Intelligent quantity and price decision-making system and method based on virtual power plant

Abstract

The invention discloses an intelligent quantity and price decision-making system and method based on a virtual power plant, which comprehensively analyzes complex data of a power market and a carbon market, performs deep learning and prediction by using an artificial intelligence technology, provides a more accurate and prospective transaction strategy for a power generation enterprise, and improves the efficiency of the power generation enterprise. The price trend is predicted by analyzing the electricity-carbon market price dynamic state and the market supply and demand condition in real time, and the optimal transaction scheme is automatically generated in combination with the operation characteristics and the power generation cost of the enterprise, so that the competitiveness and the profitability of the power generation enterprise in the electricity-carbon market are improved, and the economic benefit of the power generation enterprise is improved. Therefore, challenges confronted by quantity-price intelligent decision-making in the electric carbon energy transaction process can be handled.

Description

Technical Field

[0001] This invention relates to the field of decision support technology for carbon energy trading, specifically to a quantity and price intelligent decision-making system and method based on a virtual power plant. Background Technology

[0002] The global energy industry is facing an unprecedented transformation to address climate change and achieve the Sustainable Development Goals. Power generation companies, as key players in this energy transition, participate in both the electricity and carbon markets. As a major source of carbon emissions, the power generation industry faces immense pressure to transform and upgrade. They must adapt to the new market environment of linked electricity and carbon prices. This environment demands that companies respond quickly to market and policy changes and optimize their trading strategies.

[0003] Existing technologies have included some research on multi-market strategy behaviors that jointly consider the electricity and carbon markets. However, existing decision-making systems have limitations in handling the complexity of electricity-carbon price linkages and are slow to react to market changes, particularly in real-time data analysis and forward-looking forecasting, making it difficult to adapt to the demands of rapidly changing markets. Especially under the electricity-carbon price linkage mechanism, market price fluctuations are not only influenced by supply and demand but also closely related to carbon emission policies and market expectations. This increases the uncertainty and complexity for power generation companies when formulating trading strategies.

[0004] Electricity carbon energy trading decision support is an intelligent processing approach that combines the dynamic correlation between the electricity market and the carbon market. It aims to help market participants (such as power generation companies, electricity sales companies, and electricity consumers) make scientific decisions in complex energy trading and carbon quota management through data integration, model prediction, and optimization algorithms. Existing electricity carbon energy trading typically adopts a layered architecture, including: a data layer, a model layer, an application layer, and a technology stack. Specifically, the data layer integrates multi-source heterogeneous data (such as electricity load, power generation output, carbon price, policy rules, and meteorological data), storing and managing it through data lakes or data warehouses (such as Hadoop and Spark). The model layer builds predictive and optimization models based on algorithms such as machine learning, reinforcement learning, and operations research. The application layer provides interactive visualization interfaces (such as dashboards), real-time alerts, and strategy recommendations. The technology stack includes Python / R (modeling), TensorFlow / PyTorch (deep learning), Gurobi / CPLEX (mathematical optimization), Kafka (real-time data streaming), and Docker / Kubernetes (containerized deployment).

[0005] The core functional modules of electric carbon energy trading include data fusion and preprocessing, predictive models, trading optimization and strategy generation, market simulation and risk assessment, and visualization and interaction.

[0006] Specifically, data fusion and preprocessing include data acquisition, data cleaning, and feature engineering. Data acquisition involves obtaining real-time and historical data through APIs, Internet of Things (IoT) devices, and publicly available market data interfaces. Data cleaning involves handling missing values ​​and outliers (such as outlier detection) and performing normalization / standardization. Feature engineering involves extracting key features related to the electricity carbon market (such as electricity price volatility, carbon emission intensity, and renewable energy penetration rate).

[0007] The forecasting models include electricity load forecasting, carbon price forecasting, and renewable energy output forecasting. Electricity load forecasting uses time series models (ARIMA, Prophet) or deep learning models (LSTM, Transformer) to predict short-term / long-term electricity demand. Carbon price forecasting combines policy changes (such as tightening carbon quotas), market supply and demand (such as corporate emission reduction costs), and economic indicators, and uses XGBoost, LightGBM, or graph neural networks (GNN) for modeling. Renewable energy output forecasting is based on meteorological data (wind speed, sunshine) and physical models (such as numerical weather prediction NWP), and uses a CNN-LSTM hybrid model to predict wind and solar power generation.

