Urban virtual power plant operation method and device based on agent architecture

Abstract

The invention provides an urban virtual power plant operation method and device based on an intelligent agent architecture, relates to the field of artificial intelligence, and solves the problem that in the prior art, operation strategies related to a virtual power plant are mostly based on static rules and manually set optimization models. And rapid and efficient response is difficult to realize in a multi-participant, multi-constraint condition and dynamic electricity market environment. The method comprises the following steps: collecting multiple pieces of first data in an urban energy system in an urban virtual power plant; performing arrangement and task scheduling on the first data on the basis of data types required by the plurality of agents to execute tasks, determining the first data required to be processed by each agent, and calling the plurality of agents to process the first data required to be processed; and generating a regulation and control decision and a market transaction strategy of the urban virtual power plant based on the processing results output by the plurality of agents and the operation constraint conditions. The method is used in the operation process of the urban virtual power plant.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and in particular to a method and apparatus for operating a virtual power plant in a city based on an intelligent agent architecture. Background Technology

[0002] Existing technologies for virtual power plant operation strategies are mostly based on static rules and manually set optimization models, which make it difficult to achieve rapid and efficient responses in environments with multiple participants, multiple constraints, and dynamic electricity markets. Especially when facing complex scenarios such as real-time electricity price fluctuations, drastic changes in electricity load, and sudden grid events, traditional methods often lack flexibility and intelligence, making it difficult to fully leverage the resource synergy potential of virtual power plants. Therefore, there is an urgent need for an intelligent virtual power plant operation method to solve the above problems. Summary of the Invention

[0003] This application provides a method and apparatus for operating a virtual power plant in a city based on an intelligent agent architecture, which solves the technical problem that existing technologies are unable to achieve rapid and efficient responses in virtual power plant operation strategies under multiple participants, multiple constraints, and dynamic electricity market environments.

[0004] Firstly, a method for operating a city virtual power plant based on an intelligent agent architecture is provided, comprising: collecting multiple first data from the city energy system in the city virtual power plant, including distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environmental data; arranging and scheduling the first data based on the data types required for multiple intelligent agents to perform tasks, determining the first data to be processed by each intelligent agent, and calling multiple intelligent agents to process the required first data; and generating control decisions and market trading strategies for the city virtual power plant based on the processing results output by multiple intelligent agents and operating constraints.

[0005] In conjunction with the first aspect mentioned above, in one possible implementation, multiple intelligent agents include: a power load forecasting intelligent agent, used to forecast future time load and renewable energy output; a power supply scheduling optimization intelligent agent, used to output optimized task execution plans; a market trading intelligent agent, used to output market trading decisions for day-ahead markets, real-time markets, or ancillary service markets; and an equipment operation and maintenance intelligent agent, used to detect system operating status. Based on the data types required for multiple intelligent agents to execute tasks, the first data is orchestrated and task scheduled, including: classifying and storing the first data in a database; calling a large model inference engine to parse task requirements and determining the data to be processed by each intelligent agent based on task requirements; distributing the corresponding tasks and data to each intelligent agent; activating the intelligent agents to execute corresponding tasks; calling the large model inference engine to receive the results returned by each intelligent agent executing tasks; and optimizing the process and strategy of each intelligent agent executing tasks based on the returned results.

[0006] In conjunction with the first aspect mentioned above, in one possible implementation, the power load forecasting agent is specifically used to perform the following processes: receiving meteorological environmental data and adjustable load data from the first data; calling an industry-wide model to predict the load and renewable energy output within the first time period in the future; determining the prediction confidence and prediction accuracy based on the load and renewable energy output; and optimizing the prediction accuracy of the power load forecasting agent through the prediction confidence and prediction accuracy.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, the prediction of renewable energy power satisfies the following formula:

[0008]

[0009] in, Let I(t) represent the output power of the renewable energy source at time t, μ represent the rated conversion efficiency of the photovoltaic module, A represent the total light-receiving area of ​​the photovoltaic module, I(t) represent the solar irradiance at the current time, and T(t) represent the surface temperature of the module or the ambient temperature at the current time. LSTM(I,T,P) hist () is the output of the LSTM prediction module;

[0010] Load forecasting Satisfy the following formula:

[0011]

[0012] Where g(t) represents the growth trend term, s(t) represents the seasonal cycle term, and h(t) represents the holiday effect;

[0013] The prediction confidence level satisfies the following formula:

[0014]

[0015] Where RSME represents the root mean square error of the current prediction, and Max... Historical This represents the maximum RMSE value recorded for similar prediction tasks in the current task scenario.

[0016] In conjunction with the first aspect mentioned above, in one possible implementation, the power supply scheduling optimization agent includes: establishing optimization tasks based on the prediction results of the power load prediction agent and the data and tasks issued by the built-in large model; optimizing the optimization tasks based on a multi-objective optimization algorithm, and outputting the optimized task execution plan.

[0017] In conjunction with the first aspect mentioned above, in one possible implementation, the optimization objectives of the multi-objective optimization algorithm include: maximizing market revenue, ensuring operational stability and functionality meet power balance constraints, ensuring the energy storage system's state of charge (SOC) meets dynamic SOC constraints, and minimizing trading deviations.

[0018] The objective function for maximizing market returns satisfies:

[0019]

[0020] Where T represents the total optimization period. This indicates the current market electricity price. This represents the power that the virtual power plant buys or sells from the grid at time t. Represents the real-time market electricity price, ΔP t This indicates the power deviation between the real-time market and the day-ahead market. R represents the revenue coefficient of the ancillary services market. t This represents the capacity provided by the virtual power plant to the ancillary services market at time t. Represents the scheduling cost at time t;

[0021] The power balance constraints are satisfied as follows:

[0022]

[0023] in, This represents the photovoltaic power generation at time t. This represents the wind power generation capacity at time t. L represents the discharge power of the energy storage system. t The load at time t is represented. Indicates the charging power of the energy storage system. This indicates the power purchased from or sold to the power grid;

[0024] SOC dynamic constraints satisfy:

[0025]

[0026] Among them, SOC t η represents the state of charge of the energy storage system at time t. ch Indicates the energy storage charging efficiency, η dis E represents the energy storage discharge efficiency, Δt represents the time interval, and E cap Indicates the rated capacity of the energy storage system;

[0027] The transaction deviation constraint is satisfied:

[0028]

[0029] in, This represents the trading power reported to the day-ahead market at time t. The actual scheduling power, ΔP, represents the power actually executed. max This indicates the maximum tolerance range for deviations allowed by the market.

