New energy field group collaborative power distribution method and system, electronic equipment, storage medium and program product

Abstract

The invention belongs to the technical field of new energy access and control, and discloses a new energy field group collaborative power distribution method and system, electronic equipment, a storage medium and a program product. The new energy field group cooperative power distribution method comprises the following steps: acquiring local observation data required by an intelligent agent corresponding to a selected field station; performing a power distribution decision through a greedy strategy generated by an action value function of the intelligent agent corresponding to the selected station; in the training process of the intelligent agent corresponding to each station, firstly, key state variables used for representing dynamic constraint and regulation potential are extracted, then, an interactive observation space is constructed, and finally, based on the constructed interactive observation space and a designed multi-target collaborative reward function, centralized learning-distributed execution multi-agent training is carried out. In the technical scheme disclosed by the invention, new energy field group collaborative power distribution is realized based on multi-agent reinforcement learning, and the power control capability of the new energy field group can be effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of new energy access and control technology, and specifically relates to a method, system, electronic equipment, storage medium and program product for coordinated power allocation of new energy power fields. Background Technology

[0002] As the scale of new energy power plants continues to expand, their ability to respond quickly and accurately to grid power dispatch has become crucial for improving system control capabilities.

[0003] Currently, traditional centralized optimization methods face technical challenges such as high computational complexity, model mismatch, and response delays when dealing with renewable energy clusters that are widely distributed and have diverse output characteristics. Explained, wind farms and photovoltaic power plants within a cluster exhibit drastically different output characteristics and controllability due to differences in resource characteristics and geographical location. For example, photovoltaic power plants are governed by diurnal cycles, exhibiting significant start-stop characteristics and short-term fluctuations; wind farms, on the other hand, show significant differences in their generating capacity and ramp-up potential due to variations in their micro-topography. Given these circumstances, existing control strategies struggle to dynamically adapt to these complex characteristics, resulting in an inability to achieve an optimal balance between tracking speed, allocation accuracy, and operational stability when responding to grid dispatch commands. This hinders the realization of the potential of renewable energy clusters as a unified and flexible resource.

[0004] In summary, how to further improve the rapid response capability of new energy power plants to grid power dispatch under various operating conditions has become an urgent technical challenge. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, electronic device, storage medium, and program product for coordinated power allocation among renewable energy power plants, in order to solve one or more of the aforementioned technical problems. Specifically, the technical solution disclosed in this invention is a multi-field coordinated control scheme, which achieves coordinated power allocation among renewable energy power plants based on multi-agent reinforcement learning, effectively improving the power control capability of renewable energy power plants.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for coordinated power allocation among new energy power plants, comprising the following steps: Based on selected power stations in the new energy power cluster, obtain the local observation data required by the corresponding intelligent agent of the selected power station; wherein, the local observation data includes its own state information, global information and peer interaction information; Based on the acquired local observation data, a greedy strategy is used to generate the power allocation decision by selecting the action value function of the corresponding intelligent agent of the field station. In this system, each power station within the new energy power cluster corresponds to an intelligent agent. During the training process of the intelligent agents corresponding to each power station, key state variables are extracted based on the operational characteristics of different types of power stations within the new energy power cluster to characterize dynamic constraints and control potential. Based on the extracted key state variables, an interactive observation space is constructed. Based on the constructed interactive observation space and the designed multi-objective collaborative reward function, multi-agent training is carried out through centralized learning and distributed execution.

[0007] A further improvement to the technical solution of this invention lies in the step of extracting key state variables for characterizing dynamic constraints and control potential based on the operational characteristics of different types of power stations within a new energy power cluster. Based on the photovoltaic power station, extract the dispatchable status flag to distinguish whether it is in an effective output period, and the output volatility index to characterize the current stability level. Based on the wind farm, extract the geographic location attribute identifier to represent the category of the region; The real-time ramp-up capability of each station is calculated based on both photovoltaic and wind power stations.

[0008] A further improvement of the technical solution of the present invention is that, based on the photovoltaic power station, the standard deviation of recent power output data is used as an output volatility index to characterize the current stability level.

