Single-plant-to-multipoint virtual power plant optimal regulation and control method

Abstract

The invention relates to the technical field of power system regulation and control, and provides a single-plant-to-multipoint virtual power plant optimal regulation and control method, which comprises the steps of collecting virtual power plant information and power grid node information; based on the virtual power plant information and the power grid node information, constructing a single-plant-to-multipoint virtual power plant adjustable capability optimal reporting model; establishing an optimal reporting strategy for the adjustable capability of the virtual power plant, and solving the optimal reporting model for the adjustable capability of the single-plant-to-multipoint virtual power plant; establishing an internal instruction decomposition method of the virtual power plant, and decomposing the optimal reporting strategy of the adjustable capability of the virtual power plant; according to the invention, the real-time capability of the virtual power plant can be accurately reflected, and the reporting precision and the resource utilization rate are improved; efficient coordination of multi-point resources is realized, and the resource utilization efficiency is maximized; the competitiveness and flexibility of the virtual power plant are enhanced, and the diversified requirements of the power grid can be quickly responded; the conversion from passive response to active optimization is realized, and the development of intelligent regulation and control is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of power system control, and in particular to an optimal control method for a single-plant to multi-point virtual power plant. Background Art

[0002] The optimization of the control method of virtual power plants has the following significance: the optimization control method can coordinate the use of distributed energy, energy storage equipment and controllable loads to the maximum extent, reduce idle and redundant resources; achieve the optimal allocation of resources, reduce operating costs, and improve the overall economic benefits of virtual power plants; reduce energy loss in the control process, optimize the charging and discharging cycle of energy storage equipment, and reduce equipment maintenance costs. Improve the operational flexibility and intelligence of virtual power plants. The optimization strategy can achieve the coordinated scheduling of wind power, photovoltaics, energy storage and loads, adapt to the changes in demand at different times and in different spaces, and enhance the adaptability and flexibility of virtual power plants; by introducing optimization algorithms such as artificial intelligence, reinforcement learning, genetic algorithms, etc., the control method of virtual power plants can shift from passive response to active prediction and optimization. Improve the competitiveness of virtual power plants. The optimization control method can make virtual power plants more advantageous in terms of resource allocation and market response, enhance the core competitiveness of virtual power plants in the power market, and attract more users to join virtual power plants.

[0003] However, the existing methods have the following shortcomings: 1) Insufficient reporting accuracy. Existing methods are usually based on static evaluation or simple prediction, which makes it difficult to accurately reflect the real-time adjustable capabilities of multi-point resources, especially when resources change dynamically, such as load fluctuations and changes in energy storage capacity; the response characteristics, reliability and coordination of different resources are ignored, resulting in a deviation between the reporting results and the actual control capabilities. 2) Insufficient multi-objective optimization capabilities. Reporting methods often only focus on economic efficiency and ignore multi-objective trade-offs such as control reliability and flexibility; the matching degree between virtual power plants and grid control requirements is not fully considered, which may lead to poor control execution effects. 3) Simplified decomposition methods. Existing decomposition methods mostly use fixed rules or simple optimization models, which lack adaptability to the dynamic characteristics of resources within virtual power plants and multi-objective control requirements; they fail to fully utilize the synergy potential within virtual power plants, resulting in low resource control efficiency. 4) Poor execution flexibility. The instruction decomposition process does not consider the coupling relationship and dynamic response characteristics between multi-point resources, which may lead to delayed or over-adjusted resource responses. Therefore, it is necessary to develop a new optimal control method for single-plant virtual power plants to multiple points to solve existing problems. Summary of the invention

[0004] The present invention provides a single-plant-to-multipoint virtual power plant optimal control method, aiming to solve the existing problems.

[0005] The present invention is implemented as follows: a method for optimal control of a single-plant virtual power plant for multiple points, comprising:

[0006] Collect virtual power plant information and grid node information;

[0007] Based on the virtual power plant information and grid node information, a single-plant-to-multi-point virtual power plant adjustable capacity optimal reporting model is constructed;

[0008] Establish an optimal reporting strategy for the adjustable capacity of a virtual power plant based on a multi-objective raccoon optimization algorithm, and solve the optimal reporting model for the adjustable capacity of a virtual power plant from a single plant to multiple points;

[0009] A virtual power plant internal instruction decomposition method based on a deep incomplete multi-view clustering algorithm is established to decompose the optimal reporting strategy for the adjustable capacity of the virtual power plant.

