Virtual power plant economic dispatch method and device based on neural network under two-layer architecture

Abstract

The present invention relates to the technical field of electric power dispatching, and discloses a virtual power plant economic dispatching method and device based on a neural network under a two-layer architecture. The present invention selects data such as the supply-demand ratio of the entire network from the early to the late period of the historical decision period, the clearing price of the entire network, the supply-demand ratio of the virtual power plant, the storage capacity of the energy storage device, and the response times of the controllable load as input signals, uses the normalized value of the input signal as the input of the upper neural network intelligent body, and uses the normalized value of the exchange power of the energy storage device and the response power of the controllable load as the output to perform neural network training; inputs the actual input signal set into the trained upper neural network intelligent body, uses the original data value of the output signal obtained as the boundary condition, and constructs a lower-level decision optimization model with the goal of maximizing the expected benefit, and solves the model to obtain the corresponding virtual power plant economic dispatching plan. The present invention can effectively improve the efficiency and accuracy of solving the economic dispatching problem of virtual power plants.

Description

Technical Field

[0001] The present invention relates to the technical field of power dispatching, and in particular to a method and device for economic dispatching of a virtual power plant based on a neural network under a double-layer architecture. Background Art

[0002] Virtual power plants are essentially a special type of load aggregation that has a certain degree of electricity flexibility. They can change their electricity demand in response to the power supply demand of the large power grid, and present the external characteristics of power generation of power plants to the outside world. There are many types of adjustable resources in virtual power plants, such as wind power, photovoltaics, energy storage, gas units, etc. In order to maximize the expected benefits, economic scheduling is required. The economic scheduling problem of virtual power plants is essentially a process of optimizing the operation plan of adjustable resources based on the prediction results of the operating status during the decision-making period, with the goal of maximizing the expected benefits.

[0003] The main difficulty of the virtual power plant economic dispatch problem is that its optimization decision is based on the prediction of uncertain factors such as electricity price, demand-side response, load, and new energy. It is essentially a highly nonlinear and uncertain planning problem, and its efficient solution faces challenges. In recent years, energy storage devices have been widely promoted and applied, and their operating status decisions have become more complex, resulting in the need to add mixed integer decision variables to the above virtual power plant economic dispatch problem, which further increases the difficulty of solving it.

[0004] There are two main technical paths for solving the economic dispatch problem of virtual power plants. The first technical path is to use classical mathematical programming methods for direct solution. However, nonlinear mixed integer programming problems have always been difficult to solve mathematical problems, and it is difficult to achieve efficient and high-precision solutions to the above problems. The current problem of low solution efficiency when using this method to solve the economic dispatch problem of virtual power plants is particularly prominent, especially in real-time economic dispatch, which is difficult to meet the actual application needs of virtual power plants. Summary of the invention

[0005] The present invention provides a method and device for economic dispatching of a virtual power plant based on a neural network under a two-layer architecture, which solves the technical problem that the efficiency and accuracy of existing methods for solving the economic dispatching problem of virtual power plants cannot meet the actual application needs of virtual power plants.

[0006] A first aspect of the present invention provides a virtual power plant economic dispatch method based on a neural network under a two-layer architecture, the method comprising:

[0007] Select historical input signals and perform data normalization processing, and construct a corresponding historical input signal set based on the normalized data, wherein the selected historical input signals include the network supply-demand ratio data from the early stage to the late stage of the historical decision period, the network clearing price data, the virtual power plant supply-demand ratio data, the energy storage capacity data of the energy storage device, and the response times data of the controllable load;

[0008] Constructing an upper neural network intelligent agent, taking the historical input signal set as the input of the upper neural network intelligent agent, taking the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network intelligent agent, training the upper neural network intelligent agent, and obtaining a trained upper neural network intelligent agent;

[0009] Input the actual input signal set into the trained upper neural network agent, restore the obtained normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, and construct a lower-level decision optimization model based on the virtual power plant economic dispatch problem with the original data values ​​as boundary conditions and the expected profit maximization as the goal;

[0010] The classical mathematical programming method is used to solve the lower-level decision-making optimization model to obtain the corresponding virtual power plant economic dispatch plan.