[0008] In terms of trading optimization and strategy generation, this includes multi-objective optimization, objective functions, constraints, and reinforcement learning. Multi-objective optimization is used to balance economic benefits (electricity price difference, carbon quota trading revenue) with carbon emission constraints, constructing mixed integer programming (MIP) or dynamic programming models. An example of an objective function is: maximize total revenue = electricity trading revenue + carbon quota sales revenue - carbon emission costs. Constraints include grid security constraints, total carbon quota limits, renewable energy integration rates, etc. Reinforcement learning (RL) is used in dynamic market environments (such as real-time electricity price fluctuations) to train agents to generate adaptive trading strategies through Q-learning or PPO algorithms.

[0009] Market simulation and risk assessment include game theory modeling and Monte Carlo simulation. Game theory modeling is used to simulate the bidding behavior of market participants (such as generators and users) and analyze market equilibrium under Nash equilibrium or cooperative game. Monte Carlo simulation is used to assess the impact of uncertainties such as carbon price fluctuations and extreme weather on returns and calculate VaR (Value at Risk) or CVaR (Conditional Value at Risk).

[0010] Visualization and interaction include dynamic dashboards and scenario simulations. Dynamic dashboards are used to display key indicators such as real-time electricity prices, carbon prices, carbon emissions, and trading strategy returns. Scenario simulations allow users to adjust parameters (such as carbon allowance prices and the proportion of renewable energy) and intuitively compare the effects of different strategies.

[0011] Typical application scenarios include power generation companies optimizing unit combinations (the ratio of thermal power to renewable energy) and dynamically adjusting carbon allowance trading strategies; electricity consumers reducing electricity costs through demand response while purchasing carbon allowances or investing in carbon offset projects; and trading platforms designing electricity-carbon joint trading products to provide price discovery and liquidity support.

[0012] However, most existing power trading decision-making systems lack in-depth analysis and forecasting capabilities for carbon market dynamics. The establishment and development of carbon markets have brought new trading dimensions and risk management needs to power generation companies. However, price fluctuations and policy changes in carbon emission rights have a direct impact on power generation costs and market competitiveness, factors that are often not fully considered in traditional decision-making models. The main technical challenges currently faced include: multi-market coupled modeling, as the power market and carbon market are strongly correlated (e.g., rising high-carbon power generation costs affect electricity prices), requiring the construction of cross-market joint optimization models; and real-time performance and computational efficiency, as low-latency responses are required in high-frequency trading scenarios (e.g., intraday power markets), necessitating the integration of edge computing or lightweight model processing.

[0013] In summary, the core of decision support for carbon trading is to transform complex market rules, dynamic supply and demand relationships, and carbon emission constraints into quantifiable decision-making solutions through data-driven and intelligent algorithms. The technical challenges lie in the fusion and modeling of multi-dimensional data and the balance between real-time performance and robustness. Summary of the Invention

[0014] In view of this, the present invention provides a quantity and price intelligent decision-making system and method based on virtual power plants. By analyzing the dynamics of electricity-carbon market prices and market supply and demand in real time, predicting price trends, and combining the enterprise's operating characteristics and power generation costs, the optimal trading scheme is automatically generated.

[0015] This invention provides a quantity-price intelligent decision-making system based on a virtual power plant, comprising: an information monitoring module, a data management module, a configuration management module, an algorithm management module, a decision management module, a human-computer interaction interface, and a database. The information monitoring module monitors real-time operational data, equipment status, alarm information, application information, and contract execution status of various entities. The data management module configures, statistically manages, and manages the database, enabling data exchange between the database, the human-computer interface, and the algorithm model. The configuration management module provides the configuration of features and parameters required for intelligent decision-making and algorithm execution, including general time-series features, bidding space features, supply-demand ratios, supply-demand sums, and aggregation statistical rules. The algorithm management module manages the built-in quantity-price prediction algorithm model after its implementation through code. The decision management module is used for the formation and execution of intelligent decisions by different roles in the virtual power plant. The human-computer interaction interface provides users with system usage functions. The database includes a historical database, a model and feature database, and a time-series database.

