Distributed energy system control method, apparatus, device, and medium

Abstract

This invention provides a control method, apparatus, device, and medium for a distributed energy system, relating to the field of energy system operation control technology. The method includes: constructing a multi-dimensional operating characteristic vector of producers and consumers; quantifying operating state observation and parameter sharing behavior based on a communication topology graph model; establishing an operating state observation mechanism to adjust actual net power; establishing a parameter sharing mechanism and updating neighborhood coordination factors in conjunction with quasi-static coupling; dividing multi-operator mutual aid modes; constructing a graph neural network-master-slave distributed collaborative control model, with the upper layer adjusting control parameters and the lower layer using a graph convolutional network to extract neighborhood features and optimize power allocation and point-to-point mutual aid plans; introducing a trusted distributed mirror learning framework; and performing privacy protection verification through static graph convolutional embedding compression, splitting computation, and zero-knowledge proof; iteratively solving the Nash equilibrium to determine the net power and output power of each producer and consumer. This invention can realize collaborative optimization control of multi-entity energy systems.

Description

Technical Field

[0001] This invention relates to the field of energy system operation and control technology, and in particular to a method, apparatus, equipment and medium for controlling distributed energy systems. Background Technology

[0002] With the widespread integration of distributed energy resources, the traditional power load side is gradually transforming into a prosumer, possessing both power generation and consumption capabilities. Prosumers refer to energy terminal nodes equipped with distributed generation units, energy storage devices, and adjustable loads. Physical forms include intelligent building energy management systems, microgrid node controllers, electric vehicle charging and discharging units, and integrated energy systems for industrial users. Multi-entity energy systems thus exhibit a new characteristic of high uncertainty on both the source and load sides. How to improve the utilization efficiency of distributed resources while maintaining system power balance and operational stability through effective collaborative control mechanisms has become a key technical issue in the operation and control of new power systems. However, current research on distributed control for multi-prosumer systems still faces many challenges. On the one hand, prosumers exhibit significant individual differences in operational characteristics such as power regulation rate, contribution of new energy consumption, and control response delay. Traditional unified control parameters or single-type node models are difficult to accurately adapt to this heterogeneity, resulting in insufficient targeting and effectiveness of control strategies. On the other hand, the power allocation decisions of producers and consumers are not only driven by system scheduling signals, but also greatly influenced by the communication topology between them and other producers and consumers. Distributed collaborative behaviors such as operation status observation and parameter sharing are playing an increasingly prominent role in the field of energy control, but existing distributed control models have not yet systematically quantified and integrated these communication interaction mechanisms into the control framework.

[0003] The mutual support relationships between multiple operator control areas and the privacy protection requirements in the collaborative control solution process further exacerbate the difficulty of designing distributed control mechanisms. In multi-entity systems composed of multiple community energy operators, there are both collaborative needs for power mutual support and independent autonomous operation requirements among operator control areas. This complex mutual support mode directly affects the power allocation pattern and the balance of power exchange between areas. However, existing master-slave distributed control models often treat operators as independent control units, ignoring the dynamic changes in the mutual support mode and its endogenous impact on control parameters. At the same time, the iterative solution of distributed collaborative control usually requires producers and consumers to expose individual operating parameters and real-time power data, which raises serious risks of operational data privacy leakage. Existing privacy protection methods often sacrifice convergence speed or control optimality. Therefore, there is an urgent need for an efficient distributed collaborative control method that can integrate the multidimensional operational characteristics differences of producers and consumers and the multiple communication interactions between entities, while taking into account both multi-operator mutual support control and operational data privacy protection. Summary of the Invention

[0004] This invention provides a distributed energy system control method, device, equipment, and medium. Through a series of technical means, including multi-dimensional operational characteristic modeling of producers and consumers, communication topology graph construction, operational status observation and parameter sharing quantification, multi-operator mutual assistance mode division, graph neural network-master-slave distributed collaborative control, distributed sequential update, quasi-static coupling, and trusted distributed mirror learning privacy-preserving solution, it effectively guides producers and consumers to participate in system power regulation according to their own operational characteristics, achieves a balance between collaborative mutual assistance and independent autonomy among multi-operator control areas, and efficiently solves distributed control strategies while protecting the privacy of operational data of each entity, thus realizing collaborative optimization control of multi-entity energy systems.

[0005] In a first aspect, the present invention provides a control method for a distributed energy system, the method comprising: Historical operating data of each producer-consumer is obtained. Based on the historical operating data, a multi-dimensional operating characteristic vector of the producer-consumer is constructed from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay. The basic power plan of each producer-consumer is generated based on the multi-dimensional operating characteristic vector. Based on the communication topology graph model, the operating status observation and parameter sharing behavior between producers-consumers are quantified according to the communication link status between each producer-consumer to obtain the producer-consumer control model. Based on the producer-consumer control model and the basic power plan, an operational status observation mechanism for each producer-consumer is established. The actual net power of each producer-consumer is adjusted according to the operational status observation mechanism to obtain the adjusted net power data. A parameter sharing mechanism is established. The neighborhood coordination factor is updated through comprehensive operational status evaluation according to the parameter sharing mechanism. The neighborhood coordination factor is updated across control cycles using a quasi-static coupling mechanism to obtain the updated neighborhood coordination factor. The mutual support relationships between the control areas of various multi-energy community operators are obtained, and these relationships are divided into three mutual support modes: collaborative mutual support, independent operation, and mutual backup. A graph neural network-master-slave distributed collaborative control model is constructed based on these mutual support modes. In the graph neural network-master-slave distributed collaborative control model, the upper-layer multi-operators adjust the control parameters according to the mutual support modes to obtain the adjusted upper-layer control parameters. The lower-layer producers and consumers use a graph convolutional network to extract the communication neighborhood operation features to obtain node embedding features, and optimize their own power allocation and P2P power mutual support plan based on the node embedding features and the updated neighborhood collaboration factor. A trusted distributed mirror learning framework is introduced to compress the operational characteristic parameters of prosumers using static graph convolutional embedding to obtain compressed embedding vectors. Split computation is used to map the local operational state vectors of each prosumer to sub-values, and zero-knowledge proof is used to verify the availability compliance of the communication link for P2P power mutual assistance, thus obtaining compliance verification results. Under the premise of protecting the privacy of the operational data of each entity, the Nash equilibrium of the graph neural network-master-slave distributed collaborative control model is iteratively solved based on the compressed embedding vector, the sub-value, and the compliance verification result to obtain the net power of each producer and consumer in the current time period and the power sent to the communication neighbor, thereby realizing the collaborative optimization control of the multi-entity energy system.