[0030] In conjunction with the first aspect mentioned above, in one possible implementation, the market trading agent includes: constructing a multi-dimensional state feature vector based on the scheduling scheme and real-time acquired electricity market trading data, and using time series modeling methods to predict future market electricity price trends; based on the prediction results, learning and generating an optimal bidding strategy through a multi-agent game algorithm based on reinforcement learning, the strategy including bidding timing, bidding volume, and bidding price; and outputting market trading decisions for the day-ahead market, real-time market, or ancillary service market, the market trading decisions including: bid volume, bid price, and adjustable capacity configuration.

[0031] In conjunction with the first aspect mentioned above, in one possible implementation, the day-ahead market bid price Satisfy the following formula:

[0032]

[0033] in, The current expected forecast of the day-ahead market price, σ t This represents the uncertainty of current market price forecasts, and k represents the risk adjustment factor, which is controlled through reinforcement learning.

[0034] Where k satisfies the following formula:

[0035]

[0036] k t+1 This represents the updated risk parameter, used in the next round of bidding, k. t The risk adjustment factor at the current moment is represented by α, and the learning rate is represented by r. t This represents the profit value in the current transaction round. This represents the long-term average or reference value of returns.

[0037] In conjunction with the first aspect mentioned above, in one possible implementation, the equipment operation and maintenance intelligent agent includes: constructing feature vectors based on multi-dimensional status data of operating equipment such as photovoltaic modules, energy storage devices, and load terminals; inputting these vectors into the model to identify anomalies in the operating status and outputting anomaly scores; matching the anomaly scores and key feature indicators with a preset fault rule base to determine the anomaly type and classify the fault; estimating the state of charge, health status, and remaining service life of the energy storage devices to form a status score; and generating maintenance suggestions based on the anomaly scores and feeding them back to the large model.

[0038] Secondly, an electronic device is provided, comprising: a communication unit and a processing unit; the communication unit is used to collect various types of data from the urban energy system as first data, including distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environmental data. The processing unit is used to arrange and schedule the first data based on the data types required for multiple intelligent agents to perform tasks, determine the first data to be processed by each intelligent agent, and call multiple intelligent agents to process the required first data; the processing unit is also used to generate control decisions and market trading strategies for the urban virtual power plant based on the processing results output by multiple intelligent agents and operating constraints.

[0039] Thirdly, this application provides an electronic device, including: a processor and a storage medium; the storage medium includes instructions, and the processor is configured to execute the instructions to implement the methods described in the first aspect and any possible implementation thereof. This electronic device may be an electronic device or a chip within an electronic device.

[0040] Fourthly, this application provides an urban virtual power plant operation system based on an intelligent agent architecture, comprising: sensors and electronic devices; wherein, the sensors are used to collect various types of data in the urban energy system as first data, the first data including distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environmental data; and the electronic devices are used to arrange and schedule the first data based on the data types required for multiple intelligent agents to perform tasks, determine the first data to be processed by each intelligent agent, and call multiple intelligent agents to process the first data to be processed; the electronic devices are also used to generate control decisions and market trading strategies for the urban virtual power plant based on the processing results output by multiple intelligent agents and operating constraints.

[0041] Fifthly, this application provides a computer-readable storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0042] In a sixth aspect, this application provides a computer program product containing instructions that, when run on an electronic device, cause the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0043] This application introduces a city virtual power plant operation method and device based on an intelligent agent architecture, constructing a multi-level, multi-role intelligent agent collaborative mechanism to achieve unified management and optimized scheduling of heterogeneous resources such as distributed energy, energy storage, and adjustable loads. Various intelligent agents possess autonomy, learning, and collaborative capabilities, enabling real-time perception and intelligent response to dynamic information such as grid status, load demand, and electricity price fluctuations. This improves operational efficiency and market returns while maintaining system stability. It addresses the problems of slow response and lack of decision-making flexibility in existing technologies under multi-participant, multi-constraint, and dynamic electricity market environments.

[0044] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0045] Figure 1 A system architecture diagram of a city virtual power plant operation system based on an intelligent agent architecture is provided for embodiments of this application;

[0046] Figure 2 A flowchart illustrating a method for operating a virtual power plant in a city based on an intelligent agent architecture, provided as an embodiment of this application;

[0047] Figure 3 A flowchart illustrating another method for operating a virtual power plant in a city based on an intelligent agent architecture, provided in an embodiment of this application;

[0048] Figure 4 A flowchart illustrating another method for operating a virtual power plant in a city based on an intelligent agent architecture, provided in an embodiment of this application;

[0049] Figure 5A flowchart illustrating another method for operating a virtual power plant in a city based on an intelligent agent architecture, provided in an embodiment of this application;

[0050] Figure 6 A flowchart illustrating another method for operating a virtual power plant in a city based on an intelligent agent architecture, provided in an embodiment of this application;

[0051] Figure 7 A flowchart illustrating another method for operating a virtual power plant in a city based on an intelligent agent architecture, provided in an embodiment of this application;

[0052] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application;

[0053] Figure 9 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0054] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0055] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0056] The urban virtual power plant operation method based on an intelligent agent architecture provided in this application embodiment can be applied to, for example... Figure 1 In the city virtual power plant operation system 100 based on the intelligent agent architecture shown, such as Figure 1 As shown, the communication system includes a sensor 101 and an electronic device 102.

[0057] Among them, sensor 101 is used to collect various types of data in the urban energy system as first data, including distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environment data; and electronic device 102 is used to arrange and schedule the first data based on the data types required for multiple intelligent agents to perform tasks, determine the first data that each intelligent agent needs to process, and call multiple intelligent agents to process the first data that needs to be processed; electronic device 102 is also used to generate control decisions and market trading strategies for the urban virtual power plant based on the processing results output by multiple intelligent agents and operating constraints.

[0058] Figure 2 A method for operating a virtual power plant in a city based on an intelligent agent architecture, as provided in the embodiments of the application, is as follows: Figure 2 As shown, the method includes:

[0059] S201. Sensors collect multiple primary data points from the urban energy system within the urban virtual power plant.

[0060] The first set of data includes distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological and environmental data.

[0061] In one possible implementation, the data acquisition layer of the urban virtual power plant platform collects, in real time, primary data including but not limited to, the following types through various sensor devices deployed at various nodes of the urban energy system: distributed generation data, such as the output power and operating status of photovoltaic and wind power; energy storage device status data, such as battery state of charge (SOC), temperature, and voltage; adjustable load data, such as the power consumption and start / stop status of controllable industrial, commercial, and residential equipment; overall urban electricity load data and electricity market data, such as day-ahead market clearing prices and spot price curves; and meteorological and environmental data, such as solar radiation intensity, wind speed, temperature, and humidity.

[0062] As an example, in this embodiment, the first data is collected by IoT sensors installed at power generation equipment, energy storage units, user-side load terminals, and meteorological stations, and uploaded to the virtual power plant platform via an edge gateway. The collected data includes both real-time operating parameters and historical operating records, ensuring that the system has a comprehensive and continuous perception of the energy status.

