A two-stage self-organizing optimization aggregation method and system for distributed resources of a virtual power plant

Abstract

The application discloses a kind of virtual power plant distributed resource two-stage self-organizing optimization aggregation method and system, the natural physical cluster of DERs in distribution area is formed as first level, and is aggregated by edge computing server deployed in distribution area department;Generalized substation load model containing wind power, photovoltaic and load is constructed;For the generator of distributed gas turbine and small hydropower station in distribution area, it is aggregated into a unified virtual synchronous generator model;For the distributed energy storage in distribution area, it is aggregated into a centralized virtual energy storage model, that is, the first level aggregation for distribution area can be completed;By uploading all parameters of the generalized substation load model, the virtual synchronous generator model and the virtual energy storage model to the cloud, the second level aggregation across distribution area is completed.The application can fully regulate and control the distributed power resources of distribution network, and can improve energy utilization.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of virtual power plants, and particularly relates to a two-stage self-organizing optimization aggregation method and system for distributed resources of a virtual power plant. BACKGROUND

[0002] With the jump-like growth of the number of DERs (distributed power resources) such as wind power, photovoltaic power and electric vehicles, in the future, there will be a large number of distributed power resources in various forms accessing the distribution network. Although the proportion of DERs in the distribution network is increasing, the advantages thereof are difficult to be fully and effectively exerted. This is because the DERs have high dispersion, uncertainty and heterogeneity, and a large number of small-capacity DERs such as distributed power sources, controllable loads and electric vehicles are difficult to directly participate in the regulation and control of the power grid. For an indirect regulation mode based on electricity price, since the capacity of each DER individual is small, the influence on the power system is usually low, and the economic benefits brought by the participation of the DERs in the regulation and control of the power grid are usually small, so the enthusiasm of the DERs for participating in the regulation and control is low. For a direct regulation mode in which the DERs directly participate in the dispatching of the system, the system operator needs to solve a complex high-dimensional optimization model, which brings a large calculation burden and thus reduces the operation efficiency, and cannot guarantee the efficient and orderly operation of the power system. Therefore, the aggregation of resources is the key to realizing the regulation of the distributed power resources of the distribution network and improving the energy utilization rate. As an effective means of aggregating DERs, the virtual power plant (VPP) can realize the aggregation, energy supply, energy use and energy storage of the DERs by means of advanced metering, communication and control technologies, without changing the grid connection mode and geographical position of the DERs, effectively connecting the DERs and the power system, realizing the integration, distribution and recombination of resources, and directly participating in the dispatching operation of the power system as an aggregation entity, and is an important way for the smart grid to realize interaction and intelligentization on the energy supply side. SUMMARY

[0003] Therefore, the application provides a two-stage self-organizing optimization aggregation method and system for distributed resources of a virtual power plant.

[0004] The technical solution adopted by the application to solve the technical problem is as follows:

[0005] A two-stage self-organizing optimization aggregation method for distributed resources of a virtual power plant, comprising the following steps:

[0006] 1) First-stage aggregation oriented to a distribution area, specifically comprising:

[0007] S100, a natural physical cluster of DERs in a distribution area is taken as a first stage, and an edge computing server deployed in the distribution area is used for aggregation;

[0008] S200, based on the historical data retrieved from the cloud, the uncertainty modeling of the wind energy, the solar energy and the load curve of the entire distribution area is performed in the edge computing server based on the deep Bayesian network, and a hourly prediction model is constructed;

[0009] S300, according to the hourly prediction model, a generalized load model of the distribution area containing wind power, photovoltaic and load is constructed;

[0010] S400, for the distributed gas turbine and the generator of the small hydropower station in the distribution area, they are aggregated into a unified mathematical model of the virtual synchronous generator output;

[0011] S500, for the distributed energy storage in the distribution area, it is aggregated into a centralized mathematical model of the virtual energy storage model capacity;

[0012] 2) Second-level aggregation across the distribution area: upload all parameters of the generalized load model, the virtual synchronous generator model and the virtual energy storage model to the cloud for second-level aggregation.

