refer to Figure 1 , a method for responding to the demand of the distribution network provided by the embodiment of the present invention.

[0061] In this embodiment, the process of the method for responding to the demand of the distribution network in the present application is described in detail through steps S01 - S05.

[0062] The information interaction process of the method for responding to the demand of the distribution network in the present application is as Figure 2 shown. The present application can orchestrate the decomposed instances of the market clearing algorithm on the available computing infrastructure. The main components of the proposed demand response architecture are:

[0063] The energy management system module of each facility is responsible for: 1) monitoring and controlling the flexibility assets of the facility; 2) allowing end-users to declare their electricity consumption preferences through a user interface; 3) communicating the demand response capabilities of the facility and receiving dispatch instructions.

[0064] The service orchestrator module provides the necessary interfaces between the platform and the facilities and the distribution system operator (DSO). It receives the requirements and specifications of the demand response market clearing problem through its interfaces.

[0065] The resource orchestrator module decomposes the clearing problem of the demand response market into smaller sub-problems and assigns subtasks to the most suitable edge or cloud computing resources.

[0066] The infrastructure management module handles the interaction with the resource orchestrator module on various computing resources, models its capabilities, and monitors them.

[0067] S01: Obtain the demand response requests sent by the distribution system operator, the models of each facility, and the quotation information of each facility.

[0068] As a preferred embodiment of Embodiment 1, the obtaining of the demand response requests sent by the distribution system operator, the models of each facility, and the quotation information of each facility is specifically as follows:

[0069] The service orchestrator obtains the demand response requests sent by the distribution system operator, the models of each facility, and the quotation information of each facility.

[0070] S02: Construct a demand response model based on the demand response request, the models of each facility, and the quotation information of each facility.

[0071] As a preferred embodiment of Embodiment 1, the constructing of the demand response model based on the demand response request, the models of each facility, and the quotation information of each facility is specifically as follows:

[0072] The service orchestrator establishes a demand response model based on the obtained information and sends it to the resource orchestrator; the demand response model is specifically as follows:

[0073] A flexibility market consists of a set of flexible facilities N = {0, 1, 2,..., |N|} (such as smart buildings, electric vehicle charging stations, energy storage facilities, etc.) and a distribution system operator (represented by 0 in the set N). Each flexible facility is located at a specific node in the distribution network and can perform demand response operations. In subsequent research and analysis, continuous time is divided into a set of time slots, denoted as T. Each facility can control the power consumption of its flexibility assets through its own energy management system (EMS: Energy management system). The set of flexibility assets of facility n is denoted as Γ n The total energy consumption of each facility in time slot t is denoted as xn,t , where a specific asset γ ∈ Γ n The energy consumption is denoted as p γ,t , from which we can obtain:

[0074]

[0075] Each facility has a set of local variables y n , including x n,t , p γ,t and other local variables depending on the specific model of the flexibility assets of the facility. Each facility has a number of feasible operating points C n , the number of which is determined by the combination of all local variables y n : to determine:

[0076]

[0077] The distribution network operator is responsible for operating the distribution network within a safe range. Suppose a distribution network has a network with A nodes and B branches, where one node a ∈ A, and Ω p (a) (or Ω d (a)) represents the upstream node (or downstream node) connected to node a. Each distribution network node a has a certain power consumption and a certain amount of renewable energy generation The distribution network operator can reduce the generation power by adjusting the coefficient θ a,t ∈ [0, 1]. When θ a,t = 0, it means no reduction of renewable energy generation; when θ a,t = 1, it means that all the renewable energy generation at node a is reduced. Let N a represent the set of facilities at node a ∈ A. To meet the distribution network power flow security constraints, a distribution network power flow model based on DistFlow is established, including equations (3) - (8):

[0078]

[0079]

[0080] In the formula: f n,t is a parameter that relates the active power and reactive power (through the power factor), P ia,t and Q ia,t are the active power and reactive power flowing from node i ∈ A to node a ∈ A, respectively. Equations (3) and (4) constrain the active and reactive power balance of each node in the distribution network, where and respectively represent the total active power and total reactive power flowing out of node a, which are equal to the sum of the powers flowing into the node of each branch minus the net consumption of active and reactive power of the node or Equation (5) describes the voltage drop relationship between each adjacent node i and a, where V a,t and V i,t respectively represent the voltages of node a and node i at time t. R ia and X ia are the resistance and reactance of branch i-a respectively. Equation (6) ensures that the voltages of all nodes are maintained within the safety margin, and equations (7) and (8) limit the active and reactive power flows of all lines within reasonable ranges

