Resource scheduling method, device and equipment for virtual power plant and storage medium

Abstract

The invention provides a resource scheduling method and device for a virtual power plant, equipment and a storage medium, and the method comprises the steps: collecting power grid data of a source network load storage scene, and carrying out the preprocessing of the power grid data; performing operation prediction based on the preprocessed power grid data through a generative AI large model to obtain a prediction result of a source grid load storage scene in a future preset time period; generating a plurality of candidate scheduling strategies based on the prediction result through a multi-objective optimization model; evaluating the plurality of candidate scheduling strategies to obtain corresponding evaluation results; and determining a target scheduling strategy based on the evaluation result, and executing a resource scheduling task based on the target scheduling strategy. According to the resource scheduling scheme of the virtual power plant provided by the invention, the energy utilization efficiency can be improved, the power resource distribution can be optimized, and meanwhile, the stability and economy of a power system are ensured, so that the resource scheduling scheme adapts to a complex energy system, deals with the uncertainty of a market environment and meets the requirement of high real-time performance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy scheduling, in particular to a resource scheduling method and device of a virtual power plant, an electronic device and a storage medium. BACKGROUND

[0002] With the transformation of global energy structure and the development of smart grid technology, virtual power plant (VPP) as a new type of energy management and scheduling mode has become a key technology to improve energy efficiency and cope with the volatility of renewable energy. Virtual power plant integrates and optimizes distributed energy resources such as solar energy, wind energy, energy storage devices, etc., to realize flexible scheduling and economic operation of the power system.

[0003] However, with the increasing complexity of energy structure and the increasing uncertainty of market environment, the traditional virtual power plant resource scheduling technology has the following defects: the existing technology mainly relies on rules and experience, lacks adaptability to the dynamic changes of complex energy systems, and is difficult to realize efficient coordination of multiple energy resources; the fluctuation of power market price and the intermittency of new energy output increase the uncertainty of scheduling, and the existing technology is difficult to effectively respond to the rapid changes of market and technical environment; in addition, virtual power plant needs to respond quickly to the dispatching instructions of the power grid, which puts higher requirements on the real-time performance of the system, and the existing technology often faces the problems of large amount of calculation and low efficiency when dealing with the scheduling of large-scale and multiple types of energy resources.

[0004] The preceding description is to provide general background information and does not necessarily constitute the prior art. SUMMARY

[0005] To solve the above technical problems, the present application provides a resource scheduling method and device of a virtual power plant, an electronic device and a storage medium, which solves the problems of low energy utilization efficiency and insufficient adaptability of the existing technology to complex energy systems and uncertain market environment.

[0006] To solve the above technical problems, the present application provides a resource scheduling method of a virtual power plant, comprising the following steps: Collecting power grid data of a source-grid-load-storage scenario and preprocessing the power grid data; Performing operation prediction based on the preprocessed power grid data through a generative AI large model to obtain a prediction result of the source-grid-load-storage scenario in a future preset time period; Generating a plurality of candidate scheduling strategies based on the prediction result through a multi-objective optimization model; Evaluating the plurality of candidate scheduling strategies to obtain corresponding evaluation results; Determining a target scheduling strategy based on the evaluation results and executing a resource scheduling task based on the target scheduling strategy.

[0007] Further, in some embodiments of the present application, the power grid data of the source grid load storage scene is collected, and the power grid data is preprocessed, including: collecting real-time power data, historical data and weather data of the source grid load storage scene; performing timestamp alignment processing and / or data preprocessing on the real-time power data, the historical data and the weather data.

[0008] Further, in some embodiments of the present application, the operation prediction is performed based on the preprocessed power grid data by the generative AI large model, to obtain a prediction result of the source grid load storage scene in a future preset time period, including: performing standardization processing on the preprocessed power grid data; inputting the standardized power grid data into a pre-constructed generative AI large model to perform operation prediction, and outputting to obtain load prediction results, power generation prediction results and electricity price prediction results of the source grid load storage scene in a future preset time period.

[0009] Further, in some embodiments of the present application, the generative AI large model is constructed in the following manner, including: integrating the collected power grid data to obtain integrated power grid data; performing data set segmentation on the integrated power grid data according to the time sequence diagram analysis result, to obtain a training data set and a test data set; constructing an initial generative AI large model by machine learning technology and deep learning technology; performing model training on the initial generative AI large model based on the training data set and a machine learning framework and a deep learning framework, to obtain a trained generative AI large model; adjusting the trained generative AI large model by using an optimization parameter adjustment technology, to obtain an adjusted generative AI large model; evaluating the prediction accuracy of the adjusted generative AI large model based on the test data set.

[0010] Further, in some embodiments of the present application, a plurality of candidate scheduling strategies are generated based on the prediction result by a multi-objective optimization model, including: defining optimization objectives and constraint conditions of the multi-objective optimization model; analyzing resource scheduling tasks to obtain resource scheduling objectives; analyzing adjustment parameters corresponding to the resource scheduling objectives according to the prediction result, the optimization objectives and the constraint conditions by the multi-objective optimization model; generate a plurality of candidate scheduling strategies based on the adjustment parameters.