[0006] In one possible design, the multidimensional operational characteristic vector of the prosumer is: ,in The producer-consumer power regulation rate coefficient, Parameters for the contribution of new energy consumption. To control the response delay factor; The power regulation rate coefficient Determined by the following formula: (1) in: For producers and consumers i The change in power, For producers and consumers i The rated power is the power of the producer and consumer. i Maximum allowable power of grid connection interface To adjust the time window; The system regulation demand signal perceived by prosumers is calculated using the following formula: (2) in: For producers and consumers i During the period t Actual perceived regulatory demand signals; For operators During the period t The control signals sent out For producers and consumers i The penetration rate of new energy sources in the controlled area As a reference signal; The basic power plan is derived from the power allocation algorithm under operating constraints. The basic power plan is positively correlated with the adjustment demand signal and also positively correlated with the contribution parameter of new energy consumption.

[0007] In one possible design, the communication topology graph model is an undirected weighted graph. ,in For the set of producer-consumer nodes, N For the number of producer-consumer nodes, ; Let be the set of edges, representing the communication links between producers and consumers; Let be an adjacency matrix, where elements Indicates producer-consumer i and j The communication link quality weights between them are determined, and they satisfy the properties of being undirected and free of self-loops. It is the set of real numbers; P2P power sharing is constrained by the availability of communication links as follows: (3) in: and Time periods t Producers and consumers i To consumers j The transmitted power and the received power, This is the upper limit for a single power exchange. The operational status observation mechanism adjusts the actual net power of producer-consumer according to the following formula. : (4) (5) (6) in: Based on power planning, For operational status observation coefficients, For communication neighbor set, For neighboring observable power states, For power allocation preference parameters, For producers and consumers i exist t The neighborhood average observed power at time 10:00. For communication neighbors j exist t Actual net operating power at any given time For communication neighbors j exist t The average observed power of the neighborhood at time -1.

[0008] In one possible design, the parameter sharing mechanism updates the comprehensive operating status assessment and adjusts the neighborhood cooperation factor according to the following formula: (7) (8) in: As a weight for evaluating its own local operating status, This is a local operating status assessment value. This is a comprehensive operational status assessment value. For parameter sharing sensitivity coefficient, As an effective neighborhood collaboration factor, For producers and consumersi exist t The original neighborhood cooperation factor at time; The update rule of the quasi-static coupling mechanism is expressed as follows: (9) in: For slowly varying coefficients, The baseline value is the scaled value of the graph neural network embedding. As a neighborhood collaboration factor, k This is the iteration number of the distributed collaborative control cycle. This update is only performed across control cycles. Within a single distributed collaborative control cycle... It is approximately a constant.

[0009] In one possible design, the mutual assistance mode is implemented through a mutual assistance coefficient matrix. Description, in which These respectively represent mutual backup, independent operation, and coordinated support. M Total number of operators; Operators m During the period t Based on its own control parameters The mutual adjustment term adjusts the control strategy, wherein the mutual adjustment term is defined as: (11) (12) in: The mutual adjustment coefficient, To control the coupling strength of parameters, For operators n During the period t The control parameters, For operators m During the period t Mutual adjustment items, For all operators during the time period t The maximum value of the lower control parameter.

[0010] In one possible design, the lower-layer prosumer utilizes a two-layer graph convolutional network to extract communication neighborhood operational features, and the node embedding update method is as follows: (13) in: l For graph convolutional layer index, For weighted degrees, For the first l Layer trainable weight matrix, It is the ReLU activation function. For nodes i In the l+1 layer output embedding features, For producer-consumer nodes i itself, For nodes j The weighting degree, For nodes j In the l The layer's input embedding features; Final Embedding Used for adaptive scaling of neighborhood cooperability factor Shared sensitivity coefficient with parameters The mapping function is shown in the following formula: (14) in: , The initial baseline values ​​correspond to the producer-consumer nodes respectively. i Before introducing graph convolutional embedding features, the preset basic neighborhood collaborative weights and basic parameter sharing sensitivity thresholds are used. The parameter vector is a learnable vector, and the sigmoid function is a sigmoid function. The lower-level prosumers adopt a distributed sequential update protocol, where a random permutation of the prosumer set is... For current update consumers The average power of its neighbors is calculated using the following formula: (15) Among them: updated neighbors use the latest power allocation value for the current time period, while unupdated neighbors use the value for the previous time period.

[0011] In one possible design, the static graph convolutional embedding compression employs an information functional. Local running information set The vector is compressed into an embedding vector, as shown in equation (16): (16) in: e i For producers and consumers i Embedded vector, θ i For multidimensional operational characteristic vectors, d This is the embedding dimension; the embedding is calculated only once during system initialization and is not exchanged during subsequent control processes. In the splitting calculation, each producer-consumer locally calculates its sub-value. The calculation formula is: (17) in: For the private projection vector of the prosumer, For producers and consumersi Local runtime state vector, It is the transpose symbol; The zero-knowledge proof is used for compliance verification of the availability of communication links in P2P power exchange. The sender generates the proof using the following formula. : (18) in: "Random number" and "public" indicate publicly available information. For random number based For initial parameters The value obtained by performing the calculation Here, com represents the initial parameter values ​​related to the sender's private information, private represents the sender's private information, and Prove represents the zero-knowledge proof generation function. zk Evidence for generating zero-knowledge proofs; Receiver verification: (19) in: These are initial parameter values ​​related to the sender's private information; If they are equal, it proves that the sender meets the communication link availability constraint, and the compliance verification result is obtained. The Nash equilibrium of the graph neural network-master-slave distributed collaborative control model is iteratively solved based on the compressed embedding vector, the sub-values, and the compliance verification results, ultimately determining each producer-consumer. i During the period t Net power and to communication neighbors j P2P output power This enables coordinated and optimized control of multiple energy systems.