[0011] According to an achievable manner of the first aspect of the present invention, the data normalization processing includes:

[0012] The historical input signal is processed using a maximum-minimum value normalization method.

[0013] According to an achievable manner of the first aspect of the present invention, the processing of the historical input signal by using a maximum-minimum normalization method includes:

[0014] Taking the maximum supply-demand ratio and the minimum supply-demand ratio in the network-wide supply-demand ratio data as a benchmark, converting the network-wide supply-demand ratio data into a value between 0 and 1;

[0015] Based on the highest clearing price and the lowest clearing price of the system, convert the network-wide clearing price data into a value between 0 and 1;

[0016] Based on the maximum supply-demand ratio and the minimum supply-demand ratio in the virtual power plant supply-demand ratio data, convert the virtual power plant supply-demand ratio data into a value between 0 and 1;

[0017] Based on the maximum storage capacity of the energy storage device, convert the storage capacity data into a value between 0 and 1;

[0018] Based on the maximum response times of the controllable load, the response times data is converted into a value between 0 and 1.

[0019] According to an achievable manner of the first aspect of the present invention, the normalized value of the energy storage device exchange power and the controllable load response power is used as the output signal of the upper neural network agent, including:

[0020] Obtaining the energy storage device exchange power and controllable load response power corresponding to the historical input signal;

[0021] Based on the maximum charge and discharge power of the energy storage device, the maximum-minimum normalization method is used to convert the exchange power of the energy storage device into a value between -1 and 1; and based on the maximum response power of the controllable load, the maximum-minimum normalization method is used to convert the response power of the controllable load into a value between 0 and 1.

[0022] A second aspect of the present invention provides a virtual power plant economic dispatch device based on a neural network under a two-layer architecture, the device comprising:

[0023] A historical input signal set construction module is used to select historical input signals and perform data normalization processing, and construct a corresponding historical input signal set based on the normalized data, wherein the selected historical input signals include the network supply-demand ratio data from the early stage to the late stage of the historical decision period, the network clearing price data, the virtual power plant supply-demand ratio data, the energy storage device storage capacity data, and the controllable load response times data;

[0024] An upper neural network agent training module is used to construct an upper neural network agent, use the historical input signal set as the input of the upper neural network agent, use the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network agent, train the upper neural network agent, and obtain a trained upper neural network agent;

[0025] A lower-level decision optimization model construction module is used to input the actual input signal set into the trained upper-level neural network agent, restore the obtained normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, and construct a lower-level decision optimization model based on the virtual power plant economic dispatch problem with the original data values ​​as boundary conditions and the expected profit maximization as the goal;

[0026] The economic dispatch problem solving module is used to solve the lower-level decision optimization model using a classical mathematical programming device to obtain a corresponding virtual power plant economic dispatch plan.

[0027] According to an achievable manner of the second aspect of the present invention, the historical input signal set construction module includes a data normalization unit, and the data normalization unit is used to:

[0028] The historical input signal is processed using a maximum-minimum value normalization device.

[0029] According to an achievable manner of the second aspect of the present invention, the data normalization unit is specifically used for:

[0030] Taking the maximum supply-demand ratio and the minimum supply-demand ratio in the network-wide supply-demand ratio data as a benchmark, converting the network-wide supply-demand ratio data into a value between 0 and 1;

[0031] Based on the highest clearing price and the lowest clearing price of the system, convert the network-wide clearing price data into a value between 0 and 1;

[0032] Based on the maximum supply-demand ratio and the minimum supply-demand ratio in the virtual power plant supply-demand ratio data, convert the virtual power plant supply-demand ratio data into a value between 0 and 1;

[0033] Based on the maximum storage capacity of the energy storage device, convert the storage capacity data into a value between 0 and 1;

[0034] Based on the maximum response times of the controllable load, the response times data is converted into a value between 0 and 1.

[0035] According to an achievable manner of the second aspect of the present invention, when the upper neural network agent training module uses the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network agent, it is specifically used to:

[0036] Obtaining the energy storage device exchange power and controllable load response power corresponding to the historical input signal;

[0037] Based on the maximum charge and discharge power of the energy storage device, a maximum-minimum value normalization device is used to convert the obtained exchange power of the energy storage device into a value between -1 and 1; and based on the maximum response power of the controllable load, a maximum-minimum value normalization device is used to convert the obtained controllable load response power into a value between 0 and 1.