[0081] The demand response model is used to clear the distribution network demand response market, that is, to determine the output conditions of flexibility facilities, distributed new energy, etc. The distribution network operator decides the reduction amount θ a,t , and the cost of reducing 1MW of power generation is x 0,t represents the power exchange between this distribution network and the main network. The cost of the distribution network operator's power interaction with the main network and reducing renewable power generation can be calculated by the following formula

[0082]

[0083] In the formula: π t represents the price of input or output power, θ a,t is the reduction amount of renewable power generation decided by the distribution network operator, is the cost of reducing 1MW of power generation, x 0,t represents the power exchange between this distribution network and the main network, is the cost of the distribution network operator's power interaction with the main network and reducing renewable power generation, T represents the set of continuous time slots, t represents the t-th time slot, and the distribution network operator decides the reduction amount θ a,t . The adjustment coefficient θ a,t ∈[0,1] is used to reduce the power generation. When θ a,t =0, it indicates that renewable energy power generation is not reduced; when θ a,t =1, it indicates that all renewable energy power generation at node a is reduced. P ia,t and Q ia,t are the active power and reactive power flowing from node i∈A to node a∈A respectively. P ai,t and Q ai,t are the active power and reactive power flowing from node a∈A to node i∈A respectively. x 0,t represents the power exchange between the distribution network and the main network. V a,tDenotes the voltage of node a. The subscript t represents the time slot t, i.e., the t-th time slot.

[0084] The demand response cost of each facility is denoted as d n (y n ) and the function d n (·) represents the general model of the demand response cost of the corresponding assets of the facility, and this model depends on the specific assets.

[0085] The goal of the market clearing model is to ensure that the network operates within the constraints of safe operation while minimizing the total system cost (i.e., the cost of demand response and the cost of power exchange with the main grid):

[0086]

[0087] In this preferred embodiment, the present application constructs an accurate demand response model by comprehensively considering demand response requests, facility models, and bid information. First, the market clearing model is determined based on the demand response request, and then the allocation status of each facility and distributed new energy is calculated. Next, the costs of power interaction between the distribution network operator and the main grid and the curtailment of renewable energy generation are calculated using the facility models. Then, in combination with the facility bid information, the demand response cost of each facility is clarified. Finally, based on the above allocation, costs, and expenses, the objective function is set, and the solution is obtained by following the principle of minimizing the total cost, thereby obtaining an optimized demand response model. This process not only improves the accuracy of resource allocation but also effectively reduces the operating cost, enhances the economy and stability of the distribution network, and realizes the efficiency and optimization of demand response.

[0088] S03: According to the demand response model, decompose the demand response model into multiple sub-problems for parallel processing.

[0089] As a preferred embodiment of Embodiment 1, the step of decomposing the demand response model into multiple sub-problems for parallel processing according to the demand response model is specifically as follows:

[0090] It is very difficult to solve the demand response model shown in Equation (10) in a centralized manner. All models of the facilities (the cost function and operation constraints of demand response) need to be communicated to a central entity (cloud computing node), which raises concerns about security and privacy. The second problem is that a large number of variables make the computational amount of the problem very large. The Lagrangian decomposition method is used to decompose the problem (10) in a distributed manner. In the proposed method, each facility solves a local optimization problem to determine the value of its local variable y n , while the distribution network operator solves an optimal power flow problem. The augmented Lagrangian function form of the problem (10) for iterative operation using the alternating direction multiplier method is:

[0091]

[0092] In the formula:

[0093]

[0094] Define an iterative method for solving problem (10) based on the following variable update rules, corresponding to different computing tasks and the entities initiating the tasks, respectively, which are:

[0095] 1) Each facility entity:

[0096]

[0097] 2) Distribution network operator entity:

[0098]

[0099] 3) Coordination entity:

[0100]

[0101] In the formula: λ a,t and μ a,t are a set of Lagrange multipliers related to the power balance of node a; ρ1 and ρ2 are update step coefficients, generally taking values in

[0102] It can be seen that after the problem decomposition, the complex high-dimensional solution problem can be decomposed into sub-problems corresponding to each entity, and each sub-problem corresponds to a computing task. Thus, the computing of each task can be realized near the computing resource infrastructure close to each entity, or selectively uploaded to the cloud for computing and solution, which provides a basis for further optimizing the allocation of computing resources, enabling the reduction of computing costs and the improvement of demand response efficiency by formulating a reasonable allocation strategy.