[0016] Furthermore, the algorithm management module includes a volume and price prediction submodule, an optimization control submodule, and a self-learning submodule. The volume and price prediction submodule combines a static database, a dynamic database, and an expert knowledge base, integrates probability models and feature model configurations, performs dynamic combination optimization, and constructs an intelligent decision-making model using control equations. The intelligent decision-making model includes two types of decisions: one is market operation forecasting, including medium- and long-term electricity holding planning and D+1 spot trading forecast information; the other is intraday operation control strategies for resources within the virtual power plant and D+1 volume and price reporting strategies for the virtual power plant.

[0017] Furthermore, the human-computer interaction interface includes three forms: Web, mobile, and PC.

[0018] This invention provides a quantity-price intelligent decision-making method based on a virtual power plant, which specifically includes the following steps:

[0019] Step 1: Collect electricity market data to form a training sample set;

[0020] Step 2: Construct a smart trading decision model for power generation enterprises with electricity-carbon price linkage as the transmission variable. Use the training sample set constructed in Step 1 to complete the training of the smart trading decision model for power generation enterprises.

[0021] Step 3: In actual use, the AI ​​large model analysis engine is used to predict the electricity-carbon market price and optimize the trading strategy through the intelligent trading decision model of power generation enterprises.

[0022] Step 4: Based on the analysis results obtained in Step 3, provide intelligent recommendations on trading timing, price setting, and trading volume to form trading strategy opinions;

[0023] Step 5: Display real-time data, recommended trading strategies, and relevant information on historical trading analysis using a user interface, and provide an operation interface for power generation company management and operators.

[0024] Furthermore, the training sample set is constructed by collecting key data from the transaction system or the enterprise's internal system to form the training sample set.

[0025] Furthermore, the electricity market data includes historical transaction electricity prices, historical transaction carbon prices, historical declared electricity volume, declared electricity prices, actual power generation, actual transaction volume, power loss, cost per kilowatt-hour, and operation and maintenance costs.

[0026] Furthermore, the electricity market data undergoes pre-cleaning to process missing and invalid data before being calculated to obtain secondary feature data.

[0027] Furthermore, the secondary characteristic data includes the market supply-demand ratio, the marginal cost of average power generation capacity, the monthly electricity volume available for market trading within the safety constraints, seasonal price factors, and time-series price trends.

[0028] Furthermore, the intelligent trading decision-making model for power generation enterprises includes an upper-level multi-objective optimization model for multiple market trading strategies of power generation enterprises and a lower-level data-driven prediction model for different markets.

[0029] Furthermore, the constraints of the market trading strategy optimization model include: physical constraints, technical constraints, market constraints, or regulatory constraints; the objective function of the market trading strategy optimization model includes: maximizing trading revenue, minimizing carbon emissions, or minimizing costs.

[0030] Beneficial effects:

[0031] This invention provides power generation companies with more accurate and forward-looking trading strategies by comprehensively analyzing complex data from the electricity and carbon markets and using artificial intelligence technology for in-depth learning and prediction. By analyzing real-time price dynamics and market supply and demand in the electricity-carbon market, predicting price trends, and combining the company's operating characteristics and power generation costs, it automatically generates the optimal trading plan, thereby improving the competitiveness and profitability of power generation companies in the electricity-carbon market and addressing the challenges of intelligent decision-making on quantity and price in the electricity-carbon energy trading process. Attached Figure Description

[0032] Figure 1 This invention provides a schematic diagram of the system architecture of a quantity and price intelligent decision-making system based on a virtual power plant.

[0033] Figure 2 This is a schematic diagram of the structure of a quantity and price prediction submodule in a quantity and price intelligent decision-making system based on a virtual power plant, provided by the present invention.

[0034] Figure 3 This is a schematic diagram of the processing flow of a quantity-price intelligent decision-making method based on a virtual power plant provided by the present invention. Detailed Implementation

[0035] The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0036] A Virtual Power Plant (VPP) is a power resource coordination and management system that uses advanced information and communication technologies and software systems to aggregate and coordinate distributed energy resources (DERs) such as distributed generators (DG), energy storage systems, controllable loads, and electric vehicles, allowing them to participate as a special power plant in the electricity market and grid operation. The core concept of a virtual power plant can be summarized as communication and aggregation. Key technologies of virtual power plants mainly include coordination and control technology, smart metering technology, and information and communication technology, enabling DERs to participate in the operation of the electricity market and ancillary services market, providing management and ancillary services for distribution and transmission networks.