[0009] A further improvement to the technical solution of this invention lies in the step of constructing an interactive observation space based on the extracted key state variables. Using a "partially observable" setting, the observation space for each agent at each site is represented as follows: =[ ]; In the formula, For the first i The observation space of the intelligent agent corresponding to each station; For its own state; This is global information; For interactive information; in, Use the extracted key state variables; This includes the total power demand issued by the power grid and the current total output of the renewable energy cluster; the interactive information includes the adjustment margin of other intelligent agents in the renewable energy cluster.

[0010] A further improvement of the technical solution of the present invention lies in the multi-objective collaborative reward function. Represented as: ; In the formula, They are respectively , , , Weighting coefficients; , , , These are respectively: tracking accuracy reward, fair allocation reward, motion smoothness reward, and constraint terms; in, ; In the formula, For a moment t Total power demand; For station i At any moment t Actual power output; It is a pre-defined, extremely small constant; N Total number of stations; ; In the formula, It is a station The change in power command. It is the change in total power demand; For station i At any moment t Available power, For station j At any moment t Available power; ; In the formula, For station i At any moment t Actual power output; For station i At any moment t -1 actual power output; For station i Maximum power output; ; In the formula, It is an indicator function; its value is 1 when the condition is true, and 0 otherwise. The penalty coefficient is... It is a station i Maximum gradeability; For station i The change in power.

[0011] In a second aspect, the present invention provides a new energy power distribution system for coordinated power allocation among power plants, comprising: The observation data acquisition unit is used to acquire the local observation data required by the corresponding intelligent agent of the selected station in the new energy field cluster; wherein, the local observation data includes its own state information, global information and peer interaction information; The allocation decision acquisition unit is used to make power allocation decisions based on the acquired local observation data and through a greedy strategy generated by the action value function of the selected intelligent agent corresponding to the station. In this system, each power station within the new energy power cluster corresponds to an intelligent agent. During the training process of the intelligent agents corresponding to each power station, key state variables are extracted based on the operational characteristics of different types of power stations within the new energy power cluster to characterize dynamic constraints and control potential. Based on the extracted key state variables, an interactive observation space is constructed. Based on the constructed interactive observation space and the designed multi-objective collaborative reward function, multi-agent training is carried out through centralized learning and distributed execution.

[0012] A further improvement to the technical solution of this invention is that it further includes: a state variable extraction unit, wherein the state variable extraction unit is configured to extract key state variables for characterizing dynamic constraints and control potential based on the operating characteristics of different types of power stations within the new energy power cluster; wherein, The state variable extraction unit includes: The first extraction module is used to extract, based on the photovoltaic power station, a dispatchable status flag to distinguish whether it is in an effective power output period, and a power output fluctuation index to characterize the current stability level. The second extraction module is used to extract the geographic location attribute identifier that represents the region category of the wind farm. The calculation module is used to calculate the real-time ramp-up capability of both photovoltaic and wind power stations.

[0013] A further improvement of the technical solution of the present invention is that, in the first extraction module, the standard deviation of recent output data is used as an output fluctuation index to characterize the current stability level.

[0014] A further improvement to the technical solution of this invention is that it further includes: an observation space construction unit, which is configured to construct an interactive observation space based on extracted key state variables; wherein, a "partially observable" setting is adopted, and the observation space corresponding to each agent of each station is represented as follows: =[ ]; In the formula, For the first i The observation space of the intelligent agent corresponding to each station; For its own state; This is global information; For interactive information; in, Use the extracted key state variables; This includes the total power demand issued by the power grid and the current total output of the renewable energy cluster; the interactive information includes the adjustment margin of other intelligent agents in the renewable energy cluster.

[0015] A further improvement of the technical solution of the present invention lies in the multi-objective collaborative reward function. Represented as: ; In the formula, They are respectively , , , Weighting coefficients; , , , These are respectively: tracking accuracy reward, fair allocation reward, motion smoothness reward, and constraint terms; in, ; In the formula, For a moment t Total power demand; For station i At any moment t Actual power output; It is a pre-defined, extremely small constant; N Total number of stations; ; In the formula, It is a station The change in power command. It is the change in total power demand; For station i At any moment t Available power, For station j At any moment t Available power; ; In the formula, For station i At any moment t Actual power output; For station i At any moment t -1 actual power output; For station i Maximum power output; ; In the formula, It is an indicator function; its value is 1 when the condition is true, and 0 otherwise. The penalty coefficient is... It is a station i Maximum gradeability; For station i The change in power.

[0016] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the new energy field cluster coordinated power allocation method as described in any one of the first aspects of the present invention.