[0103] In this preferred embodiment, the present application accurately extracts key information by analyzing the demand response model, including the power consumption of each facility asset, the reduction of renewable energy generation, the power exchange between the distribution network operator and the main network, and the security constraints of the distribution network flow. Based on these detailed data, the model is cleverly decomposed into multiple sub-problems for parallel processing, and each sub-problem is specifically associated with the demand response operations of one or more facilities. This decomposition strategy greatly improves the computing efficiency because it allows multiple sub-problems to be processed simultaneously, making full use of the advantages of parallel computing. At the same time, by optimizing each sub-problem separately, the solution accuracy and response speed of the overall model are significantly improved, thereby enhancing the rapid adaptation ability of the distribution network to demand changes, ensuring the stability and economy of the grid operation, and realizing the efficient execution of demand response operations.

[0104] S04: According to the preset resource performance and the demand response request, send the sub-problem to the edge computing resource or the cloud computing resource for solution to obtain each solution result.

[0105] As a preferred embodiment of Embodiment 1, the step of sending the sub-problem to the edge computing resource or the cloud computing resource for solution according to the preset resource performance and the demand response request to obtain each solution result is specifically as follows:

[0106] Each computational problem of demand response is decomposed into N + 1 sub-problems, and each sub-problem corresponds to a local optimization problem of a facility (i.e., Equation 14), which can be regarded as different computational tasks. Thus, each demand response request r ∈ R is characterized by its arrival time α r , its task set (i.e., the set of facilities), and the upper limit of the waiting time allowed for each iteration.

[0107] Each task needs to transmit input data and output data These tasks can be executed in the edge computing resource, which is close to or located at the facility that generates these tasks; or be forwarded to be executed in the cloud computing, and a delay lat will be generated in this process f . Aggregating multiple tasks in the cloud computing resource can reduce the total cost of operations, but will increase the overall execution time of the tasks due to the introduction of network latency.

[0108] Use G = (V, E) to jointly represent the flexible facilities, computing, and network infrastructure. The node set V = V c ∪V f includes the nodes V with computing resources c and the nodes V with demand response facilities f : f ∈ F r , r ∈ R, F r is the set of computational sub-tasks of each facility in the demand response request r as shown in Equation (14). Simply put, G is a graph model of a communication network, including V c and V f two types of nodes. Each facility node V f will also be configured with certain computing resources. Therefore the set of edges E in the graph is used to represent the virtual links that interconnect the nodes by wired and wireless communication paths. Use M to represent the set of types of communication resources. Different facility types located at the node v ∈ V have different task computing processing overheads Each communication link (i - j) ∈ E has different network overheads and its transmission rate is denoted as tr i,j .

[0109] The optimization objective of computing resource allocation is to minimize the processing cost per iteration and the sum of the delays of all demand response requests in the demand response request set R, while satisfying the time constraint of each request. The index f ∈ F is used to represent the facilities of the demand response requests and their computing tasks, where F = ∪ r∈R {F r}, and thus the optimization objective model of computing resources is as follows:

[0110]

[0111] In the formula: where the integer variable ψ v,m,f represents the time slot when the computing resource type m of the demand response facility located at node v starts to process task f (the resource types include edge computing and cloud computing, and different types have different computing capabilities, and the computing power of cloud computing is greater than that of edge computing); the binary variable ζ v,m,f represents whether there is m-type computing resource at node v to process computing task f; the binary variable represents whether the facility corresponding to task f has a communication connection with node n; the binary variable represents whether task f is executed before task . pr v,m,f represents the computing time when computing resource m processes computing task f. w is an optimization objective weight coefficient, w ∈ [0, 1]. When w = 0, the demand response request has the minimum processing delay; when w = 1, the computing cost of each iteration calculation is the smallest.

[0112] During the computing task allocation process, in order to allocate task f to the computing resource (v, m), a connection path must be selected, that is:

[0113]

[0114] The multiplier update task is performed by the distribution network operator. Let represent the node where the distribution network operation of the demand response request r is located, represent the multiplier update computing task, and each facility can send data to the distribution network operator, that is:

[0115]

[0116] Node v cannot execute task f after receiving the input data of f , so the constraint needs to be satisfied:

[0117]

[0118] In the formula: Φ represents the propagation speed of data on the communication link, represents node v and vf The shortest path length between them, where Q is a sufficiently large value. The DSO cannot update the multiplier until it receives the responses of each computing task. Therefore:

[0119]

[0120] After the operator update is completed, a distributed computing iteration needs to be completed. Therefore, the time constraint for each iteration is:

[0121]

[0122] In the formula: represents the maximum acceptable delay of the demand response request r.

