Flexible resource aggregation scheduling method based on double-layer consistency algorithm

Abstract

The application discloses a flexible resource aggregation scheduling method based on a double-layer consistency algorithm, which comprises the following steps: S1. An aggregation dispersion scheduling model suitable for flexible resources is established, a target function is established with the minimum generation cost as the target, and constraint conditions of the aggregation dispersion scheduling model are determined; S2. A double-layer consistency algorithm suitable for aggregation dispersion is constructed according to the power grid architecture and the distribution of the flexible resources, the upper layer is a decision layer, the lower layer is a decomposition layer, the lower layer decomposes by receiving the decision information sent by the upper layer, and the upper layer makes decisions by receiving the decomposition information returned by the lower layer; S3. An asynchronous iteration strategy of the double-layer consistency algorithm is set, the iteration of the double-layer consistency algorithm is carried out according to the asynchronous iteration strategy, until a convergence value is obtained by the algorithm convergence, a dispersion scheduling result of the power grid node is obtained, and each node of the power grid is controlled according to the dispersion scheduling result. The aggregation scheduling can significantly accelerate the convergence speed of the algorithm and keep high accuracy.

Description

Technical Field

[0001] This invention relates to power system dispatching methods, and more specifically, to an aggregated dispatching method adapted to large-scale flexible resources. Background Technology

[0002] Flexible resources are characterized by their large quantity, small capacity, and wide distribution, and can be uniformly managed and scheduled by aggregators. However, existing centralized scheduling methods suffer from significant communication burdens and information privacy issues, and cannot guarantee decision independence, making it difficult to meet the scheduling needs of large-scale flexible resources. To address this challenge, researchers have proposed distributed optimization algorithms such as Analysis Target Cascading (ATC), Alternating Direction Method of Multipliers (ADMM), and Auxiliary Principle Problem (APP). However, these distributed algorithms are mainly used for distributed economic scheduling problems in multiple regions and are ill-suited to handle frequent on / off-grid switching of flexible resources.

[0003] Consensus algorithms, as classic point-to-point decentralized distributed optimization algorithms, have been successfully applied in the economic dispatch of power in microgrids. This algorithm is flexible and can meet the requirements of plug-and-play operation. However, existing consensus algorithms are all single-level and are not suitable for the aggregated dispatch mode of flexible resources. In the aggregated mode, distributed dispatch mainly includes two levels: the inner level is the dispersion among individual flexible resources, and the outer level is the dispersion among aggregators, with interaction between the two levels through the aggregators. Therefore, it is necessary to propose a new two-level consensus algorithm to meet the requirements of large-scale flexible resource aggregated dispatch. Summary of the Invention

[0004] The purpose of this invention is to solve the problem that traditional consensus algorithms are difficult to implement distributed scheduling in the aggregator mode, and to provide a flexible resource aggregation scheduling method based on a two-layer consensus algorithm to achieve two-layer distributed scheduling in the flexible resource aggregator mode.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A flexible resource aggregation scheduling method based on a two-layer consensus algorithm includes the following steps:

[0007] S1. Establish an aggregated-distributed scheduling model that adapts to flexible resources. The aggregated-distributed scheduling model establishes an objective function with the goal of minimizing power generation costs and defines the constraints of the aggregated-distributed scheduling model.

[0008] S2. Based on the power grid architecture and the distribution of flexible resources, a two-layer consensus algorithm adapted to aggregation and dispersion is constructed. The upper layer is the decision layer and the lower layer is the decomposition layer. The lower layer decomposes the data by receiving decision information sent by the upper layer, and the upper layer makes decisions by receiving decomposition information returned by the lower layer.

[0009] S3. Set the asynchronous iteration strategy of the two-level consensus algorithm, iterate the two-level consensus algorithm according to the asynchronous iteration strategy until the algorithm converges and obtains the convergence value, obtain the distributed scheduling result of the power grid nodes, and control each node of the power grid according to the distributed scheduling result.

[0010] Furthermore, the objective function of the aggregated distributed scheduling model is:

[0011]

[0012] In the formula, a i,n b is the quadratic term of the cost coefficient of the flexibility resource n connected to node i; i,n c is a linear term representing the cost coefficient of the flexibility resource n connected to node i. i,n P is a constant term for the cost coefficient of the flexibility resource n connected to node i; i,n The power of the flexibility resource n connected to node i; Ω G,i The set of all flexibility resources connected to node i;

[0013] The constraints of the aggregated distributed scheduling model include power flow constraints, power constraints, and line capacity constraints.

[0014] Current constraints are:

[0015]

[0016] In the formula, PL i Let θ be the load at node i; i Let θ be the phase angle of node i; j X is the phase angle of node j; ij This represents the line impedance connecting node i and node j; Ω i Let i be the set of nodes that are connected to node i.