[0017] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the new energy field cluster coordinated power allocation method as described in any one of the first aspects of the present invention.

[0018] In a fifth aspect, the present invention provides a computer program product comprising computer instructions which, when executed by a processor, implement the steps of the new energy field cluster coordinated power allocation method as described in any one of the first aspects of the present invention.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The new energy power grid collaborative power allocation method disclosed in this invention extracts the dynamic operating characteristics of different types of power plants into key state variables and designs intelligent agents with interactive observation spaces. This enables each power plant agent to not only perceive the global scheduling needs when making decisions, but also to understand the operating status and constraints of its peers. Ultimately, the following objectives are achieved: the new energy power grid can respond quickly and accurately to the total power demand of the grid, while minimizing the power adjustment amplitude of each power plant in the dynamic process. This ensures that the allocation strategy achieves optimal overall performance while meeting the safety constraints of the entire power grid, thereby effectively improving the power control capability of the power system. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for coordinated power allocation among new energy power plants, as described in an embodiment of the present invention. Figure 2This is a schematic diagram of a distributed control framework in an embodiment of the present invention; Figure 3 This is a schematic diagram of the multi-agent training algorithm in an embodiment of the present invention; Figure 4 This is a schematic diagram of a new energy field cluster coordinated power distribution system in an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention; obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0023] Based on the technical solutions disclosed in the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0024] Please see Figure 1 The present invention provides a method for coordinated power allocation among new energy power plants, comprising the following steps: Step 1: Based on the selected stations in the new energy field cluster, obtain the local observation data required by the corresponding intelligent agent of the selected station; the local observation data includes its own state, global information and peer interaction information; Step 2: Based on the local observation data obtained in Step 1, power allocation decisions are made using a greedy strategy generated by the action value function of the selected intelligent agent. In this system, each power station in the new energy power cluster corresponds to an intelligent agent. During the training process of the intelligent agents corresponding to each power station, key state variables for characterizing dynamic constraints and control potential are first extracted based on the operational characteristics of different types of power stations within the new energy power cluster. Then, an interactive observation space is constructed based on the extracted key state variables. Finally, based on the constructed interactive observation space, multi-agent training is carried out using a centralized learning-distributed execution method according to the designed multi-objective collaborative reward function. The multi-objective collaborative reward function includes tracking accuracy reward, allocation fairness reward, action smoothness reward, and constraint terms.

[0025] The technical solution provided in this invention is a multi-field cooperative control scheme based on a multi-agent reinforcement learning framework. Its core lies in designing an agent observation space and cooperative mechanism that can perceive the dynamic operating characteristics of power stations, thereby realizing adaptive distributed power optimization allocation and improving the power control capability of the power system.

[0026] Please see Figure 2 and Figure 3 In a specific exemplary technical solution of this invention, the provided multi-agent training algorithm includes the following steps: Step (1): Dynamic characteristics analysis and key state variable extraction of the station.

[0027] In a specific exemplary technical solution, this step first extracts key state variables to characterize the dynamic constraints and control potential of different types of stations within the field cluster, based on their operational characteristics. These variables will subsequently serve as part of the observation inputs for the intelligent agent. Specifically: 1) For photovoltaic power plants, extract their dispatchable status flags. This refers to the day / night cycle, used to distinguish whether the system is in an effective power output period. ; In the formula, For the current time, , These are the local sunrise and sunset times, Solar irradiance, The irradiance threshold that enables power generation; The values ​​of the dispatchable status flag must meet the requirements for photovoltaic power generation. The value of the dispatchable status flag is determined when the conditions for photovoltaic power generation are not met. Output volatility index The standard deviation of recent output data is used to characterize its current stability. ; In the formula, The length of the time window; For the past Photovoltaic output at each time step This represents the average output within that time window; 2) For wind farms, extract their geographic location attribute identifiers. Based on historical data, the wind farm is divided into high wind speed / high fluctuation zone and low wind speed / stable zone (i.e., this label is used to indicate the regional category to which the wind farm belongs). "High wind speed / high fluctuation zone", "Low wind speed / stable zone" ; 4) Calculate the real-time climbing capacity of each station. (Available power adjustments based on current output and forecast data): ; In the formula, For wind farm stations at any time Real-time hill-climbing capability; These are the wind farm's uphill ramp capability (i.e., the power that the farm can increase, also known as available uphill power) and downhill ramp capability (i.e., the power that the farm can decrease, also known as available downhill power). These represent the upper and lower limits of the technical output of wind farms; For wind farm stations at any time The current actual output; Contributing to ultra-short-term forecasting for wind farms The time interval for prediction; A confidence coefficient less than 1; Indicates taking a positive value; For photovoltaic power station time Real-time hill-climbing capability; These refer to the photovoltaic power station's ability to climb uphill and its ability to climb downhill. These represent the upper and lower limits of the technical contribution of photovoltaic power stations; For photovoltaic power station time The current actual output; Contribute to ultra-short-term forecasting for photovoltaic power plants.