[0123] S05: Generate each scheduling instruction according to the respective solution results;

[0124] Send the respective scheduling instructions to the energy management systems of each facility, so that the energy management systems execute demand response according to the respective scheduling instructions and control the allocation of resources.

[0125] This application constructs a demand response model by obtaining the demand response requests of the distribution network operator, the models and quotation information of each facility. This model can accurately reflect the operating state of the distribution network and the accuracy of each facility. Further, the model is decomposed into multiple sub-problems for parallel processing, and these sub-problems are allocated to edge computing resources or cloud computing resources for efficient solution according to the preset resource performance and demand response requests. This decomposition and allocation strategy not only improves the computing efficiency but also reduces the communication delay, ensuring a quick response. The solution results are used to generate scheduling instructions, which are sent to the energy management systems of each facility to guide them to perform demand response operations and optimize resource allocation. This process significantly improves the operating efficiency and stability of the distribution network, while reducing the operating cost, achieving efficient, accurate and cost-effective demand response. This application solves the problem in the prior art that the demand of the distribution network cannot be accurately and efficiently responded to.

[0126] Embodiment 2

[0127] Please refer to Figure 3 , a device for responding to the demand of the distribution network provided by the embodiment of this application.

[0128] In this embodiment, the device for responding to the demand of the distribution network includes an acquisition module 10, a construction module 20, a decomposition module 30, a solution module 40 and a response module 50.

[0129] The information interaction process of the method for responding to the demand of the distribution network in this application is as Figure 2As shown, the present application can orchestrate decomposed instances of market clearing algorithms on available computing infrastructure. The main components of the proposed demand response architecture are:

[0130] The energy management system module of each facility is responsible for: 1) monitoring and controlling the flexibility assets of the facility; 2) allowing end-users to declare their electricity consumption preferences through a user interface; 3) communicating the demand response capabilities of the facility and receiving scheduling instructions.

[0131] The service orchestrator module provides the necessary interfaces between the platform, the facilities, and the distribution system operator (DSO). It receives the requirements and specifications of the demand response market clearing problem through its interfaces.

[0132] The resource orchestrator module decomposes the clearing problem of the demand response market into smaller sub-problems and assigns subtasks to the most suitable edge or cloud computing resources.

[0133] The infrastructure management module handles the interaction with the resource orchestrator module on various computing resources, models their capabilities, and monitors them.

[0134] The acquisition module 10 is used to acquire the demand response requests sent by the distribution system operator, the models of each facility, and the quotation information of each facility.

[0135] As a preferred embodiment of the second embodiment, the acquisition of the demand response requests sent by the distribution system operator, the models of each facility, and the quotation information of each facility is specifically:

[0136] The service orchestrator acquires the demand response requests sent by the distribution system operator, the models of each facility, and the quotation information of each facility.

[0137] The construction module 20 is used to construct a demand response model according to the demand response requests, the models of each facility, and the quotation information of each facility.

[0138] As a preferred embodiment of the second embodiment, the construction of the demand response model according to the demand response requests, the models of each facility, and the quotation information of each facility is specifically:

[0139] The service orchestrator establishes a demand response model based on the acquired information and sends it to the resource orchestrator; the demand response model is specifically:

[0140] A flexibility market consists of a set of flexible facilities \(N = \{0, 1, 2,\cdots,|N|\}\) (such as smart buildings, electric vehicle charging stations, energy storage facilities, etc.) and a distribution network operator (represented by 0 in the set \(N\)). Each flexible facility is located at a specific node in the distribution network and can perform demand response operations. In subsequent research analysis, continuous time is divided into a set of time slots, denoted as \(T\). Each facility can control the power consumption of its flexibility assets through its own energy management system (EMS: Energy management system). The set of flexibility assets of facility \(n\) is denoted as \(\Gamma\). n The total energy consumption of each facility within time slot \(t\) is denoted as \(x\). n,t Among them, the energy consumption of a specific asset \(\gamma\in\Gamma\). n is denoted as \(p\). γ,t Thus, we can obtain:

[0141]

[0142] Each facility has a set of local variables \(y\). n , including \(x\). n,t , \(p\). γ,t and other local variables that depend on the specific model of the facility's flexibility assets. Each facility has a number of feasible operating points \(C\). n The number of which is determined by the combination of all local variables \(y\). n : as follows:

[0143]