[0012] In a second aspect, the present invention provides a distributed energy system control device, the device comprising: The characteristic modeling and quantification module is configured to acquire historical operating data of each producer-consumer, construct a multi-dimensional operating characteristic vector of the producer-consumer based on the historical operating data from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay, and generate a basic power plan for each producer-consumer based on the multi-dimensional operating characteristic vector; based on the communication topology graph model, quantify the operating status observation and parameter sharing behavior between producers-consumers according to the communication link status between each producer-consumer, and obtain the producer-consumer control model; The state observation and collaborative update module is configured to establish an operational state observation mechanism for each producer and consumer based on the producer-consumer control model and the basic power plan; adjust the actual net power of each producer and consumer based on the operational state observation mechanism to obtain the adjusted net power data; establish a parameter sharing mechanism; update the neighborhood collaboration factor through comprehensive operational state evaluation based on the parameter sharing mechanism; and use a quasi-static coupling mechanism to update the neighborhood collaboration factor across control cycles to obtain the updated neighborhood collaboration factor. The mutual assistance graph neural network optimization module is configured to acquire the mutual assistance relationships between the control areas of various multi-energy community operators, and divide the mutual assistance relationships into three types of mutual assistance modes: collaborative mutual assistance, independent operation, and mutual backup. Based on the mutual assistance modes, a graph neural network-master-slave distributed collaborative control model is constructed. In the graph neural network-master-slave distributed collaborative control model, the upper-layer multi-operators adjust the control parameters according to the mutual assistance modes to obtain the adjusted upper-layer control parameters. The lower-layer producers and consumers use a graph convolutional network to extract the communication neighborhood operation features to obtain node embedding features, and optimize their own power allocation and P2P power mutual assistance plan based on the node embedding features and the updated neighborhood collaboration factor. The embedded split zero-knowledge module is configured to introduce a trusted distributed mirror learning framework to perform static graph convolutional embedding compression on the operational characteristic parameters of prosumers to obtain compressed embedding vectors; splitting computation is used to map the local operational state vectors of each prosumer to sub-values, and zero-knowledge proof is used to perform compliance verification of the availability of communication links for P2P power mutual assistance to obtain compliance verification results. The privacy-preserving Nash solution allocation module is configured to iteratively solve the Nash equilibrium of the graph neural network-master-slave distributed collaborative control model based on the compressed embedding vector, the sub-values, and the compliance verification results, while protecting the privacy of the operational data of each entity, to obtain the net power of each producer and consumer in the current time period and the power sent to the communication neighbor, thereby realizing the collaborative optimization control of the multi-entity energy system.

[0013] Thirdly, embodiments of the present invention provide an electronic device, comprising: at least one processor and a memory; the memory storing computer execution instructions; the at least one processor executing the computer execution instructions stored in the memory, causing the at least one processor to perform the distributed energy system control method as described in the first aspect and various possible designs of the first aspect.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the distributed energy system control method described in the first aspect and various possible designs of the first aspect.

[0015] Fifthly, embodiments of the present invention provide a computer program product, including a computer program that, when executed by a processor, implements the distributed energy system control method described in the first aspect and various possible designs of the first aspect.

[0016] The distributed energy system control method, apparatus, equipment, and medium provided by this invention have at least the following beneficial effects: This invention not only characterizes the individual operational differences of each producer-consumer in terms of power regulation rate, contribution of new energy consumption, and control response delay by constructing a multi-dimensional operational characteristic vector of the producers-consumers, but also quantifies the operational status observation and parameter sharing behavior among producers-consumers based on a communication topology graph model. It achieves dynamic adjustment of operational status observation and parameter sharing by establishing an operational status observation mechanism to adjust actual net power and establishing a parameter sharing mechanism in conjunction with quasi-static coupling to update neighborhood coordination factors. Simultaneously, it constructs a graph neural network-master-slave distributed collaborative control model, dividing it into multi-operator mutual aid modes. The upper layer adjusts control parameters, while the lower layer uses a graph convolutional network to extract communication neighborhood features and optimize power allocation and P2P power mutual aid plans, achieving collaborative optimization between the upper and lower layers. Furthermore, it introduces a trusted distributed mirror learning framework, using static graph convolutional embedding compression, split computation, and zero-knowledge proof for privacy protection and compliance verification results. Under the premise of protecting the privacy of operational data of each entity, iteratively solves the Nash equilibrium, ultimately determining the net power of each producer-consumer and the power sent to communication neighbors, realizing collaborative optimization control of a multi-entity energy system. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0018] Figure 1 A flowchart of a distributed energy system control method provided in an embodiment of the present invention; Figure 2 This is a diagram illustrating the overall control architecture of a multi-productive-consumer community energy system provided in an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the classification and operational characteristics of producer-consumer communication topologies provided in an embodiment of the present invention. Figure 4The system topology and distributed resource configuration diagram of the test case provided in the embodiments of the present invention are as follows: (a) the producer-consumer communication link quality weight matrix at the moment of the first P2P power exchange; (b) the communication link quality weight matrix at the moment of the last P2P power exchange in scenario 3; (c) the communication link quality weight change matrix after all P2P power exchanges are completed in scenario 3; (d) the communication link quality weight matrix at the moment of the last P2P power exchange in scenario 4; (e) the communication link quality weight change matrix after all P2P power exchanges are completed in scenario 4; (f) the communication link quality weight matrix at the moment of the last P2P power exchange in scenario 5; and (g) the communication link quality weight change matrix after all P2P power exchanges are completed in scenario 5. Figure 5 This is a heatmap evolution sequence diagram of the communication link quality weight matrix in scenarios 3 to 5 provided in the embodiments of the present invention; Figure 6 Lorentz curves of P2P power mutual assistance in scenarios 3, 4, and 5 provided in the embodiments of the present invention; Figure 7 The present invention provides a heatmap showing the dynamic switching of the operator mutual assistance mode coefficient matrix in scenarios 4 and 5 during multiple time periods within a day. Subgraphs (a)-(c) show the changes in the mutual assistance mode coefficient matrix of scenario 4 under three different operating periods, while subgraphs (d)-(f) show the changes in the mutual assistance mode coefficient matrix of scenario 5 under three corresponding operating periods. Figure 8 This is a structural diagram of a distributed energy system control device provided in an embodiment of the present invention.

[0019] The accompanying drawings have illustrated specific embodiments of the invention, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0021] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision and disclosure of relevant data and information comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0022] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of the present invention. However, they do not mean that the inventor has used or necessarily used the solution.

[0023] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0024] Example 1: This invention provides a control method for a distributed energy system, such as... Figure 1 As shown, the control method for the distributed energy system includes the following steps S10-S50.

[0025] S10: Obtain historical operating data for each producer-consumer, construct a multi-dimensional operating characteristic vector for the producer-consumer based on the historical operating data from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay, and generate a basic power plan for each producer-consumer based on the multi-dimensional operating characteristic vector; based on the communication topology graph model, quantify the operating status observation and parameter sharing behavior between producers-consumers according to the communication link status between each producer-consumer, and obtain the producer-consumer control model.

[0026] In this embodiment, a multi-dimensional operating characteristic vector of producers and consumers is constructed from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay. Based on the communication topology graph model, the operating status observation and parameter sharing behavior among producers and consumers are quantified to form a producer and consumer control model that integrates multi-dimensional operating characteristics and communication topology influence.

[0027] In some embodiments, the multidimensional operational characteristic vector of prosumers is: The power regulation rate coefficient Defined according to formula (1): (1) in: For producers and consumers i The change in power, For producers and consumers i The rated power is the power of the producer and consumer. i Maximum allowable power of grid connection interface To adjust the time window; The parameter for the contribution of new energy consumption, with a range of values. ; To control the response delay factor, the unit is hours, and the value range is... .