[0011] Further, in some embodiments of the present application, the multi-objective optimization model is constructed in the following manner: obtain and label pre-balance energy data and post-balance energy data corresponding to the energy system, and construct training data; An initial energy system balance model is constructed using a Transformer-like architecture, which extracts and compresses key information and multi-dimensional relationships corresponding to the pre-balance energy data through an encoder, and converts the key information extracted by the encoder into a balanced energy system of a predetermined balance target through a decoder; The initial energy system balance model is pre-trained using a self-supervised learning method based on the training data to obtain deep features of the energy system; Use the training data of a specific project as a label for supervised learning to obtain an energy system balance model; evaluate the energy system balance model and adjust the energy system balance model based on the evaluation results.

[0012] Further, in some embodiments of the present application, the evaluation of the plurality of candidate scheduling strategies to obtain corresponding evaluation results comprises: Each of the candidate scheduling strategies is evaluated using a simulation environment to obtain machine learning evaluation results and multi-dimensional benefit evaluation results. Based on the machine learning evaluation results and the multi-dimensional benefit evaluation results, determine the comprehensive evaluation results of each of the candidate scheduling strategies.

[0013] Correspondingly, the present application provides a resource scheduling device of a virtual power plant, comprising: a data acquisition module for acquiring power grid data of a source network load storage scene and preprocessing the power grid data; An operation prediction module is used for generating an operation prediction based on the preprocessed power grid data through a generative AI large model to obtain a prediction result of the source network load storage scene in a future preset time period; A strategy generation module is used for generating a plurality of candidate scheduling strategies based on the prediction result through a multi-objective optimization model; A strategy evaluation module is used for evaluating the plurality of candidate scheduling strategies to obtain corresponding evaluation results; A strategy execution module is used for determining a target scheduling strategy based on the evaluation results, and executing a resource scheduling task based on the target scheduling strategy.

[0014] The application further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the resource scheduling method of the virtual power plant when executing the program.

[0015] The application further provides a storage medium storing a computer program capable of being loaded and executed by a processor to perform the resource scheduling method of the virtual power plant.

[0016] Implementing the embodiments of the application has the following beneficial effects: As described above, the application provides a resource scheduling method, device, electronic device, and storage medium of a virtual power plant. The resource scheduling method of the virtual power plant comprises: collecting power grid data of a source network load storage scene and preprocessing the power grid data; performing operation prediction based on the preprocessed power grid data through a generative AI large model to obtain a prediction result of the source network load storage scene in a future preset time period; generating a plurality of candidate scheduling strategies based on the prediction result through a multi-objective optimization model; evaluating the plurality of candidate scheduling strategies to obtain corresponding evaluation results; determining a target scheduling strategy based on the evaluation results, and performing a resource scheduling task based on the target scheduling strategy. The resource scheduling scheme of the virtual power plant provided by the application accurately predicts the supply and demand of the future preset source network load storage scene and provides an optimized scheduling strategy to comprehensively coordinate and schedule the complex energy system, thereby effectively improving the energy utilization efficiency and the flexibility of resource allocation, reducing the power grid load fluctuation, and enhancing the stability and reliability of the power grid. In addition, the model is continuously updated through adaptive learning, effectively improving the real-time performance and flexibility of the scheduling strategy, effectively adapting to the complex energy system and coping with the uncertainty of the market environment, and meeting the high real-time requirement. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings incorporated into the specification and forming a part thereof illustrate embodiments consistent with the application and, together with the specification, serve to explain the principles of the application. In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, for those of ordinary skill in the art, other drawings can also be obtained without creative labor based on these drawings.

[0018] Figure 1 is an application scenario diagram of the resource scheduling method of the virtual power plant provided by the embodiments of the application; Figure 2 is a flow diagram of the resource scheduling method of the virtual power plant provided by the embodiments of the application; Figure 3 is a structural diagram of the resource scheduling device of the virtual power plant provided by the embodiments of the application; Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0019] The realization of the objectives, functional features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and textual descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0021] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0022] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0023] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.

[0024] With the widespread integration of distributed energy resources, the energy structure of virtual power plants has become more complex, bringing new challenges to dispatching. Furthermore, the significant fluctuations in electricity market prices and the intermittent nature of renewable energy output increase dispatching uncertainty. Virtual power plants also need to respond quickly to grid dispatching commands, placing higher demands on the system's real-time performance. However, traditional virtual power plant resource dispatching methods are mostly based on rules and experience, making them difficult to adapt to complex energy systems and rapidly changing market environments. In addition, traditional methods often face challenges such as high computational load and low efficiency when dealing with large-scale, multi-type energy resource dispatching problems.

[0025] To address the aforementioned technical problems, this application provides a resource scheduling method, apparatus, electronic device, and storage medium for a virtual power plant.

[0026] Specifically, the resource scheduling device of the virtual power plant can be integrated into an electronic device, which can be a smartphone, tablet, laptop, or desktop computer, but is not limited to these. The electronic device can be directly or indirectly connected to the server via wired or wireless communication. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. This application does not impose any restrictions on these aspects.