[0017] The power constraint is:

[0018]

[0019] In the formula, The upper limit of the power of the flexibility resource n connected to node i; P i,n The lower power limit for the flexibility resource n connected to node i;

[0020] The line capacity constraint is:

[0021]

[0022] In the formula, This represents the upper limit of the transmission capacity of the line connecting node i and node j; Ω L This is the set of all routes.

[0023] Furthermore, the principle of the consensus algorithm is to converge variables to the target value through iteration. The iteration process includes a consensus term and a target asymptotic term. The consensus term is assigned the sum of the consensus term and the target asymptotic term. When the target asymptotic term tends to zero, the consensus term converges to a stable value. Specifically:

[0024]

[0025] In the formula, ξ k H is the consistency term at the k-th iteration. *k Let ξ be the objective asymptotic term in the k-th iteration. k+1 This is the consistency term at the (k+1)th iteration;

[0026] In DC power flow, define local variables. For consistency items, It represents the set of all variables related to node i in the k-th iteration; Let be the Lagrange multiplier corresponding to the power flow constraint of node i in the kth iteration; Let i be the phase angle of node i in the k-th iteration; The Lagrange multiplier corresponding to the capacity constraint of the line connecting node i and node j in the kth iteration; The power of the flexibility resource n connected to node i in the kth iteration;

[0027] The update strategy is as follows:

[0028]

[0029] In the formula, the function g(·) is the target asymptotic term; Φ is the adjustment parameter corresponding to the target asymptotic term; T(·) is the constraint function, which is responsible for adjusting the variables. Project onto the defined feasible region; It represents the set of all variables related to node i in the (k+1)th iteration.

[0030] Furthermore, the adaptive aggregation-dispersion two-layer consensus algorithm is divided into two layers: the upper layer represents the dispersion among aggregators, and the lower layer represents the dispersion among individuals within an aggregator. The principle of the adaptive aggregation-dispersion two-layer consensus algorithm is as follows:

[0031] First, in the upper-level model, the upper-level aggregator nodes make decisions in a decentralized, point-to-point manner, and then pass the decision variables to the lower-level aggregator nodes. Next, in the lower-level model, the individuals within the aggregator decompose the decision variable in a decentralized manner, and then the lower-level aggregator nodes return the decomposed price information to the upper-level aggregator nodes. Finally, the upper-level aggregator nodes continue the upper-level iteration based on the returned price information.

[0032] Furthermore, in the constructed adaptive aggregation-dispersion two-layer consensus algorithm, the Lagrange multipliers corresponding to the power flow constraints are transmitted to the upper layer as price information. In the upper-layer model, aggregators participate in the upper-layer iteration based on the cost information from the lower layer, and the update strategy for the power of aggregator nodes is as follows:

[0033]

[0034] In the formula, Let be the Lagrange multiplier corresponding to the power flow constraint of node i in the kth iteration; ρ represents the price information returned by node i from the lower layer during the k-th iteration of the upper layer; ρ is an adjustment parameter, and the value of ρ is greater than 0. Let be the injected power of node i in the (k-1)th iteration; Let be the injected power of node i in the kth iteration;

[0035] The decision variable of the upper-level aggregator node represents the power that the node needs to output to the grid, i.e., the power injected by the aggregator node during iteration, which can be equivalent to the load demand of the lower level. In the lower-level model, this decision variable is used as the load at the aggregator node in the lower-level iteration, specifically as follows:

[0036]

[0037] In the formula, PL i Let be the load of node i. The corrected load for node i;

[0038] Regarding the phase angle of a node, the phase angle of the aggregator node can be divided into two parts. and θ, θ participates in the iterations of the upper and lower layers respectively; after both the upper and lower layers converge, the phase angles of all nodes in the lower layer are processed as follows:

[0039] θ′ i,o =θ i,o +Δ i

[0040] In the formula, θ i,o The convergence value of the phase angle of node o, which is in the same lower layer as node i; Δ i The difference between the convergence phase angle values ​​of node i and the upper and lower layers. Let θ be the convergence value of the phase angle for node i participating in the upper-level iteration. i θ′ is the convergence value of the phase angle for node i to participate in the next iteration. i,o The phase angle convergence value is the corrected value for node o, which is in the same lower layer as node i.

[0041] Furthermore, the upper and lower layers are decoupled through an interaction pattern, based on variables. The update strategy for the Lagrange multipliers corresponding to the power flow constraints in the next iteration is as follows:

[0042]

[0043] In the formula, α and β are adjustment parameters, and the values ​​of α and β are both greater than 0; Let i be the phase angle of node i in the k-th iteration; Let L be the Lagrange multiplier corresponding to the power flow constraint of node i in the (k+1)th iteration; L is the Lagrange function of the scheduling model.