[0038] The third aspect of the present invention provides a virtual power plant economic dispatching device based on a neural network under a two-layer architecture, comprising:

[0039] A memory for storing instructions; wherein the instructions are instructions for implementing the virtual power plant economic dispatch method based on a neural network under a two-layer architecture as described in any of the above implementation methods;

[0040] A processor is used to execute instructions in the memory.

[0041] A fourth aspect of the present invention is a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a virtual power plant economic dispatch method based on a neural network under a two-layer architecture as described in any of the above implementation methods.

[0042] It can be seen from the above technical solutions that the present invention has the following advantages:

[0043] The present invention selects the network-wide supply-demand ratio data, network-wide clearing price data, virtual power plant supply-demand ratio data, energy storage device power storage data, and controllable load response times data from the early to late stages of the historical decision period as input signals, uses the normalized value of the input signal as the input of the upper neural network intelligent agent, and uses the normalized values ​​of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network intelligent agent to train the upper neural network intelligent agent; inputs the actual input signal set into the trained upper neural network intelligent agent, restores the obtained normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, uses the original data values ​​as boundary conditions, and takes the expected benefit maximization as the goal, to construct a virtual power plant economic system based on the virtual power plant. The lower-level decision optimization model of the scheduling problem is further solved to obtain the corresponding virtual power plant economic scheduling plan; the present invention splits the mixed integer decision items and continuous variable decision items in the virtual power plant economic scheduling model into upper and lower decision problems, the upper layer is the upper-level neural network intelligent body for mixed integer decision items such as energy storage, and the lower layer is the lower-level decision optimization model based on the operating status of a given energy storage device. By solving the mixed integer decision variables in the upper-level neural network intelligent body, the original model of the economic scheduling of the virtual power plant is simplified, so that the existing classical mathematical programming method can be used to directly solve the lower-level decision optimization model, which can effectively improve the efficiency and accuracy of solving the economic scheduling problem of the virtual power plant and meet the actual application needs of the virtual power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 A flowchart of a virtual power plant economic dispatch method based on a neural network under a two-layer architecture provided by an optional embodiment of the present invention;

[0046] Figure 2 A structural connection block diagram of a virtual power plant economic dispatching device based on a neural network in a two-layer architecture is provided as an optional embodiment of the present invention.

[0047] Reference numerals:

[0048] 1-Historical input signal set construction module; 2-Upper-level neural network agent training module; 3-Lower-level decision optimization model construction module; 4-Economic scheduling problem solving module. DETAILED DESCRIPTION

[0049] The embodiments of the present invention provide a virtual power plant economic dispatch method and device based on a neural network under a two-layer architecture, which are used to solve the technical problem that the efficiency and accuracy of the existing methods for solving the economic dispatch problem of virtual power plants cannot meet the actual application needs of virtual power plants.

[0050] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The present invention provides a virtual power plant economic dispatching method based on a neural network under a double-layer architecture.

[0052] See also Figure 1 , Figure 1 A flowchart of a virtual power plant economic dispatch method based on a neural network under a two-layer architecture provided by an embodiment of the present invention is shown.

[0053] An embodiment of the present invention provides a virtual power plant economic dispatch method based on a neural network under a two-layer architecture, comprising:

[0054] Step S1, select historical input signals and perform data normalization processing, and construct a corresponding historical input signal set based on the normalized data, wherein the selected historical input signals include the supply-demand ratio data of the entire network from the early to the late period of the historical decision period, the clearing price data of the entire network, the supply-demand ratio data of the virtual power plant, the storage capacity data of the energy storage device, and the response times data of the controllable load.

[0055] The purpose of this step is to construct the input signal of the upper neural network agent and process the data according to the actual economic dispatch problem of the virtual power plant. Although the real-time economic dispatch stage only optimizes the operation plan for the next decision cycle (generally 15 minutes), it is necessary to coordinate the entire decision day, especially the operation in the later period of the decision day.