[0027] Please see Figure 1 , Figure 1 This is an application environment diagram of a resource scheduling method for a virtual power plant in one embodiment. (Refer to...) Figure 1 The resource scheduling method of this virtual power plant can be applied to the resource scheduling system of a virtual power plant. This system can include a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 can be a desktop terminal or a mobile terminal, specifically a mobile phone, tablet, laptop, or other similar devices. The server 120 can be a standalone server or a server cluster composed of multiple servers. The terminal 110 is used to collect grid data from the source-grid-load-storage scenario and preprocess the data; it then uses a generative AI model to perform operational predictions based on the preprocessed grid data, obtaining prediction results for the source-grid-load-storage scenario over a preset time period; a multi-objective optimization model generates multiple candidate scheduling strategies based on the prediction results; the candidate strategies are evaluated to obtain corresponding evaluation results; a target scheduling strategy is determined based on the evaluation results, and resource scheduling tasks are executed based on the target scheduling strategy.

[0028] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the priority of the embodiments.

[0029] Please see Figure 2 , Figure 2 This is a flowchart illustrating a resource scheduling method for a virtual power plant according to an embodiment of this application. The resource scheduling method for a virtual power plant provided in this embodiment may specifically include the following steps: S1. Collect grid data from the source-grid-load-storage scenario and preprocess the grid data; Specifically, for step S1, grid data from the power generation, grid, load, and storage scenarios are collected. This data includes, but is not limited to, real-time power data, historical data, and weather data. This data is crucial for understanding the current state of the power grid and predicting future trends. The collected grid data undergoes preprocessing, including timestamp alignment and data preprocessing, to ensure data consistency and availability. This preprocessing step is key to ensuring that subsequent prediction and optimization models receive high-quality, accurate data. In addition to basic data acquisition, this step may also include data integration and fusion techniques to integrate data from different sources and formats, improving the comprehensiveness and accuracy of the data.

[0030] S2. Based on preprocessed power grid data, a generative AI large model is used to make operational predictions and obtain prediction results for the source-grid-load-storage scenario in the future preset time period; Specifically, in step S2, a pre-built generative AI model is used to perform operational forecasts based on preprocessed grid data, yielding predictions for the power generation, grid, load, and storage scenarios over a future period. This includes forecasts of grid load, power generation, and electricity prices, providing decision support for resource scheduling. In addition to the generative AI model, other advanced prediction techniques, such as deep learning and reinforcement learning, can be further used to improve the accuracy and robustness of the predictions.

[0031] S3. Based on the prediction results, generate multiple candidate scheduling strategies using a multi-objective optimization model; Specifically, in step S3, based on the prediction results, an optimization objective and constraints are defined using a multi-objective optimization model. Multi-objective optimization techniques are then used to find the optimal balance among multiple objectives, generating multiple candidate scheduling strategies. Within the multi-objective optimization model, new optimization algorithms, such as evolutionary algorithms and mixed-integer programming, can be explored to improve the efficiency and effectiveness of strategy generation.

[0032] S4. Evaluate multiple candidate scheduling strategies and obtain the corresponding evaluation results; Specifically, in step S4, all generated candidate scheduling strategies are evaluated to obtain the evaluation result for each candidate strategy. The evaluation process includes machine learning evaluation metrics and multi-dimensional benefit metrics evaluation to determine the performance and benefits of each strategy.

[0033] S5. Determine the target scheduling strategy based on the evaluation results, and execute the resource scheduling task based on the target scheduling strategy; Specifically, in step S5, based on the evaluation results of each candidate scheduling strategy, the optimal candidate scheduling strategy is selected as the final target scheduling strategy. Resource scheduling tasks are then executed based on this target scheduling strategy to ensure that the selected strategy achieves optimal economic benefits and system stability in actual power grid operation. This embodiment can also utilize a real-time feedback mechanism and adaptive adjustment strategy to enable the system to dynamically adjust the scheduling strategy based on real-time data and feedback, thereby improving the system's response speed and adaptability.

[0034] As can be seen, this embodiment can utilize power grid resources more effectively, reduce waste, and improve energy efficiency through accurate prediction and optimized scheduling; the optimized scheduling strategy helps to balance supply and demand, reduce power grid load fluctuations, and enhance the stability and reliability of the power grid; through prediction and optimization, unnecessary energy purchases and waste can be reduced, power grid operating costs can be lowered, and the economic benefits of the power grid can be improved.

[0035] Furthermore, in some embodiments, step S1, "collecting grid data in a source-grid-load-storage scenario and preprocessing the grid data," may specifically include: S11. Collect real-time power data, historical data, and weather data from the power generation, grid, load, and storage scenarios; S12. Perform timestamp alignment and / or data preprocessing on real-time power data, historical data, and weather data.

[0036] Specifically, step S1 mainly includes data acquisition and data preprocessing. Data acquisition includes collecting real-time power data, historical data, and weather data from the power generation, grid, load, and storage scenarios. Real-time power data collection includes, but is not limited to, load, power generation, and electricity price. First, the input and output load of the power system is tracked in real-time using devices such as smart meters and secondary anti-reverse flow meters to ensure data real-time performance and accuracy. Real-time data from distributed energy facilities is collected, such as the output power of photovoltaic panels, the speed and power output of wind turbine generators, and the charging status and capacity of energy storage batteries. Sensors and monitoring systems are used to monitor the health and efficiency of equipment in real-time to ensure data accuracy and reliability. Third-party meteorological services are accessed to obtain key meteorological parameters such as sunlight, wind speed, and temperature in real time. For data preprocessing, the timestamps of the collected energy system data are aligned for subsequent analysis and model training. Load data is preprocessed, such as through denoising, interpolation, and smoothing, to improve data quality. Meteorological data is supplemented according to the timestamps of the load data for better integration with the model.