[0013] As a further improvement, the step S400 comprises the following steps:

[0014] S401, the upward ramp rate Ramp i，up and the downward ramp rate Ramp i，down of the generator with serial number i are respectively accumulated to obtain the upward ramp rate Ramp sum，up and the downward ramp rate Ramp sum，down of the virtual synchronous generator, wherein, N is a positive integer greater than 0;

[0015] S402, according to the upward ramp rate and the downward ramp rate of each generator, the upper limit and the lower limit of the output of the corresponding virtual synchronous generator at time t are calculated, wherein, i，max (t)=P i (t-1)+Δt×Ramp i，up , P i，min (t)=P i (t-1)-Δt×Ramp i，down , P i，max (t)≤P i，max , P i，min (t)≤P i，min , P i，max (t) represents the upper limit of the output of the i-th virtual synchronous generator at time t, P i，min (t) represents the upper limit of the output of the i-th virtual synchronous generator at time t, Δt represents the time difference between the previous time and the current time, P i(t-1) represents the output of the virtual synchronous generator at the previous moment, P i，max and P i，min are the maximum upper limit and the maximum lower limit of the output of the i-th virtual synchronous generator, respectively;

[0016] S403, the upper limit and the lower limit of the output of each generator at time t are added up to obtain the total output limit value of the corresponding virtual synchronous generator at time t: wherein P max (t) represents the upper limit of the output of the virtual synchronous generator at time t, P min (t) represents the lower limit of the output of the virtual synchronous generator at time t, and finally the mathematical model of the output of the virtual synchronous generator is obtained.

[0017] As a further improvement, the step S500 comprises the following process:

[0018] S501, the rated charging power Pess j，char_N and the rated discharging power Pess j，disc_N of the energy storage device with serial number j are added up respectively to obtain the maximum charging power Pess char，max (t) and Pess disc，max (t) of the virtual energy storage model, wherein, M is a positive integer greater than 0;

[0019] S502, the upper limit of the capacity of the energy storage device at time t is set as E j，max (t) = E j (t-1) + Δt × Pess j，char_N , wherein E j (t-1) represents the upper limit of the capacity of the energy storage device at the previous moment, E j，max (t) ≤ E j，max , E j，max is the maximum value of the capacity of the energy storage device; the lower limit of the capacity of the energy storage device at time t is set as E j，min (t) = E j (t-1) - Δt × Pess j，disc_N , wherein E j，min (t) ≥ E j，min , E j，min is the minimum value of the capacity of the energy storage device;

[0020] S503, the upper limit and the lower limit of the capacity of each energy storage device at time t are added up to obtain the total capacity limit value E max (t) and E min (t) of the virtual energy storage at time t, wherein, Finally, a centralized mathematical model of the capacity of the virtual energy storage model is obtained.

[0021] As a further improvement, the second-level aggregation across the distribution substation area specifically includes:

[0022] S601, taking the minimum internal operation cost of the virtual power plant as the optimization target, constructing a virtual power plant internal supply-demand interaction optimization scheduling model:

[0023]

[0024] Wherein, Cost VS，k (t) represents the operation cost of the virtual synchronous generator in the kth distribution substation area, Cost ESS，k (t) is the operation cost of the virtual energy storage model of the kth distribution substation area, Cost Grid (t) is the cost of purchasing electricity from the external network by the virtual power plant as a whole, Cost Grid (t) is positive, indicating purchasing electricity, Cost Grid (t) is negative, indicating selling electricity, T represents the total time length of the time statistics, and K represents the total number of distribution substation areas participating in aggregation;

[0025] S602, obtaining an optimized virtual power plant operation data set through the virtual power plant internal supply-demand interaction optimization scheduling model, and saving the output of each generalized substation load model, virtual energy storage model and virtual synchronous generator in the data set as a preset value;

[0026] S603, subtracting the output of all power generation units in the virtual power plant from the total internal load demand to obtain the total remaining active power output and the remaining energy storage capacity, calculating the inertia and damping coefficient of the virtual synchronous generator corresponding to the active power output capacity according to the total remaining active power output and the remaining energy storage capacity, and constructing a virtual synchronous generator mathematical model based on the inertia and damping coefficient, taking the total active power output value of the virtual power plant internal supply-demand interaction optimization scheduling model at different capacity levels as the input of the virtual synchronous generator mathematical model, and combining the input and the output of the virtual synchronous generator mathematical model to form a training data set;

[0027] S604, constructing a deep reinforcement learning model using a deep Q learning algorithm, and obtaining a virtual power plant aggregation data model simulating the characteristics of a real large virtual synchronous generator group with adaptive capacity through training;

[0028] S605, uploading the virtual power plant aggregation data model as a virtual power plant model to a cloud scheduling platform for second-level aggregation.