[0037] This invention provides a quantity and price intelligent decision-making system and method based on a virtual power plant. The core idea is that different entities such as operators, energy storage, photovoltaics, and charging stations log into the system with different roles in the virtual power plant. Then, they view different information content according to their different role permissions. Historical databases and time-series databases are configured according to the permissions of different roles and their own needs to support the realization of resource operation regulation and information monitoring functions of the virtual power plant. In the configuration management stage, the main tasks are to complete the parameter configuration of various data aggregation, time-series characteristics, bidding optimization, etc., to support intelligent decision-making and algorithm prediction and inference of roles with different permissions. Finally, the system functions of the roles are realized through human-computer interaction.

[0038] This invention provides a quantity-price intelligent decision-making system based on a virtual power plant, the system architecture of which is as follows: Figure 1 As shown, the system includes an information monitoring module, a data management module, a configuration management module, an algorithm management module, a decision management module, a human-machine interface, and a database. The database includes a historical database, a model and feature database, and a time-series database. The information monitoring module monitors real-time operational data, equipment status, alarm information, reporting information, and contract execution status for various entities. The data management module configures, statistically analyzes, and manages the database, ensuring data security, validity, and reliability, and enabling data exchange between the database, the human-machine interface, and the algorithm model. The configuration management module provides configurations for general time-series features, bidding space features, supply-demand ratios, supply-demand sums, and aggregation statistical rules required for intelligent decision-making and algorithm execution. The algorithm management module manages the built-in quantity-price prediction algorithm model after its implementation in code, facilitating access by different roles with varying permissions. The decision management module is used for the formation and execution of intelligent decisions by different roles in the virtual power plant. The human-machine interface meets the usage needs of users with different system architectures.

[0039] Furthermore, the information monitoring module mainly monitors the real-time operation data, equipment status, alarm information, reporting information, and contract execution of entities including energy storage, photovoltaics, power distribution networks, wind power, and charging stations.

[0040] The data management module is used for the configuration, statistics and management of historical databases and time-series databases, and ensures the security, validity and reliability of data in historical databases and time-series databases, and realizes data interoperability between historical databases, time-series databases and human-computer interfaces and algorithm models.

[0041] The configuration management module includes a unified field configuration submodule and a supply and demand item configuration submodule, which support the configuration of features and parameters such as general time-series features, bidding space features, supply and demand ratios, supply and demand sums, and aggregation statistical rules required for intelligent decision-making and algorithm execution.

[0042] The algorithm management module manages the built-in volume and price prediction algorithm models through code implementation, facilitating access by different roles and permissions. It includes a volume and price prediction submodule, an optimization control submodule, and a self-learning submodule. Typical self-learning algorithms include convolutional neural networks and recurrent neural networks, while typical volume and price prediction algorithms include grey prediction and Markov prediction.

[0043] The decision management module can meet the intelligent decision-making and execution needs of different roles in the virtual power plant, such as energy storage application power prediction, application price prediction, revenue prediction, and deviation level assessment, including generating strategies such as short-term revenue maximization strategy and medium-to-long-term revenue maximization strategy according to different configurations.

[0044] Human-computer interaction interfaces mainly include three forms: Web, mobile, and PC, which can meet the needs of users with different system architectures to realize the use of the system architecture.

[0045] The quantity and price prediction submodule within the algorithm management module primarily functions to achieve high-quality, accurate collection and governance of both static and dynamic data, supporting the training of full-scenario power prediction models and their participation in electricity market reporting. The structure of the quantity and price prediction submodule is as follows: Figure 2 As shown, the volume and price forecasting submodule combines static and dynamic databases with an expert knowledge base, integrates probabilistic and feature model configurations, performs dynamic combinatorial optimization, and constructs an intelligent decision-making model using control equations. The output intelligent decision-making model mainly includes two levels. First, it forecasts market operations from the perspective of a virtual power plant participating in the electricity market, including medium- and long-term power holding plans and D+1 spot trading. Second, it develops intraday operation control strategies for resources within the virtual power plant and D+1 volume and price reporting strategies for the virtual power plant. During the formation of intelligent strategies, it continuously performs output analysis, historical revenue analysis, and day-ahead strategy optimization for D-day, D+1-day, and D+2-day.