[0063] It should be noted that, in this embodiment of the application, the data is accessed to the system through a standardized interface module, which has plug-and-play capability, facilitating the rapid access and expansion of subsequent resources.

[0064] S202. The electronic device, based on the data types required for multiple intelligent agents to perform tasks, organizes and schedules the first data, determines the first data that each intelligent agent needs to process, and calls multiple intelligent agents to process the first data that needs to be processed.

[0065] In one possible implementation, the large model within the city's virtual power plant platform manages business processes based on the functional divisions and task requirements of each agent, orchestrating tasks and distributing data from the initial data provided by sensors. The large model first identifies the multiple agents involved in the current business process and their respective required data types. Then, it classifies, organizes, and labels the collected initial data, constructing a data task flow. After completing the data orchestration and mapping, it automatically invokes the corresponding agent modules, sending tasks and data to the target agent via an interface, where the agent executes the corresponding data processing, analysis, or inference tasks.

[0066] As an example, in this embodiment, multiple intelligent agents include: a power load forecasting agent, used to forecast future load and renewable energy output; a power supply scheduling optimization agent, used to output optimized task execution plans; a market trading agent, used to output market trading decisions for day-ahead markets, real-time markets, or ancillary service markets; and an equipment operation and maintenance agent, used to detect system operating status. The invocation process supports concurrent processing and data caching mechanisms to improve task scheduling efficiency and system response speed.

[0067] S203. Based on the processing results output by multiple intelligent agents and the operating constraints, the electronic equipment generates control decisions and market trading strategies for the city's virtual power plant.

[0068] In one possible implementation, the large model uniformly summarizes and comprehensively analyzes the processing results from multiple intelligent agent modules. It then combines this with the current grid operating status, the electricity market environment, and system operating constraints, such as power balance, power limits, energy storage capacity, and safety redundancy, to automatically generate control decision schemes and market trading strategies for the city's virtual power plant. Through an embedded knowledge rule engine and optimization algorithm module, based on the integrated intelligent agent results and combined with externally input operating objectives and system constraints, it models and solves multi-objective control schemes to form specific execution plans.

[0069] To address the issues of slow response and lack of decision-making flexibility in existing technologies under multi-participant, multi-constraint, and dynamic electricity market environments, this application employs a real-time data acquisition mechanism. This mechanism can collect primary data, including distributed generation data, energy storage status, and adjustable load, in real time and accurately, enabling comprehensive perception of the urban energy system's operational status. Utilizing distributed sensors and edge computing gateways, it supports high-frequency, low-latency data reporting and possesses plug-and-play scalability. Through a large-model-driven data orchestration and task scheduling mechanism, it can accurately classify, distribute, and map massive amounts of data according to the professional task requirements of various intelligent agents, ensuring that each agent processes target data highly relevant to its responsibilities. An agent collaborative output and large-model fusion decision-making mechanism generates multi-objective control decision schemes and market trading strategies. This solves the problems of slow response and lack of decision-making flexibility in existing technologies under multi-participant, multi-constraint, and dynamic electricity market environments.

[0070] In one possible implementation, combining Figure 2 ,like Figure 3 As shown, the above-mentioned S202, based on the data types required for multiple agents to perform tasks, arranges and schedules the first data. This can be specifically implemented through the following S301, S302, and S303, which are explained in detail below:

[0071] S301. The electronic device classifies and stores the first data, calls the large model inference engine to analyze the task requirements, and determines the data that each intelligent agent needs to process based on the task requirements, and sends the corresponding tasks and data to each intelligent agent.

[0072] In one possible implementation, the electronic device categorizes, organizes, and stores the collected initial data. The data is structured and stored according to dimensions such as data source, data type, and timestamp, constructing an efficient and searchable data warehouse. After categorization, the initial data is labeled with different data tags, such as "photovoltaic output," "load curve," and "energy storage SOC," facilitating subsequent task processing and retrieval. Next, the electronic device invokes its built-in large-scale model inference engine to perform semantic parsing and target modeling for the current business process. Based on the current operating scenario, external input requests, or scheduling objectives, the large-scale model automatically identifies the intelligent agent roles and their processing divisions required for the current business process.

[0073] As an example, in the real-time example of this application, the large model distributes the task requests and target data corresponding to each agent to each agent node through the internal task scheduler or agent interface module, triggering them to execute subsequent processing operations. The data and task distribution process supports asynchronous calls, concurrent execution, and caching mechanisms to ensure high system throughput and low latency.

[0074] S302, The electronic device activates the intelligent agent to execute the corresponding task.

[0075] In one possible implementation, after the large model in the electronic device completes task distribution, it actively activates the corresponding intelligent agent module, enabling it to start executing the specified task according to the received task instructions and data content.

[0076] As an example, in the embodiments of this application, the activation process can be controlled by a task scheduler or a large model scheduling engine, and can be implemented through event triggering mechanism, time scheduling mechanism or external instruction call, etc., and this application does not limit it.

[0077] S303. The electronic device calls the large model inference engine to receive the results returned by each agent in performing the task, and optimizes the process and strategy of each agent in performing the task based on the returned results.

[0078] In one possible implementation, after each agent completes its task, the electronic device invokes a large-scale model inference engine to uniformly receive, parse, and evaluate the processing results returned by the agents. As the core analysis and decision-making hub of the system, the large-scale model inference engine is capable of understanding, correlating, and comprehensively modeling multi-source heterogeneous results. After receiving the results, the large model automatically triggers process evaluation and strategy optimization mechanisms to evaluate and adjust the task execution processes, collaborative logic, and decision-making strategies of each agent.

[0079] As an example, in an embodiment of this application, the large model can identify whether the error of the power load prediction agent is too large. If the prediction error is too large, it can automatically adjust the power load prediction agent model structure or reconfigure the model training parameters.

[0080] This application embodiment utilizes electronic devices to perform structured classification, warehousing, and tagging management of the collected first data, improving data retrieval and retrieval efficiency, and providing a fundamental guarantee for the efficient invocation of intelligent agents. By calling a large-model inference engine for task semantic parsing and target recognition, accurate matching and distribution of tasks and data are achieved, significantly improving the intelligence and automation level of task scheduling in multi-agent parallel operation scenarios. By calling the large-model inference engine to uniformly receive and correlate the results returned by each intelligent agent, and implementing process reconstruction and strategy optimization based on the analysis results, an intelligent agent collaborative execution system with closed-loop feedback capability is constructed. This effectively enhances the system's self-learning, self-adaptation, and self-evolution capabilities, thereby improving the prediction accuracy of each intelligent agent and the robustness of the system.

[0081] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 4As shown, the specific execution process of the above-mentioned S203 power load forecasting agent includes: S401 to S404, which are explained in detail below:

[0082] S401, The electronic device receives meteorological environmental data and adjustable load data from the first data.