[0010] Preferably, the virtual power plant information includes: adjustable capacity and sustainable adjustment time, and the grid node information includes: node electricity price, node carbon emission factor and node green electricity factor of each node;

[0011] The collection of the virtual power plant information and the grid node information is shown in formula (1):

[0012] X(t)=[ψ i (t),C i (t),L i (t), ΔP(t), ΔT(t)] formula (1);

[0013] Where: X(t) is the virtual power plant and grid node information collected at time t, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of the grid node i collected at time t, ΔP(t) is the adjustable capacity of the virtual power plant collected at time t, and ΔT(t) is the sustainable adjustment time of the virtual power plant collected at time t.

[0014] Preferably, the construction of an optimal reporting model for the adjustable capacity of a single-plant to multiple-point virtual power plant based on the collected data includes:

[0015] f C (P i (t),T i (t)) Formula (2)

[0016]

[0017] Formula (2) and Formula (3) are the objective function and calculation process of a single virtual power plant constructed; where: F C (P i (t),T i(t)) is the objective function of a single virtual power plant, P i (t) is the adjustable capacity reported by the virtual power plant to node i, T i (t) is the sustainable adjustment time reported by the virtual power plant to node i; F C (t) is the fitness value of the virtual power plant, N is the number of grid nodes, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t, k ψ , k C and k L They are compensation coefficient, carbon emission coefficient and green electricity coefficient respectively.

[0018] Preferably, the constraint conditions of the optimal reporting model of the adjustable capacity of the virtual power plant from a single plant to multiple points include: the adjustable capacity constraint reported by the virtual power plant to each node of the power grid and the sustainable adjustment time constraint reported by the virtual power plant to each node of the power grid;

[0019]

[0020] max(T i (t))≤ΔT(t) Formula (5)

[0021] Formula (4) is the adjustable capacity constraint reported by the virtual power plant to each node in the power grid; where: ΔP(t) is the adjustable capacity of the virtual power plant collected at time t, and N is the number of power grid nodes;

[0022] Formula (5) is the sustainable regulation time constraint reported by the virtual power plant to each node of the power grid; where: ΔT(t) is the sustainable regulation time of the virtual power plant collected at time t.

[0023] Preferably, the establishment of the optimal reporting strategy for the adjustable capacity of the virtual power plant based on the multi-objective raccoon optimization algorithm includes: randomly generating an initial population in the reporting solution space of the virtual power plant; constructing a fitness function according to the objective function of a single virtual power plant, and performing optimization based on the multi-objective raccoon optimization algorithm according to the fitness of the fitness function, wherein the optimization includes crossover operation, mutation operation and constraint modification, and repeating the above steps until a preset number of iterations G is reached. max Or the population converges and outputs the optimal solution;

[0024] X MO = {P i (t),T i (t)},i=1,2,3,…,NFormula (6)

[0025]

[0026] P i new (t) = λP i (1) (t)+(1-λ)P i (2) (t) Formula (8)

[0027] T i new (t) = λT i (1) (t)+(1-λ)T i (2) (t) Formula (9)

[0028] P i mut (t) = P i new (t)+η·δ p Formula (10)

[0029] T i mut (t) = T i new (t)+η·δ T Formula (11)

[0030] P i mut (t) = min(P i mut (t),P max (t)) Formula (12)

[0031] T i mut (t) = min(T i mut (t),T max (t)) Formula (13)

[0032] And there is at least one

[0033]

[0034] Formula (6) represents the initial population, which is randomly generated in the virtual power plant reported solution space, and each individual represents a possible virtual power plant reported solution; where: X MO To initialize the population, P i (t) is the adjustable capacity reported by the virtual power plant to node i, T i (t) is the sustainable adjustment time reported by the virtual power plant to node i, and N is the number of grid nodes;