[0029] As a further improvement, the distribution substation area is a 400V distribution substation area including buildings, communities, factories and schools.

[0030] The system for implementing the two-stage self-organizing optimization aggregation method of the virtual power plant distributed resources described above comprises a power distribution area, a first-stage aggregation module, a second-stage aggregation module, an edge computing server and a cloud, the edge computing server is deployed in the power distribution area, the first-stage aggregation module comprises a generalized load module, a centralized generator module and a centralized energy storage module:

[0031] A plurality of DERs are arranged in the power distribution area, and the plurality of DERs form a natural physical cluster, and the natural physical cluster is aggregated by the edge computing server deployed in the power distribution area as a first stage.

[0032] The edge computing server utilizes historical data retrieved from the cloud to construct a hourly prediction model in the edge computing server based on an uncertainty modeling method of deep Bayesian network learning for the output of wind energy and solar energy and the load curve of the power distribution area.

[0033] The generalized load module is used to construct a generalized power distribution area load model containing wind power, photovoltaic and load according to the hourly prediction model.

[0034] The centralized generator module is used to aggregate the distributed gas turbine and small hydropower station generators in the power distribution area into a unified virtual synchronous generator model.

[0035] The centralized energy storage module is used to aggregate the distributed energy storage in the power distribution area into a centralized virtual energy storage model.

[0036] The second-stage aggregation module is used to upload all parameters of the generalized power distribution area load model, the virtual synchronous generator model and the virtual energy storage model to the cloud for second-stage aggregation.

[0037] A computer device comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the two-stage self-organizing optimization aggregation method of the virtual power plant distributed resources when executing the computer program.

[0038] A computer readable storage medium having a computer program stored thereon, the computer program is executed by a processor to implement the steps of the two-stage self-organizing optimization aggregation method of the virtual power plant distributed resources.

[0039] The virtual power plant distributed resource two-level self-organizing optimization aggregation method and system first aggregates the natural physical cluster of DERs in the power distribution area as the first level by the edge computing server deployed in the power distribution area; secondly, the historical data is called to the cloud, the uncertainty modeling is performed on the output of wind energy and solar energy and the load curve of the power distribution area in the edge computing server based on the deep Bayesian network, the hour-level prediction model is constructed, the generalized substation load model containing wind power, photovoltaic and load is constructed according to the hour-level prediction model; then, the distributed gas turbine and the generator of the small hydropower station in the power distribution area are aggregated into a unified virtual synchronous generator output mathematical model; again, the distributed energy storage in the power distribution area is aggregated into a centralized virtual energy storage model capacity mathematical model, and the first-level aggregation for the power distribution area is completed through the above process; finally, all parameters of the generalized substation load model, the virtual synchronous generator model and the virtual energy storage model are uploaded to the cloud for the second-level aggregation, and the second-level aggregation across the power distribution area is completed. The present application can fully regulate the distributed power resources of the power distribution network, and can improve the energy utilization rate. BRIEF DESCRIPTION OF DRAWINGS

[0040] The present application is further described by means of the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the following drawings.

[0041] Figure 1 It is a flow chart of a virtual power plant distributed resource two-level self-organizing optimization aggregation method.

[0042] Figure 2 It is a model diagram of a virtual power plant distributed resource two-level self-organizing optimization aggregation system. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solutions of the present application, the present application will be further described in detail below with reference to the accompanying drawings.

