Emergency demand response resource optimal allocation method under hierarchical aggregation architecture

Abstract

The invention discloses an emergency demand response resource optimal allocation method under a hierarchical aggregation architecture, which comprises the following steps of: by constructing the hierarchical aggregation architecture, dividing flexible resources into a plurality of regions according to geographic positions and communication characteristics, and enabling a region coordination center to be responsible for local power allocation and load scheduling in each region; a power grid dispatching center is responsible for coordinated management of global demand targets, dispatching complexity is remarkably reduced, real-time performance is improved, an optimization model adopts a multi-target optimization method, power response cost, response reliability and time delay penalty are integrated, and through power balance, capability constraint and time constraint, accuracy and economical efficiency of an allocation scheme are ensured, and the allocation efficiency is improved. According to the method, rapid adjustment of resource state abnormity, response deviation or node faults is realized through dynamic adjustment and a distributed consensus algorithm, the surplus capacity in the region is preferentially utilized to compensate the insufficient demand, and redundant power support is scheduled through an inter-region coordination mechanism when necessary.

Description

Technical Field

[0001] The present invention relates to the technical field of power system regulation and allocation, and in particular to an emergency demand response resource optimization allocation method under a layered aggregation architecture. Background Art

[0002] With the development of new power systems, the large-scale access of renewable energy has significantly increased the uncertainty and volatility of power system operation. The backup power supply and peak shaving methods in traditional power systems are difficult to effectively deal with the power shortage problem in emergency demand response scenarios. Flexible resources (industrial loads, residential loads, energy storage equipment and distributed power sources) have gradually become an important part of demand response in power systems due to their flexible regulation and rapid response. However, the existing emergency demand response has the following shortcomings:

[0003] First, the current centralized dispatching architecture requires the grid dispatching center to directly manage all flexible resources. The dispatching complexity increases linearly with the resource scale, resulting in increased response delays, making it difficult to meet the high requirements for real-time performance and accuracy in emergency demand response scenarios.

[0004] Second, flexible resources have significant differences in physical characteristics, responsiveness, and communication delays. Existing methods fail to fully consider the heterogeneous characteristics of resources in power allocation, resulting in a lack of accuracy and economy in allocation schemes.

[0005] 3. During the emergency demand response process, the resource status undergoes sudden changes such as communication interruption, node failure or load fluctuation. The existing technology lacks efficient adjustment in dealing with sudden changes, which can easily lead to insufficient power response or imbalanced distribution, thus affecting the safe operation of the power grid.

[0006] Based on this, an optimal allocation method for emergency demand response resources under a hierarchical aggregation architecture is needed. Summary of the invention

[0007] To achieve the above object, the present invention provides the following solution: a method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture, comprising the following steps:

[0008] Step S1: Partition and hierarchical architecture construction:

[0009] Divide the flexible resources into several regions according to their geographical locations and communication characteristics. Set up a regional coordination center in each region. The regional coordination center is connected to the power grid dispatching center through a communication interface to form a hierarchical and partitioned power distribution architecture.

[0010] Step S2: Receiving power demand information and real-time data collection:

[0011] The power grid dispatching center sends power shortage demand to each regional coordination center, and each regional coordination center collects the available power range, response time, operating status and historical load operation data of the local load node;

[0012] Step S3: Optimization model building:

[0013] According to the power response requirements, an optimization model is established with the goals of minimizing power response cost, minimizing response time delay, and maximizing response reliability, and power balance constraints, resource capacity constraints, and dynamic response constraints are set to ensure the practical feasibility of the power allocation solution;

[0014] Step S4: Execution of the distributed consensus algorithm:

[0015] In each regional coordination center, a distributed consensus algorithm is used to solve the optimization model and generate a preliminary power allocation plan. The dynamic initialization strategy is used to reduce the deviation of the initial plan, and the learning rate is adjusted in combination with gradient prediction to quickly achieve algorithm convergence. In the event of communication interruption or node failure, the power allocation result is dynamically adjusted through a fault-tolerant mechanism.