[0144] The distribution network operator is responsible for operating the distribution network within a safe range. Suppose a distribution network has a network with \(A\) nodes and \(B\) branches. Among them, a node \(a\in A\), denoted by \(\Omega\). p (a) (or \(\Omega\). d (a)) represents the upstream node (or downstream node) connected to node \(a\). Each distribution network node \(a\) has a certain power consumption and a certain amount of renewable energy generation The distribution network operator can reduce the generation power by adjusting the coefficient \(\theta\). a,t \(\in[0, 1]\). When \(\theta\). a,t = 0, it indicates that no renewable energy generation is reduced; when \(\theta\). a,t = 1, it indicates that all renewable energy generation at node \(a\) is reduced. Denote the set of facilities at node \(a\in A\) by \(N\). a To satisfy the distribution network power flow security constraints, a distribution network power flow model based on DistFlow is established, including equations (26) - (31):

[0145]

[0146] In the formula: f n,t is a parameter that relates the active power and the reactive power (through the power factor), P ia,t and Q ia,t are respectively the active power and the reactive power flowing from node i ∈ A to node a ∈ A. Equations (26) and (27) constrain the active and reactive power balance of each node in the distribution network, where and respectively represent the total outgoing active power and the total outgoing reactive power of node a, which are equal to the sum of the powers flowing into the node of each branch minus the net consumption of active and reactive power of the node or Equation (28) describes the voltage drop relationship between adjacent nodes i and a, where V a,t and V i,t respectively represent the voltages of node a and node i at time t. R ia and X ia are respectively the resistance and reactance of branch i-a. Equation (29) ensures that the voltages of all nodes are maintained within the safety margin, and equations (30) and (31) respectively limit the active and reactive power flows of all lines within a reasonable range.

[0147] The demand response model is used to clear the distribution network demand response market, that is, to determine the output conditions of flexibility facilities, distributed new energy, etc. The distribution network operator decides the reduction amount θ a,t of the renewable power generation. The cost of reducing 1 MW of power generation is x 0,t represents the power exchange between this distribution network and the main grid. The cost of the distribution network operator's power interaction with the main grid and reducing the renewable power generation can be calculated by the following formula:

[0148]

[0149] In the formula: π t represents the price of the input or output power, θ a,t is the reduction amount of the renewable power generation decided by the distribution network operator, is the cost of reducing 1 MW of power generation, x 0,t represents the power exchange between this distribution network and the main grid, is the cost of the distribution network operator's power interaction with the main grid and reducing the renewable power generation. T represents the set of continuous time slots, t represents the t-th time slot, and the distribution network operator decides the reduction amount θ a,t of the renewable power generation. The adjustment coefficient θ a,t ∈ [0,1] is used to reduce the power generation. When θ a,t = 0, it indicates that the renewable energy power generation is not reduced; when θ a,tWhen = 1, it indicates that all the renewable energy generation of node a is curtailed. P ia,t and Q ia,t are the active power and reactive power flowing from node i ∈ A to node a ∈ A respectively. P ai,t and Q ai,t are the active power and reactive power flowing from node a ∈ A to node i ∈ A respectively. x 0,t represents the power exchange between the distribution network and the main network. V a,t represents the voltage of node a. The subscript t represents the time slot t, i.e., the t-th time slot.

[0150] The demand response cost of each facility is denoted as d n (y n ), and the function d n (·) represents the general model of the demand response cost of the corresponding assets of the facility, and this model depends on the specific assets.

[0151] The goal of the market clearing model is to ensure that the network operates within the safety operation constraints while minimizing the total system cost (i.e., the cost of demand response and the cost of power exchange with the main network:

[0152]

[0153] In this preferred embodiment, the present application constructs an accurate demand response model by comprehensively considering demand response requests, facility models, and quotation information. First, the market clearing model is determined based on the demand response request, and then the allocation status of each facility and distributed new energy is calculated. Next, the cost of power interaction between the distribution network operator and the main network and the cost of curtailing renewable energy generation are calculated using the facility models. Then, in combination with the facility quotation information, the demand response cost of each facility is determined. Finally, based on the above allocation, cost, and cost, the objective function is set and solved following the principle of minimizing the total cost, so as to obtain the optimized demand response model. This process not only improves the accuracy of resource allocation, but also effectively reduces the operating cost, enhances the economy and stability of the distribution network, and realizes the efficiency and optimization of demand response.

[0154] The decomposition module 30 is used to decompose the demand response model into multiple sub-problems for parallel processing according to the demand response model.