[0046] Among them, D+1 spot trading usually refers to the strategy design of trading for the next day (Day-ahead+1, or D+1) in the electricity spot market. That is, in the day-ahead market, based on the prediction of market parameters such as electricity price and carbon price for the next 24 hours (or more subdivided time periods), the trading decisions are made for energy or carbon quotas in different time periods to achieve arbitrage, cost optimization or risk hedging.

[0047] The database of the quantity and price forecasting submodule is the core foundation for forecast accuracy. The dynamic data list is shown in Table 1, which includes real-time data, meteorological data, and status data. Real-time data includes total power generation data, 15S-level inverter operation data, and sample project reference data. Meteorological data includes environmental information data, disaster warnings, satellite cloud images, multi-source high-quality data, conventional forecast data, local sensor data, and meteorological station and wind tower data. Status data includes shutdown maintenance data, system alarms, control command issuance and feedback data. The static database contains detailed historical operation data, historical meteorological data, and geographical data of each participating entity, and at least covers historical power generation data, historical maintenance information, power curtailment information, high-precision 3D terrain, historical satellite radiation data, theoretical power data, and historical inverter data.

[0048] Table 1. List of Regular Dynamic Data for the Body Volume and Price Prediction Engine

[0049]

[0050] The working principle of the quantity and price prediction submodule is to make intelligent decisions by combining data and expert experience with mathematical algorithms. First, it ensures accurate prediction of power generation and consumption. Then, it uses AI tools to achieve real-time analysis of power data and uses the coordinated control of power generation, grid, load and storage to achieve planned and expected resource scheduling and management.

[0051] Based on the intelligent quantity-price decision-making system based on virtual power plants provided by this invention, this invention also provides an intelligent quantity-price decision-making method based on virtual power plants, the processing flow of which is as follows: Figure 3 As shown, after logging into the system with different roles, the ontology is instantiated simultaneously with identity verification. Regular ontology data retrieval can be completed directly through information browsing and human-computer interaction. When the ontology domain intelligent decision engine is invoked, output and load data for more than a set number of days can be imported manually or manually first, and preprocessed according to user-configured rules. Then, based on the characteristics of high and low electricity prices, the XGB (eXtremeGradient Boosting) regression model or the optimized classification model is invoked. Node modeling is adopted, and multi-model fusion of electricity price and electricity volume is used to carry out segmented prediction and optimization. Finally, a decision dashboard is generated, and the user completes the entire process through the human-computer interface.

[0052] The purpose of this invention is to provide an AI-driven intelligent trading decision-making system and method for power generation enterprises, linking electricity and carbon prices. Based on a real-time market data-driven electricity-carbon price linkage mechanism, this system analyzes electricity and carbon prices in real time, predicts market trends, and guides power generation enterprises to conduct energy transactions at different times, achieving optimal transfer of electricity-carbon value. This method establishes an intelligent decision-making model that considers the operating revenue and market risks of power generation enterprises, using electricity-carbon prices as the key variable to guide the dynamic balance of transaction pricing and trading volume, enabling power generation enterprises to find a balance between economic and environmental benefits.

[0053] Specifically, the present invention provides a quantity-price intelligent decision-making method based on a virtual power plant, comprising the following steps:

[0054] Step 1: Based on the collected multi-dimensional data such as electricity price, carbon price, power generation, energy consumption, and cost, establish a comprehensive data system architecture.

[0055] Key data is collected and managed from the trading system or internal enterprise systems, including historical electricity prices (10-30 trading days), historical carbon prices (3-30 trading days), historical declared electricity volume, declared electricity price, actual power generation, actual traded electricity volume, power loss, cost per kilowatt-hour, and operation and maintenance costs for the same type of entity. This data undergoes pre-cleaning via a data governance module, removing missing and invalid data to standardize the data and ensure quality. Then, a data calculation module is used to calculate secondary characteristic data, such as the market supply-demand ratio, the marginal cost of average power generation capacity, the monthly market-trading volume within safety constraints, seasonal price factors, and time-series price trends.

[0056] Step 2: Construct a smart trading decision-making model for power generation enterprises with electricity-carbon price linkage as the transmission variable.

[0057] The upper-level model is a multi-objective optimization model for power generation companies' trading strategies in multiple markets, while the lower-level model is a data-driven prediction model for different markets.