[0056] Among them, the supply and demand ratio data of the entire network include the maximum supply and demand ratio in the early stage of the entire network decision-making period, the minimum supply and demand ratio in the early stage of the entire network decision-making period, the supply and demand ratio of the entire network decision-making period, the maximum supply and demand ratio in the later stage of the entire network decision-making period and the minimum supply and demand ratio in the later stage of the entire network decision-making period.

[0057] The network-wide clearing price data includes the highest clearing price before the network-wide decision-making period and the lowest clearing price before the network-wide decision-making period.

[0058] The virtual power plant supply and demand ratio data includes the supply and demand ratio during the virtual power plant decision period, the maximum supply and demand ratio in the later period of the virtual power plant decision period, and the minimum supply and demand ratio in the later period of the virtual power plant decision period.

[0059] From the perspective of object and time, the above input signals can be divided into different types.

[0060] From the object perspective, the above input signals can be divided into full-grid type signals, virtual power plant type signals, and control object type signals, as shown in Table 1:

[0061] Table 1: Input signal type classification

[0062]

[0063] From a timing perspective, the above input signal can be divided into three timings: early decision period, decision period, and late decision period.

[0064] The above indicators are selected as input signals mainly from the perspective of their influence on the adjustable power supply corresponding to the mixed integer decision variable. The specific reasons are:

[0065] (1) The change in the supply-demand ratio of the entire network is the key variable that determines the trend of the transaction clearing price. Therefore, seven indicators, including the maximum supply-demand ratio in the early decision period, the minimum supply-demand ratio in the early decision period, the highest clearing price in the early decision period, the lowest clearing price in the early decision period, the supply-demand ratio in the decision period, the minimum supply-demand ratio in the late decision period, and the maximum supply-demand ratio in the late decision period, are selected as the decision information for the intelligent agent to analyze the system transaction clearing price;

[0066] (2) Similar to the situation of the entire grid, the supply-demand ratio of the virtual power plant itself reflects the changes in its supply situation and is also an important reference for decision-making on energy storage devices and controllable loads;

[0067] (3) Control object information, including the storage capacity of the energy storage device and the response times of the controllable load of the virtual power plant. These information directly restrict the operating status of the system.

[0068] As an achievable manner, when performing data normalization processing, the historical input signal may be processed using a maximum-minimum value normalization method.

[0069] In specific implementation, for the supply-demand ratio data of the entire network, taking the maximum supply-demand ratio and the minimum supply-demand ratio in the supply-demand ratio data of the entire network as the benchmark, the maximum supply-demand ratio in the early stage of the decision-making period of the entire network, the minimum supply-demand ratio in the early stage of the decision-making period of the entire network, the supply-demand ratio in the decision-making period of the entire network, the maximum supply-demand ratio in the late stage of the decision-making period of the entire network, and the minimum supply-demand ratio in the late stage of the decision-making period of the entire network are all converted into coefficients between 0 and 1;

[0070] For the network-wide clearing price data, the highest clearing price in the previous period of the network-wide decision-making period and the lowest clearing price in the previous period of the network-wide decision-making period are used as the benchmarks to convert the highest clearing price in the previous period of the network-wide decision-making period and the lowest clearing price in the previous period of the network-wide decision-making period into values ​​between 0 and 1;

[0071] For the supply-demand ratio data of the virtual power plant, based on the maximum supply-demand ratio and the minimum supply-demand ratio in the supply-demand ratio data of the virtual power plant, the supply-demand ratio of the virtual power plant decision period, the maximum supply-demand ratio in the late stage of the virtual power plant decision period, and the minimum supply-demand ratio in the late stage of the virtual power plant decision period are converted into values ​​between 0 and 1;

[0072] For the storage capacity data, the maximum storage capacity of the energy storage device is used as a benchmark to convert the storage capacity data into a value between 0 and 1;

[0073] As for the controllable load response times data, the response times data is converted into a value between 0 and 1 based on the maximum response times of the controllable load.

[0074] Step S2, constructing an upper neural network intelligent agent, taking the historical input signal set as the input of the upper neural network intelligent agent, taking the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network intelligent agent, training the upper neural network intelligent agent, and obtaining a trained upper neural network intelligent agent.

[0075] It should be noted that, in the embodiment of the present invention, the upper neural network agent can be trained based on the existing neural network training method. The embodiment of the present invention does not limit the neural network training method.