[0035] Formula (7) is the fitness function; where: is the fitness value of the kth solution, P i (k) (t) is the adjustable capacity of the kth solution, T i (k) (t) sustainable adjustment time of the kth solution, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t, k ψ , k C and k L They are compensation coefficient, carbon emission coefficient and green electricity coefficient respectively;

[0036] Formula (8) and Formula (9) are crossover operations, where two individuals are randomly selected from the population. Perform crossover operations to generate new virtual power plant reporting solutions; where: P i new (t), T i new (t) are the new adjustable capacity and sustainable adjustment time generated after the crossover operation, respectively. i (1) (t), T i (1) (t) are the adjustable capacity and sustainable adjustment time of individual 1, P i (2) (t), T i (2) (t) are the adjustable capacity and sustainable adjustment time of individual 2, respectively, and λ is the crossover coefficient;

[0037] Formula (10) and Formula (11) are mutation operations, which randomly mutate the reported solutions of the virtual power plant to increase diversity; where: P i mut (t), T i mut (t) are the adjustable capacity and sustainable adjustment time after the variation operation, η is the coefficient of variation, δ P and δ T is the random disturbance of adjustable capacity and sustainable adjustment time;

[0038] Formula (12) and Formula (13) are constraint corrections. Constraint checks are performed on the reported solutions of the mutated virtual power plant. If the constraints are not met, they are corrected to meet the constraints. Where: P max (t), T max(t) are the optimal adjustable capacity and optimal sustainable adjustment time retained in the current population, respectively;

[0039] Formula (14) shows the solution Dominate Perform selection update, select the better solution based on the fitness value using non-dominated sorting, and update the population; where: are the fitness functions corresponding to individual 1 and individual 2 respectively;

[0040] Formula (15) represents the optimal solution of the output; where: is the optimal solution for the final output, P i * (t) is the optimal adjustable capacity, T i * (t) is the optimal duration.

[0041] Preferably, the method for establishing an internal instruction decomposition method of a virtual power plant based on a deep incomplete multi-view clustering algorithm includes: dividing users of the virtual power plant based on a deep incomplete multi-view clustering algorithm, using the load characteristics P of each internal user at multiple time points j (t) and transaction time characteristics T j (t), cluster analysis is performed on the multi-view data of each view, deep representation learning is performed on the data of each view, and a low-dimensional feature representation of the view is obtained; the users are divided into i categories using a clustering loss function, and the adjustable capacity P of the optimal solution is issued to each category of users i (t) and sustainable adjustment time T i (t), complete the decomposition of the strategy.

[0042] Preferably, the virtual power plant internal instruction decomposition method of the deep incomplete multi-view clustering algorithm includes:

[0043]

[0044] Formula (16) is to use neural network to learn the features of each perspective; where: is the low-dimensional feature of user j under view v, is the original data of user j under perspective v, M is the number of perspectives, that is, the number of users in the virtual power plant, is the feature extraction network function of the viewing angle v, with parameter θ v ;

[0045] Formula (17) is the cross-view alignment model; where: L cluster For C v is the cluster center of view v, S v is the set of available samples under the viewing angle v;

[0046] Formula (18) is a clustering optimization model, which uses the clustering loss function to classify users into i categories; where: L cluster Classify collections for users, is the cluster center of category i under perspective v, and || ||2 is the Euclidean distance, which is used to measure the similarity between the user and the center of his category.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention proposes an optimal control method for a single plant to multiple points of a virtual power plant, which improves reporting accuracy and resource utilization. The optimal reporting strategy for adjustable capacity based on MOCOA can accurately reflect the real-time capacity of the virtual power plant and reduce control deviations; optimize internal resource coordination, and the instruction decomposition method based on DIMC realizes efficient coordination of multi-point resources and maximizes resource utilization efficiency; enhances the competitiveness and flexibility of virtual power plants, supports multi-objective optimization, and can quickly respond to the diversified needs of power grids. The present invention realizes the transition from passive response to active optimization, promotes the development of intelligent control, and realizes the transition from passive response to active optimization. It is the key to the development of virtual power plants towards intelligence; it can optimize the reporting strategy of a single virtual battery, report the optimal virtual power plant control strategy to different power grid nodes, improve the compensation obtained from power grid nodes, reduce carbon emissions, and increase the proportion of green electricity, which is of great significance to improving the benefits of virtual power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 It is an overall principle flow chart of an embodiment of the present invention; DETAILED DESCRIPTION