[0044] The formula for calculating the Lagrange function L is:

[0045]

[0046] In the formula, The Lagrange multiplier corresponding to the capacity upper limit constraint of the flexibility resource n connected to node i; The Lagrange multiplier corresponding to the capacity lower bound constraint of the flexibility resource n connected to node i; λ i Let be the Lagrange multiplier corresponding to the power flow constraint at node i.

[0047] Furthermore, to reduce the number of variables, a constraint function is used, which is expressed through the Lagrange multiplier λ. i Update power P i,n The update strategy is as follows:

[0048]

[0049] In the formula, The power of the flexibility resource n connected to node i in the (k+1)th iteration; T i,n (·) represents the power P i,n The constraint function, T i,n (·) determined by the upper limit of power P i,n and power lower limit Definition; if constraint function T i,n (·) The value obtained from the internal update is less than the lower power limit. P i,n Then it will be fixed at the lower power limit. P i,nConversely, it is fixed at the upper limit of power.

[0050] Furthermore, the update strategy for the Lagrange multiplier μ corresponding to the line capacity constraint is as follows:

[0051]

[0052] In the formula, δ is the adjustment parameter, and the adjustment parameter δ > 0; μ k μ is the Lagrange multiplier corresponding to the line capacity constraint at the k-th iteration; k+1 T represents the Lagrange multiplier corresponding to the line capacity constraint in the (k+1)th iteration; l (·) is the constraint function for the Lagrange multiplier μ; when the constraint function T l When μ obtained from the update in (·) is less than 0, the constraint function T l (·) Fix it to 0 to satisfy the constraint that μ is not less than 0;

[0053] The update strategy for the node phase angle θ is as follows:

[0054]

[0055] In the formula, γ is the adjustment parameter, and the adjustment parameter γ > 0; Let be the phase angle of node i at the k-th iteration; Let be the phase angle of node i in the (k+1)th iteration.

[0056] Furthermore, the objective asymptotic terms for all variables in the local variables are the first-order optimality conditions of the objective function and constraints of the aggregated distributed scheduling model, specifically:

[0057]

[0058]

[0059]

[0060]

[0061] In the formula, λ j μ is the Lagrange multiplier corresponding to the power flow constraint at node j; ij The Lagrange multiplier is the line capacity constraint corresponding to the connection between node i and node j.

[0062] Furthermore, in the two-level consensus algorithm, when the changes in all variables satisfy the first convergence criterion, it indicates that the lower-level iteration has converged. The first convergence criterion is:

[0063]

[0064] In the formula, εs The allowable error for the first convergence criterion;

[0065] In addition to satisfying the first convergence criterion, the convergence condition of the upper-level iteration must also satisfy the second convergence criterion to ensure that the upper and lower levels achieve consistency. The second convergence criterion is as follows:

[0066]

[0067] In the formula, ε r The allowable error for the second convergence criterion;

[0068] The asynchronous iteration strategy improves the lower-level iteration of the two-level consensus algorithm. The initial value of the lower-level iteration is taken as the convergence value of the previous iteration, specifically:

[0069]

[0070] In the formula, This is a local variable of node i in the lower layer during the k-th iteration of the upper layer and the m-th iteration of the lower layer; This is a local variable of node i in the lower layer when the upper layer is in the (k+1)th iteration and the lower layer is in the 0th iteration;

[0071] In the asynchronous iteration strategy, the lower-level iteration does not need to converge; it only needs to iterate a certain number of times to transmit information to the upper level. The value of is the Lagrange multiplier corresponding to the aggregator node at the end of the next iteration, i.e. This represents the Lagrange multiplier corresponding to the power flow constraint of node i in the lower layer at the k-th iteration of the upper layer and the m-th iteration of the lower layer. The convergence criterion for the end of the lower layer iteration is:

[0072]

[0073] In the formula, m represents the number of iterations in the lower layer; p represents the number of iterations set in the lower layer; and ∨ represents the OR operation.

[0074] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0075] 1. The two-layer consensus algorithm proposed in this invention supports distributed scheduling under the aggregator model, which is beneficial for the management of massive flexible resources, enhances the flexibility of the power system, and helps in the construction of new power systems.

[0076] 2. The two-layer consensus algorithm proposed in this invention interacts between the two layers through aggregators, adapting to the two-layer distributed scheduling mode. The distributed optimization between the upper-layer aggregators ensures the decision independence of the aggregators, while the distributed optimization between the lower-layer individuals enables the plug-and-play flexibility of the individual resources within the aggregator.

[0077] 3. The asynchronous iteration strategy of the two-layer consensus algorithm proposed in this invention allows the upper layer to proceed to the next iteration without waiting for the lower layer to finish its iteration, resulting in faster convergence speed and reduced algorithm convergence time. Attached Figure Description

[0078] Figure 1 This is a flowchart of a flexible resource aggregation scheduling method based on a two-level consensus algorithm.

[0079] Figure 2 This is a schematic diagram illustrating the construction principle of a two-level consensus algorithm.