[0083] In one possible implementation, electronic devices in the city virtual power plant platform receive a subset of data related to meteorological environment and adjustable load from the initial data through various IoT sensors and edge gateways deployed in the city's energy system, serving as an important input source for subsequent intelligent analysis and predictive modeling.

[0084] As an example, the meteorological environmental data in this embodiment includes information such as solar radiation intensity, wind speed, and temperature, provided by meteorological stations connected to the platform or external meteorological service systems. Adjustable load data includes real-time load power, start / stop status, response capability level, and equipment operation strategy parameters of controllable electrical equipment on the industrial, commercial, or residential side, collected by terminals such as smart meters, load controllers, and energy management systems.

[0085] It should be noted that the data reception process supports multi-protocol parsing and data preprocessing mechanisms. Through standardized data acquisition interfaces, real-time data access, cleaning, and preliminary labeling are achieved from meteorological equipment and load control systems. This data is then uploaded to a data center or cache pool, providing input support for subsequent predictive and scheduling optimization agents.

[0086] S402. Electronic equipment uses industry-wide models to predict load and renewable energy output in the first time period of the future.

[0087] In one possible implementation, based on received meteorological and environmental data and adjustable load data, an industry-wide model connected to the city's virtual power plant platform is invoked to predict the power load and renewable energy output for the first time period in the future.

[0088] As an example, in some embodiments, the first time period can be the next 24 hours, 96 hours, or any arbitrarily defined scheduling period window. In this embodiment, the first time period is the next 24 hours, and this application does not limit it.

[0089] It should be noted that the power load forecasting agent can perform forecasting tasks by accessing and calling existing industry-wide models, or by having a built-in forecasting model.

[0090] S403. Electronic equipment determines the prediction confidence and prediction accuracy based on load and new energy output.

[0091] In one possible implementation, the power load forecasting agent calls upon a large industry model to forecast the load and renewable energy output for the first time period in the future. Then, it further evaluates the confidence level and verifies the accuracy of the forecast results to determine their reliability and credibility for subsequent decision-making.

[0092] As an example, the prediction of renewable energy power satisfies the following formula:

[0093]

[0094] in, Let I(t) represent the output power of the renewable energy source at time t, μ represent the rated conversion efficiency of the photovoltaic module, A represent the total light-receiving area of ​​the photovoltaic module, I(t) represent the solar irradiance at the current time, and T(t) represent the surface temperature of the module or the ambient temperature at the current time. LSTM(I,T,P) hist () is the output of the LSTM prediction module;

[0095] As an example, load forecasting Satisfy the following formula:

[0096]

[0097] Where g(t) represents the growth trend term, s(t) represents the seasonal cycle term, and h(t) represents the holiday effect;

[0098] As an example, the prediction confidence level satisfies the following formula:

[0099]

[0100] Where RSME represents the root mean square error of the current prediction, and Max... Historical This represents the maximum RMSE value recorded for similar prediction tasks in the current task scenario.

[0101] S404. Electronic devices optimize the prediction accuracy of power load prediction agents by improving prediction confidence and prediction accuracy.

[0102] In one possible implementation, the power load forecasting agent is model optimized and its parameters adjusted based on the prediction confidence and prediction accuracy results obtained in S403, so as to improve its prediction accuracy and stability in future tasks.

[0103] As an example, in this embodiment, by statistically analyzing the error distribution between predicted and actual values ​​under different time periods and operating conditions, high-error periods, error types, and possible external factors associated with the errors are identified to uncover the causes of prediction deviations. For the identified errors, the power load forecasting agent automatically invokes the model training component to structurally adjust the industry-wide model used by the agent. For example, training parameters such as the learning rate, regularization parameters, and moving average window are adjusted to improve the model's fitting and generalization abilities.

[0104] As an example, when multiple built-in candidate models exist, the power load forecasting agent can automatically select the optimal model by comparing the prediction performance of different models on the same dataset, or build a multi-model fusion mechanism to improve the overall prediction robustness.

[0105] As an example, once the model is optimized, it will be applied to the next prediction cycle, continuing to receive actual results and prediction error data, forming a closed-loop online learning and dynamic iteration mechanism, enabling the power load prediction agent to have continuous self-improvement capabilities.

[0106] This application constructs a closed-loop intelligent agent for power load forecasting, enabling precise access to meteorological and environmental data and adjustable load data, accurate forecasting of future load and renewable energy output, and confidence assessment and dynamic optimization of forecast results. This enhances the system's adaptability and forecasting stability in complex operating environments, and improves the operational intelligence level and dispatching decision reliability of the virtual power plant.

[0107] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 5 As shown, the specific execution process of the power supply scheduling optimization agent in S203 includes: S501 and S502, which are explained in detail below:

[0108] S501: The electronic device establishes optimization tasks based on the prediction results of the power load prediction agent and the data and tasks issued by the built-in large model.

[0109] In one possible implementation, after the electronic device completes the prediction of future power load and renewable energy output, it combines the prediction results provided by the power load prediction agent with the task instructions and data issued by the built-in large model inference engine to construct specific optimization tasks that can be used for scheduling optimization or transaction optimization.

[0110] As an example, in this embodiment, the power supply dispatch optimization agent receives forecast data from the power load forecasting agent, including time-of-use load curves and predicted output values ​​of new energy power generation within a specified future time period. Simultaneously, it receives task description information such as dispatch objectives, constraint parameters, and market instructions generated by the large model inference engine. The agent parses the task type issued by the large model, such as whether it is day-ahead planning, real-time dispatch response, or market pricing strategy generation, and identifies relevant operational constraints, including power balance and energy storage capacity limitations, to construct the optimization task.

[0111] S502: The electronic device optimizes the task based on a multi-objective optimization algorithm and outputs the optimized task execution plan.

[0112] In one possible implementation, an optimization task is constructed based on the analysis of tasks issued by the large model. The constructed optimization task is then solved and optimized using a multi-objective optimization algorithm to generate an optimized task execution plan to guide the scheduling of urban virtual power plants or market participation.

[0113] As an example, in this embodiment of the application, the optimization objectives of the multi-objective optimization algorithm include: maximizing market revenue, ensuring operational stability and functionality meet power balance constraints, ensuring the energy storage system's state of charge (SOC) meets dynamic SOC constraints, and minimizing transaction deviations.