[0016] Step S5: Generation and issuance of allocation instructions:

[0017] The regional coordination center generates specific power allocation instructions to each load node in the region based on the optimization results of the distributed consensus algorithm, and sends the instructions to each load node for execution through the communication network;

[0018] Step S6: Real-time monitoring and feedback optimization of execution status:

[0019] After the load node executes the power allocation instruction, the regional coordination center collects the response data of each node in real time and compares it with the expected response result; if deviation is found, the power allocation plan is adjusted through dynamic optimization strategy, and the revised allocation instruction is reissued to ensure that the final response meets the real-time needs of the power grid.

[0020] Further preferably, the hierarchical partitioning architecture includes a three-layer structure of a power grid dispatching center, a regional coordination center and local load nodes, wherein the power grid dispatching center is responsible for calculating the global power shortage demand and sending it to each regional coordination center, and at the same time receiving the operating status and response results fed back from each region. The regional coordination center analyzes the operating status of the local load nodes, performs local optimization based on the demand sent by the power grid dispatching center, and sends power allocation instructions to the load nodes. After receiving the instructions, the local load nodes adjust the power according to the actual capacity, and feed back the response status to the regional coordination center in real time.

[0021] Further preferably, the multi-objective optimization model takes minimization of power response cost, minimization of response time delay and maximization of response reliability as optimization objectives, and its objective function is:

[0022]

[0023] Where: Power response cost C i (P i ) is used to quantify the power response cost of resource i and is defined as where c i and k i are the linear and quadratic cost coefficients, respectively;

[0024] Response reliability R i Indicates the actual power response value of the resource and target power response value The deviation between them is expressed as:

[0025]

[0026] The time delay penalty term Penalty(t) is used to reflect the resource response time exceeding the target time t target The resulting cost is defined by the formula:

[0027]

[0028] Among them, k is the time penalty coefficient, α i and λ are the weights of response reliability and time penalty, respectively, which are used to adjust the priority relationship between optimization objectives.

[0029] Further preferably, the multi-objective optimization model is subject to the following constraints, including power balance constraints, resource capacity constraints and dynamic response constraints:

[0030] The power balance constraint ensures that the sum of the power responses of all flexible resources is equal to the power deficit requirement set by the grid dispatch center, which is defined as

[0031] The resource capacity constraint is used to limit the power response of each resource to not exceed its maximum physical capacity, which is defined as 0≤P i ≤P i,max , where P i,max is the maximum response power of resource i;

[0032] The dynamic response constraint: t response,i ≤T max , stipulates that the response time of all flexible resources shall not exceed the preset time limit T max .

[0033] Further preferably, the distributed consensus algorithm realizes the rapid solution of the power allocation scheme through a dynamic initialization strategy, a gradient prediction and step size initialization strategy, and a fault tolerance mechanism, wherein the dynamic initialization strategy is used to generate an initial power allocation scheme by combining historical data with real-time load status, and its formula is: in is the initial value of power allocation predicted based on historical data, δ represents the correction value calculated according to the current load state, which is used to correct the deviation between the historical prediction and the actual load;

[0034] The gradient prediction and step size initialization strategy calculates the initial step size by analyzing the initial gradient value of the objective function. The formula is:

[0035]

[0036] in, Represents the initial gradient norm, ∈ is a positive number to avoid zero denominator. The fault-tolerant mechanism monitors the status of load nodes in real time. When a node failure or communication interruption is detected, it dynamically reallocates the power of the faulty node to the normal node and re-optimizes the power allocation after the fault is recovered.

[0037] Further preferably, the design of the dynamic initialization strategy generates an initial power allocation scheme by statistically analyzing the correlation between historical load data and real-time operating status, including:

[0038] The initial power response value of each load node is fitted through historical operation data Reflects the expected response capability of the node under similar operating conditions;

[0039] The correction value δ is calculated based on the real-time operating status of the load node. The correction value is obtained by comprehensive calculation based on the real-time load change rate, the node available power range and the historical response deviation range. Differences from actual operating status;

[0040] Using the formula Generate the final initial power allocation plan.