[0080] Figure 3 This is a flowchart of a two-level consensus algorithm.

[0081] Figure 4 This is a flowchart of an asynchronous iterative strategy based on a two-level consensus algorithm. Detailed Implementation

[0082] The flexible resource aggregation and scheduling method based on the two-layer consensus algorithm of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0083] Please see Figure 1 This invention discloses a flexible resource aggregation and scheduling method based on a two-layer consensus algorithm, comprising the following steps:

[0084] S1. Establish an aggregated-distributed scheduling model that adapts to flexible resources. The aggregated-distributed scheduling model establishes an objective function with the goal of minimizing power generation costs and defines the constraints of the aggregated-distributed scheduling model.

[0085] S2. Based on the power grid architecture and the distribution of flexible resources, a two-layer consensus algorithm adapted to aggregation and dispersion is constructed. The upper layer is the decision layer and the lower layer is the decomposition layer. The lower layer decomposes the data by receiving decision information sent by the upper layer, and the upper layer makes decisions by receiving decomposition information returned by the lower layer.

[0086] S3. Set the asynchronous iteration strategy of the two-level consensus algorithm, iterate the two-level consensus algorithm according to the asynchronous iteration strategy until the algorithm converges and obtains the convergence value, obtain the distributed scheduling result of the power grid nodes, and control each node of the power grid according to the distributed scheduling result.

[0087] The objective function of the aggregated distributed scheduling model is:

[0088]

[0089] In the formula, a i,n b is the quadratic term of the cost coefficient of the flexibility resource n connected to node i; i,n c is a linear term representing the cost coefficient of the flexibility resource n connected to node i. i,nP is a constant term for the cost coefficient of the flexibility resource n connected to node i; i,n The power of the flexibility resource n connected to node i; Ω G,i This is the set of all flexibility resources connected to node i.

[0090] The constraints of the aggregated distributed scheduling model include power flow constraints, power constraints, and line capacity constraints.

[0091] The power flow constraint is:

[0092]

[0093] In the formula, PL i Let θ be the load at node i; i Let θ be the phase angle of node i; j X is the phase angle of node j; ij This represents the line impedance connecting node i and node j; Ω i Let i be the set of nodes that are connected to node i.

[0094] The power constraint is:

[0095]

[0096] In the formula, The upper limit of the power of the flexibility resource n connected to node i; P i,n The lower power limit for the flexibility resource n connected to node i.

[0097] The line capacity constraint is as follows:

[0098]

[0099] In the formula, This represents the upper limit of the transmission capacity of the line connecting node i and node j; Ω L This is the set of all routes.

[0100] The two-layer consensus algorithm and asynchronous iterative strategy proposed in this invention complete the decentralized solution for the aggregated scheduling of power grid flexibility resources. When the algorithm converges, all constraints are satisfied, and the objective function reaches its optimum; the convergence value is the result of the aggregated scheduling of flexibility resources.

[0101] The basic principle of consensus algorithms is to converge variables to a target value through iteration, primarily by updating the state itself by acquiring information from neighbors. The iterative process includes a consensus term and a target asymptotic term. Generally, the consensus term and the target asymptotic term are superimposed and assigned to the consensus term, so that when the target asymptotic term approaches zero, the consensus term converges to a stable value. Specifically:

[0102]

[0103] In the formula, ξ k H is the consistency term at the k-th iteration. *k Let ξ be the objective asymptotic term in the k-th iteration. k+1 This is the consistency term at the (k+1)th iteration.

[0104] In DC power flow, define local variables. For consistency items, Let represent the set of all variables related to node i in the k-th iteration. Let be the Lagrange multiplier corresponding to the power flow constraint of node i in the kth iteration; Let i be the phase angle of node i in the k-th iteration; The Lagrange multiplier corresponding to the capacity constraint of the line connecting node i and node j in the kth iteration; The power of the flexibility resource n connected to node i in the kth iteration.

[0105] The general expression for the update method is:

[0106]

[0107] In the formula, the function g(·) is the objective asymptotic term; Φ is the corresponding adjustment parameter; and T(·) is the constraint function, which is responsible for controlling the variables. Project onto the defined feasible region; It represents the set of all variables related to node i in the (k+1)th iteration.

[0108] The adaptive aggregation-dispersion two-layer consensus algorithm proposed in this invention consists of two layers. The upper layer represents the dispersion among aggregators, aiming to ensure decision independence through this dispersion. The second layer represents the dispersion among individuals within each aggregator, aiming to handle the frequent joining and leaving of various flexible resources through a plug-and-play approach. The principle of the adaptive aggregation-dispersion two-layer consensus algorithm is as follows: Figure 2 As shown.