[0076] When obtaining the normalized value of the exchange power of the energy storage device and the controllable load response power, the following can be performed:

[0077] Obtaining the energy storage device exchange power and controllable load response power corresponding to the historical input signal;

[0078] Based on the maximum charge and discharge power of the energy storage device, the maximum-minimum normalization method is used to convert the exchange power of the energy storage device into a value between -1 and 1; and based on the maximum response power of the controllable load, the maximum-minimum normalization method is used to convert the response power of the controllable load into a value between 0 and 1.

[0079] When outputting a signal, the output signal can be converted into a data signal set form, which can be expressed as:

[0080] S 0 = {P N,S,D ,P N,CL,D}

[0081] In the formula, S 0is the output signal set, P N,S,D is the normalized value of the energy storage device exchange power, P N,CL,D It is the normalized value of the controllable load response power.

[0082] Step S3, input the actual input signal set into the trained upper neural network intelligent agent, restore the normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, use the original data values ​​as boundary conditions, and take maximizing the expected benefits as the goal to construct a lower-level decision-making optimization model based on the economic dispatch problem of the virtual power plant.

[0083] The actual input signal set can be obtained according to the power grid operation data during the real-time operation process.

[0084] Before the actual input signal set is input into the trained upper neural network agent, the actual input signal set data needs to be normalized. Specifically, the maximum supply-demand ratio of the entire network decision period in the early stage, the minimum supply-demand ratio of the entire network decision period in the early stage, the highest clearing price of the entire network decision period in the early stage, the lowest clearing price of the entire network decision period in the early stage, the supply-demand ratio of the entire network decision period, the minimum supply-demand ratio of the entire network decision period in the late stage, the maximum supply-demand ratio of the entire network decision period in the late stage, the supply-demand ratio of the virtual power plant decision period, the maximum supply-demand ratio of the virtual power plant decision period in the late stage, the minimum supply-demand ratio of the virtual power plant decision period in the late stage, the energy storage capacity of the virtual power plant, the number of controllable load response periods of the virtual power plant, etc. of the actual input signal set can be converted into a normalized form according to the processing method of the historical input signal in step S1.

[0085] The trained upper neural network agent can automatically calculate according to the above actual input signal set to obtain the normalized results of the exchange power of the energy storage device and the controllable load response power.

[0086] Energy storage devices and controllable load operating status are the direct reasons why the current economic dispatch model of virtual power plants with energy storage is a mixed integer programming problem and is difficult to solve. By solving the mixed integer decision variables in the upper-level neural network agent, the original model is simplified, and the lower-level decision optimization model can be directly solved using existing classical mathematical programming methods.

[0087] Step S4, using classical mathematical programming methods to solve the lower-level decision optimization model to obtain the corresponding virtual power plant economic dispatch plan.

[0088] Since the exchange power of the energy storage device and the controllable load response power in the lower-level decision optimization model have been determined, the model only has continuous decision variables. The model can be efficiently solved using current classical mathematical programming methods.

[0089] The present invention also provides a virtual power plant economic dispatching device based on a neural network under a two-layer architecture, which can be used to implement the above-mentioned virtual power plant economic dispatching method based on a neural network under a two-layer architecture.

[0090] See also Figure 2 , Figure 2 A structural connection block diagram of a virtual power plant economic dispatching device based on a neural network under a two-layer architecture provided by an embodiment of the present invention is shown.

[0091] The embodiment of the present invention provides a virtual power plant economic dispatching device based on a neural network under a two-layer architecture, comprising:

[0092] A historical input signal set construction module 1 is used to select historical input signals and perform data normalization processing, and construct a corresponding historical input signal set according to the normalized data, wherein the selected historical input signals include the supply-demand ratio data of the entire network from the early stage to the late stage of the historical decision period, the clearing price data of the entire network, the supply-demand ratio data of the virtual power plant, the storage capacity data of the energy storage device, and the response number data of the controllable load;

[0093] The upper neural network agent training module 2 is used to construct an upper neural network agent, use the historical input signal set as the input of the upper neural network agent, use the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network agent, train the upper neural network agent, and obtain a trained upper neural network agent;

[0094] The lower-level decision optimization model construction module 3 is used to input the actual input signal set into the trained upper-level neural network agent, restore the normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, and use the original data values ​​as boundary conditions and the expected profit maximization as the goal to construct a lower-level decision optimization model based on the economic dispatch problem of the virtual power plant;

[0095] The economic dispatch problem solving module 4 is used to solve the lower-level decision optimization model using a classical mathematical programming device to obtain a corresponding virtual power plant economic dispatch solution.