[0041] Further preferably, the gradient prediction and dynamic step size adjustment strategy accelerates the convergence of the optimization algorithm by calculating the objective function gradient value in real time and dynamically adjusting the step size. First, in each iteration process, according to the current power allocation scheme P (t) Calculate the gradient of the objective function The gradient value is used to determine the current optimization direction; secondly, the step size is dynamically adjusted according to the gradient change rate. The step size adjustment formula is:

[0042]

[0043] where η (t) is the current step length; represents the change between the current gradient and the previous iteration gradient, β is the step size adjustment coefficient, and then, by setting the step size boundary control range [η min ,η max ], ensuring that the optimization algorithm can quickly converge to the global optimal solution in dynamically changing scenarios.

[0044] Further preferably, the regional coordination center collects the actual response data of the local load nodes in real time, including the actual power response value of the node Response time t response,i and operating status data, comparing the actual response results with the expected response values ​​of the optimization model and the target response time t target A comparison is performed, and when it is found that the response deviation exceeds the preset threshold or the node operation status is abnormal, the strategy is dynamically adjusted to reallocate the power response plan.

[0045] Further preferably, the dynamic adjustment strategy includes giving priority to using the remaining power in the region for compensation. If the demand cannot be met in the region, redundant power support is requested through an inter-regional coordination mechanism, and max Complete within the time limit.

[0046] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0047] 1. The present invention divides flexible resources into several areas according to geographical location and communication characteristics. A regional coordination center is set up in each area to be responsible for local power scheduling. The regional coordination center independently handles the optimal allocation and real-time adjustment of local loads, communicates with the power grid dispatching center, receives global demands and feedbacks regional status. This architecture effectively reduces the complexity of centralized dispatching to directly control all resources, reduces communication delays, and greatly improves the overall efficiency in emergency demand response scenarios.

[0048] 2. The present invention adopts a multi-objective optimization method to construct a resource allocation model, with the goals of minimizing power response cost, maximizing response reliability and minimizing delay penalty. Through power balance constraints, capacity constraints and time constraints, the economy and practical feasibility of the optimization results are ensured, the accuracy of the allocation plan is improved, and the participation needs of different types of flexible resources are adapted.

[0049] 3. During the power response process, the present invention monitors the status of load nodes in real time. When response deviation, delay or abnormality is found, a dynamic adjustment mechanism is triggered to give priority to using the remaining capacity in the area for compensation. If it is still insufficient, redundant power support is mobilized through the inter-regional coordination mechanism to achieve cross-regional collaboration. At the same time, the fault-tolerant mechanism is used to re-optimize the allocation plan when a node fails or communication is interrupted. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 It is the overall flow chart of the present invention;

[0052] Figure 2 This is a schematic diagram of the hierarchical aggregation architecture of the present invention;

[0053] Figure 3 This is a flowchart of the distributed consensus algorithm execution of the present invention;

[0054] Figure 4 This is a timing diagram of the operation of the system of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Example

[0058] like Figure 1-4 As shown, a method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture in this embodiment includes the following steps:

[0059] Step S1: divide the flexible resources into multiple regions according to their geographical location, physical characteristics and communication capabilities, wherein each region contains a number of local load nodes, each of which includes industrial loads, residential loads, energy storage equipment or controllable distributed power sources; the regional division standard is based on communication delay, geographical distribution density between nodes and power response capability classification, and a regional coordination center is established in each region. The regional coordination center maintains connection with the local load nodes in its region through a high-reliability communication interface, and establishes a cross-layer communication link with the power grid dispatching center for two-way information interaction;

[0060] After receiving the power shortage demand issued by the power grid dispatching center, the regional coordination center is responsible for optimizing the local power distribution according to the real-time operating status of the load nodes in the region. After the optimization result is generated, the regional coordination center sends the specific power distribution instruction to the local load node through the communication interface for execution. The local load node adjusts its power response behavior according to the instruction and feeds back the response execution result in real time, including the actual power response value of the node. Current operating status and dynamic change trend data, feedback data is summarized by the regional coordination center and uploaded to the power grid dispatching center;

[0061] The power grid dispatching center is the global control layer and is responsible for calculating the global power shortage demand P demand Specifically: The power grid dispatch center allocates a specific power target value P to each regional coordination center based on the historical response records, current status and demand urgency of each region. target,region ; The regional coordination center receives the P target,region , perform local allocation optimization based on the real-time status of nodes in the region;

[0062] The communication link within the architecture adopts a two-way interactive mode, with real-time information updates and feedback from the regional coordination center and local load nodes. Each regional coordination center connects with the power grid dispatching center through a standardized communication protocol to ensure the security and integrity of data exchange. The data packet includes power allocation requirements, node operating status, and response execution result information; when communication anomalies or node status anomalies occur, the regional coordination center automatically triggers the fault-tolerant mechanism, maintains power balance in the region by dynamically adjusting the allocation plan, and reports the abnormal status to the power grid dispatching center.