[0155] As a preferred embodiment of the second embodiment, the decomposing the demand response model into multiple sub-problems for parallel processing according to the demand response model is specifically:

[0156] It is very difficult to solve the demand response model shown in Equation (33) in a centralized manner. All models of the facilities (the cost function and operation constraints of demand response) need to be communicated to a central entity (cloud computing node), which raises concerns about security and privacy. The second problem is that a large number of variables make the computational workload of the problem very large. The Lagrangian decomposition method is used to decompose Problem (33) in a distributed manner. In the proposed method, each facility solves a local optimization problem to determine its local variable y n value, while the distribution network operator solves an optimal power flow problem. The alternating direction multiplier method is used for iterative operation, and the augmented Lagrangian function form of Problem (33) is:

[0157]

[0158] where:

[0159]

[0160] An iterative method for solving Problem (33) is defined based on the following variable update rules, corresponding to different computational tasks and the entities initiating the tasks, respectively, as follows:

[0161] 1) Each facility entity:

[0162]

[0163] 2) Distribution network operator entity:

[0164]

[0165] 3) Coordination entity:

[0166]

[0167] where: λ a,t and μ a,t are a set of Lagrangian multipliers related to the power balance of node a; ρ1 and ρ2 are update step coefficients, generally taking values in

[0168] It can be seen that after the problem decomposition, the complex high-dimensional solution problem can be decomposed into sub-problems corresponding to each entity, and each sub-problem corresponds to a computational task. Thus, the calculations of their respective tasks can be realized near the computational resource infrastructure close to each entity, or selectively uploaded to the cloud for computational solution, which provides a basis for further optimizing the allocation of computational resources, enabling the reduction of computational costs and the improvement of demand response efficiency by formulating a reasonable allocation strategy.

[0169] In this preferred embodiment, the present application accurately extracts key information by analyzing the demand response model, including the power consumption of each facility asset, the reduction in renewable energy generation, the power exchange between the distribution network operator and the main grid, and the distribution network power flow security constraints. Based on this detailed data, the model is ingeniously decomposed into multiple sub-problems for parallel processing, and each sub-problem is specifically associated with the demand response operations of one or more facilities. This decomposition strategy greatly improves the computational efficiency because it allows multiple sub-problems to be processed simultaneously, fully leveraging the advantages of parallel computing. At the same time, by optimizing each sub-problem separately, the solution accuracy and response speed of the overall model are significantly improved, thereby enhancing the rapid adaptability of the distribution network to demand changes, ensuring the stability and economy of the power grid operation, and achieving the efficient execution of demand response operations.

[0170] The solving module 40 is used to send the sub-problems to edge computing resources or cloud computing resources for solution according to the preset resource performance and the demand response request, and obtain each solution result.

[0171] As a preferred embodiment of the second embodiment, the step of sending the sub-problems to edge computing resources or cloud computing resources for solution according to the preset resource performance and the demand response request, and obtaining each solution result is specifically as follows:

[0172] Each computational problem of demand response is decomposed into N + 1 sub-problems, and each sub-problem corresponds to a local optimization problem of a facility (i.e., Equation 37), which can be regarded as different computational tasks. Thus, each demand response request r ∈ R is characterized by its arrival time α r , its task set (i.e., the set of facilities), and the upper limit of the waiting time allowed for each iteration.

[0173] Each task needs to transmit input data and output data These tasks can be executed in edge computing resources, which are close to or located at the facilities that generate these tasks; or be forwarded to be executed in the cloud computing, and a delay lat f will be generated in this process. Aggregating multiple tasks in cloud computing resources can reduce the total cost of operations, but will increase the overall execution time of the tasks due to the introduction of network latency.

[0174] Use G = (V, E) to jointly represent flexible facilities, computing, and network infrastructure. The node set V = V c ∪V f contains nodes V with computing resources c and nodes V with demand response facilities f : f ∈ F r , r ∈ R, F rIt is the set of sub - tasks for each facility calculation in the demand response request r as shown in Equation (37). Briefly, G is a graph model of a communication network, including V c and V f two types of nodes. Each facility node V f will also be configured with a certain amount of computing resources. Therefore the set of edges E in the graph is used to represent the virtual links that interconnect nodes through wired and wireless communication paths. Let M represent the set of types of communication resources. Different facility types located at node v ∈ V have different task - computing processing overheads Each communication link (i - j) ∈ E has different network overheads and its transmission rate is denoted as tr i,j .