[0028] The system regulation demand signal perceived by producers and consumers is calculated according to equation (2): (2) in: For producers and consumers i During the period t Actual perceived regulatory demand signals; For operators During the period t The control signals sent out For producers and consumers i The penetration rate of new energy sources in the controlled area.

[0029] S20: Based on the producer-consumer control model and the basic power plan, establish an operation status observation mechanism for each producer-consumer, adjust the actual net power of each producer-consumer according to the operation status observation mechanism, and obtain the adjusted net power data; establish a parameter sharing mechanism, update the neighborhood coordination factor through comprehensive operation status evaluation according to the parameter sharing mechanism, and use a quasi-static coupling mechanism to update the neighborhood coordination factor across control cycles, and obtain the updated neighborhood coordination factor.

[0030] In some embodiments, a communication topology diagram is defined. ,in For the set of producer-consumer nodes, ; Let be the set of edges, representing the communication links between producers and consumers; For an adjacency matrix, the elements are... Indicates producer-consumer i and j The communication link quality weights between them, and satisfying undirectedness and no self-loops, i.e. , P2P power exchange is constrained by the availability of the communication link as shown in equation (3): (3) in: and Time periods t Producers and consumers i Towards j The transmitted power and the received power, This is the upper limit for a single power exchange.

[0031] The operational status observation mechanism adjusts the actual net power of producers and consumers according to equations (4) to (6). : (4) (5) (6) in: Based on power planning, For operational status observation coefficients, As a neighborhood collaboration factor, For power allocation preference parameters, For communication neighbor set, For the observability state of neighbors, For producers and consumers i exist t The neighborhood average observed power at time 10:00. For communication neighbors j exist t Actual net operating power at any given time For communication neighbors j exist t The average observed power of the neighborhood at time -1.

[0032] In some embodiments, the parameter sharing mechanism updates the comprehensive operating status assessment and adjusts the neighborhood collaboration factor according to equations (7) and (8): (7) (8) in: As a weight for evaluating its own local operating status, This is a local operating status assessment value. For parameter sharing sensitivity coefficient, As an effective neighborhood collaboration factor, For producers and consumers i exist t The original neighborhood cooperating factor at time.

[0033] The neighborhood coordination factor adopts a quasi-static coupling mechanism, and the update rule is shown in equation (9): (9) in: For slowly varying coefficients, The baseline value is the scaled value of the graph neural network embedding. As a neighborhood collaboration factor, k This is the iteration number of the distributed collaborative control cycle. This update is only performed across control cycles. Within a single distributed collaborative control cycle... It is approximately a constant.

[0034] S30: Obtain the mutual support relationship between the control areas of various multi-energy community operators, and divide the mutual support relationship into three types of mutual support modes: collaborative mutual support, independent operation, and mutual backup. Construct a graph neural network-master-slave distributed collaborative control model based on the mutual support mode. In the graph neural network-master-slave distributed collaborative control model, the upper-layer multi-operators adjust the control parameters according to the mutual support mode to obtain the adjusted upper-layer control parameters. The lower-layer producers and consumers use a graph convolutional network to extract the communication neighborhood operation features to obtain node embedding features, and optimize their own power allocation and P2P power mutual support plan based on the node embedding features and the updated neighborhood collaboration factor.

[0035] In this embodiment, the mutual support relationship between the control areas of multi-energy community operators is divided into three types of mutual support modes: collaborative mutual support, independent operation, and mutual backup. A graph neural network-master-slave distributed collaborative control model is constructed. In this model, the upper-layer multi-operators adjust the control parameters according to the mutual support mode, and the lower-layer producers and consumers use graph convolutional networks to extract the operating characteristics of the communication neighborhood and optimize their own power allocation and P2P power mutual support plan.

[0036] In some embodiments, the operator mutual assistance mode uses a mutual assistance coefficient matrix. Description, in which These respectively represent mutual backup, independent operation, and collaborative support; operators m During the period t Based on its own control parameters Adjustment strategies for mutual assistance and adjustment items.

[0037] The mutual assistance adjustment term is defined as: (11) (12) in: The mutual adjustment coefficient, To control the coupling strength of parameters, For operators m During the period t Mutual adjustment items, For all operators during the time period t The maximum value of the lower control parameter.

[0038] In some embodiments, the lower-layer prosumer uses a two-layer graph convolutional network to extract communication neighborhood operation features, and the node embedding update rule is shown in Equation (13): (13) in: For weighted degrees, For the first Layer trainable weight matrix, It is the ReLU activation function. For nodes i In the l +1 layer output embedding features, For producer-consumer nodes i itself, For nodes j The weighting degree, For nodes j In the l Input embedding features of the layer.

[0039] Final Embedding Used for adaptive scaling of neighborhood cooperability factor Shared sensitivity coefficient with parameters The mapping function is shown in equation (14): (14) in: , The initial baseline values ​​correspond to the producer-consumer nodes respectively. i Before introducing graph convolutional embedding features, the preset basic neighborhood collaborative weights and basic parameter sharing sensitivity thresholds are used. Let be a learnable parameter vector, and sigmoid be a sigmoid function.

[0040] The lower-level prosumers adopt a distributed sequential update protocol. Let a random permutation of the prosumer set be... For current update consumers The average power of its neighbors is calculated according to formula (15): (15) Among them: updated neighbors use the latest power allocation value for the current time period, while unupdated neighbors use the value for the previous time period.

[0041] S40: Introduce a trusted distributed mirror learning framework to compress the operational characteristic parameters of prosumers using static graph convolutional embedding to obtain compressed embedding vectors; use split computation to map the local operational state vectors of each prosumer into sub-values, and use zero-knowledge proof to perform compliance verification of the availability of the communication link for P2P power mutual assistance, and obtain compliance verification results.

[0042] In this embodiment, a trusted distributed mirror learning framework is introduced to compress the operational characteristic parameters of producers and consumers using static graph convolutional embedding, and a split computation and zero-knowledge proof method is adopted.

[0043] In some embodiments, static graph convolutional embedding compression employs an information functional. Local running information set The vector is compressed into an embedding vector, as shown in equation (16): (16) In the decomposition calculation method, each producer and consumer calculates the sub-value locally. As shown in equation (17): (17) in: For the private projection vector of the prosumer, For producers and consumers i Local runtime state vector, This is the transpose symbol. This embedding is calculated only once during system initialization and is not swapped during subsequent control processes.