[0051] In order to more fully understand the technical content of the present invention, the technical solution of the present invention is further introduced and illustrated in conjunction with specific embodiments below, but is not limited thereto. The technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0052] refer to Figure 1The optimal control method of a single-plant to multiple-point virtual power plant includes: virtual power plant and power grid node information collection technology, a single-plant to multiple-point virtual power plant adjustable capacity optimal reporting model, a virtual power plant adjustable capacity optimal reporting strategy based on MOCOA, and a virtual power plant internal instruction decomposition method based on the DIMC algorithm.

[0053] To achieve the above objectives, the present invention proposes a method for optimal control of a single-plant virtual power plant for multiple points. The method comprises the following steps:

[0054] Step 1: Propose a technology for collecting virtual power plant and grid node information. The collected information includes virtual power plant information and grid node information. The collected virtual power plant information mainly includes adjustable capacity and sustainable adjustment time; the collected grid node information includes node electricity price, node carbon emission factor, and node green power factor of each node.

[0055] For step 1, the collected virtual power plant information and grid node information are as shown in formula (1). The collected virtual power plant information includes: adjustable capacity and sustainable adjustment time. The collected grid node information includes the node compensation electricity price, node carbon emission factor, and node green electricity factor of each node.

[0056] X(t)=[ψ i (t),C i (t),L i (t),ΔP(t),ΔT(t)] Formula (1)

[0057] Where: X(t) is the virtual power plant and grid node information collected at time t. i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t. ΔP(t) is the adjustable capacity of the virtual power plant collected at time t, and ΔT(t) is the sustainable adjustment time of the virtual power plant collected at time t.

[0058] Step 2: Construct an optimal reporting model for the adjustable capacity of a single-plant virtual power plant to multiple points.

[0059] The present invention assumes that a virtual power plant serves multiple grid nodes, and the compensation electricity price, carbon emission factor, and green electricity factor of different grid nodes are different. The model takes the highest compensation electricity price, the lowest carbon emission, and the highest green electricity of each grid node as the objective function, and gives the adjustable capacity of a single virtual power plant to different grid nodes and reports it to the corresponding different grid nodes. Constraints include: the adjustable capacity constraint reported by the virtual power plant to each grid node, that is, the total adjustable capacity reported to each grid node shall not exceed the adjustable capacity of the virtual power plant itself; the sustainable adjustment time constraint reported by the virtual power plant to each grid node, that is, the maximum sustainable adjustment time reported to each grid node shall not exceed the sustainable adjustment time of the virtual power plant itself.

[0060] For step 2, the optimal reporting model for the adjustable capacity of a single-plant virtual power plant to multiple points is constructed as shown in formula (2)-formula (5). The present invention assumes that a virtual power plant can serve multiple grid nodes, and the compensation electricity price, carbon emission factor, and green electricity factor of different grid nodes are different. The model takes the highest compensation electricity price, the lowest carbon emission, and the highest green electricity of each grid node as the objective function, and gives the adjustable capacity of a single virtual power plant to different grid nodes and reports them to the corresponding different grid nodes.

[0061] Formula (2) is the objective function of a single virtual power plant. The objective function aims to maximize the electricity price compensated by each grid node, minimize carbon emissions, and maximize green electricity. The specific calculation process is shown in formula (3).

[0062] Formula (4) is the adjustable capacity constraint reported by the virtual power plant to each grid node, that is, the total adjustable capacity reported to each grid node shall not exceed the adjustable capacity of the virtual power plant itself.

[0063] Formula (5) is the sustainable regulation time constraint reported by the virtual power plant to each grid node, that is, the maximum sustainable regulation time reported to each grid node shall not exceed the sustainable regulation time of the virtual power plant itself.