[0037] Furthermore, this embodiment can also integrate data from different sources and time scales through advanced data fusion technology, improving data consistency and usability. In the data preprocessing stage, an automated anomaly detection algorithm is integrated to identify and process anomalous data points, further improving data quality. For certain hard-to-obtain data, such as data from extreme weather conditions, data augmentation techniques can be used to supplement the deficiencies of actual data by simulating and generating additional data.

[0038] This embodiment improves data accuracy through real-time data acquisition and precise timestamp alignment, providing a reliable data foundation for subsequent forecasting and scheduling. Preprocessing and supplementing load and meteorological data enhances data integrity, enabling the model to more comprehensively understand and predict the energy system's operating status. High-quality preprocessed data improves the performance of generative AI models, resulting in more accurate predictions and thus enhancing the effectiveness of scheduling strategies. Automated anomaly detection and data augmentation technologies allow the system to better adapt to different operating conditions and environmental changes, improving its robustness. Accurate data acquisition and preprocessing make resource scheduling strategies more precise, contributing to optimized energy allocation and improved energy utilization efficiency.

[0039] Furthermore, in some embodiments, step S2, "using a generative AI large model to perform operational forecasting based on preprocessed grid data to obtain prediction results for the source-grid-load-storage scenario in the future preset time period," may specifically include: S21. Standardize the preprocessed power grid data; S22. Input the standardized grid data into the pre-built generative AI model for operation prediction, and output the load prediction results, power generation prediction results, and electricity price prediction results for the source-grid-load-storage scenario within a preset time period.

[0040] Specifically, in step S2, the preprocessed power grid data is standardized to ensure data consistency and comparability. Since different data sources may have different dimensions and distributions, standardization ensures that these data have the same scale before being input into the AI ​​model, thereby improving the model's training efficiency and prediction accuracy. The standardized power grid data is then input into a pre-built generative AI model for prediction. Based on historical and current data, the generative AI model uses a sliding window recursive approach to output prediction results for the next several days, yielding load predictions, power generation predictions, and electricity price predictions for a preset future time period.

[0041] Furthermore, feature selection and engineering are performed before data is input into the AI ​​model to extract the most useful information and reduce model complexity. This can be achieved through automated feature selection algorithms or guided by domain expert knowledge. Model ensemble techniques, such as random forests or gradient boosting machines, are employed to combine the predictions of multiple generative AI models, improving prediction accuracy and robustness. The model can be updated in real time to adapt to rapid changes in power grid conditions. This can be achieved through online learning or incremental learning algorithms, enabling the model to quickly adapt to new data without requiring retraining from scratch.

[0042] This embodiment improves the model's prediction accuracy through standardization and feature engineering, making resource scheduling strategies more based on actual grid operation conditions. Model ensemble technology enhances the model's generalization ability, ensuring high accuracy of prediction results under different grid conditions. Real-time updates and adaptive learning technologies enable the model to quickly respond to changes in grid conditions, improving the system's response speed and adaptability. Accurate load, power generation, and electricity price predictions provide crucial decision support for resource scheduling, helping to optimize resource allocation and improve energy efficiency.

[0043] Furthermore, in some embodiments, the construction method of generative AI large models may specifically include: The collected power grid data is integrated and processed to obtain integrated power grid data; Based on the time series diagram analysis results, the integrated power grid data is split into training and testing datasets. Build an initial generative AI model using machine learning and deep learning techniques; Based on the training dataset and machine learning and deep learning frameworks, the initial generative AI large model is trained to obtain the trained generative AI large model for modeling. The trained generative AI model was adjusted using optimization and parameter tuning techniques to obtain the adjusted generative AI model. The prediction accuracy of the adjusted generative AI large model is evaluated based on the test dataset.

[0044] Specifically, the construction and training process of the generative AI large-scale model in this embodiment is as follows: The collected power grid data is integrated to obtain integrated power grid data. For example, multi-source data such as real-time power data, historical data, and weather data are fused, and the input data is efficiently processed through regularization and normalization to form a unified dataset for model training. Based on time series analysis of volatility and periodicity, the integrated power grid data is split into training and testing datasets to evaluate the model's performance on unseen data. An initial generative AI large-scale model is constructed using machine learning and deep learning techniques. A suitable model architecture, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), or a Transformer, is selected to meet the needs of time series forecasting. Based on the training dataset, the constructed model is trained using frameworks such as sklearn and pytorch to obtain the trained generative AI large-scale model. For example, model parameters are adjusted to minimize prediction errors. The trained generative AI model was tuned using hyperparameter optimization techniques, including learning rate and batch size optimization. Grid search and random search were employed to significantly improve the model's capabilities. The prediction accuracy of the tuned generative AI model was evaluated using a test dataset. Various evaluation metrics, such as mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), were used to quantify the model's predictive performance.