[0044] In one embodiment, as shown in Figure 1 A virtual power plant distributed resource two-level self-organizing optimization aggregation method includes the following steps:

[0045] 1) The first-level aggregation for the power distribution area, specifically including:

[0046] S100, the natural physical cluster of DERs in the power distribution area is aggregated as the first level by the edge computing server deployed in the power distribution area;

[0047] It should be noted that the above power distribution area is preferably a 400V power distribution area including buildings, communities, factories, and schools; the natural physical cluster is a cluster of distributed photovoltaic, energy storage, electric vehicles, and distributed wind power generation, which is a cluster of distributed power resources (DERs).

[0048] S200, based on the historical data retrieved to the cloud, the uncertainty modeling of the wind energy and solar energy output and the load curve of the entire power distribution area is carried out in the edge computing server based on the deep Bayesian network, and a hourly prediction model is constructed;

[0049] Specifically, the historical data includes load baseline data, wind power, and photovoltaic power generation data,

[0050] S300, according to the hourly prediction model, a generalized load model of the power distribution area containing wind power, photovoltaic and load is constructed;

[0051] S400, for the distributed gas turbine and the generator of the small hydropower station in the power distribution area, they are aggregated into a unified virtual synchronous generator output mathematical model;

[0052] S500, for the distributed energy storage in the power distribution area, it is aggregated into a centralized virtual energy storage model capacity mathematical model;

[0053] 2) Second-level aggregation across the power distribution area: upload all parameters of the generalized load model, the virtual synchronous generator model and the virtual energy storage model to the cloud for second-level aggregation.

[0054] Specifically, all parameters of the virtual energy storage model include capacity, active power output, reactive power output, inertia coefficient and damping coefficient of the virtual synchronous generator.

[0055] This invention achieves self-organized aggregation of widely distributed distributed power resources (DERs) in a power distribution network based on resource aggregation. According to the geographical distribution characteristics, load density, and electricity consumption levels of DERs, a two-level aggregation model for massive distributed power resources is proposed. First, the natural physical clusters formed by DERs in 400V distribution substations such as buildings, residential areas, factories, and schools serve as the first level, and are aggregated by edge computing servers deployed in the 400V distribution substations. Second, considering the randomness of DER resources and the volatility of load, historical data retrieved from the cloud is used to perform deep Bayesian networks on the edge computing servers to analyze the wind and solar power output and substation load curves of the entire distribution substation. An uncertainty modeling method for learning is used to construct an hourly prediction model. Based on this hourly prediction model, a generalized distribution area load model including wind power, photovoltaic power, and loads is constructed. Next, for distributed gas turbines and small hydropower generators within the distribution area, they are aggregated into a unified virtual synchronous generator model. Then, for distributed energy storage within the distribution area, they are aggregated into a centralized virtual energy storage model. Through the above process, the first-level aggregation for the distribution area is completed. Finally, by uploading all parameters of the generalized distribution area load model, the virtual synchronous generator model, and the virtual energy storage model to the cloud, a second-level aggregation is performed, completing the second-level aggregation across distribution areas. This invention can fully regulate distributed power resources in the distribution network and improve energy utilization efficiency.

[0056] In one embodiment, step S400, which aggregates the distributed gas turbines and generators of small hydropower stations within the distribution substation area into a unified virtual synchronous generator model, includes the following steps:

[0057] S401, the ramp rate (Ramp) of generator i. i，up and downhill ramp rate i，down The ramp rate (Ramp) of the virtual synchronous generator is obtained by summing the results separately. sum，up and downhill ramp rate sum，down ,in, N is a positive integer greater than 0;

[0058] S402. Based on the upward and downward ramp rates of each generator, calculate the upper and lower limits of the output of the corresponding virtual synchronous generator at time t, where P i，max (t)=P i (t-1)+Δt×Ramp i，up P i，min (t)=P i (t-1)-Δt×Ramp i，dowm P i，max (t)≤P i，max Pi，min (t)≤P i，min , P i，max (t) represents the upper limit of the output of the i th virtual synchronous generator at time t, P i，min (t) represents the upper limit of the output of the i th virtual synchronous generator at time t, Δt represents the time difference between the previous time and the current time, P i (t-1) represents the output of the virtual synchronous generator at the previous time, P i，max and P i，min are the maximum upper limit and lower limit of the output of the i th virtual synchronous generator, respectively;

[0059] S403, the upper limit and lower limit of the output of each generator at time t are accumulated to obtain the total output limit of the corresponding virtual synchronous generator at time t: wherein, P max (t) represents the upper limit of the output of the virtual synchronous generator at time t, P min (t) represents the lower limit of the output of the virtual synchronous generator at time t, and finally the mathematical model of the output of the virtual synchronous generator is obtained.