[0064] min F C (P i (t),T i (t)) Formula (2)

[0065]

[0066] max(T i (t))≤ΔT(t) Formula (5)

[0067] Where: F C (P i (t),T i (t)) is the objective function of a single virtual power plant, P i (t) is the adjustable capacity reported by the virtual power plant to node i, Ti (t) is the sustainable adjustment time reported by the virtual power plant to node i. C (t) is the fitness value of the virtual power plant, N is the number of grid nodes, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t. ψ , k C and k L They are compensation coefficient, carbon emission coefficient and green power coefficient respectively. ΔP(t) is the adjustable capacity of the virtual power plant collected at time t, and ΔT(t) is the sustainable adjustment time of the virtual power plant collected at time t.

[0068] Step 3: Propose an optimal reporting strategy for the adjustable capacity of virtual power plants based on the multi-objective raccoon optimization algorithm (MOCOA).

[0069] The present invention uses MOCOA to solve the optimal reporting model of the adjustable capacity of the virtual power plant constructed in step 2, and gives the final optimal adjustable capacity and sustainable control time reported to each grid node. First, initialize the MOCOA virtual power plant population, randomly generate the initial population in the virtual power plant reporting solution space, build a fitness function according to the single virtual power plant objective function in step 2, and optimize according to the fitness of the fitness function based on the MOCOA algorithm. The optimization includes operations such as crossover operation, mutation operation and constraint correction, and repeat the above steps until the preset number of iterations G is reached. max Or the population converges and outputs the optimal solution, that is, the optimal adjustable capacity and sustainable adjustment time reported by the virtual power plant to each grid node.

[0070] For step 3, the optimal reporting strategy for the adjustable capacity of the virtual power plant based on MOCOA is shown in formula (6)-formula (15). MOCOA is a multi-objective optimization algorithm. The following steps describe its application in virtual power plant optimization. First, initialize the population and randomly generate the initial population in the virtual power plant reporting solution space. Each individual represents a possible virtual power plant reporting solution. The initial population size is N, and each individual in the population Corresponding to a set of decision variables {P i (k) (t),T i (k) (t)}.

[0071] Formula (7) is a fitness function, and the present invention searches for the optimum based on the MOCOA algorithm according to the fitness.

[0072] Formulas (8) and (9) are crossover operations. Two individuals are randomly selected from the population and crossover operations are performed to generate new virtual power plant reporting solutions.

[0073] Formulas (10) and (11) are mutation operations, which randomly mutate the reported solutions of the virtual power plant to increase diversity.

[0074] Formulas (12) and (13) are constraint corrections. Constraint checks are performed on the reported solutions of the mutated virtual power plant. If the constraints are not met, they are corrected to meet the constraints.

[0075] Perform selection update, select the better solution according to the fitness value using non-dominated sorting, and update the population, as shown in formula (14). Formula (14) represents the solution Dominate

[0076] Repeat the above steps until the preset number of iterations is reached or the population converges. Output the optimal solution, as shown in formula (15).

[0077] X MO = {P i (t),T i (t)},i=1,2,3,…,NFormula (6)

[0078]

[0079] P i new (t) = λP i (1) (t)+(1-λ)P i (2) (t) Formula (8)

[0080] T i new (t) = λT i (1) (t)+(1-λ)T i (2) (t) Formula (9)

[0081] P i mut (t) = P i new (t)+η·δ p Formula (10)

[0082] T i mut (t) = T i new (t)+η·δ T Formula (11)

[0083] P i mut (t) = min(P i mut (t),P max (t)) Formula (12)

[0084] T i mut (t) = min(T i mut (t),T max (t)) Formula (13)

[0085] And there is at least one

[0086]