[0045] In specific embodiments, model interpretability techniques, such as LIME or SHAP, can be introduced to provide interpretability of model predictions and help understand the model's decision-making process; multi-model fusion strategies can be implemented to combine the prediction results of different models to improve the robustness and accuracy of predictions.

[0046] This embodiment improves the accuracy of the model's predictions of future grid load, power generation, and electricity prices through meticulous data preprocessing and model training; enhances the model's generalization ability to new data by splitting the dataset and evaluating the model on independent test sets; optimizes parameter tuning and model evaluation steps to ensure that the model can adapt to changes in grid data, improving the model's adaptability and flexibility; and provides powerful decision support for resource scheduling with a high-performance generative AI model, making scheduling strategies more scientific and accurate.

[0047] Furthermore, in some embodiments, step S3, "generating multiple candidate scheduling strategies based on the prediction results using a multi-objective optimization model," may specifically include: S31. Define the optimization objective and constraints of the multi-objective optimization model; S32. Analyze the resource scheduling task to obtain the resource scheduling objective; S33. Analyze the adjustment parameters corresponding to the resource scheduling objectives based on the prediction results, optimization objectives, and constraints using a multi-objective optimization model; S34. Generate multiple candidate scheduling strategies based on the adjusted parameters.

[0048] Specifically, for step S3, the optimization objectives of the multi-objective optimization model are defined, including maximizing economic benefits, maximizing system stability, and maximizing the utilization of renewable energy. Simultaneously, the constraints that the model must adhere to are defined, such as equipment operating limitations, safety standards, and environmental regulations. A thorough analysis of the resource scheduling task is conducted to clarify scheduling objectives, including identifying key resources, predicting resource demand, and assessing resource availability, to ensure that the scheduling strategy meets the system's operational needs. Using the multi-objective optimization model, based on the prediction results, optimization objectives, and constraints, the adjustment parameters corresponding to the resource scheduling objectives are analyzed, including power generation allocation, energy storage device charging and discharging strategies, and load management. Based on the adjustment parameters, multiple candidate scheduling strategies are generated using the multi-objective optimization model. These strategies will consider different optimization objectives and constraints to provide diverse scheduling options.

[0049] In a specific embodiment, based on the trained model, predictions are made for a future period, and multiple candidate scheduling strategies are generated based on the prediction results. The resource scheduling task is decomposed using a large energy model, and the objectives, corresponding inputs, and outputs are set for each sub-task. The core task is energy system balancing. The energy system, composed of energy block data generated in the previous steps (preprocessing, prediction, etc.), is input into the energy balancing system model, and then the model-balanced energy balancing system is output. The parameters that need to be adjusted are extracted from the model-balanced energy balancing system and transformed into strategies that can be deployed and parameters that specific devices need to receive.

[0050] The multi-objective optimization model provided in this embodiment can comprehensively consider multiple objectives and constraints to achieve optimal resource allocation and improve energy utilization efficiency. Through the optimization model, economic benefits can be maximized, such as reducing operating costs and increasing energy sales revenue. The optimization model helps maintain grid stability by reducing the risks of grid load fluctuations and power supply instability through reasonable resource scheduling. Through multi-scenario simulation and real-time optimization adjustments, the adaptability of the scheduling strategy to uncertainties and changes is improved, enhancing the robustness of the system. The multi-objective optimization model can balance economic benefits, system stability, and environmental impact, achieving comprehensive optimization.

[0051] Furthermore, in some embodiments, the construction method of the multi-objective optimization model may specifically include: Acquire and label the energy data before and after the energy system is in equilibrium, and construct training data. An initial energy system balance model is constructed using a Transformer-like architecture. The encoder extracts and compresses key information and multi-dimensional relationships corresponding to the energy data before balance. The decoder transforms the key information extracted by the encoder into a balanced energy system with a predetermined balance target. A self-supervised learning method is used to pre-train the initial energy system balance model based on training data to obtain the deep features of the energy system; Supervised learning is performed using training data from specific projects as labels to obtain an energy system balance model. The energy system balance model is evaluated, and adjustments are made based on the evaluation results.

[0052] Specifically, for the construction of the multi-objective optimization model in this embodiment, the pre-equilibrium and post-equilibrium energy data of the energy system are acquired, labeled, and used to construct training data. This includes energy consumption data, energy storage data, and new energy power generation data, which will be used to train the model to understand and predict the equilibrium state of the energy system. An initial energy system equilibrium model is constructed using a Transformer-like architecture. This architecture extracts and compresses key information and multi-dimensional relationships from the pre-equilibrium energy data through an encoder, and transforms this information into a balanced energy system with a predetermined equilibrium target through a decoder. A self-supervised learning method is used to pre-train the initial energy system equilibrium model based on the training data to obtain deep features of the energy system. This step helps the model learn the inherent laws and patterns of energy data. Supervised learning is performed using training data from specific projects as labels to obtain the energy system equilibrium model, enabling the model to adapt to the special needs of energy system equilibrium in specific scenarios. The energy system equilibrium model is evaluated, and the model is adjusted based on the evaluation results to ensure the effectiveness and accuracy of the model in practical applications.