[0060] In one embodiment, step S500 includes the following processes:

[0061] S501, the rated charging power Pess j，char_N and the rated discharging power Pess j，disc_N of the energy storage device with serial number j are respectively accumulated to obtain the maximum charging power Pess char，max (t) and Pess disc，max (t) of the virtual energy storage model, wherein, M is a positive integer greater than 0;

[0062] S502, the upper limit of the capacity of the energy storage device at time t is set as E j，max (t) = E j (t-1) + Δt × Pess j，char_N , wherein E j (t-1) represents the upper limit of the capacity of the energy storage device at the previous time, E j，max (t) ≤ E j，max , E j，max is the maximum value of the capacity of the energy storage device; the lower limit of the capacity of the energy storage device at time t is set as E j，min (t) = E j (t-1) - Δt × Pess j，disc_N , wherein E j，min (t) ≥ E j，min , E j，min is the minimum value of the capacity of the energy storage device;

[0063] S503, add the upper limit and the lower limit of the capacity of each energy storage device at time t to obtain the total capacity limit E of the virtual energy storage at time t max (t) and E min (t), wherein, Finally, a centralized mathematical model of the virtual energy storage model capacity is obtained.

[0064] In an embodiment, the second-level aggregation across the distribution areas specifically includes:

[0065] S601, an internal supply-demand interaction optimization scheduling model of the virtual power plant is constructed with the minimum internal operation cost of the virtual power plant as the optimization target:

[0066]

[0067] wherein, Cost VS，k (t) represents the operation cost of the virtual synchronous generator in the kth distribution area, Cost ESS，k (t) is the operation cost of the virtual energy storage model of the kth distribution area, Cost Grid (t) is the cost of purchasing power from the external network by the virtual power plant as a whole, Cost Grid (t) is positive, indicating that the power is purchased, Cost Grid (t) is negative, indicating that the power is sold, T represents the total time length counted at time t, and K represents the total number of distribution areas participating in the aggregation;

[0068] S602, an optimized data set of the virtual power plant operation is obtained through the internal supply-demand interaction optimization scheduling model of the virtual power plant, and the output of each generalized area load model, virtual energy storage model and virtual synchronous generator in the data set is saved as a preset value;

[0069] S603, the output of all power generation units in the virtual power plant is subtracted from the total load demand to obtain the total active power output and the remaining energy storage capacity, the inertia and damping coefficient of the virtual synchronous generator corresponding to the active power output capacity are calculated according to the total active power output and the remaining energy storage capacity, and a virtual synchronous generator mathematical model is constructed based on the inertia and damping coefficient, the total active power output value of the internal supply-demand interaction optimization scheduling model of the virtual power plant at different capacity levels is taken as the input of the virtual synchronous generator mathematical model, and the input and the output of the virtual synchronous generator mathematical model are combined to form a training data set.

[0070] Specifically, the power generation units include gas turbines and small hydropower generators, and the total load demand is the generalized load demand data obtained by adding the load baseline data and the wind power and photovoltaic power generation output.

[0071] S604, a deep Q learning algorithm is used to construct a deep reinforcement learning model, and a virtual power plant aggregation data model with adaptive capacity and simulating real large virtual synchronous generator group characteristics is obtained through training;

[0072] S605, the virtual power plant aggregation data model is uploaded to a scheduling platform in the cloud as a virtual power plant model for second-level aggregation.

[0073] In summary, the two-level aggregation mode is used to realize the distributed power resource regulation of the power distribution network, and the energy utilization rate is high.

[0074] In one embodiment, a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the two-level self-organizing optimization aggregation method of the virtual power plant distributed resources when executing the computer program.