[0058] Multiple markets include: carbon trading, electricity trading, and ancillary services trading.

[0059] Predictive models primarily output the fluctuation trends of key data such as electricity prices and carbon prices. By selecting and configuring appropriate predictive models based on the data characteristics of different markets, such as ARIMA, seasonal decomposition time series predictive models, and machine learning models (random forest, gradient boosting tree, etc.), the models output the probability distribution of future electricity clearing prices, as well as the trends of carbon emission prices, frequency regulation (reserve) prices, and winning bid K-value trends.

[0060] This paper proposes an optimization model for multiple market trading strategies of power generation enterprises with a multi-objective linkage between electricity and carbon prices. The model defines objective functions such as maximizing trading revenue, minimizing carbon emissions, and minimizing costs, assigns weights to each objective, and configures event factors and constraints. These constraints include physical constraints (e.g., maximum power generation), technical constraints (e.g., equipment capacity, unit ramp-up parameters), market constraints (e.g., price limits, market supply and demand), and regulatory constraints (e.g., emission standards). A linear programming solution technique is selected, and the Pareto optimization method is used to find the Pareto front.

[0061] Step 3: Utilize the AI ​​large model analysis engine, combined with deep learning, reinforcement learning and other technologies, to predict electricity-carbon market prices and optimize trading strategies.

[0062] Step 4: Develop an intelligent decision support system. Based on the analysis results of the AI ​​model, provide intelligent recommendations on trading timing, price setting, and trading volume, and output trading strategy opinions.

[0063] Step 5: Provide a user interface to display real-time data, trading strategy recommendations, historical trading analysis and other information, providing an operating interface for power generation company management and operators.

[0064] Specifically, taking the actual operation of a VPP in a certain province as the target, and using data from 2023-2024 as the basis, the VPP participants include 5 photovoltaic power plants and 3 energy storage systems, with specific configurations shown in Tables 1 and 2.

[0065] Table 1 Configuration parameters for photovoltaic power plants

[0066]

[0067] Table 2 Configuration parameters for energy storage power stations

[0068]

[0069] Within the experimental period, the data sources included actual settlement data, MTL-DNN prediction data, and prediction data from other models. By analyzing and comparing these three types of data, the accuracy of the predictions can be intuitively judged.

[0070] For example, taking PV and ESS power generation data of a typical day in June 2024 as shown in Table 3, the MTL-DNN model considers factors such as PV operation mode and transmission characteristics. Compared with the prediction data of other models, the prediction accuracy of MTL-DNN-LSTM is significantly closer to the actual settlement data. The prediction accuracy of PV peak power generation at 12 o'clock is improved by 0.8%, and the prediction accuracy at 16 o'clock considering weather and other factors is improved by 4%.

[0071] Table 3 Typical Daily Data for Photovoltaic Power Plants in June

[0072]

[0073] Table 4 shows the data for a typical day in June for the energy storage power station. All three energy storage units (1-3) underwent a charge-discharge process on the same day. Compared with the actual settlement data, both models are close to the actual data, with the accuracy of the charging power prediction improved by 1.7% and the accuracy of the discharging power prediction improved by 1.3%.

[0074] Table 4 Typical Daily Data of Energy Storage Power Stations in June

[0075]

[0076] During actual operation in 2024, most photovoltaic power generation occurred through participation in VPP (Vehicle Power Plant), with no significant changes in settlement prices. The regulatory role of energy storage systems (ESS) in VPP operation was significant. Analysis of settlement prices on three typical days (June 1st, 15th, and 23rd) showed that the actual settlement charging and discharging prices for ESS and VPP were RMB 0.15 / kWh and RMB 0.55 / kWh, respectively, as shown in Table 5. The MTL-DNN bidirectional LSTM model significantly outperformed other models in predicting electricity prices, improving the accuracy of charging price prediction by 5% and discharging price prediction by 4%.

[0077] Table 5. Typical daily data for energy storage electricity prices in June (RMB / kWh)

[0078]

[0079] Finally, it should be noted that those skilled in the art can obviously make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0080] The above description is merely one embodiment of the present invention, and should not be construed as limiting the scope of the invention. Any structural changes made based on the present invention, as long as they do not depart from the essence of the invention, should be considered as falling within the protection scope of the present invention and subject to its restrictions. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the method described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0081] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / method.