[0087] Where: X MO To initialize the population, P i (t) is the adjustable capacity reported by the virtual power plant to node i, T i (t) is the sustainable adjustment time reported by the virtual power plant to node i, and N is the number of grid nodes. is the fitness value of the kth solution, P i (k) (t) is the adjustable capacity of the kth solution, T i (k) (t) is the sustainable adjustment time of the kth solution, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t. ψ , k C and k L They are compensation coefficient, carbon emission coefficient and green electricity coefficient respectively. i new (t), T i new (t) are the new adjustable capacity and sustainable adjustment time generated after the crossover operation; P i (1) (t), T i (1) (t) are the adjustable capacity and sustainable adjustment time of individual 1, P i (2) (t), T i (2) (t) are the adjustable capacity and sustainable adjustment time of individual 2, respectively, and λ is the crossover coefficient, which is usually a random number. i mut(t), T i mut (t) are the adjustable capacity and sustainable adjustment time after the variation operation, η is the coefficient of variation, δ P and δ T is the random disturbance of adjustable capacity and sustainable adjustment time. max (t), T max (t) are the optimal adjustable capacity and optimal sustainable adjustment time retained in the current population, respectively. are the fitness functions corresponding to individual 1 and individual 2 respectively. is the optimal solution for the final output, P i * (t) is the optimal adjustable capacity, T i * (t) is the optimal duration.

[0088] Step 4: Propose a virtual power plant internal instruction decomposition method based on the Deep Incomplete Multi-view Clustering (DIMC) algorithm.

[0089] For step 4, the division of virtual power plant users is based on DIMC, making full use of the load characteristics P of each internal user at multiple time points. j (t) and transaction time characteristics T j (t) and other multi-view data, and perform cluster analysis to divide users into i categories, and issue the power control strategy P given in step 3 to each category of users i (t) and T i (t).

[0090] Deep Incomplete Multi-view Clustering (DIMC) completes user segmentation through the following steps: deep representation learning is performed on the data of each view to obtain a low-dimensional feature representation of the view; cross-view alignment technology is used to solve the problems of missing and inconsistency data between different views; and a clustering loss function is used to complete user category segmentation.

[0091] The internal instruction decomposition method of a virtual power plant based on the DIMC algorithm proposed in the present invention is shown in formula (16) to formula (18).

[0092] First, multi-view feature representation is performed. A neural network is used to learn the features representing each view, as shown in formula (16).

[0093] Formula (17) is a cross-view alignment model, which mainly eliminates the inconsistency between views through prototype alignment technology.

[0094] Formula (18) is a clustering optimization model, which uses the clustering loss function to classify users into i categories.

[0095] For each type of user i, the power control strategy P given in step 3 is issued i (t) and T i (t).

[0096]

[0097]

[0098] Where: is the low-dimensional feature of user j under view v; is the original data (adjustable capacity, sustainable adjustment time) of user j under perspective v; M is the number of perspectives, that is, the number of users in the virtual power plant; is the feature extraction network function of the viewing angle v, with parameter θ v ; C v is the cluster center of view v; S v is the set of available samples under the viewing angle v; L cluster Classify collections for users, is the cluster center of category i under perspective v; || ||2 is the Euclidean distance, which is used to measure the similarity between the user and the center of his category.

[0099] Based on the above steps, the reporting strategy of a single virtual battery can be optimized, and the optimal virtual power plant control strategy can be reported to different grid nodes, thereby improving the compensation obtained from grid nodes, reducing carbon emissions, and increasing the proportion of green electricity, which is of great significance to improving the efficiency of virtual power plants.

[0100] The optimal control method of a single plant for multi-point virtual power plants improves reporting accuracy and resource utilization. The optimal reporting strategy based on MOCOA's adjustable capacity can accurately reflect the real-time capabilities of the virtual power plant and reduce control deviations; optimizes internal resource coordination, and the instruction decomposition method based on DIMC realizes efficient coordination of multi-point resources and maximizes resource utilization efficiency; enhances the competitiveness and flexibility of virtual power plants, supports multi-objective optimization, and can quickly respond to the diversified needs of power grids; realizes the transformation from passive response to active optimization, promotes the development of intelligent control, and realizes the transformation from passive response to active optimization, which is the key to the intelligent development of virtual power plants.

[0101] The embodiments described above are only part of the embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any ordinary technician in the field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, and they should be covered by the protection scope of the present invention.