[0063] Step S2: Receiving power demand information and real-time data collection:

[0064] After the hierarchical partitioning architecture is built, the power grid dispatching center calculates the overall power shortage demand P by monitoring the operation status and load fluctuation of the entire network in real time. demand , and the demand target is divided into regional target power P target,regionAfter receiving the target power demand, the regional coordination center starts the data acquisition module in the region to obtain the operating status of the local load node in real time, including the following data items:

[0065] Actual power response value The actual power response value currently output by each load node reflects the actual response effect of the node after receiving the allocation instruction;

[0066] Response time t response,i : The time from each node receiving the allocation instruction to completing the power adjustment is used to monitor the timeliness of the node response;

[0067] Available power range: The upper and lower limits of the power that can be adjusted for each node [P i,min ,P i,max ], reflecting the current remaining power regulation capability of the node;

[0068] Historical load data: The historical operation records of each node, including past load change trends, response accuracy, and time delay, are used to assist in predicting the reliability of the current node;

[0069] The regional coordination center aggregates the collected node data and compares it with the expected response value set in the optimization model. and the target response time t target Compare one by one; during the comparison process, analyze the response status according to the following judgment conditions:

[0070] If the actual power response value of the node Deviation from expected response value If the error exceeds the set error threshold δP, the node response effect is judged to be abnormal;

[0071] If the node response time t response,i Exceed target time t target , it is recorded as a delayed response;

[0072] If the available power range of the node cannot meet the current allocation instruction, it is determined that the node has insufficient regulation capability.

[0073] When any node status is abnormal or the overall power response is insufficient, the regional coordination center automatically triggers a dynamic adjustment strategy to reallocate the power response plan. The specific adjustment process is as follows:

[0074] Prioritize the use of nodes with larger residual power capacity in the region for compensation, sort the nodes according to their adjustment capabilities, and allocate the residual power in turn to ensure the regional target power P target,region balance;

[0075] If the node regulation capacity in a region is insufficient, redundant power support from other regions is requested through communication with the grid dispatching center, which allocates and coordinates resources based on the power status of each region;

[0076] After the adjustment plan is completed, the new allocation instructions will be sent to the corresponding nodes. At the same time, the execution status of the nodes will be monitored and the real-time data will be updated to ensure the effectiveness of the dynamic adjustment.

[0077] Step S3: Optimization model construction:

[0078] After collecting load node data in real time and receiving the power shortage demand from the power grid dispatching center, the regional coordination center starts the construction and solution of the optimization model and establishes the following objective function:

[0079]

[0080] in:

[0081] Power response cost C i (P i ): c i is the linear cost coefficient, which represents the economic cost of unit power adjustment, k i is the quadratic cost coefficient, reflecting the additional cost of nonlinear power regulation;

[0082] For example, for an industrial load node, the linear cost coefficient c i It can be 0.5 yuan / kW, and k i 0.1 yuan / kW 2 , used to describe the nonlinear increase cost during high power adjustment;

[0083] Response reliability R i : Used to quantify the deviation between the actual power response value and the expected value;

[0084] For example, when the desired power response value is is 50kW, and the actual response value For 47kW, the response reliability (i.e. 6% deviation);

[0085] Time delay penalty Penalty(t): Penalty(t) is used to penalize response time exceeding the target time t target The situation is defined as:

[0086]

[0087] Where k is the time penalty coefficient, which represents the cost of delay per unit time;

[0088] For example, if the target response time t target The actual response time t is 2 seconds, the actual response time t is 3 seconds, and the penalty coefficient k is 10 yuan / second. The delay penalty is Penalty(t)=10·(3-2)=10 yuan.