[0175] The optimization goal of computing resource allocation is to minimize the sum of the processing cost per iteration and the latency of all demand response requests in the demand response request R, while meeting the time constraints of each request. Use the index f ∈ F to represent the facilities and their computing tasks in the demand response request, where F = ∪ r∈R {F r}. Thus, the optimization goal model of computing resources can be obtained as follows:

[0176]

[0177] In the formula: where the integer variable ψ v,m,f represents the time slot when the computing resource type m of the demand response facility located at node v starts to process task f (the resource types include edge computing and cloud computing. Different types have different computing capabilities, and the computing power of cloud computing is greater than that of edge computing); the binary variable ζ v,m,f represents whether there is an m - type computing resource at node v to process the computing task f; the binary variable represents whether the facility corresponding to task f has a communication connection with node n; the binary variable represents whether task f is executed before task . pr v,m,f represents the computing time when computing resource m processes computing task f. w is an optimization goal weight coefficient, w ∈ [0, 1]. When w = 0, the demand response request has the minimum processing latency; when w = 1, the computing cost overhead of each iteration calculation is the smallest.

[0178] During the computing task allocation process, in order to allocate task f to the computing resource (v, m), a connection path must be selected, that is:

[0179]

[0180] The multiplier update task is carried out by the distribution network operator. Let The node where the distribution network operation corresponding to the demand response request r is located Denote the multiplier update calculation task, and each facility can send data to the distribution network operator, that is:

[0181]

[0182] After node v receives the input data of f It cannot execute task f, so the constraint needs to be satisfied:

[0183]

[0184] In the formula: Φ represents the propagation speed of data on the communication link Denote the shortest path length between node v and v f DSO cannot update the multiplier before receiving the responses of each calculation task, so:

[0185]

[0186] After the operator update is completed, a distributed computing iteration needs to be completed. Therefore, the time constraint for each iteration is:

[0187]

[0188] In the formula: Denote the maximum acceptable delay of the demand response request r

[0189] The response module 50 is used to generate each scheduling instruction according to the respective solution results;

[0190] Send the respective scheduling instructions to the energy management systems of each facility, so that the energy management systems execute demand response according to the respective scheduling instructions and control the allocation of resources

[0191] This application constructs a demand response model by obtaining the demand response requests of distribution network operators, the models and quotation information of various facilities. This model can accurately reflect the operating status of the distribution network and the accuracy of each facility. Further, the model is decomposed into multiple sub-problems for parallel processing. These sub-problems are allocated to edge computing resources or cloud computing resources for efficient solution according to the preset resource performance and demand response requests. This decomposition and allocation strategy not only improves the computing efficiency but also reduces the communication delay, ensuring a quick response. The solution results are used to generate scheduling instructions, which are sent to the energy management systems of each facility to guide them to perform demand response operations and optimize resource allocation. This process significantly improves the operating efficiency and stability of the distribution network, while reducing the operating costs, achieving efficient, accurate and cost-effective demand response. This application solves the problem in the prior art that the demand of the distribution network cannot be accurately and efficiently responded to.

[0192] The specific embodiments described above further elaborate on the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. In particular, for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for responding to distribution network demand, characterized in that: include: Obtain demand response requests sent by distribution network operators, models of each facility, and quotation information of each facility; constructing a demand response model according to the demand response request, the models of each facility and the quotation information of each facility; According to the demand response model, decomposing the demand response model into a plurality of sub-problems to be processed in parallel; According to the preset resource performance and the demand response request, the sub-problems are sent to edge computing resources or cloud computing resources for solving, and various solution results are obtained; Generate various scheduling instructions according to the various solution results; The various scheduling instructions are sent to the energy management system of each facility, so that the energy management system executes demand response and controls resource allocation according to the various scheduling instructions.

2. The method for responding to distribution network demand according to claim 1, characterized in that: The demand response model is constructed according to the demand response request, the model of each facility and the quotation information of each facility, specifically: Determining a demand response market clearing model according to the demand response request; According to the demand response market clearing model, the allocation of each facility and each preset distributed renewable energy source is calculated; Based on the model of each facility, the cost of power interaction between the distribution network operator and the main grid and the cost of reducing renewable power generation are calculated; Determine the demand response expenses of each facility based on the quotation information of each facility; Setting the objective function based on the allocation of each facility and each preset distributed renewable energy source, the cost of power interaction between the distribution network operator and the main grid, the cost of reducing renewable power generation and the demand response overhead of each facility; According to the preset principle of minimizing the total cost, the objective function is solved to obtain a demand response model.