[0028] Step (2), construction of interactive observation space.

[0029] In a specific exemplary technical solution, this step constructs an interactive observation space based on the key state variables extracted in step (1); wherein, the "partially observable" setting is adopted, and the observation space of each agent (corresponding to each station) is... It includes not only its own state, but also key information obtained from other intelligent agents for collaborative decision-making, represented as: =[ ]; In the formula, This refers to its own state, i.e., the key state variables extracted in the above steps; This is global information; For interactive information.

[0030] For example, for photovoltaic power stations and wind farm station They are respectively: Global information may include the total power demand issued by the power grid. The current total output of the field cluster Interaction information may include the regulatory margin of other agents. Specifically, it is expressed as: = ;in, , Stations The power can be adjusted up or down. This is the maximum power.

[0031] Step (3) Design of multi-objective collaborative reward function.

[0032] In a specific exemplary technical solution, this step, targeting the goal of improving power control capabilities, designs a specific multi-objective cooperative reward function. , is represented as: ; In the formula, They are respectively , , , Weighting coefficients; , , , These are tracking accuracy reward, allocation fairness reward, motion smoothness reward, and constraint terms.

[0033] In the specific exemplary technical solutions of the embodiments of the present invention, ; in, It is an extremely small constant; For a moment t Total power demand; For station i At any moment t Actual power output; ; in, It is a station The change in power command. It is the change in total power demand; For station i At any moment t Available power, For station j At any moment t Available power, (when )or (when ); ; in, For station i At any moment t Actual power output; For station i At any moment t -1 actual power output; For station i Maximum power output; ; in, It is an indicator function; its value is 1 when the condition is true, and 0 otherwise. The penalty coefficient is... It is a station i Maximum gradeability; For station i The change in power.

[0034] Step (4) is multi-agent training based on centralized learning and distributed execution.

[0035] In the specific example's technical solution, this step, based on the specific optimization objectives of the above design, involves centralized learning and distributed execution of multi-agent training; whereby... During the network initialization phase, a centralized Critic network is used. The input to this network is the joint observation of all agents. and joint actions The training objective is to decompose the network into the action-value function of each agent. The mixture is represented as: ; in, It is a mixed function, ensuring and Consistency.

[0036] During the network parameter update phase, and during training, based on global reward... and the next state The parameters of all agents are updated simultaneously using gradient descent. and After training is complete, remove the centralized Critic network. Each agent relies solely on its own action value function. Generated greedy strategy Make decisions.

[0037] In the subsequent application phase, each agent will base its actions on local observations. (Including its own state, global information, and interaction information with its companions) Independently allocate power commands This enables power control.

[0038] In a specific embodiment of the present invention, a single photovoltaic power station is used. and two wind farms Taking a new energy power plant cluster as an example, the application process of the technical solution of the present invention is further illustrated, and the specific steps are as follows: 1. Extraction of key state variables for photovoltaic power plants: During the effective daytime power output period, However, its output fluctuates due to the passing of clouds, resulting in volatility indicators. High; for wind farms Located in an area of ​​high wind speed and high fluctuation, its geographical location attribute is identified. Its output varies drastically, but it has a large potential for adjustment; for wind farms Located in a low-wind-speed, stable area, its geographical location attribute is identified. Its output is relatively stable, but its adjustable range is small. In summary, the status of each station is as follows. .

[0039] 2. Calculate real-time climbing ability The calculation results are shown in Table 1.

[0040] Table 1. Calculation results of real-time climbing ability

[0041] 3. Read the total power demand issued by the power grid. The current total output of the field cluster To obtain global information In agent-based distributed decision-making, interaction information is calculated based on the data from steps 1 and 2. ,renew Each of these power stations has a rated power of 100MW and a confidence coefficient of [missing information]. The results of the interactive information calculation are shown in Table 2.