[0053] In a specific implementation, the construction of the energy system balance model begins with the processing of training data. This involves using a large amount of previously labeled energy system balance data, including substantial pre-balancing energy data (energy consumption, energy storage, new energy generation, etc.) and post-balancing energy data generated by the energy balance optimization model and further refined by the operation and maintenance team. This data constitutes the training data for the energy system balance model. The energy system balance model's structure is based on a transformer-like architecture. The encoder is responsible for extracting and compressing information from the pre-balancing energy data and their multi-dimensional relationships. The decoder uses the information extracted by the encoder to transform the unbalanced energy system into a balanced energy system based on a predetermined balance target. Both the input and output are energy systems composed of multi-dimensional energy blocks. Model training utilizes the PyTorch framework, combined with distributed training technology, to accelerate the processing of massive amounts of data and model iteration. The training phase employs a "multi-stage training" strategy: the first stage uses a self-supervised learning method to pre-train the model based on the energy system composed of large-scale pre- and post-balancing multi-dimensional energy block data, capturing the deep features of the energy system. The second stage uses multidimensional energy block data from specific projects as labels for supervised learning, enabling it to adapt to the special needs of energy system balance in specific scenarios.

[0054] This embodiment improves the accuracy of the model's prediction of the energy system balance state by combining self-supervised learning and supervised learning; the Transformer-like architecture and ensemble learning method enhance the model's generalization ability to different power grid environments; the efficient energy system balance model can quickly respond to power grid changes, improving the efficiency and response speed of resource scheduling; accurate energy system balance prediction helps optimize energy allocation and utilization, and improve energy utilization efficiency; through model evaluation and adjustment, as well as reinforcement learning techniques, the robustness of the system in the face of uncertainty and change is enhanced.

[0055] Furthermore, in some embodiments, step S4, "evaluating multiple candidate scheduling strategies and obtaining corresponding evaluation results," may specifically include: S41. Using a simulation environment, each candidate scheduling strategy is evaluated using machine learning evaluation metrics and multi-dimensional benefit metrics to obtain machine learning evaluation results and multi-dimensional benefit evaluation results. S42. Based on the machine learning evaluation results and the multi-dimensional benefit evaluation results, determine the comprehensive evaluation result for each candidate scheduling strategy.

[0056] Specifically, for step S4, a simulation environment is used to evaluate each candidate scheduling strategy using machine learning metrics and multi-dimensional benefit metrics. The simulation environment provides a platform to simulate actual operation, allowing the effectiveness of the strategy to be tested and evaluated without actually affecting the grid operation. Machine learning evaluation metrics, such as accuracy, recall, and F1 score, are used to quantitatively evaluate the performance of the candidate scheduling strategies. These metrics help assess the effectiveness of the strategy in prediction and scheduling tasks. Multi-dimensional benefit metrics are evaluated, including economic benefits, system stability, and environmental impact; for example, the overall performance of the strategy in different aspects is evaluated to determine its overall benefit. Based on the machine learning evaluation results and the multi-dimensional benefit evaluation results, a comprehensive evaluation result for each candidate scheduling strategy is determined, taking into account the strategy's performance across various dimensions, to select the optimal scheduling strategy.

[0057] In specific embodiments, a simulation environment is used to evaluate each candidate strategy, calculating its performance under different objectives, such as economic benefits and system stability. Specifically, the evaluation can be conducted in two aspects: First, from a machine learning perspective, the output of the energy system balance model is essentially the output of a continuous numerical task, and can be evaluated using common evaluation metrics in the field of machine learning (MAE, MSE) to ensure the model can grasp the basic fit of energy industry operational optimization capabilities. Second, from a business perspective, a performance-oriented evaluation system is designed. The model is evaluated through multi-dimensional benefit indicators (such as improved economic benefits, enhanced safety, and increased green energy ratio) to ensure it not only possesses excellent theoretical performance but also brings considerable benefits in actual operation. Furthermore, simulations of several typical process scenarios are used to test whether the model can balance economic efficiency and system stability under complex production constraints.

[0058] This embodiment improves the accuracy of selecting the optimal scheduling strategy through comprehensive evaluation, ensuring the efficiency and effectiveness of resource scheduling; real-time data integration evaluation and reinforcement learning evaluation methods improve the system's adaptability to dynamic changes in the power grid; accurate evaluation helps optimize resource allocation and utilization, improve energy efficiency and reduce costs; multi-dimensional benefit index evaluation ensures the economic performance of the scheduling strategy, helping to improve the economic benefits of the virtual power plant; by evaluating the impact of the strategy on system stability, it can be ensured that the selected strategy will not have a negative impact on the stability and security of the power grid.

[0059] Furthermore, in some embodiments, step S5, "determining the target scheduling strategy based on the evaluation results and executing the resource scheduling task based on the target scheduling strategy," may specifically include: S51. Convert the selected target scheduling strategy into the corresponding control command; S52. Issue control commands to each energy device to execute resource scheduling tasks.