[0109] like Figure 2As shown, nodes A and B represent aggregator nodes, and nodes A′ and B′ represent replicas of aggregator nodes A and B. The replication of aggregator nodes here is for ease of illustrating the coupling relationship between the upper and lower layers; in reality, both are the same node. Compared with traditional consensus algorithms, the innovation of the proposed two-layer consensus algorithm lies mainly in the interaction mode between the upper and lower layers. First, in the upper-layer model, aggregators A′ and B′ make decisions in a point-to-point decentralized manner, and then pass the decision variables to the lower-layer nodes A and B. Next, in the lower-layer model, individuals within the aggregator (lower layer 1 and lower layer 2) decompose the decision variable in a decentralized manner, and then aggregator nodes A and B return the decomposed price information to the upper layers A′ and B′. Finally, A′ and B′ continue the upper-layer iteration based on the returned price information.

[0110] In economics, Lagrange multipliers can be interpreted as shadow prices. Therefore, the Lagrange multipliers corresponding to power flow constraints can be transmitted as price information to the upper layer. In the upper-layer model, aggregators participate in the upper-layer iteration based on the cost information from the lower layer. The update strategy for aggregator node power is as follows:

[0111]

[0112] In the formula, Let be the Lagrange multiplier corresponding to the power flow constraint of node i in the kth iteration; ρ represents the price information returned by node i from the lower layer during the k-th iteration of the upper layer; ρ is an adjustment parameter, and the value of ρ is greater than 0. Let be the injected power of node i in the (k-1)th iteration; Let be the injected power of node i in the k-th iteration. In this embodiment, the Lagrange multiplier corresponding to the power flow constraint in the lower-level iteration represents the price information returned from the lower level.

[0113] The decision variable of the upper-level aggregator node represents the power that the node needs to output to the grid, i.e., the power injected by the aggregator node during iteration, which can be equivalent to the load demand of the lower level. Therefore, in the lower-level model, this decision variable can be used as the load at the aggregator node to participate in the lower-level iteration, specifically:

[0114]

[0115] In the formula, PL i Let be the load of node i. This is the corrected load for node i.

[0116] For the phase angle of a node, the phase angle of the aggregator node can be divided into two. θ participates in the iterations of the upper and lower layers, respectively. After both the upper and lower layers converge, the phase angles of all nodes in the lower layer can be processed as follows:

[0117] θ′ i,o =θ i,o +Δ i (9)

[0118] In the formula, θ i,o The convergence value of the phase angle of node o, which is in the same lower layer as node i; Δ i The difference between the convergence phase angle values ​​of node i and the upper and lower layers. Let θ be the convergence value of the phase angle for node i participating in the upper-level iteration. i θ′ is the convergence value of the phase angle for node i to participate in the next iteration. i,o The phase angle convergence value is the corrected value for node o, which is in the same lower layer as node i.

[0119] At this point, the upper and lower layers have been decoupled through the above interaction mode. According to the update strategy shown in equation (6), the update strategy for the Lagrange multipliers corresponding to the power flow constraints in the lower layer iteration is:

[0120]

[0121] In the formula, α and β are adjustment parameters, and the values ​​of α and β are both greater than 0; Let i be the phase angle of node i in the k-th iteration; Let L be the Lagrange multiplier corresponding to the power flow constraint of node i in the (k+1)th iteration; L is the Lagrange function of the scheduling model.

[0122] The formula for calculating the Lagrange function L is:

[0123]

[0124] In the formula, The Lagrange multiplier corresponding to the capacity upper limit constraint of the flexibility resource n connected to node i; The Lagrange multiplier corresponding to the capacity lower bound constraint of the flexibility resource n connected to node i; λ i Let be the Lagrange multiplier corresponding to the power flow constraint at node i.

[0125] To reduce the number of variables, a constraint function is used, which is obtained through the Lagrange multiplier λ. i Update power P i,n The update strategy is as follows:

[0126]

[0127] In the formula, The power of the flexibility resource n connected to node i in the (k+1)th iteration; T i,n (·) represents the power P i,n The constraint function, T i,n(·) determined by the upper limit of power P i,n and power lower limit Definition. If the value obtained from the update within the constraint function is less than the lower power limit. P i,n Then it will be fixed at the lower power limit. P i,n Conversely, it is fixed at the upper limit of power.

[0128] Furthermore, the update strategy for the Lagrange multiplier μ corresponding to the line capacity constraint is as follows:

[0129]

[0130] In the formula, δ is the adjustment parameter, and the adjustment parameter δ > 0; μ k μ is the Lagrange multiplier corresponding to the line capacity constraint at the k-th iteration; k+1 T represents the Lagrange multiplier corresponding to the line capacity constraint in the (k+1)th iteration; l (·) is the constraint function of the Lagrange multiplier μ; when the constraint function T of equation (13) l When μ obtained from the update in (·) is less than 0, the constraint function T l (·) Fix it to 0 to satisfy the constraint that μ is not less than 0.