[0044] In some embodiments, zero-knowledge proof methods are used for compliance verification of the availability of communication links in P2P power exchange, with the sender generating evidence. As shown in equation (18): (18) in: "Random number" and "public" indicate publicly available information. For random number based For initial parameters The value obtained by performing the calculation Here, `com` represents the initial parameter values ​​related to the sender's private information, `private` represents the generated commitment value, and `Prove` represents the zero-knowledge proof generation function. zk This is the evidence for generating zero-knowledge proofs.

[0045] Receiver verification (19) in: Initial parameter values ​​related to the sender's private information.

[0046] If they are equal, it proves that the sender satisfies the communication link availability constraint. .

[0047] S50: Under the premise of protecting the privacy of the operating data of each entity, the Nash equilibrium of the graph neural network-master-slave distributed collaborative control model is iteratively solved according to the compressed embedding vector, the sub-value and the compliance verification result to obtain the net power of each producer and consumer in the current time period and the power sent to the communication neighbor, so as to realize the collaborative optimization control of the multi-entity energy system.

[0048] In some embodiments, the Nash equilibrium of the distributed cooperative control model is iteratively solved while protecting the privacy of the operational data of each entity, ultimately determining the producers and consumers. i During the period t Net power and to communication neighborsj P2P output power This enables coordinated and optimized control of multiple energy systems.

[0049] Example 2: To further verify the effectiveness and advancement of the distributed energy system control method provided in Embodiment 1, this embodiment further provides a specific implementation of the distributed energy system control method. It should be noted that the following implementation is merely an example of the method of the present invention and does not constitute a limitation on the invention. For ease of understanding, this embodiment provides an application of the method to a specific distributed energy system (…). Figure 2 The specific implementation steps of the multi-prosumer community energy system shown are as follows: Steps 1 to 6. Step 1 corresponds to step S10 in Embodiment 1; Steps 2 and 3 correspond to step S20 in Embodiment 1; Steps 4 and 5 correspond to step S30 in Embodiment 1; the first half of Step 6 (static graph convolutional embedding compression, splitting calculation, and zero-knowledge proof) corresponds to S40 in Embodiment 1; and the second half of Step 6 (solving Nash equilibrium and outputting results) corresponds to S50 in Embodiment 1. The following steps are explained using specific data. For specific calculation methods and principles involved, please refer to the corresponding steps in Embodiment 1 above; these will not be repeated in this embodiment.

[0050] Specifically, to verify the effectiveness of this method, a test case was constructed based on the IEEE 33-node system. The system is configured with 5 operators and 50 prosumers, with each operator managing 10 prosumers. Prosumers are categorized by peak load as follows: multi-source type (55%), peak load 25-40kW, net load usually negative; and load type (45%), peak load 40-55kW, net load usually positive. Internally, they integrate photovoltaic power generation units, wind power generation units, gas turbines, energy storage devices, heat pumps, and electric vehicles. Heat energy trading is limited to within each community, with operators coordinating waste heat from heat pumps and gas turbines, as well as heat storage, to achieve community-level self-balancing, without exchange with the upper-level heat network.

[0051] Step 1: Construct a multi-dimensional operating characteristic vector of producers and consumers from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay, and quantify the operating status observation and parameter sharing behavior based on the communication topology graph model.

[0052] Step 1 is used to model the multi-dimensional operational characteristics of prosumers. For example... Figure 2As shown, the multi-prosumer community energy system consists of a hierarchical control architecture comprised of operator-controlled areas and prosumers. First, a multi-dimensional operating characteristic vector is generated for each prosumer based on historical operating data, including a power regulation rate coefficient, a renewable energy consumption contribution parameter, and a control response delay factor. The power regulation rate coefficient is defined as the ratio of the power change rate to the rated power divided by the regulation time window, thereby classifying prosumers into fast-regulation, medium-regulation, and slow-regulation types. The renewable energy consumption contribution parameter ranges from 0 to 1, categorized as high-contribution, medium-contribution, and low-contribution types. The control response delay factor ranges from 0 to 4 hours, categorized as immediate-response, delayed-response, and lagging-response types. The system regulation demand signal perceived by the prosumer is obtained by adding the product of the renewable energy consumption contribution, renewable energy penetration rate, and a baseline signal to the control signal issued by the operator. The basic power plan for the prosumer is derived from a power allocation algorithm under operating constraints, and is positively correlated with both the regulation demand signal and the renewable energy consumption contribution.

[0053] Step 2: Define the communication topology and establish an operational status observation mechanism to adjust the net power of producers and consumers to the average of their neighbors.

[0054] Step 2 is used to execute the communication topology graph model and P2P power balance constraints. For example... Figure 3 As shown, the communication topology is divided into three categories: strong connections, ordinary connections, and weak connections, each corresponding to different operational characteristics. The communication topology graph model adopts an undirected weighted graph, where nodes represent producers and consumers, and edge weights represent the quality of the communication link. P2P power sharing is constrained by the availability of communication links; power sharing can only occur between producers and consumers who have communication links, and the upper limit of the sharing amount is proportional to the link quality weight.

[0055] Step 3: Establish a parameter sharing mechanism, update the neighborhood collaboration factor through comprehensive operational status evaluation, and achieve slow-change update by quasi-static coupling.

[0056] Step 3 is used to implement the operational status observation mechanism and parameter sharing mechanism. The operational status observation mechanism adjusts the actual net power of producers and consumers towards the average power of their neighbors, with the adjustment magnitude controlled by the operational status observation coefficient. The average power of neighbors is calculated weighted by communication link quality; updated neighbors use the power allocation value for the current time period, while unupdated neighbors use the value from the previous time period. The parameter sharing mechanism updates the neighborhood coordination factor through a comprehensive operational status assessment, which is obtained by weighted fusion of the local operational status assessment and the operational status assessment of neighbors. The neighborhood coordination factor adopts quasi-static coupling and is updated slowly only across control cycles.

[0057] Step 4: Divide the mutual support relationship between operators into three categories: collaborative support, independent operation, and mutual backup, and construct a graph neural network-master-slave distributed collaborative control model.

[0058] Step 4 is used to perform operator mutual assistance mode modeling and upper-level control parameter adjustment. The mutual assistance mode between operator control areas is described by a mutual assistance coefficient matrix, with a value of 1 indicating coordinated mutual assistance, -1 indicating mutual backup, and 0 indicating independent operation. The comprehensive control objectives of the operators include indicators such as power allocation deviation and control parameter fluctuation of each producer and consumer, and a mutual assistance adjustment term is introduced. The mutual assistance adjustment coefficient is multiplied by the mutual assistance mode coefficient, the control parameter coupling strength, and the square of the difference between the control parameter vectors of the two operators, and then summed. The control parameter coupling strength is defined as the absolute value of the difference between the control parameter vectors of the two operators divided by the maximum value of the control parameter vectors of all operators on that day, multiplied by 2 and then subtracted from the absolute value of 0.5. The mutual assistance coefficient is slowly updated according to the cumulative counter of control parameter deviation during intraday rolling optimization.