Claims

1. A method for optimal control of a single-plant virtual power plant to multiple points, characterized in that: include: Collect virtual power plant information and grid node information; Based on the virtual power plant information and grid node information, a single-plant-to-multi-point virtual power plant adjustable capacity optimal reporting model is constructed; Establish an optimal reporting strategy for the adjustable capacity of a virtual power plant based on a multi-objective raccoon optimization algorithm, and solve the optimal reporting model for the adjustable capacity of a virtual power plant from a single plant to multiple points; A virtual power plant internal instruction decomposition method based on a deep incomplete multi-view clustering algorithm is established to decompose the optimal reporting strategy for the adjustable capacity of the virtual power plant.

2. The optimal control method of a single-plant to multi-point virtual power plant according to claim 1 is characterized in that: The virtual power plant information includes: adjustable capacity and sustainable adjustment time, and the grid node information includes: node electricity price, node carbon emission factor and node green power factor of each node; The collection of the virtual power plant information and the grid node information is shown in formula (1): X(t) = [ψ i , i (t), C i (t), L i (t), ΔP(t), ΔT(t)] Equation (1); Where: X(t) is the virtual power plant and grid node information collected at time t, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of the grid node i collected at time t, ΔP(t) is the adjustable capacity of the virtual power plant collected at time t, and ΔT(t) is the sustainable adjustment time of the virtual power plant collected at time t.

3. The optimal control method of a single-plant-to-multi-point virtual power plant according to claim 1, characterized in that: The construction of the optimal reporting model for the adjustable capacity of a single-plant to multiple-point virtual power plant includes: min F C (P i (t),T i (t)) Formula (2); Formula (2) and Formula (3) are the objective function and calculation process of a single virtual power plant constructed; where: F C (P i (t),T i (t)) is the objective function of a single virtual power plant, P i (t) is the adjustable capacity reported by the virtual power plant to node i, T i (t) is the sustainable adjustment time reported by the virtual power plant to node i; F C (t) is the fitness value of the virtual power plant, N is the number of grid nodes, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t, k ψ , k C and k L They are compensation coefficient, carbon emission coefficient and green electricity coefficient respectively.

4. The optimal control method of a single-plant to multi-point virtual power plant according to claim 3 is characterized in that: The constraint conditions of the optimal reporting model of the adjustable capacity of the virtual power plant from a single plant to multiple points include: the adjustable capacity constraint reported by the virtual power plant to each node of the power grid and the sustainable adjustment time constraint reported by the virtual power plant to each node of the power grid; max(T i (t))≤ΔT(t) Formula (5) Formula (4) is the adjustable capacity constraint reported by the virtual power plant to each node in the power grid; where: ΔP(t) is the adjustable capacity of the virtual power plant collected at time t, and N is the number of power grid nodes; Formula (5) is the sustainable regulation time constraint reported by the virtual power plant to each node of the power grid; where: ΔT(t) is the sustainable regulation time of the virtual power plant collected at time t.