[0096] In one achievable manner, the historical input signal set construction module 1 includes a data normalization unit, and the data normalization unit is used to:

[0097] The historical input signal is processed using a maximum-minimum value normalization device.

[0098] In one achievable manner, the data normalization unit is specifically used for:

[0099] Taking the maximum supply-demand ratio and the minimum supply-demand ratio in the network-wide supply-demand ratio data as a benchmark, converting the network-wide supply-demand ratio data into a value between 0 and 1;

[0100] Based on the highest clearing price and the lowest clearing price of the system, convert the network-wide clearing price data into a value between 0 and 1;

[0101] Based on the maximum supply-demand ratio and the minimum supply-demand ratio in the virtual power plant supply-demand ratio data, convert the virtual power plant supply-demand ratio data into a value between 0 and 1;

[0102] Based on the maximum storage capacity of the energy storage device, convert the storage capacity data into a value between 0 and 1;

[0103] Based on the maximum response times of the controllable load, the response times data is converted into a value between 0 and 1.

[0104] In one achievable manner, when the upper neural network agent training module 2 uses the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network agent, it is specifically used to:

[0105] Obtaining the energy storage device exchange power and controllable load response power corresponding to the historical input signal;

[0106] Based on the maximum charge and discharge power of the energy storage device, a maximum-minimum value normalization device is used to convert the obtained exchange power of the energy storage device into a value between -1 and 1; and based on the maximum response power of the controllable load, a maximum-minimum value normalization device is used to convert the obtained controllable load response power into a value between 0 and 1.

[0107] The present invention also provides a virtual power plant economic dispatching device based on a neural network under a double-layer architecture, comprising:

[0108] A memory for storing instructions; wherein the instructions are instructions that can implement the virtual power plant economic dispatch method based on a neural network under a two-layer architecture as described in any one of the above embodiments;

[0109] A processor is used to execute instructions in the memory.

[0110] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the virtual power plant economic dispatch method based on a neural network under a two-layer architecture as described in any of the above embodiments is implemented.

[0111] Technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working processes of the devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments, and the specific beneficial effects of the devices and modules described above can refer to the corresponding beneficial effects in the aforementioned method embodiments, which will not be repeated here.

[0112] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0113] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0114] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0115] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0116] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A virtual power plant economic dispatch method based on neural network under a two-layer architecture. It is characterized in that The method comprises: Select historical input signals and perform data normalization processing, and construct a corresponding historical input signal set based on the normalized data, wherein the selected historical input signals include the network supply-demand ratio data from the early stage to the late stage of the historical decision period, the network clearing price data, the virtual power plant supply-demand ratio data, the energy storage capacity data of the energy storage device, and the response times data of the controllable load; Constructing an upper neural network intelligent agent, taking the historical input signal set as the input of the upper neural network intelligent agent, taking the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network intelligent agent, training the upper neural network intelligent agent, and obtaining a trained upper neural network intelligent agent, including: obtaining the energy storage device exchange power and the controllable load response power corresponding to the historical input signal; taking the maximum charge and discharge power of the energy storage device as a reference, using the maximum-minimum value normalization method to convert the obtained energy storage device exchange power into a value between -1 and 1; and taking the maximum response power of the controllable load as a reference, using the maximum-minimum value normalization method to convert the obtained controllable load response power into a value between 0 and 1; Input the actual input signal set into the trained upper neural network agent, restore the obtained normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, and construct a lower-level decision optimization model based on the virtual power plant economic dispatch problem with the original data values ​​as boundary conditions and the expected profit maximization as the goal; The classical mathematical programming method is used to solve the lower-level decision-making optimization model to obtain the corresponding virtual power plant economic dispatch plan.