[0059] Step 5: The lower-level prosumers use a two-layer graph convolutional network to extract communication neighborhood features and use a distributed sequential update protocol to update the net power allocation sequentially.

[0060] Step 5 executes the lower-layer graph convolutional network and the distributed sequential update protocol. The lower-layer prosumer uses a two-layer graph convolutional network to extract the operational features of the communication neighborhood. When updating node embeddings, for each neighbor, the embedding from the previous layer is multiplied by the communication link quality weight and divided by the geometric mean of the weighted degrees of the two endpoints. The sum is then processed by a trainable weight matrix and the ReLU activation function to obtain a new embedding. The final embedding is used to adaptively scale the neighborhood collaboration factor and parameter sharing sensitivity coefficient. The upper-layer operator adjusts control parameters according to the mutual assistance mode to optimize the overall system control indicators. The lower-layer prosumer uses a graph neural network to optimize the power allocation plan; the two constitute a master-slave distributed collaborative control architecture. The lower-layer prosumer adopts a distributed sequential update protocol. Assuming a random permutation of the prosumer set, the position of the currently updating prosumer in the permutation is k. The average power of its neighbors is calculated by weighting the communication link quality. If a neighbor has been updated, the power allocation value for the current time period is used; otherwise, the value from the previous time period is used. This mechanism allows the later-updating prosumer to utilize the latest operational status information of its updated neighbors, naturally reflecting the operational status observation effect.

[0061] Step 6: Introduce a trusted distributed mirror learning framework, perform static graph convolutional embedding compression on the prosumer-consumer running characteristic parameters, and use split computation and zero-knowledge proof methods to solve the Nash equilibrium of the model while protecting data privacy.

[0062] Step 6 is used to perform privacy-preserving solution and simulation verification for trusted distributed mirror learning. Static graph convolutional embedding compression uses an information functional to compress the local runtime information set into a low-dimensional embedding vector. This embedding is calculated only once during system initialization and is not exchanged during subsequent control processes. In the split calculation method, each producer-consumer maps sensitive runtime parameters through a private projection vector and takes the hyperbolic tangent to obtain a one-dimensional sub-value. The operator collects the sub-values ​​for global runtime status monitoring but cannot reverse-engineer the original data. The zero-knowledge proof method is used for compliance verification of communication link availability in P2P power mutual aid. The sender generates a commitment and a zero-knowledge proof, and the receiver recalculates and compares them. Figure 4 The system topology and distributed resource configuration diagram for the test case.

[0063] To verify the incremental contribution of each innovation element, five progressive comparison scenarios were set up, and their functional support is shown in Table 1.

[0064] Table 1: Supported Scene Functions

[0065] The technical indicators for each scenario are compared in Table 2.

[0066] Table 2 Comparison of technical indicators under different scenarios

[0067] As shown in Table 2, the introduction of the communication topology and P2P power reconciliation mechanism increased the power reconciliation amount in Scenario 3 from zero to 74.09 kWh, and the average power adjustment range increased from 0.1896 to 0.2234, enhancing the power regulation capability. The operator reconciliation mode and graph neural network and privacy protection framework further optimized power allocation, making the net load and peak-valley ratio more reasonable, and the system power allocation more balanced.

[0068] like Figure 5 As shown, the heatmap evolution of the communication link quality weight matrix in scenarios 3 to 5 reveals rich and valuable information. Initially, the link quality weights are evenly distributed, but a significant polarization phenomenon emerges as mutual assistance activities progress. High-frequency mutual assistance links, due to their frequent participation in mutual assistance activities, continuously increase their weights, indicating that these links are gradually taking on a more critical role in the system's power transmission and interaction; conversely, idle links, due to their minimal participation in mutual assistance, gradually decrease in weight. By scenario 5, this polarization reaches a significant level, with a small number of links having weights close to 1, indicating that they have become extremely important communication channels in the system, while the weights of most links drop below 0.3, indicating a relatively weakened role in the system. This dynamic change in link weights provides important fundamental support for the system's power allocation and mutual assistance behavior.

[0069] Figure 6The Lorentz curve reveals the system's characteristics from another dimension. Power mutual assistance is significantly concentrated in strongly connected links across all scenarios, especially in scenario 5, where the top 20% of links contribute over 85% of the mutual assistance. This result is consistent with... Figure 5 The polarization characteristics of the link weights corroborate each other, strongly confirming the positive feedback mechanism of "mutual reinforcement of links and links promoting mutual reinforcement." In other words, mutual reinforcement further strengthens links that were already of good quality and strong connection, and these strengthened links can in turn better promote mutual reinforcement, forming a virtuous cycle and thus improving the power mutual reinforcement efficiency of the entire system.

[0070] Figure 7 The focus is on the dynamic switching of the mutual assistance mode coefficient matrix in scenarios 4 and 5. Initially, operators 1 and 3, and operators 2 and 3 exhibit competitive relationships, while operators 2 and 5, and operators 3 and 5 are cooperative. As time progresses and the system state changes, these relationships evolve, with scenario 5 showing a higher switching frequency. This dynamic evolution process profoundly demonstrates that the mutual assistance mode is not a pre-set fixed label, but rather an adaptive adjustment result endogenously derived from differences in control parameters. The system can flexibly adjust the mutual assistance mode between operators according to different operating states and control parameters to better adapt to complex and changing operating environments, achieving more efficient power allocation and collaborative control.

[0071] The above results fully demonstrate that this method can effectively integrate the differences in multidimensional operating characteristics and multiple communication interactions between producers and consumers, coordinate multi-operator mutual control, achieve efficient solution of distributed control strategies under the premise of strictly protecting the privacy of operating data, and realize the collaborative optimization control of multi-entity energy systems.