[0060] Specifically, for step S5, after evaluating candidate scheduling strategies in the simulation environment, comprehensive evaluation results, including machine learning evaluation metrics and multi-dimensional benefit metrics, are collected. These results provide a performance overview of each candidate strategy, including economic benefits, system stability, and environmental impact. Based on the comprehensive evaluation results, one or more optimal candidate scheduling strategies are selected as the target scheduling strategy. Factors considered in the selection include the overall benefit of the strategy, risk level, cost-effectiveness ratio, and impact on system stability and security. After determining the target scheduling strategy, it needs to be fine-tuned to adapt to the specific requirements of actual power grid operation. This includes adjusting the parameters in the strategy to better adapt to the actual operating conditions of the power grid and market changes. The target scheduling strategy is converted into specific control commands and issued to various energy devices. This includes adjusting the output of power generation equipment, managing the charging and discharging of energy storage devices, and optimizing load allocation. During the execution of resource scheduling tasks, the power grid status and equipment performance are monitored in real time to ensure that the implementation effect of the strategy meets expectations. Data during the execution process is collected to provide feedback for subsequent strategy adjustments and model optimization.

[0061] This embodiment improves the efficiency and accuracy of resource scheduling through comprehensive evaluation and optimization, ensuring that the power grid operates in its optimal state; the real-time monitoring and feedback mechanism enhances the system's adaptability to emergencies and market changes, improving the system's robustness; by selecting the optimal scheduling strategy, economic benefits can be maximized, such as reducing operating costs and increasing energy sales revenue; the execution of the target scheduling strategy helps maintain the stability and security of the power grid, reducing the risk of power grid failures and accidents.

[0062] In a specific embodiment, after step S5, "determine the target scheduling strategy based on the evaluation results and execute the resource scheduling task based on the target scheduling strategy," this embodiment also provides a feedback and learning process, designing a user-friendly human-computer interaction interface to facilitate manual viewing, modification, and confirmation of strategy recommendations. Manual personnel can modify and adjust the recommended strategy based on actual conditions and experience, and provide feedback to the system. The system uses manual feedback and actual execution results as new data to continuously train and optimize the model, improving the accuracy of decision-making. As the market environment and system state change, the system can adaptively adjust its strategy, improving the system's flexibility and robustness.

[0063] The adaptive adjustment strategy can be achieved in two ways: First, industry experts and operations personnel revise the daily strategies initially, and then re-input the revised energy balance system data into the relevant models for fine-tuning, so that the models can better adapt to the current scenario. Second, by using reinforcement learning and other techniques based on multi-faceted benefit feedback, the models can update parameters in a timely manner according to the benefit changes caused by small changes in parameters in the real-time scenario, and better allocate weights to understand the current scenario.

[0064] In summary, the resource scheduling method for a virtual power plant provided in this embodiment collects grid data from a source-grid-load-storage scenario and preprocesses the grid data; it then uses a generative AI large-scale model to perform operational predictions based on the preprocessed grid data, obtaining prediction results for the source-grid-load-storage scenario over a preset future time period; a multi-objective optimization model generates multiple candidate scheduling strategies based on the prediction results; these candidate scheduling strategies are evaluated to obtain corresponding evaluation results; a target scheduling strategy is determined based on the evaluation results, and resource scheduling tasks are executed based on the target scheduling strategy. This embodiment provides a resource scheduling scheme for a virtual power plant that accurately predicts the supply and demand of a preset future source-grid-load-storage scenario and provides optimized scheduling strategies for comprehensive and coordinated scheduling of complex energy systems. This effectively improves energy utilization efficiency and resource allocation flexibility, reduces grid load fluctuations, and enhances grid stability and reliability. Furthermore, by continuously updating the model through adaptive learning, the real-time performance and flexibility of the scheduling strategy are effectively improved, adapting to complex energy systems and responding to market uncertainties, thus meeting high real-time requirements.

[0065] To facilitate better implementation of the resource scheduling method for virtual power plants according to the embodiments of this application, the embodiments of this application also provide a resource scheduling device for virtual power plants. The meanings of the terms used are the same as in the resource scheduling method for virtual power plants described above, and specific implementation details can be found in the descriptions in the method embodiments.

[0066] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a resource scheduling device for a virtual power plant provided in an embodiment of this application. Specifically, the resource scheduling device for the virtual power plant may include a data acquisition module 201, an operation prediction module 202, a strategy generation module 203, a strategy evaluation module 204, and a strategy execution module 205, as follows: The data acquisition module 201 is used to acquire power grid data in the source-grid-load-storage scenario and to preprocess the power grid data. The operation forecasting module 202 is used to perform operation forecasting based on preprocessed power grid data using a generative AI big model to obtain the forecast results of the source-grid-load-storage scenario for a future preset time period. The strategy generation module 203 is used to generate multiple candidate scheduling strategies based on the prediction results through a multi-objective optimization model. The strategy evaluation module 204 is used to evaluate multiple candidate scheduling strategies and obtain the corresponding evaluation results; The strategy execution module 205 is used to determine the target scheduling strategy based on the evaluation results and to execute resource scheduling tasks based on the target scheduling strategy.