5. The method for optimal control of a single-plant-to-multi-point virtual power plant according to claim 1, characterized in that: The establishment of the optimal reporting strategy for the adjustable capacity of a virtual power plant based on the multi-objective raccoon optimization algorithm includes: randomly generating an initial population in the reporting solution space of the virtual power plant; constructing a fitness function according to the objective function of a single virtual power plant, and performing optimization based on the fitness of the fitness function based on the multi-objective raccoon optimization algorithm, wherein the optimization includes crossover operation, mutation operation and constraint modification, and repeating the above steps until a preset number of iterations G is reached. max Or the population converges and outputs the optimal solution; X MO = {P i (t),T i (t)},i=1,2,3,…,NFormula (6); P i new (t) = λP i (1) (t)+(1-λ)P i (2) (t) Formula (8); T i new \(\tau(t)=\lambda T\) i (1) +(1 - \(\lambda\))T i (2) \(\tau(t)\) formula (9); P i mut P(t) = i new P(t) + η·δ p Formula (10); T i mut (t) = T i new (t) + η·δ T Formula (11); P i mut (t) = min(P i mut (t),P max (t)) Formula (12); T i mut (t) = min(T i mut (t),T max (t)) Formula (13); Formula (6) represents the initial population, which is randomly generated in the virtual power plant reported solution space, and each individual represents a possible virtual power plant reported solution; where: X MO To initialize the population, P i (t) is the adjustable capacity reported by the virtual power plant to node i, T i (t) is the sustainable adjustment time reported by the virtual power plant to node i, and N is the number of grid nodes; Formula (7) is the fitness function; where: is the fitness value of the kth solution, P i (k) (t) is the adjustable capacity of the kth solution, T i (k) (t) sustainable adjustment time of the kth solution, ψ i (t) is the node compensation electricity price of grid node i collected at time t, C i (t) is the carbon emission factor of grid node i collected at time t, L i (t) is the green power factor of grid node i collected at time t, k ψ , k C and k L They are compensation coefficient, carbon emission coefficient and green electricity coefficient respectively; Formula (8) and Formula (9) are crossover operations, where two individuals are randomly selected from the population. Perform crossover operations to generate new virtual power plant reporting solutions; where: P i new (t), T i new (t) are the new adjustable capacity and sustainable adjustment time generated after the crossover operation, respectively. i (1) (t), T i (1) (t) are the adjustable capacity and sustainable adjustment time of individual 1, P i (2) (t), T i (2) (t) are the adjustable capacity and sustainable adjustment time of individual 2, respectively, and λ is the crossover coefficient; Formula (10) and Formula (11) are mutation operations, which randomly mutate the reported solution of the virtual power plant; where: P i mut (t), T i mut (t) are the adjustable capacity and sustainable adjustment time after the variation operation, η is the coefficient of variation, δ P and δ T is the random disturbance of adjustable capacity and sustainable adjustment time; Formula (12) and Formula (13) are constraint corrections. Constraint checks are performed on the reported solutions of the mutated virtual power plant. If the constraints are not met, they are corrected to meet the constraints. Where: P max (t), T max (t) are the optimal adjustable capacity and optimal sustainable adjustment time retained in the current population, respectively; Formula (14) shows the solution Dominate Perform selection update, select the better solution based on the fitness value using non-dominated sorting, and update the population; where: are the fitness functions corresponding to individual 1 and individual 2 respectively; Formula (15) represents the optimal solution of the output; where: is the optimal solution for the final output, P i * (t) is the optimal adjustable capacity, T i * (t) is the optimal duration.

6. The method for optimal control of a single-plant-to-multi-point virtual power plant according to claim 1, characterized in that: The method for establishing an internal instruction decomposition method of a virtual power plant based on a deep incomplete multi-view clustering algorithm includes: dividing users of the virtual power plant based on a deep incomplete multi-view clustering algorithm, using the load characteristics P of each internal user at multiple time points j (t) and transaction time characteristics T j (t), cluster analysis is performed on the multi-view data of each view, and deep representation learning is performed on the data of each view to obtain a low-dimensional feature representation of the view; the users are divided into i categories using the clustering loss function, and the optimal solution adjustable capacity P is issued to each category of users. i (t) and sustainable adjustment time T i (t), complete the decomposition of the strategy.

7. The method for optimal control of a single-plant-to-multi-point virtual power plant according to claim 1 or 6, characterized in that: The virtual power plant internal instruction decomposition method of the deep incomplete multi-view clustering algorithm includes: Formula (16) is to use neural network to learn the features of each perspective; where: is the low-dimensional feature of user j under view v, is the original data of user j under view v, M is the number of view points, i.e. the number of users in the virtual power plant, and f θv is the feature extraction network function of the viewing angle v, with parameter θ v ; Formula (17) is the cross-view alignment model; where: C v is the cluster center of view v, S v is the set of available samples under the viewing angle v; Formula (18) is a clustering optimization model, which uses the clustering loss function to classify users into i categories; where: L cluster Classify collections for users, is the cluster center of category i under perspective v, and || ||2 is the Euclidean distance, which is used to measure the similarity between the user and the center of his category.

Images