[0114] As an example, the objective function for maximizing market returns satisfies:

[0115]

[0116] Where T represents the total optimization period. This indicates the current market electricity price. This represents the power that the virtual power plant buys or sells from the grid at time t. Represents the real-time market electricity price, ΔP t This indicates the power deviation between the real-time market and the day-ahead market. R represents the revenue coefficient of the ancillary services market. t This represents the capacity provided by the virtual power plant to the ancillary services market at time t. Represents the scheduling cost at time t;

[0117] As an example, the power balance constraint satisfies:

[0118]

[0119] in, This represents the photovoltaic power generation at time t. This represents the wind power generation capacity at time t. L represents the discharge power of the energy storage system.t The load at time t is represented. Indicates the charging power of the energy storage system. This indicates the power purchased from or sold to the power grid;

[0120] As an example, the SOC dynamic constraints satisfy:

[0121]

[0122] Among them, SOC t η represents the state of charge of the energy storage system at time t. ch Indicates the energy storage charging efficiency, η dis E represents the energy storage discharge efficiency, Δt represents the time interval, and E cap Indicates the rated capacity of the energy storage system;

[0123] As an example, the transaction deviation constraint satisfies:

[0124]

[0125] in, This represents the trading power reported to the day-ahead market at time t. The actual scheduling power, ΔP, represents the power actually executed. max This indicates the maximum tolerance range for deviations allowed by the market.

[0126] This application automatically constructs structured optimization tasks based on prediction results and scheduling objectives, and uses a multi-objective optimization algorithm to generate specific task execution plans. It achieves intelligent transformation from prediction data to optimization decisions, improves the flexibility of task modeling and the automation level of system scheduling, dynamically balances multiple objectives, and enhances the scheduling efficiency, responsiveness, and operational benefits of the virtual power plant in complex operating environments.

[0127] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 6 As shown, the specific execution process of the S203 market trading intelligent agent includes: S601 to S603, which are explained in detail below:

[0128] S601, based on the scheduling scheme and real-time acquired electricity market transaction data, the electronic equipment constructs a multi-dimensional state feature vector and uses time series modeling methods to predict future market electricity price trends.

[0129] In one possible implementation, the market trading agent combines the current scheduling scheme with real-time acquired electricity market trading data to construct a multi-dimensional state feature vector for market forecasting, and uses time series modeling methods to predict the electricity price trend for one or more future periods, providing data support for the formulation of trading strategies for urban virtual power plants.

[0130] As an example, in this embodiment, the market trading agent receives real-time electricity market trading-related data, including day-ahead market prices, real-time market prices, load-side declarations, generation-side clearing volumes, ancillary service transaction records, and regional supply-demand balance indicators, and simultaneously reads the current dispatch plan. Based on the input data, a feature vector set is extracted, and the extracted feature vectors are used as a dataset. A built-in time series modeling method is invoked to model the constructed feature vector set, capturing short-term fluctuations and long-term trends in electricity price movements, and outputting market electricity price predictions for several future time periods. This embodiment uses LSTM for time series modeling, but this embodiment is not limited to this approach.

[0131] S602. Based on the prediction results, the electronic device learns and generates the optimal bidding strategy through a multi-agent game algorithm based on reinforcement learning. The strategy includes the timing of bidding, the amount of electricity to be bid, and the price to be bid.

[0132] In one possible implementation, based on the predicted future market electricity price trend, a multi-agent game algorithm based on reinforcement learning is invoked to simulate the trading behavior of a virtual power plant in the city under different market conditions. Through continuous learning and strategy evolution, the optimal pricing strategy applicable to various market scenarios is generated.

[0133] As an example, in this embodiment, a simulated trading environment is constructed using predicted market electricity price curves, scheduling plans, and resource availability as environmental states; a reward function and constraints are defined. For the multiple players in the market, such as other aggregators, power plants, and user clusters, the system introduces multiple learning agents to simulate the bidding behavior of each player under conditions of incomplete information, achieving local strategy adaptation and global strategy game theory. Through the Deep Q-Network (DQN) algorithm, each agent continuously learns in multiple rounds of market simulation and optimizes its bidding behavior to form a stable and optimal bidding strategy.

[0134] S603, Electronic equipment outputs are used for market transaction decisions in day-ahead markets, real-time markets, or ancillary service markets.

[0135] Market transaction decisions include: bidding for electricity volume, bidding for price, and adjustable capacity configuration.

[0136] In one possible implementation, based on the optimal pricing strategy, formal trading decision instructions for different electricity markets are generated and output to guide virtual power plants to participate in market transactions, thereby maximizing revenue and efficiently utilizing resources.

[0137] As an example, based on the market operating cycle and time window, the target transaction type is identified, and the optimal pricing strategy is adapted to the corresponding market rules and transaction format. Combining the transaction type and strategy output, complete market transaction decision content is generated, including: declared electricity volume, declared price, and adjustable capacity configuration. The generated market transaction instructions are submitted to a manual review process. After approval, they are submitted via a standard interface to the power trading platform, market dispatch system, or virtual power plant aggregator platform as formal market participation input.

[0138] As an example, the market's quoted prices Satisfy the following formula:

[0139]

[0140] in, The current expected forecast of the day-ahead market price, σ t This represents the uncertainty of current market price forecasts, and k represents the risk adjustment factor, which is controlled through reinforcement learning.

[0141] Where k satisfies the following formula:

[0142]

[0143] k t+1 This represents the updated risk parameter, used in the next round of bidding, k. t The risk adjustment factor at the current moment is represented by α, and the learning rate is represented by r. t This represents the profit value in the current transaction round. This represents the long-term average or reference value of returns.

[0144] This application constructs an intelligent closed-loop mechanism from market trend prediction to transaction decision generation. By integrating scheduling schemes and real-time market data, it builds a multi-dimensional state feature vector and introduces time series modeling methods to predict future electricity price trends. Utilizing reinforcement learning and multi-agent game theory algorithms, it simulates the dynamic competitive relationships between different market participants, continuously learns and optimizes bidding behavior, and automatically generates optimal bidding strategies, including bidding timing, bidding volume, and bidding price. It also automatically generates transaction declaration schemes that conform to different market rules, enabling virtual power plants to participate in the day-ahead market, real-time market, and ancillary service market, thereby enhancing the responsiveness and competitive advantage of urban virtual power plants in multi-market environments.

[0145] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 7 As shown, the specific execution process of the above-mentioned S203 equipment operation and maintenance intelligent agent includes: S701 to S704, which are explained in detail below:

[0146] S701: Electronic equipment constructs feature vectors based on multi-dimensional status data of operating equipment such as photovoltaic modules, energy storage devices, and load terminals. The input vectors are used to identify anomalies in the operating status and output anomaly scores.

[0147] In one possible implementation, by accessing real-time operating status data uploaded by various operating devices in a virtual power plant, a multi-dimensional status feature vector is constructed, including but not limited to device voltage, current, temperature, power output, communication status, environmental parameters, historical operating records, etc. The operating status is identified through the model, and anomaly scores are output.