[0075] In one embodiment, a computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the two-level self-organizing optimization aggregation method of the virtual power plant distributed resources.

[0076] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer readable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer usable program code. The present application is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The functions specified in one or more flows and / or blocks These computer program instructions can also be stored in a computer readable memory that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0077] The above provides a virtual power plant distributed resource two-level self-organizing optimization aggregation method and system. The principle and implementation of the present application are described by applying specific examples. The above examples are only used to help understand the core idea of the present application. It should be pointed out that for those skilled in the art, without departing from the principle of the present application, the present application can be improved and modified, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A two-stage self-organizing optimization aggregation method for distributed resources of a virtual power plant, characterized in that, The method comprises the following steps: 1) first-level aggregation of a power distribution area, specifically comprising: S100, a natural physical cluster of DERs in a power distribution area is aggregated as a first level by an edge computing server deployed in the power distribution area; the power distribution area is a 400V power distribution area including buildings, communities, factories, and schools; S200, based on a deep Bayesian network, an uncertainty model is established for the output of wind energy and solar energy and the load curve of the entire power distribution area in the edge computing server using historical data retrieved from the cloud, and an hourly-level prediction model is constructed; S300, a generalized area load model containing wind power, photovoltaic power, and load is constructed according to the hourly-level prediction model; S400, for distributed gas turbines and generators of small hydropower stations in the power distribution area, a unified mathematical model of virtual synchronous generator output is aggregated; S500, for distributed energy storage in the power distribution area, a centralized mathematical model of virtual energy storage capacity is aggregated; 2) second-level aggregation across the power distribution area: all parameters of the generalized area load model, the virtual synchronous generator model, and the virtual energy storage model are uploaded to the cloud for second-level aggregation; all parameters of the virtual energy storage model include capacity, active power output, reactive power output, inertia coefficient, and damping coefficient of the virtual synchronous generator; the second-level aggregation across the power distribution area specifically comprises: S601, a virtual power plant internal supply-demand interaction optimization scheduling model is constructed with the optimization target of minimizing the internal operation cost of the virtual power plant; wherein, Cost VS，k (t) represents the operating cost of the virtual synchronous generator in the kth distribution area, Cost ESS，k (t) is the operating cost of the virtual energy storage model in the kth distribution area, Cost Grid (t) is the cost of purchasing electricity from the external network by the virtual power plant as a whole, Cost Grid (t) is positive to represent purchasing electricity, Cost Grid (t) is negative to represent selling electricity, T represents the total time length counted at the time, and K represents the total number of distribution areas participating in aggregation. S602, an optimized data set of virtual power plant operation is obtained through the virtual power plant internal supply-demand interaction optimization scheduling model, and the output of each generalized area load model, virtual energy storage model, and virtual synchronous generator in the data set is saved as a preset value; S603, the output of all power generation units in the virtual power plant is subtracted from the total internal load demand to obtain the total active power output and the remaining energy storage capacity, the inertia and damping coefficient of the virtual synchronous generator corresponding to the active power output capacity are calculated according to the total active power output and the remaining energy storage capacity, and a virtual synchronous generator mathematical model is constructed based on the inertia and damping coefficient; the total active power output value of the virtual power plant internal supply-demand interaction optimization scheduling model at different capacity levels is taken as the input of the virtual synchronous generator mathematical model, and the input and the output of the virtual synchronous generator mathematical model are combined to form a training data set; S604, a deep reinforcement learning model is constructed by using a deep Q learning algorithm, and a virtual power plant aggregation data model simulating the characteristics of a real large virtual synchronous generator group with adaptive capacity is obtained through training; S605, the virtual power plant aggregation data model is uploaded to the cloud scheduling platform as a virtual power plant model for second-level aggregation. 2.The two-stage self-organizing optimization aggregation method for virtual power plant distributed resources according to claim 1, characterized in that, The step S400 comprises the following steps: S401、the up ramp rate Ramp of the generator with serial number i is calculated i，ip and the down ramp rate Ramp i，down are accumulated respectively, to obtain the up ramp rate Ramp of the virtual synchronous generator sum，up and the down ramp rate Ramp sum，down wherein, N is a positive integer greater than 0; S402、According to the upward ramp rate and the downward ramp rate of each generator, the upper limit and the lower limit of the corresponding virtual synchronous generator output at time t are calculated, wherein, P i，max (t)≤P i (t-1)+Δt×Ramp i，up , P i，min (t)=Pi(t-1)-Δt×Ramp i，down , P i，max (t)≤P i，max , P i，min (t)≤P i，min , P i，max (t) represents the upper limit of the i-th virtual synchronous generator output at time t, P i，min (t) represents the upper limit of the i-th virtual synchronous generator output at time t, Δt represents the time difference between the previous time and the current time, P i (t-1) represents the virtual synchronous generator output at the previous time, P i，max and P i，min are the maximum upper limit and the maximum lower limit of the corresponding output of the i-th virtual synchronous generator respectively; S403, add the upper and lower limits of the power output of each generator at time t to obtain the total power output limit of the corresponding virtual synchronous generator at time t: wherein, P max (t) represents the upper limit of the power output of the virtual synchronous generator at time t, P min (t) represents the lower limit of the power output of the virtual synchronous generator at time t, and finally the mathematical model of the power output of the virtual synchronous generator is obtained. 3.The two-stage self-organizing optimization aggregation method for virtual power plant distributed resources according to claim 2, characterized in that, The step S500 comprises the following processes: S501, the rated charging power Pess of the energy storage device with serial number j. j，char_N and rated discharge power Pess j，disc_N By summing the results separately, the maximum charging power Pess of the virtual energy storage model can be obtained. char，max (t) and Pess disc，max (t), where, M is a positive integer greater than 0; S502, set the upper limit of the capacity of the energy storage device at time t as E j，max (t) = E j (t-1) + Δt × Pess j，char_N , wherein E j (t-1) represents the upper limit of the capacity of the energy storage device at the previous time, E j，max (t) ≤ E j，max , E j，max is the maximum value of the capacity of the energy storage device; set the lower limit of the capacity of the energy storage device at time t as E j，min (t) = E j (t-1) - Δt × Pess j，disc_N , wherein E j，min (t) ≥ E j，min , E j，min is the minimum value of the capacity of the energy storage device; S503, accumulate the upper limit and lower limit of the capacity of each energy storage device at time t to obtain the total capacity limit E of the virtual energy storage at time t max (t) and E min (t), wherein Finally, a centralized mathematical model of the capacity of the virtual energy storage model is obtained.