[0089] In addition, in order to ensure the feasibility and operational safety of the optimization scheme, the optimization model sets constraints, including:

[0090] Power balance constraints:

[0091] Ensure that the sum of all allocated power meets the power shortage requirement p set by the grid dispatching center demand ;

[0092] For example, if the grid shortage demand is 200kW, the sum of the power response values ​​of all nodes must meet 200kW;

[0093] Resource capacity constraints:

[0094] 0≤P i ≤P i,max , where P i,max is the maximum power response capability of node i, ensuring that the allocation result is within the physical capacity of the node;

[0095] For example, if the maximum response capacity of a node is 30 kW, the allocated power P i Must satisfy 0≤P i ≤30;

[0096] Dynamic response constraints:

[0097] t response，i ≤T max , where T max The maximum allowed response time set for the system;

[0098] For example, in an emergency demand response scenario, setting T max = 5 seconds, all nodes must complete power adjustment within 5 seconds;

[0099] In the process of solving the optimization model of the present invention, the regional coordination center classifies the load nodes, for example, nodes with lower response cost but higher reliability are allocated more power first, while nodes with larger response delay are used as supplementary backup;

[0100] Specific example: A regional coordination center manages 5 load nodes, and its parameters are as follows:

[0101] Node 1 (industrial load): c1 = 0.5, k1 = 0.1, P 1,max =50kW,t response，1 = 3 seconds;

[0102] Node 2 (resident load): c2 = 0.7, k2 = 0.2, P 2,max =20kW, t response，2 = 2 seconds;

[0103] Node 3 (energy storage device): c3 = 0.3, k3 = 0.05, P 3,max =40kW, t response，3 = 1 second;

[0104] Node 4 (distributed power supply): c4=0.6, k4=0.15, P 4,max =30kW, t response，4 =4 seconds;

[0105] Node 5 (energy storage device): c5 = 0.4, k5 = 0.1, P 5,max =60kW,t response,5 =2 seconds.

[0106] The power grid dispatching center issues the power demand p demand =150kW, target response time t target = 2 seconds, the optimization model allocates power as follows by considering response cost, reliability and delay penalty:

[0107] Node 1 is allocated 40kW, node 2 is allocated 15kW, node 3 is allocated 50kW, node 4 is allocated 20kW, node 5 is allocated 25kW, and the total power is 150kW; the final optimization result satisfies the minimization of the objective function, and the response plan meets the power balance, capacity and time constraints.

[0108] Step S4, distributed consensus algorithm solution:

[0109] After the optimization model is built, the regional coordination center uses a distributed consensus algorithm to coordinate power allocation optimization through dynamic initialization, gradient prediction, dynamic step size adjustment strategy, and fault tolerance mechanism;

[0110] Dynamic initialization strategy:

[0111] The dynamic initialization strategy generates an initial power allocation plan based on historical load data and real-time operating status. The calculation formula is:

[0112] in, It represents the initial value of power response obtained through historical data statistics and fitting. δ is the correction value, which is used to reflect the deviation of the current load state from the historical prediction. The correction value is calculated based on the real-time load change rate, remaining power capacity and historical response trend of the node.

[0113] For example, if the historical prediction value of a node is The correction value under the current load state is δ = 5kW, so the initial distribution value is Through dynamic initialization, the optimization algorithm can iterate from an initial state close to the optimal solution, thereby reducing calculation time and error.

[0114] Gradient prediction and dynamic step size adjustment strategy:

[0115] Calculate the gradient of the objective function at each iteration And dynamically adjust the learning rate based on the gradient change, the step size adjustment formula is:

[0116]

[0117] Among them, η (t) is the current step length, represents the change between the current gradient and the previous iteration gradient, β is the step size adjustment coefficient, ∈ is a positive number to avoid the denominator being zero. The step size adjustment formula can adaptively adjust the learning rate, thereby accelerating convergence when the gradient changes greatly, and maintaining stability when the gradient changes slightly. In order to prevent the learning rate from changing too much and causing algorithm oscillation or too little and causing slow convergence, the step size range is limited to [η min ,η max ];