[0148] As an example, in this embodiment, the constructed feature vector is input into a pre-trained anomaly detection model. This model can employ supervised or unsupervised learning methods. This embodiment uses the Isolation Forest algorithm to model the behavior and calculate the deviation of the current operating state, identifying whether there are equipment anomalies, performance degradation, or operational failure trends. The anomaly score output by the model reflects the degree of anomaly in the equipment state; a higher score indicates that the device is closer to the edge of a potential failure. This embodiment does not limit the scope of the application.

[0149] S702. Electronic devices determine the anomaly type and classify the fault based on anomaly scores and key feature indicators, matching them with a preset fault rule base.

[0150] In one possible implementation, anomaly scoring is combined with key feature indicators involved in the scoring process, such as sudden temperature rise, voltage fluctuation, and decreased charging and discharging efficiency, and matched with a built-in fault rule base to determine the corresponding anomaly type and classify it according to the existing fault types.

[0151] As an example, in this embodiment, the built-in fault rule base includes a large number of equipment fault modes, feature triggering conditions, and corresponding fault type tags generated by manual summarization or historical learning. Through rule matching and similarity analysis, the fault type is determined for the current abnormal state, and specific fault tags such as "photovoltaic module power degradation", "energy storage unit internal resistance abnormality", and "load response delay" are output.

[0152] S703. Electronic equipment estimates the state of charge, health status and remaining useful life of energy storage devices to form a status score.

[0153] In one possible implementation, for energy storage units in a virtual power plant, such as lithium battery packs, sodium batteries, and supercapacitors, electronic devices call upon professional evaluation models to estimate the current state of charge (SOC), state of health (SOH), and remaining useful life (RUL) based on key operating data such as voltage, current, temperature rise, cycle count, and historical charge-discharge curves.

[0154] As an example, in this embodiment of the application, a neural network regression is used to comprehensively evaluate the battery status and generate a status score report for lifecycle management, reflecting the operational stability and sustainability of the energy storage device.

[0155] S704: Electronic devices generate maintenance suggestions based on anomaly scoring and fault diagnosis results, and feed them back to the large model.

[0156] In one possible implementation, based on the aforementioned anomaly scoring and fault classification results, combined with the equipment's operational risk level and scheduling importance, corresponding maintenance suggestions are selected from the built-in knowledge base or empirical rules. Examples include: "Recommend replacing the third string of photovoltaic modules in the short term," "Recommend voltage balancing for the second module of the energy storage battery," and "Recommend adjusting the load response strategy to reduce response lag." These generated maintenance suggestions are pushed to the maintenance system or manual inspection personnel as proactive maintenance task instructions; they are also fed back to the large-scale model inference engine for continuous optimization of anomaly judgment and maintenance decision-making strategies, achieving a knowledge reinforcement learning closed loop based on operational data.

[0157] The equipment operation and maintenance intelligent agent constructed in this application realizes real-time monitoring and quantitative evaluation of the operating status of core equipment such as photovoltaic modules, energy storage devices, and load terminals by constructing multi-dimensional state feature vectors and introducing an anomaly recognition model. Based on anomaly scores and key feature indicators, it matches the fault rule base to complete automatic fault classification, improving the accuracy and location efficiency of anomaly handling. Based on the diagnostic results, it intelligently generates maintenance suggestions and feeds them back to the large model for knowledge updates and strategy optimization, constructing a closed-loop fault management process of monitoring-diagnosis-suggestion-learning, thereby improving the maintainability, operational safety, and operation and maintenance intelligence of the system.

[0158] The foregoing mainly describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as an electronic device, includes at least one of the hardware structures and software modules corresponding to the execution of each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0159] This application embodiment can divide the electronic device into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0160] When using integrated units, Figure 8 A possible structural schematic diagram of the electronic device (referred to as electronic device 80) involved in the above embodiments is shown. The electronic device 80 includes a processing unit 801 and a communication unit 802, and may also include a storage unit 803. Figure 8 The structural diagram shown can be used to illustrate the structure of the electronic device involved in the above embodiments.

[0161] when Figure 8 The schematic diagram shown is used to illustrate the structure of the electronic device involved in the above embodiments. The processing unit 801 is used to control and manage the operation of the electronic device, the communication unit 802 is used for the electronic device to communicate with other devices, and the storage unit 803 is used to store the program code and data of the electronic device.

[0162] For example, the communication unit 802 is used to collect various types of data in the urban energy system as first data, including distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environmental data; the processing unit 801 is used to arrange and schedule the first data based on the data types required for multiple intelligent agents to perform tasks, determine the first data that each intelligent agent needs to process, and call multiple intelligent agents to process the first data that needs to be processed; the processing unit 801 is also used to generate control decisions and market trading strategies for the urban virtual power plant based on the processing results output by multiple intelligent agents and operating constraints.

[0163] In one possible implementation, multiple intelligent agents include: a power load forecasting agent, used to forecast future time load and renewable energy output; a power supply dispatch optimization agent, used to output optimized task execution plans; a market trading agent, used to output market trading decisions for day-ahead, real-time, or ancillary service markets; and an equipment operation and maintenance agent, used to detect system operating status. Based on the data types required for multiple agents to execute tasks, the first data is orchestrated and task scheduled, including: classifying and storing the first data in a database; calling a large model inference engine to parse task requirements and determining the data to be processed by each agent based on task requirements; distributing the corresponding tasks and data to each agent; activating agents to execute corresponding tasks; calling the large model inference engine to receive the results returned by each agent's task execution; and optimizing the process and strategy of each agent's task execution based on the returned results.

[0164] In one possible implementation, the processing unit, specifically used by the power load forecasting agent, performs the following process: receiving meteorological environmental data and adjustable load data from the first data; calling an industry-wide model to predict the load and renewable energy output within the first time period in the future; determining the prediction confidence and prediction accuracy based on the load and renewable energy output; and optimizing the prediction accuracy of the power load forecasting agent based on the prediction confidence and prediction accuracy.

[0165] In one possible implementation, the prediction of renewable energy power satisfies the following formula:

[0166]

[0167] in, Let I(t) represent the output power of the renewable energy source at time t, μ represent the rated conversion efficiency of the photovoltaic module, A represent the total light-receiving area of ​​the photovoltaic module, I(t) represent the solar irradiance at the current time, and T(t) represent the surface temperature of the module or the ambient temperature at the current time. LSTM(I,T,P) hist () is the output of the LSTM prediction module;

[0168] Load forecasting Satisfy the following formula:

[0169]

[0170] Where g(t) represents the growth trend term, s(t) represents the seasonal cycle term, and h(t) represents the holiday effect;

[0171] The prediction confidence level satisfies the following formula:

[0172]

[0173] Where RSME represents the root mean square error of the current prediction, and Max... Historical This represents the maximum RMSE value recorded for similar prediction tasks in the current task scenario.