3. The method for responding to distribution network demand according to claim 2, characterized in that: According to the model of each facility, the cost of power interaction between the distribution network operator and the main grid and the cost of reducing renewable power generation are calculated, specifically: In the formula, π t represents the price of input or output power, θ a,t To determine the reduction in renewable energy generation for distribution network operators, For every 1MW reduction in power generation, x 0,t Indicates the power exchange between the distribution network and the main network. To facilitate power interaction between distribution network operators and the main grid and reduce the cost of renewable power generation, T represents the set of continuous time slots, t represents the tth time slot, y0 = {P ia,t ,P aj,t ,Q ia,t ,Q aj,t ,V a,t ,x 0,t ,θ a,t }, θ a,t The distribution network operator determines the amount of reduction in renewable power generation, P ia,t is the active power flowing from node i∈A to node a∈A, Q ia,t is the reactive power flowing from node i∈A to node a∈A, x 0,t Indicates the power exchange between the distribution network and the main network, V a,t Represents the voltage at node a.

4. The method for responding to distribution network demand according to claim 2, characterized in that: The distribution of each facility and each preset distributed renewable energy source is calculated according to the demand response market clearing model, specifically: According to minimizing the total system cost of the demand response market clearing model under preset constraints, the allocation of each facility and each preset distributed renewable energy source is calculated; The calculation formula for minimizing the total system cost of the demand response market clearing model under the preset constraints is: Where, d n (y n ) is the demand response cost of each facility, N is a set of flexibility facilities, and n represents a certain flexibility facility among them.

5. The method for responding to distribution network demand according to claim 1, characterized in that: According to the demand response model, the demand response model is decomposed into a plurality of sub-problems to be processed in parallel, specifically: According to the demand response model, the power consumption of the assets of each facility, the reduction amount of renewable energy generation, the power exchange amount between the distribution network operator and the main grid, and the distribution network power flow security constraints are obtained; Decomposing the demand response model into various sub-problems to be processed in parallel according to the power consumption of the assets of the various facilities, the curtailment of renewable energy generation, the power exchange between the distribution network operator and the main grid, and the power flow security constraints of the distribution network; Each sub-question corresponds to the demand response operation of one or more facilities.

6. A distribution network demand response device, characterized in that: It includes acquisition module, construction module, decomposition module, solution module and response module; The acquisition module is used to obtain the demand response request sent by the distribution network operator, the model of each facility and the quotation information of each facility; The construction module is used to construct a demand response model according to the demand response request, the model of each facility and the quotation information of each facility; The decomposition module is used to decompose the demand response model into multiple sub-problems to be processed in parallel according to the demand response model; The solution module is used to send the sub-problems to edge computing resources or cloud computing resources for solution according to preset resource performance and the demand response request to obtain various solution results; The response module is used to generate various scheduling instructions according to the various solution results; The various scheduling instructions are sent to the energy management system of each facility, so that the energy management system executes demand response and controls resource allocation according to the various scheduling instructions.

7. The device for responding to distribution network demand according to claim 6, characterized in that: The construction module is used to construct a demand response model according to the demand response request, the model of each facility and the quotation information of each facility, specifically: Determining a demand response market clearing model according to the demand response request; According to the demand response market clearing model, the allocation of each facility and each preset distributed renewable energy source is calculated; Based on the model of each facility, the cost of power interaction between the distribution network operator and the main grid and the cost of reducing renewable power generation are calculated; Determine the demand response expenses of each facility based on the quotation information of each facility; Setting the objective function based on the allocation of each facility and each preset distributed renewable energy source, the cost of power interaction between the distribution network operator and the main grid, the cost of reducing renewable power generation and the demand response overhead of each facility; According to the preset principle of minimizing the total cost, the objective function is solved to obtain a demand response model.

8. The device for responding to distribution network demand according to claim 6, characterized in that: The decomposition module is used to decompose the demand response model into multiple sub-problems to be processed in parallel according to the demand response model, specifically: The decomposition module is used to obtain the power consumption of the assets of each facility, the reduction amount of renewable energy power generation, the power exchange amount between the distribution network operator and the main grid, and the distribution network power flow security constraints according to the demand response model; Decomposing the demand response model into various sub-problems to be processed in parallel according to the power consumption of the assets of the various facilities, the curtailment of renewable energy generation, the power exchange between the distribution network operator and the main grid, and the power flow security constraints of the distribution network; Each sub-question corresponds to the demand response operation of one or more facilities.

Images