[0042] Table 2. Results of Interactive Information Calculation

[0043] 4. Update the observation space based on the self-state, global information, and interaction information obtained from the above steps. Perform intensive training, output actions, update network parameters, and remove the intensive Critic network once training is complete. Proceed directly to step 5; otherwise, return to step 1. 5. Each station's intelligent agent relies on its own action value function. Generate a greedy strategy Based on the decision made, the total power demand decreased by 1.69MW in this control cycle, and the output actions of each station are shown in Table 3.

[0044] Table 3. Output Actions of Each Station

[0045] Finally, the calculation yields... =1, achieving complete tracking of a given power, proving the feasibility and advancement of the technical solution of the present invention.

[0046] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0047] Please see Figure 4 In this embodiment of the invention, a new energy field cluster coordinated power distribution system is provided, comprising: The observation data acquisition unit is used to acquire the local observation data required by the corresponding intelligent agent of the selected station in the new energy field cluster; wherein, the local observation data includes its own state information, global information and peer interaction information; The allocation decision acquisition unit is used to make power allocation decisions based on the acquired local observation data and through a greedy strategy generated by the action value function of the selected intelligent agent corresponding to the station. In this system, each power station within the new energy power cluster corresponds to an agent. During the training process of the agents corresponding to each power station, key state variables for characterizing dynamic constraints and control potential are first extracted based on the operational characteristics of different types of power stations within the new energy power cluster. Then, an interactive observation space is constructed based on the extracted key state variables. Finally, multi-agent training is carried out using a centralized learning-distributed execution approach based on the constructed interactive observation space and the designed multi-objective collaborative reward function. The multi-objective collaborative reward function includes tracking accuracy reward, allocation fairness reward, action smoothness reward, and constraint terms.

[0048] In one embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve corresponding method flows or corresponding functions. The processor described in this embodiment of the present invention can be used to execute the operation of a new energy field-group coordinated power allocation method.

[0049] In one embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM (Random Access Memory) or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the new energy field-group coordinated power allocation method in the above embodiments.

[0050] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0051] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations 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, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0052] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0054] 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 it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for coordinated power allocation among new energy power plants, characterized in that, Includes the following steps: Based on selected power stations in the new energy power cluster, obtain the local observation data required by the corresponding intelligent agent of the selected power station; wherein, the local observation data includes its own state information, global information and peer interaction information; Based on the acquired local observation data, a greedy strategy is used to generate the power allocation decision by selecting the action value function of the corresponding intelligent agent of the field station. In this system, each power station within the new energy power cluster corresponds to an intelligent agent. During the training process of the intelligent agents corresponding to each power station, key state variables are extracted based on the operational characteristics of different types of power stations within the new energy power cluster to characterize dynamic constraints and control potential. Based on the extracted key state variables, an interactive observation space is constructed. Based on the constructed interactive observation space and the designed multi-objective collaborative reward function, multi-agent training is carried out through centralized learning and distributed execution.

2. The method for coordinated power allocation among new energy power plants according to claim 1, characterized in that, Based on the operational characteristics of different types of power stations within a renewable energy cluster, the step of extracting key state variables to characterize dynamic constraints and control potential is as follows: Based on the photovoltaic power station, extract the dispatchable status flag to distinguish whether it is in an effective output period, and the output volatility index to characterize the current stability level. Based on the wind farm, extract the geographic location attribute identifier to represent the category of the region; The real-time ramp-up capability of each station is calculated based on both photovoltaic and wind power stations.

3. The method for coordinated power allocation among new energy power plants according to claim 2, characterized in that, Based on the photovoltaic power station, the standard deviation of recent power output data is used as an indicator of power output volatility to characterize the current level of stability.

4. The method for coordinated power allocation among new energy power plants according to claim 1, characterized in that, In the step of constructing an interactive observation space based on the extracted key state variables... Using the "partially observable" setting, the observation space for each agent at each site is represented as follows: =[ ]； In the formula, For the first i The observation space of the intelligent agent corresponding to each station; For its own state; This is global information; For interactive information; in, Use the extracted key state variables; This includes the total power demand issued by the power grid and the current total output of the renewable energy cluster; the interactive information includes the adjustment margin of other intelligent agents in the renewable energy cluster.