[0174] In one possible implementation, the processing unit is further specifically used for the power supply scheduling optimization agent, including: establishing optimization tasks based on the prediction results of the power load prediction agent and the data and tasks issued by the built-in large model; optimizing the optimization tasks based on a multi-objective optimization algorithm, and outputting the optimized task execution plan.

[0175] In one possible implementation, the optimization objectives of the multi-objective optimization algorithm include: maximizing market revenue, ensuring operational stability and functionality meet power balance constraints, ensuring the energy storage system's state of charge (SOC) meets dynamic SOC constraints, and minimizing trading deviations.

[0176] In one possible implementation, the objective function for maximizing market returns satisfies:

[0177]

[0178] Where T represents the total optimization period. This indicates the current market electricity price. This represents the power that the virtual power plant buys or sells from the grid at time t. Represents the real-time market electricity price, ΔP t This indicates the power deviation between the real-time market and the day-ahead market. R represents the revenue coefficient of the ancillary services market. t This represents the capacity provided by the virtual power plant to the ancillary services market at time t. Represents the scheduling cost at time t;

[0179] In one possible implementation, the power balance constraint satisfies:

[0180]

[0181] in, This represents the photovoltaic power generation at time t. This represents the wind power generation capacity at time t. L represents the discharge power of the energy storage system. t The load at time t is represented. Indicates the charging power of the energy storage system. This indicates the power purchased from or sold to the power grid;

[0182] In one possible implementation, the SOC dynamic constraints satisfy:

[0183]

[0184] Among them, SOC t η represents the state of charge of the energy storage system at time t. ch Indicates the energy storage charging efficiency, η dis E represents the energy storage discharge efficiency, Δt represents the time interval, and E cap Indicates the rated capacity of the energy storage system;

[0185] In one possible implementation, the transaction deviation constraint satisfies:

[0186]

[0187] in, This represents the trading power reported to the day-ahead market at time t. The actual scheduling power, ΔP, represents the power actually executed. max This indicates the maximum tolerance range for deviations allowed by the market.

[0188] In one possible implementation, the processing unit is further specifically used as a market trading agent, including: constructing a multi-dimensional state feature vector based on the scheduling scheme and real-time acquired electricity market trading data, and predicting future market electricity price trends using time series modeling methods; based on the prediction results, learning and generating an optimal bidding strategy through a multi-agent game algorithm based on reinforcement learning, the strategy including bidding timing, bidding volume, and bidding price; and outputting market trading decisions for the day-ahead market, real-time market, or ancillary service market, the market trading decisions including: bid volume, bid price, and adjustable capacity configuration.

[0189] In one possible implementation, the day-ahead market bid price Satisfy the following formula:

[0190]

[0191] in, The current expected forecast of the day-ahead market price, σ t This represents the uncertainty of current market price forecasts, and k represents the risk adjustment factor, which is controlled through reinforcement learning.

[0192] Where k satisfies the following formula:

[0193]

[0194] k t+1 This represents the updated risk parameter, used in the next round of bidding, k. t The risk adjustment factor at the current moment is represented by α, and the learning rate is represented by r. t This represents the profit value in the current transaction round. This represents the long-term average or reference value of returns.

[0195] In one possible implementation, the processing unit is further specifically used as an intelligent agent for equipment operation and maintenance, including: constructing feature vectors based on multi-dimensional state data of operating equipment such as photovoltaic modules, energy storage devices, and load terminals; inputting these vectors into the model to identify anomalies in the operating state and outputting anomaly scores; matching the anomaly scores and key feature indicators with a preset fault rule base to determine the anomaly type and classify the fault; estimating the state of charge, health status, and remaining service life of the energy storage devices to form a state score; and generating maintenance suggestions based on the anomaly scores and feeding them back to the large model.

[0196] The processing unit 801 can be a processor or a controller, and the communication unit 802 can be a communication interface, transceiver, transceiver circuit, transceiver device, etc. The term "communication interface" is a general term and may include one or more interfaces. The storage unit 803 can be a memory. When the electronic device 80 is a chip, the processing unit 801 can be a processor or a controller, and the communication unit 802 can be an input interface and / or an output interface, pins, or circuits, etc. The storage unit 803 can be a storage unit within the chip (e.g., a register, cache, etc.) or a storage unit located outside the chip (e.g., read-only memory (ROM), random access memory (RAM, etc.).

[0197] The communication unit can also be called a transceiver unit. The antenna and control circuit with transceiver functions in the electronic device 80 can be considered as the communication unit 802 of the electronic device 80, and the processor with processing functions can be considered as the processing unit 801 of the electronic device 80. Optionally, the device in the communication unit 802 used to implement the receiving function can be considered as the communication unit. The communication unit is used to execute the receiving steps in the embodiments of this application, and the communication unit can be a receiver, a receiver circuit, etc. The device in the communication unit 802 used to implement the transmitting function can be considered as the transmitting unit. The transmitting unit is used to execute the transmitting steps in the embodiments of this application, and the transmitting unit can be a transmitter, a transmitter, a transmitting circuit, etc.

[0198] Figure 8If the integrated units in the process are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of software products. These computer software products are stored in a storage medium and include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. Storage media for storing computer software products include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0199] Figure 8 The units in the process can also be called modules; for example, a processing unit can be called a processing module.

[0200] This application also provides a hardware structure diagram of an electronic device (denoted as electronic device 90), see [link to diagram]. Figure 9 The electronic device 90 includes a processor 901, and optionally, a memory 902 connected to the processor 901.

[0201] In the first possible implementation, see Figure 9 The electronic device 90 also includes a transceiver 903. The processor 901, memory 902, and transceiver 903 are connected via a bus. The transceiver 903 is used to communicate with other devices or communication networks. Optionally, the transceiver 903 may include a transmitter and a receiver. The device in the transceiver 903 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of this application. The device in the transceiver 903 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of this application.

[0202] Based on the first possible implementation method Figure 9 The structural diagram shown can be used to illustrate the structure of the electronic device involved in the above embodiments.

[0203] in, Figure 9 This can also be illustrated by a system chip in an electronic device. In this case, the actions performed by the aforementioned electronic device can be implemented by this system chip; the specific actions performed can be found above and will not be repeated here.

[0204] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0205] The processor in this application may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., and other computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a standalone semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may form a System-on-a-Chip (SoC) with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), programmable logic devices (PLDs), or logic circuits that implement dedicated logic operations.

[0206] The memory in the embodiments of this application may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0207] This application also provides a computer-readable storage medium including instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0208] This application also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0209] This application also provides a chip including a processor and an interface circuit. The interface circuit is coupled to the processor. The processor is used to run computer programs or instructions to implement the above-described method. The interface circuit is used to communicate with other modules outside the chip.