[0082] The technical solutions of the present invention have been described above with further embodiments listed. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0083] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A quantity-price intelligent decision-making system based on a virtual power plant, characterized in that, include: The system comprises an information monitoring module, a data management module, a configuration management module, an algorithm management module, a decision management module, a human-machine interface, and a database. The information monitoring module monitors real-time operational data, equipment status, alarm information, reporting information, and contract execution status for various entities. The data management module configures, statistically analyzes, and manages the database, enabling data exchange between the database, the human-machine interface, and the algorithm model. The configuration management module provides the configuration of features and parameters required for intelligent decision-making and algorithm execution, including general time-series features, bidding space features, supply-demand ratios, supply-demand sums, and aggregation statistical rules. The algorithm management module manages the built-in quantity-price prediction algorithm model after its implementation in code. The decision management module is used for the formation and execution of intelligent decisions by different roles in the virtual power plant. The human-machine interface provides users with system functionality. The database includes a historical database, a model and feature database, and a time-series database.

2. The intelligent decision-making system for quantity and price according to claim 1, characterized in that, The algorithm management module includes a volume and price prediction submodule, an optimization control submodule, and a self-learning submodule. The volume and price prediction submodule combines a static database, a dynamic database, and an expert knowledge base, integrates probability models and feature model configurations, performs dynamic combination optimization, and constructs an intelligent decision-making model using control equations. The intelligent decision-making model includes two types of decisions: one is market operation prediction, including medium- and long-term electricity holding planning and D+1 spot trading prediction information; the other is the intraday operation control strategy for resources within the virtual power plant and the D+1 volume and price reporting strategy for the virtual power plant.

3. The intelligent decision-making system for quantity and price according to claim 1, characterized in that, The human-computer interaction interface includes three forms: Web, mobile, and PC.

4. A quantity-price intelligent decision-making method based on a virtual power plant using the quantity-price intelligent decision-making system of claim 1, characterized in that, Specifically, the following steps are included: Step 1: Collect electricity market data to form a training sample set; Step 2: Construct a smart trading decision model for power generation enterprises with electricity-carbon price linkage as the transmission variable. Use the training sample set constructed in Step 1 to complete the training of the smart trading decision model for power generation enterprises. Step 3: In actual use, the AI ​​large model analysis engine is used to predict the electricity-carbon market price and optimize the trading strategy through the intelligent trading decision model of power generation enterprises. Step 4: Based on the analysis results obtained in Step 3, provide intelligent recommendations on trading timing, price setting, and trading volume to form trading strategy opinions; Step 5: Display real-time data, recommended trading strategies, and relevant information on historical trading analysis using a user interface, and provide an operation interface for power generation company management and operators.

5. The intelligent decision-making method for quantity and price according to claim 4, characterized in that, The training sample set is constructed by collecting key data from the transaction system or the enterprise's internal system to form the training sample set.

6. The intelligent decision-making method for quantity and price according to claim 4, characterized in that, The electricity market data includes historical transaction prices, historical transaction carbon prices, historical declared electricity volume, declared electricity prices, actual power generation, actual transaction volume, power loss, cost per kilowatt-hour, and operation and maintenance costs.

7. The intelligent decision-making method for quantity and price according to claim 4, characterized in that, The electricity market data undergoes pre-cleaning to remove missing and invalid data before being used to calculate secondary feature data.

8. The intelligent decision-making method for volume and price according to claim 7, characterized in that, The secondary characteristic data includes the market supply-demand ratio, the marginal cost of average power generation capacity, the amount of electricity available for market trading within the safety constraints for the current month, seasonal price factors, and time-series price trends.

9. The intelligent decision-making method for quantity and price according to claim 4, characterized in that, The intelligent trading decision-making model for power generation enterprises includes an upper-level multi-objective optimization model for multiple market trading strategies of power generation enterprises and a lower-level data-driven prediction model for different markets.

10. The intelligent decision-making method for quantity and price according to claim 9, characterized in that, The constraints of the market trading strategy optimization model include: physical constraints, technical constraints, market constraints, or regulatory constraints; the objective function of the market trading strategy optimization model includes: maximizing trading revenue, minimizing carbon emissions, or minimizing costs.

Images