5. The method for coordinated power allocation among new energy power plants according to claim 1, characterized in that, The multi-objective collaborative reward function Represented as: ； In the formula, They are respectively , , , Weighting coefficients; , , , These are respectively: tracking accuracy reward, fair allocation reward, motion smoothness reward, and constraint terms; in, ； In the formula, For a moment t Total power demand; For station i At any moment t Actual power output; It is a pre-defined, extremely small constant; N Total number of stations; ； In the formula, It is a station The change in power command. It is the change in total power demand; For station i At any moment t Available power, For station j At any moment t Available power; ； In the formula, For station i At any moment t Actual power output; For station i At any moment t -1 actual power output; For station i Maximum power output; ； In the formula, It is an indicator function; its value is 1 when the condition is true, and 0 otherwise. The penalty coefficient is... It is a station i Maximum gradeability; For station i The change in power.

6. A new energy power distribution system for coordinated power generation, characterized in that, include: The observation data acquisition unit is used to acquire the local observation data required by the corresponding intelligent agent of the selected station in the new energy field cluster; wherein, the local observation data includes its own state information, global information and peer interaction information; The allocation decision acquisition unit is used to make power allocation decisions based on the acquired local observation data and through a greedy strategy generated by the action value function of the selected intelligent agent corresponding to the station. In this system, each power station within the new energy power cluster corresponds to an intelligent agent. During the training process of the intelligent agents corresponding to each power station, key state variables are extracted based on the operational characteristics of different types of power stations within the new energy power cluster to characterize dynamic constraints and control potential. Based on the extracted key state variables, an interactive observation space is constructed. Based on the constructed interactive observation space and the designed multi-objective collaborative reward function, multi-agent training is carried out through centralized learning and distributed execution.

7. A new energy power distribution system for coordinated power allocation among power plants according to claim 6, characterized in that, Also includes: A state variable extraction unit is configured to extract key state variables characterizing dynamic constraints and control potential based on the operational characteristics of different types of power stations within the new energy power cluster; wherein, The state variable extraction unit includes: The first extraction module is used to extract, based on the photovoltaic power station, a dispatchable status flag to distinguish whether it is in an effective power output period, and a power output fluctuation index to characterize the current stability level. The second extraction module is used to extract the geographic location attribute identifier that represents the region category of the wind farm. The calculation module is used to calculate the real-time ramp-up capability of both photovoltaic and wind power stations.

8. A new energy power distribution system according to claim 7, characterized in that, In the first extraction module, the standard deviation of recent output data is used as an output volatility index to characterize the current level of stability.

9. A new energy power distribution system for coordinated power allocation among power plants according to claim 6, characterized in that, It also includes: an observation space construction unit, which is configured to construct an interactive observation space based on extracted key state variables; wherein, using a "partially observable" setting, the observation space for each agent at each station is represented as follows: =[ ]； In the formula, For the first i The observation space of the intelligent agent corresponding to each station; For its own state; This is global information; For interactive information; in, Use the extracted key state variables; This includes the total power demand issued by the power grid and the current total output of the renewable energy cluster; the interactive information includes the adjustment margin of other intelligent agents in the renewable energy cluster.

10. A new energy power distribution system according to claim 6, characterized in that, The multi-objective collaborative reward function Represented as: ； In the formula, They are respectively , , , Weighting coefficients; , , , These are respectively: tracking accuracy reward, fair allocation reward, motion smoothness reward, and constraint terms; in, ； In the formula, For a moment t Total power demand; For station i At any moment t Actual power output; It is a pre-defined, extremely small constant; N Total number of stations; ； In the formula, It is a station The change in power command. It is the change in total power demand; For station i At any moment t Available power, For station j At any moment t Available power; ； In the formula, For station i At any moment t Actual power output; For station i At any moment t -1 actual power output; For station i Maximum power output; ； In the formula, It is an indicator function; its value is 1 when the condition is true, and 0 otherwise. The penalty coefficient is... It is a station i Maximum gradeability; For station i The change in power.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the new energy field cluster coordinated power allocation method as described in any one of claims 1 to 5.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the new energy field cluster coordinated power allocation method as described in any one of claims 1 to 5.

13. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the new energy field cluster coordinated power allocation method as described in any one of claims 1 to 5.

Images