[0131] The update strategy for the node phase angle θ is as follows:

[0132]

[0133] In the formula, γ is the adjustment parameter, and the adjustment parameter γ > 0; Let i be the phase angle of node i in the k-th iteration; Let be the phase angle of node i in the (k+1)th iteration.

[0134] Among them, the objective asymptotic terms of all local variables are the first-order optimality conditions of the objective function and constraints (Equations (1)-(4)) of the aggregated distributed scheduling model, specifically:

[0135]

[0136]

[0137]

[0138]

[0139] In the formula, λ j μ is the Lagrange multiplier corresponding to the power flow constraint at node j; ijThe Lagrange multiplier is the line capacity constraint corresponding to the connection between node i and node j.

[0140] In the two-level consensus algorithm, the lower-level iteration converges when the changes in all variables satisfy the first convergence criterion. The first convergence criterion is:

[0141]

[0142] In the formula, ε s This represents the allowable error for the first convergence criterion.

[0143] The update strategy for the upper layer iteration is similar to that for the lower layer. The update strategy for the power of aggregator nodes is shown in Equation (7), the update strategy for the power of non-aggregator nodes is shown in Equation (12), the update strategy for the Lagrange multiplier corresponding to the power flow constraint is shown in Equation (10), the update strategy for the Lagrange multiplier corresponding to the line capacity constraint is shown in Equation (13), and the update strategy for the node phase angle is shown in Equation (14).

[0144] However, in addition to satisfying the first convergence criterion (Equation (19)), the convergence condition of the upper-level iteration also needs to satisfy the second convergence criterion to ensure that the upper and lower levels achieve consistency. The second convergence criterion is:

[0145]

[0146] In the formula, ε r This represents the allowable error for the second convergence criterion.

[0147] The process of the two-level consensus algorithm is as follows: Figure 3 As shown, the iterative pattern of the above two-level consensus algorithm is similar to that of the traditional one. However, the nested loops in this algorithm cause a large number of repeated calculations in the lower-level iterations, resulting in a slower convergence speed.

[0148] To address this, we further design an asynchronous iteration strategy based on a two-level consensus algorithm to reduce unnecessary iterations. The proposed asynchronous iteration strategy primarily improves the lower-level iteration (i.e., the inner loop). Compared to traditional iteration methods, the initial value of the lower-level iteration in the asynchronous iteration strategy is taken as the convergence value of the previous iteration, specifically:

[0149]

[0150] In the formula, This is a local variable of node i in the lower layer during the k-th iteration of the upper layer and the m-th iteration of the lower layer; This is a local variable of node i in the lower layer during the (k+1)th iteration of the upper layer and the 0th iteration of the lower layer.

[0151] In addition, in asynchronous iteration strategies, lower-level iterations do not need to converge; they only need to iterate a certain number of times to transmit information to the upper level. At this point, The value of is the Lagrange multiplier corresponding to the aggregator node at the end of the next iteration, i.e. This represents the Lagrange multiplier corresponding to the power flow constraint of node i in the lower layer at the k-th iteration of the upper layer and the m-th iteration of the lower layer. The convergence criterion for the end of the lower layer iteration is:

[0152]

[0153] In the formula, m represents the number of iterations in the lower layer; p represents the number of iterations set in the lower layer; and ∨ represents the OR operation.

[0154] Equation (22) indicates that each iteration of the lower layer can be determined not only by the criterion (19) but also by whether the number of iterations is greater than p. To reduce the time difference between the lower layer iterations and reduce unnecessary waiting time for the upper layer iterations, all lower layer iterations should take the same value p. This 1-p iteration strategy, where the upper layer iterates once and the lower layer iterates only p times, is called asynchronous iteration. This method reduces redundant calculations and speeds up convergence by reducing the number of lower layer iterations. The asynchronous iteration strategy flow of the two-layer consensus algorithm is as follows: Figure 4 As shown.

[0155] The two-layer consensus algorithm and asynchronous iterative strategy of this invention complete the decentralized solution of the power grid flexibility resource aggregation scheduling, which can significantly accelerate the convergence speed of the algorithm and maintain high accuracy. When the algorithm converges, all constraints are satisfied, and the objective function is optimal. Its convergence value is the result of the flexibility resource aggregation scheduling.

[0156] In summary, the present invention has the following advantages and beneficial effects:

[0157] 1. The two-layer consensus algorithm proposed in this invention supports distributed scheduling under the aggregator model, which is beneficial for the management of massive flexible resources, enhances the flexibility of the power system, and helps in the construction of new power systems.

[0158] 2. The two-layer consensus algorithm proposed in this invention interacts between the two layers through aggregators, adapting to the two-layer distributed scheduling mode. The distributed optimization between the upper-layer aggregators ensures the decision independence of the aggregators, while the distributed optimization between the lower-layer individuals enables the plug-and-play flexibility of the individual resources within the aggregator.