4. A system for implementing the two-stage self-organizing optimization aggregation method of distributed resources of a virtual power plant according to any one of claims 1 to 3, characterized in that, The system comprises a power distribution area, a first-level aggregation module, a second-level aggregation module, an edge computing server and a cloud, the edge computing server is deployed in the power distribution area, the first-level aggregation module comprises a generalized load module, a centralized generator module and a centralized energy storage module: A plurality of DERs are arranged in the power distribution area, and a plurality of the DERs constitute a natural physical cluster, the natural physical cluster is taken as a first level, and the edge computing server deployed in the power distribution area is used for aggregation; The edge computing server uses historical data called from the cloud to construct a hourly prediction model based on an uncertainty modeling method of deep Bayesian network learning for wind power, solar power and load curve of the entire power distribution area in the edge computing server; The generalized load module is used for constructing a generalized area load model containing wind power, photovoltaic power and load according to the hourly prediction model; The centralized generator module is used for aggregating distributed gas turbines and small hydropower station generators in the power distribution area into a unified virtual synchronous generator model; The centralized energy storage module is used for aggregating distributed energy storage in the power distribution area into a centralized virtual energy storage model; The second-level aggregation module is used for uploading all parameters of the generalized area load model, the virtual synchronous generator model and the virtual energy storage model to the cloud for second-level aggregation. 5.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-4 when the computer program is executed by the processor. The processor executes the computer program to realize the steps of the virtual power plant distributed resource two-level self-organizing optimization aggregation method in any one of claims 1 to 3.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the virtual power plant distributed resource two-level self-organizing optimization aggregation method in any one of claims 1 to 3.

Images