[0072] Example 3: This invention also provides a distributed energy system control device for implementing the methods described in any of the above embodiments, such as... Figure 8 As shown, the device includes: The characteristic modeling and quantification module 801 is configured to acquire historical operating data of each producer-consumer, construct a multi-dimensional operating characteristic vector of the producer-consumer from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay based on the historical operating data, and generate a basic power plan for each producer-consumer based on the multi-dimensional operating characteristic vector; based on the communication topology graph model, quantify the operating status observation and parameter sharing behavior between producers-consumers according to the communication link status between each producer-consumer, and obtain the producer-consumer control model; The state observation and collaborative update module 802 is configured to establish an operational state observation mechanism for each producer and consumer based on the producer-consumer control model and the basic power plan; adjust the actual net power of each producer and consumer based on the operational state observation mechanism to obtain the adjusted net power data; establish a parameter sharing mechanism; update the neighborhood collaboration factor through comprehensive operational state evaluation based on the parameter sharing mechanism; and use a quasi-static coupling mechanism to update the neighborhood collaboration factor across control cycles to obtain the updated neighborhood collaboration factor. The mutual assistance graph neural network optimization module 803 is configured to acquire the mutual assistance relationship between the control areas of various multi-energy community operators, divide the mutual assistance relationship into three mutual assistance modes: collaborative mutual assistance, independent operation, and mutual backup, and construct a graph neural network-master-slave distributed collaborative control model based on the mutual assistance mode. In the graph neural network-master-slave distributed collaborative control model, the upper-layer multi-operators adjust the control parameters according to the mutual assistance mode to obtain the adjusted upper-layer control parameters, and the lower-layer producers and consumers use a graph convolutional network to extract the communication neighborhood operation features to obtain node embedding features, and optimize their own power allocation and P2P power mutual assistance plan based on the node embedding features and the updated neighborhood collaboration factor. The embedded split zero-knowledge module 804 is configured to introduce a trusted distributed mirror learning framework to perform static graph convolutional embedding compression on the operational characteristic parameters of prosumers to obtain compressed embedding vectors; splitting computation is used to map the local operational state vectors of each prosumer to sub-values, and zero-knowledge proof is used to perform compliance verification of the availability of communication links for P2P power mutual assistance to obtain compliance verification results. The privacy-preserving Nash solution allocation module 805 is configured to, under the premise of protecting the privacy of the operating data of each entity, iteratively solve the Nash equilibrium of the graph neural network-master-slave distributed collaborative control model based on the compressed embedding vector, the sub-value, and the compliance verification result, to obtain the net power of each producer and consumer in the current time period and the power sent to the communication neighbor, thereby realizing the collaborative optimization control of the multi-entity energy system.

[0073] This invention provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0074] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0075] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0076] The electronic device provided in this embodiment of the invention can be the terminal device described in the above embodiments.

[0077] This invention also provides a computer-readable storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the technical solution of the distributed energy system control method described above.

[0078] This invention also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium, and when the at least one processor executes the computer program, it can implement the technical solution of the distributed energy system control method described in the above embodiments.

[0079] In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0080] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0081] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0082] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute certain steps of the methods of the various embodiments of the present invention.

[0083] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0084] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0085] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0086] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0087] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0088] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A control method for a distributed energy system, characterized in that, The method includes: Historical operating data of each producer-consumer is obtained. Based on the historical operating data, a multi-dimensional operating characteristic vector of the producer-consumer is constructed from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay. The basic power plan of each producer-consumer is generated based on the multi-dimensional operating characteristic vector. Based on the communication topology graph model, the operating status observation and parameter sharing behavior between producers-consumers are quantified according to the communication link status between each producer-consumer to obtain the producer-consumer control model. Based on the producer-consumer control model and the basic power plan, an operational status observation mechanism for each producer-consumer is established. The actual net power of each producer-consumer is adjusted according to the operational status observation mechanism to obtain the adjusted net power data. A parameter sharing mechanism is established. The neighborhood coordination factor is updated through comprehensive operational status evaluation according to the parameter sharing mechanism. The neighborhood coordination factor is updated across control cycles using a quasi-static coupling mechanism to obtain the updated neighborhood coordination factor. The mutual support relationships between the control areas of various multi-energy community operators are obtained, and these relationships are divided into three mutual support modes: collaborative mutual support, independent operation, and mutual backup. A graph neural network-master-slave distributed collaborative control model is constructed based on these mutual support modes. In the graph neural network-master-slave distributed collaborative control model, the upper-layer multi-operators adjust the control parameters according to the mutual support modes to obtain the adjusted upper-layer control parameters. The lower-layer producers and consumers use a graph convolutional network to extract the communication neighborhood operation features to obtain node embedding features, and optimize their own power allocation and P2P power mutual support plan based on the node embedding features and the updated neighborhood collaboration factor. A trusted distributed mirror learning framework is introduced to compress the operational characteristic parameters of prosumers using static graph convolutional embedding to obtain compressed embedding vectors. Split computation is used to map the local operational state vectors of each prosumer to sub-values, and zero-knowledge proof is used to verify the availability compliance of the communication link for P2P power mutual assistance, thus obtaining compliance verification results. Under the premise of protecting the privacy of the operational data of each entity, the Nash equilibrium of the graph neural network-master-slave distributed collaborative control model is iteratively solved based on the compressed embedding vector, the sub-value, and the compliance verification result to obtain the net power of each producer and consumer in the current time period and the power sent to the communication neighbor, thereby realizing the collaborative optimization control of the multi-entity energy system.

2. The distributed energy system control method according to claim 1, characterized in that, The multidimensional operational characteristic vector of the prosumer is: ,in The producer-consumer power regulation rate coefficient, Parameters for the contribution of new energy consumption. To control the response delay factor; The power regulation rate coefficient Determined by the following formula: (1) in: For producers and consumers i The change in power, For producers and consumers i The rated power is the power of the producer and consumer. i Maximum allowable power of grid connection interface To adjust the time window; The system regulation demand signal perceived by prosumers is calculated using the following formula: (2) in: For producers and consumers i During the period t Actual perceived regulatory demand signals; For operators During the period t The control signals sent out For producers and consumers i The penetration rate of new energy sources in the controlled area As a reference signal; The basic power plan is derived from the power allocation algorithm under operating constraints. The basic power plan is positively correlated with the adjustment demand signal and also positively correlated with the contribution parameter of new energy consumption.

3. The distributed energy system control method according to claim 1, characterized in that, The communication topology graph model is an undirected weighted graph. ,in For the set of producer-consumer nodes, N For the number of producer-consumer nodes, ; Let be the set of edges, representing the communication links between producers and consumers; Let be an adjacency matrix, where elements Indicates producer-consumer i and j The communication link quality weights between them are determined, and they satisfy the properties of being undirected and free of self-loops. It is the set of real numbers; P2P power sharing is constrained by the availability of communication links as follows: (3) in: and Time periods t Producers and consumers i To consumers j The transmitted power and the received power, This is the upper limit for a single power exchange. The operational status observation mechanism adjusts the actual net power of producer-consumer according to the following formula. : (4) (5) (6) in: Based on power planning, For operational status observation coefficients, For communication neighbor set, For neighboring observable power states, For power allocation preference parameters, For producers and consumers i exist t The neighborhood average observed power at time 10:

00. For communication neighbors j exist t Actual net operating power at any given time For communication neighbors j exist t The average observed power of the neighborhood at time -1.