[0118] For example, if the current gradient change is 0.2, set the initial step size η (t) =0.01, step size adjustment coefficient β = 0.5, then the updated step size is η (t+1) =0.011;

[0119] Fault-tolerance mechanism: By real-time monitoring of the operating status of load nodes, if a node is found to be faulty or unable to receive instructions, its power response task is dynamically allocated to normal nodes; after the faulty node is restored, the power allocation is re-optimized to restore the global optimal state;

[0120] For example, the total power P needs to be distributed in a certain area region =100kW. If node A fails to provide the allocated 20kW power, other nodes will reallocate the 20kW according to their remaining power capabilities to ensure the stability of the regional target power.

[0121] Specific example: A regional coordination center manages 4 load nodes, and the initial parameters are as follows:

[0122] Node 1: P 1,max =50kW, historical forecast value Correction value δ1 = 5kW;

[0123] Node 2: P 2,max=50kW, historical forecast value Correction value δ2 = 5kW;

[0124] Node 3: P 3,max =50kW, historical forecast value Correction value δ3 = 5kW;

[0125] Node 4: P 4,max =50kW, historical forecast value Correction value δ4 = 5kW;

[0126] After dynamic initialization, the initial allocation value is:

[0127]

[0128] The total power response is

[0129] The optimization algorithm calculates the gradient value according to the objective function and dynamically adjusts the learning rate to accelerate iterative convergence. If node 3 cannot provide 13kW power due to communication interruption, the fault-tolerant mechanism allocates its power task to nodes 1 and 4. The adjusted allocation result is:

[0130] The final power response meets the regional target power demand.

[0131] Step S5: Power allocation and execution feedback:

[0132] After completing the optimization solution of the distributed consensus algorithm, the regional coordination center generates the optimization results as specific power allocation instructions and sends them to the local load nodes through a highly reliable communication link. After receiving the instructions, the local load nodes adjust the power output according to their current operating status and actual capacity, and feed back the response data to the regional coordination center in real time, as follows:

[0133] The regional coordination center will optimize the allocation results after solving Converted into power adjustment instructions for each node, including node number, target power value, execution time window, and fault tolerance mechanism parameters;

[0134] For example, for node A, the allocation instruction may be specifically expressed as "node A: target power = 30 kW, execution time = within 2 seconds".

[0135] After each load node receives the allocation instruction, it performs power adjustment according to the target power value and its own operating status. The control module inside the node calculates the target power in real time. With current power The node will adjust the power according to the adjustment rate limit. If the current output power of the node is 25kW and the target power is 30kW, the node will adjust the power at the maximum adjustment rate r. adjust Complete the power boost to ensure that the target time t response Complete the adjustment.

[0136] After the power adjustment is completed, the load node will feed back the response data to the regional coordination center in real time. The feedback data includes the actual power response value of the node. Response time t response,i , current operating status and abnormal status identification;

[0137] For example, the feedback data of node A may include "actual response power = 28 kW, response time = 1.8 seconds".

[0138] The regional coordination center analyzes the feedback data and converts the actual response value of the node into The expected response value set in the optimization model Compare and check the response time t response,i Whether the target time t is exceeded target If the actual response deviation exceeds the set threshold δP, or the response time exceeds the limit T max , then it is determined that the current response plan is abnormal. If the actual power response value of node A is 28kW, and the expected value is 30kW, the deviation If it is greater than the threshold δP=1kW, a dynamic adjustment strategy needs to be triggered.

[0139] When a node response deviation or insufficient overall power response is detected, the regional coordination center gives priority to using the remaining power capacity in the region for compensation. The dynamic adjustment strategy recalculates the power allocation of other nodes and selects the nodes with the largest remaining regulation capacity in turn to allocate the insufficient power. If the response of node A is less than 2kW, and the remaining regulation capacities of nodes B and C are 3kW and 1kW respectively, the regional coordination center gives priority to allocating node B to compensate for the insufficient power. The adjusted allocation plan allocates 2kW to node B, and node C remains unchanged.