2. According to the virtual power plant economic dispatch method based on neural network under the two-layer architecture of claim 1, It is characterized in that The data normalization process comprises: The historical input signal is processed using a maximum-minimum value normalization method.

3. According to the virtual power plant economic dispatch method based on neural network under the two-layer architecture of claim 2, It is characterized in that The method of processing the historical input signal by using the maximum-minimum value normalization method includes: Taking the maximum supply-demand ratio and the minimum supply-demand ratio in the network-wide supply-demand ratio data as a benchmark, converting the network-wide supply-demand ratio data into a value between 0 and 1; Based on the highest clearing price and the lowest clearing price of the system, convert the network-wide clearing price data into a value between 0 and 1; Based on the maximum supply-demand ratio and the minimum supply-demand ratio in the virtual power plant supply-demand ratio data, convert the virtual power plant supply-demand ratio data into a value between 0 and 1; Based on the maximum storage capacity of the energy storage device, convert the storage capacity data into a value between 0 and 1; Based on the maximum response times of the controllable load, the response times data is converted into a value between 0 and 1.

4. A virtual power plant economic dispatch device based on neural network under a two-layer architecture, It is characterized in that The device comprises: A historical input signal set construction module is used to select historical input signals and perform data normalization processing, and construct a corresponding historical input signal set based on the normalized data, wherein the selected historical input signals include the network supply-demand ratio data from the early stage to the late stage of the historical decision period, the network clearing price data, the virtual power plant supply-demand ratio data, the energy storage device storage capacity data, and the controllable load response times data; The upper neural network agent training module is used to construct an upper neural network agent, take the historical input signal set as the input of the upper neural network agent, take the normalized value of the energy storage device exchange power and the controllable load response power as the output signal of the upper neural network agent, train the upper neural network agent, and obtain a trained upper neural network agent, including: obtaining the energy storage device exchange power and the controllable load response power corresponding to the historical input signal; taking the maximum charge and discharge power of the energy storage device as a reference, using a maximum-minimum value normalization device to convert the obtained energy storage device exchange power into a value between -1 and 1; and taking the maximum response power of the controllable load as a reference, using a maximum-minimum value normalization device to convert the obtained controllable load response power into a value between 0 and 1; A lower-level decision optimization model construction module is used to input the actual input signal set into the trained upper-level neural network agent, restore the obtained normalized values ​​of the energy storage device exchange power and the controllable load response power to the original data values, and construct a lower-level decision optimization model based on the virtual power plant economic dispatch problem with the original data values ​​as boundary conditions and the expected profit maximization as the goal; The economic dispatch problem solving module is used to solve the lower-level decision optimization model using a classical mathematical programming device to obtain a corresponding virtual power plant economic dispatch plan.

5. According to the virtual power plant economic dispatching device based on neural network under the double-layer architecture of claim 4, It is characterized in that The historical input signal set building module includes a data normalization unit, and the data normalization unit is used to: The historical input signal is processed using a maximum-minimum value normalization device.

6. According to the double-layer architecture of claim 5, the virtual power plant economic dispatching device based on neural network, It is characterized in that The data normalization unit is specifically used for: Taking the maximum supply-demand ratio and the minimum supply-demand ratio in the network-wide supply-demand ratio data as a benchmark, converting the network-wide supply-demand ratio data into a value between 0 and 1; Based on the highest clearing price and the lowest clearing price of the system, convert the network-wide clearing price data into a value between 0 and 1; Based on the maximum supply-demand ratio and the minimum supply-demand ratio in the virtual power plant supply-demand ratio data, convert the virtual power plant supply-demand ratio data into a value between 0 and 1; Based on the maximum storage capacity of the energy storage device, convert the storage capacity data into a value between 0 and 1; Based on the maximum response times of the controllable load, the response times data is converted into a value between 0 and 1.

7. A virtual power plant economic dispatch device based on neural network under a two-layer architecture, It is characterized in that include: A memory for storing instructions; wherein the instructions are instructions for implementing the virtual power plant economic dispatch method based on a neural network under a two-layer architecture as described in any one of claims 1 to 3; A processor is used to execute instructions in the memory.

8. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, it implements the virtual power plant economic dispatch method based on neural network under the two-layer architecture as described in any one of claims 1-3.

Images