[0210] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0211] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, the disclosure, and the appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0212] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A method for operating a virtual power plant in a city based on an intelligent agent architecture, characterized in that, include: Collect multiple primary data points from the urban energy system within the urban virtual power plant. These primary data points include distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environmental data. Based on the data types required for multiple agents to perform tasks, the first data is orchestrated and task scheduled to determine the first data that each agent needs to process, and the multiple agents are called to process the first data that needs to be processed. Based on the processing results output by the multiple intelligent agents and the operating constraints, the control decisions and market trading strategies for the city's virtual power plant are generated.

2. The method according to claim 1, characterized in that, The plurality of intelligent agents include: a power load forecasting intelligent agent, which is used to forecast future load and renewable energy output; a power supply scheduling optimization intelligent agent, which is used to output optimized task execution plans; a market trading intelligent agent, which is used to output market trading decisions for day-ahead markets, real-time markets, or ancillary service markets; and an equipment operation and maintenance intelligent agent, which is used to detect the system operating status. The arrangement and task scheduling of the first data based on the data types required for multiple intelligent agents to perform tasks includes: The first data is classified and stored in the database. The large model inference engine is called to parse the task requirements and determine the data that each agent needs to process based on the task requirements. The corresponding tasks and data are then sent to each agent. Activate the intelligent agent to execute the corresponding task; The large model inference engine is invoked to receive the results returned by each agent in performing the task, and the process and strategy of each agent in performing the task are optimized based on the returned results.

3. The method according to claim 2, characterized in that, The power load forecasting agent is specifically used to perform the following processes: Receive meteorological environmental data and adjustable load data from the first data; The industry's large-scale model is used to predict the load and renewable energy output in the first time period of the future. Determine the prediction confidence and prediction accuracy based on the load and the output of new energy sources; The prediction accuracy of the power load prediction agent is optimized by using the prediction confidence and the prediction accuracy.

4. The method according to claim 3, characterized in that, The predicted power of new energy sources satisfies the following formula: in, Let I(t) represent the output power of the renewable energy source at time t, μ represent the rated conversion efficiency of the photovoltaic module, A represent the total light-receiving area of ​​the photovoltaic module, I(t) represent the solar irradiance at the current time, and T(t) represent the surface temperature of the module or the ambient temperature at the current time. LSTM(I,T,P) hist () is the output of the LSTM prediction module; The load forecast Satisfy the following formula: Where g(t) represents the growth trend term, s(t) represents the seasonal cycle term, and h(t) represents the holiday effect; The prediction confidence level satisfies the following formula: Where RSME represents the root mean square error of the current prediction, and Max... Historical This represents the maximum RMSE value recorded for similar prediction tasks in the current task scenario.

5. The method according to claim 2, characterized in that, The power supply scheduling optimization intelligent agent includes: Based on the prediction results of the power load prediction agent, as well as the data and tasks issued by the built-in large model, an optimization task is established; The optimization task is optimized based on a multi-objective optimization algorithm, and the optimized task execution plan is output.

6. The method according to claim 5, characterized in that, The optimization objectives of the multi-objective optimization algorithm include: maximizing market revenue, ensuring operational stability and functionality meet power balance constraints, ensuring the energy storage system's state of charge (SOC) meets dynamic SOC constraints, and minimizing trading deviations. The objective function for maximizing market returns satisfies: Where T represents the total optimization period. This indicates the current market electricity price. This represents the power that the virtual power plant buys or sells from the grid at time t. Represents the real-time market electricity price, ΔP t This indicates the power deviation between the real-time market and the day-ahead market. R represents the revenue coefficient of the ancillary services market. t This represents the capacity provided by the virtual power plant to the ancillary services market at time t. Represents the scheduling cost at time t; The power balance constraints are satisfied as follows: in, This represents the photovoltaic power generation at time t. This represents the wind power generation capacity at time t. L represents the discharge power of the energy storage system. t The load at time t is represented. Indicates the charging power of the energy storage system. This indicates the power purchased from or sold to the power grid; SOC dynamic constraints satisfy: Among them, SOC t η represents the state of charge of the energy storage system at time t. ch Indicates the energy storage charging efficiency, η dis E represents the energy storage discharge efficiency, Δt represents the time interval, and E cap Indicates the rated capacity of the energy storage system; The transaction deviation constraint is satisfied: in, This represents the trading power reported to the day-ahead market at time t. The actual scheduling power, ΔP, represents the power actually executed. max This indicates the maximum tolerance range for deviations allowed by the market.

7. The method according to claim 2, characterized in that, The market trading intelligent agent includes: Based on the aforementioned scheduling scheme and real-time acquired electricity market transaction data, a multi-dimensional state feature vector is constructed, and time series modeling methods are used to predict future market electricity price trends. Based on the prediction results, the optimal bidding strategy is learned and generated through a multi-agent game algorithm based on reinforcement learning. The strategy includes the timing of bidding, the amount of electricity to be bid, and the price to be bid. The output is used for market trading decisions in the day-ahead market, real-time market, or ancillary service market, and the market trading decisions include: bid electricity volume, bid price, and adjustable capacity configuration.

8. The method according to claim 7, characterized in that, The market bid prices of the day before Satisfy the following formula: in, The current expected forecast of the day-ahead market price, σ t This represents the uncertainty of current market price forecasts, and k represents the risk adjustment factor, which is controlled through reinforcement learning. Where k satisfies the following formula: k t+1 This represents the updated risk parameter, used in the next round of bidding, k. t The risk adjustment factor at the current moment is represented by α, and the learning rate is represented by r. t This represents the profit value in the current transaction round. This represents the long-term average or reference value of returns.

9. The method according to claim 2, characterized in that, The equipment operation and maintenance intelligent agent includes: Feature vectors are constructed based on multi-dimensional status data of operating equipment such as photovoltaic modules, energy storage devices, and load terminals. These vectors are then input into the model to identify anomalies in the operating status and output anomaly scores. Based on the anomaly score and key feature indicators, a preset fault rule base is matched to determine the anomaly type and classify the fault. Estimate the state of charge, health status, and remaining useful life of energy storage devices to generate a status score; Based on the anomaly score, the fault diagnosis results generate maintenance suggestions and are fed back to the large model.

10. A city virtual power plant operation device based on an intelligent agent architecture, characterized in that, The device includes: a communication unit and a processing unit; The communication unit collects various types of data from the urban energy system as first data, which includes distributed generation data, energy storage device status data, adjustable load data, electricity consumption data, electricity market data, and meteorological environment data. The processing unit, based on the data types required for multiple agents to perform tasks, orchestrates and schedules the first data, determines the first data to be processed by each agent, and calls the multiple agents to process the required first data; based on the processing results output by the multiple agents and the operating constraints, it generates the control decision and market trading strategy for the city virtual power plant.

Images