[0140] If the remaining power capacity in this area cannot meet the compensation demand, the regional coordination center will communicate with the power grid dispatching center to request cross-regional adjustment support. The power grid dispatching center will dynamically adjust the global allocation plan according to the redundant power capacity of other areas. If the power required to be compensated in the current area is 5kW, and this area can only provide 3kW, the power grid dispatching center will dispatch 2kW from the neighboring area to meet the overall demand. The adjusted allocation result will be updated to the coordination center of the relevant area and sent to the corresponding node.

[0141] After completing dynamic adjustment or inter-regional coordination, the regional coordination center sends new allocation instructions to relevant nodes and continuously monitors the execution of nodes; through a closed-loop control mechanism of repeated feedback and comparison, it ensures that the final power response meets the grid demand and the response time t of all regions is response,i None exceeds the set limit T max .

[0142] Step S6: Real-time monitoring and dynamic adjustment

[0143] The power grid dispatching center and the regional coordination center jointly establish a dynamic monitoring system to collect and analyze real-time data on the power response of load nodes, including core data such as the actual power response value, response time, and operating status of the nodes. The regional coordination center compares the collected data with the expected response value and target time in the optimization model to determine the response deviation, time delay, and abnormal status. When it is detected that the deviation between the actual response of a node and the expected value exceeds the set threshold, or the node operation is abnormal and the power allocation cannot meet the regional target demand, the dynamic adjustment process is immediately triggered.

[0144] The adjustment strategy includes two parts: giving priority to using the remaining power capacity in the region to compensate for insufficient demand, and dynamically allocating additional power to other normal nodes to meet the current response target; if the capacity in the region is insufficient, the grid dispatching center will coordinate redundant power support from other regions. After the adjustment, new power allocation instructions will be generated and sent to relevant nodes, while the execution status will be continuously monitored to ensure that the adjusted response plan meets the real-time requirements.

[0145] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture, characterized in that: The following steps are involved: Step S1: Partition and hierarchical architecture construction: Divide the flexible resources into several regions according to their geographical locations and communication characteristics. Set up a regional coordination center in each region. The regional coordination center is connected to the power grid dispatching center through a communication interface to form a hierarchical and partitioned power distribution architecture. Step S2: Receiving power demand information and real-time data collection: The power grid dispatching center sends power shortage demand to each regional coordination center, and each regional coordination center collects the available power range, response time, operating status and historical load operation data of the local load node; Step S3: Optimization model building: According to the power response requirements, an optimization model is established with the goals of minimizing power response cost, minimizing response time delay, and maximizing response reliability, and power balance constraints, resource capacity constraints, and dynamic response constraints are set to ensure the practical feasibility of the power allocation solution; Step S4: Execution of the distributed consensus algorithm: In each regional coordination center, a distributed consensus algorithm is used to solve the optimization model and generate a preliminary power allocation plan. The dynamic initialization strategy is used to reduce the deviation of the initial plan, and the learning rate is adjusted in combination with gradient prediction to quickly achieve algorithm convergence. In the event of communication interruption or node failure, the power allocation result is dynamically adjusted through a fault-tolerant mechanism. Step S5: Generation and issuance of allocation instructions: The regional coordination center generates specific power allocation instructions to each load node in the region based on the optimization results of the distributed consensus algorithm, and sends the instructions to each load node for execution through the communication network; Step S6: Real-time monitoring and feedback optimization of execution status: After the load node executes the power allocation instruction, the regional coordination center collects the response data of each node in real time and compares it with the expected response result; if deviation is found, the power allocation plan is adjusted through dynamic optimization strategy, and the revised allocation instruction is reissued to ensure that the final response meets the real-time needs of the power grid.

2. According to the method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 1, it is characterized in that: The hierarchical partitioning architecture includes a three-layer structure of a power grid dispatching center, a regional coordination center and local load nodes, wherein the power grid dispatching center is responsible for calculating the global power shortage demand and sending it to each regional coordination center, while receiving the operating status and response results fed back by each region. The regional coordination center analyzes the operating status of the local load nodes, performs local optimization based on the demand sent by the power grid dispatching center, and sends power allocation instructions to the load nodes. After receiving the instructions, the local load nodes adjust the power according to the actual capacity, and feed back the response status to the regional coordination center in real time.

3. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 1 is characterized in that: The multi-objective optimization model takes minimizing power response cost, minimizing response time delay and maximizing response reliability as optimization objectives, and its objective function is: Where: Power response cost C i (P i ) is used to quantify the power response cost of resource i and is defined as where c i and k i are the linear and quadratic cost coefficients, respectively; Response reliability R i Indicates the actual power response value of the resource and target power response value The deviation between them is expressed as: The time delay penalty term Penalty(t) is used to reflect the resource response time exceeding the target time t target The resulting cost is defined by the formula: Among them, k is the time penalty coefficient, α i and λ are the weights of response reliability and time penalty, respectively, which are used to adjust the priority relationship between optimization objectives.

4. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 3 is characterized in that: The multi-objective optimization model is subject to the following constraints, including power balance constraints, resource capacity constraints, and dynamic response constraints: The power balance constraint ensures that the sum of the power responses of all flexible resources is equal to the power deficit requirement set by the grid dispatch center, which is defined as The resource capacity constraint is used to limit the power response of each resource to not exceed its maximum physical capacity, which is defined as 0≤P i ≤P i,max , where P i,max is the maximum response power of resource i; The dynamic response constraint: t response,i ≤T max , stipulates that the response time of all flexible resources shall not exceed the preset time limit T max .

5. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 1 is characterized in that: The distributed consensus algorithm realizes the rapid solution of power allocation scheme through dynamic initialization strategy, gradient prediction and step initialization strategy and fault tolerance mechanism. The dynamic initialization strategy is used to generate the initial power allocation scheme by combining historical data and real-time load status. Its formula is: in is the initial value of power allocation predicted based on historical data, δ represents the correction value calculated according to the current load state, which is used to correct the deviation between the historical prediction and the actual load; The gradient prediction and step size initialization strategy calculates the initial step size by analyzing the initial gradient value of the objective function. The formula is: in, Represents the initial gradient norm, ∈ is a positive number to avoid zero denominator. The fault-tolerant mechanism monitors the status of load nodes in real time. When a node failure or communication interruption is detected, it dynamically reallocates the power of the faulty node to the normal node and re-optimizes the power allocation after the fault is recovered.

6. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 5 is characterized in that: The design of the dynamic initialization strategy generates an initial power allocation scheme by statistically analyzing the correlation between historical load data and real-time operating status, including: The initial power response value of each load node is fitted through historical operation data Reflects the expected response capability of the node under similar operating conditions; The correction value δ is calculated based on the real-time operating status of the load node. The correction value is obtained by comprehensive calculation based on the real-time load change rate, the node available power range and the historical response deviation range. Differences from actual operating status; Using the formula Generate the final initial power allocation plan.

7. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 5 is characterized in that: The gradient prediction and dynamic step size adjustment strategy accelerates the convergence of the optimization algorithm by calculating the objective function gradient value in real time and dynamically adjusting the step size. First, in each iteration, according to the current power allocation scheme P (t) Calculate the gradient of the objective function The gradient value is used to determine the current optimization direction; Secondly, the step size is dynamically adjusted according to the gradient change rate. The step size adjustment formula is: where η (t) is the current step length; represents the change between the current gradient and the previous iteration gradient, β is the step size adjustment coefficient, and then, by setting the step size boundary control range [η min ,η max ], ensuring that the optimization algorithm can quickly converge to the global optimal solution in dynamically changing scenarios.

8. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 1 is characterized in that: The regional coordination center collects the actual response data of local load nodes in real time, including the actual power response value of the node Response time t response,i and operating status data, comparing the actual response results with the expected response values ​​of the optimization model and the target response time t target A comparison is performed, and when it is found that the response deviation exceeds the preset threshold or the node operation status is abnormal, the strategy is dynamically adjusted to reallocate the power response plan.

9. The method for optimizing allocation of emergency demand response resources under a hierarchical aggregation architecture according to claim 8, characterized in that: The dynamic adjustment strategy includes giving priority to using the remaining power in the region for compensation. If the demand cannot be met in the region, redundant power support is requested through the inter-regional coordination mechanism, and the power is adjusted in T max Complete within the time limit.

Images