4. The distributed energy system control method according to claim 3, characterized in that, The parameter sharing mechanism updates the comprehensive operating status assessment and adjusts the neighborhood collaboration factor according to the following formula: (7) (8) in: As a weight for evaluating its own local operating status, This is a local operating status assessment value. This is a comprehensive operational status assessment value. For parameter sharing sensitivity coefficient, As an effective neighborhood collaboration factor, For producers and consumers i exist t The original neighborhood cooperation factor at time; The update rule of the quasi-static coupling mechanism is expressed as follows: (9) in: For slowly varying coefficients, The baseline value is the scaled value of the graph neural network embedding. As a neighborhood collaboration factor, k This is the iteration number of the distributed collaborative control cycle. This update is only performed across control cycles. Within a single distributed collaborative control cycle... It is approximately a constant.

5. The distributed energy system control method according to claim 1, characterized in that, The mutual aid mode is described through a mutual aid coefficient matrix. Description, in which These respectively represent mutual backup, independent operation, and coordinated support. M Total number of operators; Operators m During the period t Based on its own control parameters The mutual adjustment term adjusts the control strategy, wherein the mutual adjustment term is defined as: (11) (12) in: The mutual adjustment coefficient, To control the coupling strength of parameters, For operators n During the period t The control parameters, For operators m During the period t Mutual adjustment items, For all operators during the time period t The maximum value of the lower control parameter.

6. The distributed energy system control method according to claim 3, characterized in that, The lower-layer producer-consumer uses a two-layer graph convolutional network to extract communication neighborhood operation features, and the node embedding update method is as follows: (13) in: l For graph convolutional layer index, For weighted degrees, For the first l Layer trainable weight matrix, It is the ReLU activation function. For nodes i In the l +1 layer output embedding features, For producer-consumer nodes i itself, For nodes j The weighting degree, For nodes j In the l The layer's input embedding features; Final Embedding Used for adaptive scaling of neighborhood cooperability factor Shared sensitivity coefficient with parameters The mapping function is shown in the following formula: (14) in: , The initial baseline values ​​correspond to the producer-consumer nodes respectively. i Before introducing graph convolutional embedding features, the preset basic neighborhood collaborative weights and basic parameter sharing sensitivity thresholds are used. The parameter vector is a learnable vector, and the sigmoid function is a sigmoid function. The lower-level prosumers adopt a distributed sequential update protocol, where a random permutation of the prosumer set is... For current update consumers The average power of its neighbors is calculated using the following formula: (15) Among them: updated neighbors use the latest power allocation value for the current time period, while unupdated neighbors use the value for the previous time period.

7. The distributed energy system control method according to claim 6, characterized in that, The static graph convolutional embedding compression employs an information functional. Local running information set The vector is compressed into an embedding vector, as shown in equation (16): (16) in: e i For producers and consumers i Embedded vector, θ i For multidimensional operational characteristic vectors, d This is the embedding dimension; the embedding is calculated only once during system initialization and is not exchanged during subsequent control processes. In the splitting calculation, each producer-consumer locally calculates its sub-value. The calculation formula is: (17) in: For the private projection vector of the prosumer, For producers and consumers i Local runtime state vector, It is the transpose symbol; The zero-knowledge proof is used for compliance verification of the availability of communication links in P2P power exchange. The sender generates the proof using the following formula. : (18) in: "Random number" and "public" indicate publicly available information. For random number based For initial parameters The value obtained by performing the calculation Here, com represents the initial parameter values ​​related to the sender's private information, private represents the sender's private information, and Prove represents the zero-knowledge proof generation function. zk Evidence for generating zero-knowledge proofs; Receiver verification: (19) in: These are initial parameter values ​​related to the sender's private information; If they are equal, it proves that the sender meets the communication link availability constraint, and the compliance verification result is obtained. The Nash equilibrium of the graph neural network-master-slave distributed collaborative control model is iteratively solved based on the compressed embedding vector, the sub-values, and the compliance verification results, ultimately determining each producer-consumer. i During the period t Net power and to communication neighbors j P2P output power This enables coordinated and optimized control of multiple energy systems.

8. A control device for a distributed energy system, characterized in that, The device includes: The characteristic modeling and quantification module is configured to acquire historical operating data of each producer-consumer, construct a multi-dimensional operating characteristic vector of the producer-consumer based on the historical operating data from three dimensions: power regulation rate, contribution of new energy consumption, and control response delay, and generate a basic power plan for each producer-consumer based on the multi-dimensional operating characteristic vector; based on the communication topology graph model, quantify the operating status observation and parameter sharing behavior between producers-consumers according to the communication link status between each producer-consumer, and obtain the producer-consumer control model; The state observation and collaborative update module is configured to establish an operational state observation mechanism for each producer and consumer based on the producer-consumer control model and the basic power plan; adjust the actual net power of each producer and consumer based on the operational state observation mechanism to obtain the adjusted net power data; establish a parameter sharing mechanism; update the neighborhood collaboration factor through comprehensive operational state evaluation based on the parameter sharing mechanism; and use a quasi-static coupling mechanism to update the neighborhood collaboration factor across control cycles to obtain the updated neighborhood collaboration factor. The mutual assistance graph neural network optimization module is configured to acquire the mutual assistance relationships between the control areas of various multi-energy community operators, and divide the mutual assistance relationships into three types of mutual assistance modes: collaborative mutual assistance, independent operation, and mutual backup. Based on the mutual assistance modes, a graph neural network-master-slave distributed collaborative control model is constructed. In the graph neural network-master-slave distributed collaborative control model, the upper-layer multi-operators adjust the control parameters according to the mutual assistance modes to obtain the adjusted upper-layer control parameters. The lower-layer producers and consumers use a graph convolutional network to extract the communication neighborhood operation features to obtain node embedding features, and optimize their own power allocation and P2P power mutual assistance plan based on the node embedding features and the updated neighborhood collaboration factor. The embedded split zero-knowledge module is configured to introduce a trusted distributed mirror learning framework to perform static graph convolutional embedding compression on the operational characteristic parameters of prosumers to obtain compressed embedding vectors; splitting computation is used to map the local operational state vectors of each prosumer to sub-values, and zero-knowledge proof is used to perform compliance verification of the availability of communication links for P2P power mutual assistance to obtain compliance verification results. The privacy-preserving Nash solution allocation module is configured to iteratively solve the Nash equilibrium of the graph neural network-master-slave distributed collaborative control model based on the compressed embedding vector, the sub-values, and the compliance verification results, while protecting the privacy of the operational data of each entity, to obtain the net power of each producer and consumer in the current time period and the power sent to the communication neighbor, thereby realizing the collaborative optimization control of the multi-entity energy system.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the distributed energy system control method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the distributed energy system control method as described in any one of claims 1-7.

Images