[0159] 3. The asynchronous iteration strategy of the two-layer consensus algorithm proposed in this invention allows the upper layer to proceed to the next iteration without waiting for the lower layer to finish its iteration, resulting in faster convergence speed and reduced algorithm convergence time.

[0160] The above description is a detailed description of the preferred embodiments of the present invention. However, the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modifications made under the technical spirit disclosed in the present invention should fall within the patent scope covered by the present invention.

Claims

1. A flexible resource aggregation and scheduling method based on a two-layer consensus algorithm, characterized in that, Includes the following steps: S1. Establish an aggregated-distributed scheduling model that adapts to flexible resources. The aggregated-distributed scheduling model establishes an objective function with the goal of minimizing power generation costs and defines the constraints of the aggregated-distributed scheduling model. S2. Based on the power grid architecture and the distribution of flexible resources, a two-layer consensus algorithm adapted to aggregation and dispersion is constructed. The upper layer is the decision layer and the lower layer is the decomposition layer. The lower layer decomposes the data by receiving decision information sent by the upper layer, and the upper layer makes decisions by receiving decomposition information returned by the lower layer. S3. Set the asynchronous iteration strategy of the two-level consensus algorithm, iterate the two-level consensus algorithm according to the asynchronous iteration strategy until the algorithm converges and obtains the convergence value, obtain the distributed scheduling result of the power grid node, and control each node of the power grid according to the distributed scheduling result. The objective function of the aggregated distributed scheduling model is: ； In the formula, The quadratic term of the cost coefficient of the flexibility resource n connected to node i; The first-order term of the cost coefficient for the flexibility resource n connected to node i; A constant term for the cost coefficient of the flexibility resource n connected to node i; The power of the flexibility resource n connected to node i; The set of all flexibility resources connected to node i; The constraints of the aggregated distributed scheduling model include power flow constraints, power constraints, and line capacity constraints. Current constraints are: ； In the formula, Let i be the load of node i; Let be the phase angle of node i; Let be the phase angle of node j; This represents the line impedance connecting node i and node j; Let i be the set of nodes that are connected to node i. The power constraint is: ； In the formula, The upper limit of the power of the flexibility resource n connected to node i; The lower power limit for the flexibility resource n connected to node i; The line capacity constraint is: ； In the formula, This represents the upper limit of the transmission capacity of the line connecting node i and node j; This is the set of all routes.

2. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 1, characterized in that, The principle of the consensus algorithm is to converge variables to a target value through iteration. The iteration process includes a consensus term and a target asymptotic term. The consensus term is assigned the sum of the consensus term and the target asymptotic term. When the target asymptotic term tends to zero, the consensus term converges to a stable value. Specifically: ； In the formula, This is the consistency term at the k-th iteration. Let be the target asymptotic term in the k-th iteration. This is the consistency term at the (k+1)th iteration; In DC power flow, define local variables. For consistency items, Indicates the first The set of all variables related to node i in the next iteration; For node i at the th The Lagrange multiplier corresponding to the power flow constraint in the next iteration; For node i in the th... Phase angle at the next iteration; The line connecting node i and node j is in the first... The Lagrange multiplier corresponding to the capacity constraint in the next iteration; For the flexibility resource n connected to node i in the first... Power at the next iteration; The update strategy is as follows: ； In the formula, the function For the goal of incremental progress; These are the adjustment parameters corresponding to the target asymptotic terms; For constraint functions, constraint functions Responsible for variables Project onto the defined feasible region; It represents the set of all variables related to node i in the (k+1)th iteration.

3. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 2, characterized in that, The adaptive aggregation-dispersion two-level consensus algorithm consists of two levels: the upper level represents the dispersion among aggregators, and the lower level represents the dispersion among individuals within an aggregator. The principle of the adaptive aggregation-dispersion two-level consensus algorithm is as follows: First, in the upper-level model, the upper-level aggregator nodes make decisions in a decentralized, point-to-point manner, and then pass the decision variables to the lower-level aggregator nodes. Next, in the lower-level model, the individuals within the aggregator decompose the decision variable in a decentralized manner, and then the lower-level aggregator nodes return the decomposed price information to the upper-level aggregator nodes. Finally, the upper-level aggregator nodes continue the upper-level iteration based on the returned price information.

4. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 3, characterized in that, In the constructed adaptive aggregation-decentralized two-layer consensus algorithm, the Lagrange multipliers corresponding to the power flow constraints are transmitted to the upper layer as price information. In the upper-layer model, aggregators participate in the upper-layer iteration based on the cost information from the lower layer. The update strategy for the power of aggregator nodes is as follows: ； In the formula, Let be the Lagrange multiplier corresponding to the power flow constraint of node i in the kth iteration; The price information returned by node i from the lower layer during the k-th iteration of the upper layer; To adjust the parameters, The value is greater than 0; Let be the injected power of node i in the (k-1)th iteration; Let be the injected power of node i in the kth iteration; The decision variable of the upper-level aggregator node represents the power that the node needs to output to the grid, i.e., the power injected by the aggregator node during iteration, which can be equivalent to the load demand of the lower level. In the lower-level model, this decision variable is used as the load at the aggregator node in the lower-level iteration, specifically as follows: ； In the formula, Let be the load of node i. The corrected load for node i; Regarding the phase angle of a node, the phase angle of the aggregator node can be divided into two parts. and , and Each node participates in the iterations of the upper and lower layers respectively; after both the upper and lower layers converge, the phase angles of all nodes in the lower layer are processed as follows: ； In the formula, Let $\frac{i}{i}$ be the convergence value of the phase angle of node $o$, which is in the same lower layer as node $i$. The difference between the convergence phase angle values ​​of node i and the upper and lower layers. , Let be the convergence value of the phase angle for node i to participate in the upper-level iteration. Let i be the convergence value of the phase angle for node i to participate in the next iteration; The phase angle convergence value is the corrected value for node o, which is in the same lower layer as node i.

5. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 4, characterized in that, The upper and lower layers are decoupled through an interaction pattern, based on variables. The update strategy for the Lagrange multipliers corresponding to the power flow constraints in the next iteration is as follows: ； In the formula, , To adjust the parameters, , All values ​​are greater than 0; Let i be the phase angle of node i in the k-th iteration; Let be the Lagrange multiplier corresponding to the power flow constraint of node i in the (k+1)th iteration; The Lagrangian function of the scheduling model; Lagrange function The calculation formula is: ； In the formula, The Lagrange multiplier corresponding to the capacity upper limit constraint of the flexibility resource n connected to node i; The Lagrange multiplier corresponding to the capacity lower bound constraint of the flexibility resource n connected to node i; Let be the Lagrange multiplier corresponding to the power flow constraint at node i.

6. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 5, characterized in that, To reduce the number of variables, a constraint function is used, employing Lagrange multipliers. Update power The update strategy is as follows: ； In the formula, The power of the flexibility resource n connected to node i in the (k+1)th iteration; For power constraint functions, Power limit and power lower limit Definition; if constraint function The value obtained from the internal update is less than the lower power limit. Then it will be fixed at the lower power limit. Conversely, it is fixed at the upper limit of power. .

7. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 6, characterized in that, Lagrange multipliers corresponding to line capacity constraints The update strategy is as follows: ； In the formula, To adjust the parameters, adjust the parameters. ; The Lagrange multiplier corresponding to the line capacity constraint at the k-th iteration; The Lagrange multiplier corresponding to the line capacity constraint at the (k+1)th iteration; Lagrange multipliers constraint functions, constraint functions Updated When less than 0, the constraint function Fix it to 0 to satisfy Constraints not less than 0; Node phase angle The update strategy is as follows: ； In the formula, To adjust the parameters, adjust the parameters. ; Let be the phase angle of node i at the k-th iteration; Let be the phase angle of node i at the (k+1)th iteration.

8. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 7, characterized in that, The objective asymptotic terms for all local variables are the first-order optimality conditions of the objective function and constraints of the aggregated distributed scheduling model, specifically: ； ； ； ； In the formula, Let Lagrange multipliers be the power flow constraints corresponding to node j; The Lagrange multiplier is the line capacity constraint corresponding to the connection between node i and node j.

9. The flexible resource aggregation and scheduling method based on a two-layer consensus algorithm according to claim 8, characterized in that, In the two-level consensus algorithm, the lower-level iteration converges when the changes in all variables satisfy the first convergence criterion. The first convergence criterion is: ； In the formula, The allowable error for the first convergence criterion; In addition to satisfying the first convergence criterion, the convergence condition of the upper-level iteration must also satisfy the second convergence criterion to ensure that the upper and lower levels achieve consistency. The second convergence criterion is as follows: ； In the formula, The allowable error for the second convergence criterion; The asynchronous iteration strategy improves the lower-level iteration of the two-level consensus algorithm. The initial value of the lower-level iteration is taken as the convergence value of the previous iteration, specifically: ； In the formula, This is a local variable of node i in the lower layer during the k-th iteration of the upper layer and the m-th iteration of the lower layer; This is a local variable of node i in the lower layer when the upper layer is in the (k+1)th iteration and the lower layer is in the 0th iteration; In the asynchronous iteration strategy, the lower-level iteration does not need to converge; it only needs to iterate a certain number of times to transmit information to the upper level. The value of is the Lagrange multiplier corresponding to the aggregator node at the end of the next iteration, i.e. , This represents the Lagrange multiplier corresponding to the power flow constraint of node i in the lower layer at the k-th iteration of the upper layer and the m-th iteration of the lower layer. The convergence criterion for the end of the lower layer iteration is: ； In the formula, m represents the number of iterations in the lower layer; p represents the number of iterations set in the lower layer; Representation or operation.

Images