[0067] Furthermore, in some embodiments, the data acquisition module 201 may specifically include: The data acquisition unit is used to collect real-time power data, historical data, and weather data in the source-grid-load-storage scenario; The data preprocessing unit is used to perform timestamp alignment and / or data preprocessing on real-time power data, historical data, and weather data.

[0068] Furthermore, in some embodiments, the operation forecasting module 202 may specifically include: Standardization unit, used to standardize preprocessed power grid data; The prediction unit is used to input standardized grid data into a pre-built generative AI model for operation prediction, and output load prediction, power generation prediction, and electricity price prediction results for the source-grid-load-storage scenario within a preset time period.

[0069] Furthermore, in some embodiments, the operation prediction module 202 further includes an AI model building unit, which is specifically used for: integrating and processing the collected power grid data to obtain integrated power grid data; dividing the integrated power grid data into a training dataset and a test dataset based on the time series analysis results; constructing an initial generative AI model using machine learning and deep learning technologies; training the initial generative AI model based on the training dataset and the machine learning and deep learning frameworks to obtain a trained generative AI model; adjusting the trained generative AI model using optimization and parameter tuning techniques to obtain an adjusted generative AI model; and evaluating the prediction accuracy of the adjusted generative AI model based on the test dataset.

[0070] Furthermore, in some embodiments, the policy generation module 203 may specifically include: Define the unit, which is used to define the optimization objective and constraints of the multi-objective optimization model; The task analysis unit is used to analyze resource scheduling tasks and obtain resource scheduling objectives; The parameter adjustment unit is used to analyze the adjustment parameters corresponding to the resource scheduling objectives based on the prediction results, optimization objectives, and constraints through a multi-objective optimization model. The strategy generation unit is used to generate multiple candidate scheduling strategies based on the adjusted parameters.

[0071] Furthermore, in some embodiments, the strategy generation module 203 may specifically include a multi-objective optimization model construction unit, which may be used to: acquire and label the pre-equilibrium energy data and post-equilibrium energy data corresponding to the energy system to construct training data; construct an initial energy system equilibrium model using a Transformer-like architecture, extract and compress key information and multi-dimensional relationships corresponding to the pre-equilibrium energy data through an encoder, and transform the key information extracted by the encoder into a equilibrium energy system with a predetermined equilibrium target through a decoder; pre-train the initial energy system equilibrium model using a self-supervised learning method based on the training data to obtain deep features of the energy system; use the training data of a specific project as labels for supervised learning to obtain an energy system equilibrium model; evaluate the energy system equilibrium model and adjust the energy system equilibrium model based on the evaluation results.

[0072] Furthermore, in some embodiments, the policy evaluation module 204 may specifically include: The first evaluation unit is used to evaluate each candidate scheduling strategy using machine learning evaluation metrics and multi-dimensional benefit metrics in a simulation environment, and to obtain machine learning evaluation results and multi-dimensional benefit evaluation results. The second evaluation unit is used to determine the comprehensive evaluation result of each candidate scheduling strategy based on machine learning evaluation results and multi-dimensional benefit evaluation results.

[0073] In summary, the resource scheduling device for the virtual power plant provided in this embodiment collects grid data from the source-grid-load-storage scenario through the data acquisition module 201 and preprocesses the grid data; the operation prediction module 202 performs operation prediction based on the preprocessed grid data using a generative AI large model to obtain prediction results for the source-grid-load-storage scenario in the future preset time period; the strategy generation module 203 generates multiple candidate scheduling strategies based on the prediction results using a multi-objective optimization model; the strategy evaluation module 204 evaluates the multiple candidate scheduling strategies to obtain corresponding evaluation results; and the strategy execution module 205 determines the target scheduling strategy based on the evaluation results and executes the resource scheduling task based on the target scheduling strategy. The resource scheduling device for the virtual power plant provided in this embodiment accurately predicts the supply and demand of the future preset source-grid-load-storage scenario and provides optimized scheduling strategies for comprehensive and coordinated scheduling of complex energy systems, thereby effectively improving energy utilization efficiency and resource allocation flexibility, reducing grid load fluctuations, and enhancing grid stability and reliability; furthermore, by continuously updating the model through adaptive learning, it effectively improves the real-time performance and flexibility of the scheduling strategy, effectively adapts to complex energy systems and responds to market uncertainties, and meets high real-time requirements.

[0074] Furthermore, embodiments of this application also provide an electronic device, such as... Figure 4 The diagram illustrates the structure of an electronic device according to an embodiment of this application. Specifically, the electronic device may include components such as a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, and an input unit 304. Those skilled in the art will understand that... Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein: The processor 301 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 302, and by calling data stored in the memory 302, thereby providing overall monitoring of the electronic device. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 301.

[0075] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and resource scheduling methods for a virtual power plant by running the software programs and modules stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device, etc. In addition, the memory 302 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 302 may also include a memory controller to provide the processor 301 with access to the memory 302.

[0076] The electronic device also includes a power supply 303 that supplies power to various components. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 303 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0077] The electronic device may also include an input unit 304, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0078] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 301 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 302 according to the following instructions, and the processor 301 runs the applications stored in the memory 302 to realize various functions, as follows: Data on power grids, loads, and storage in a source-grid-load-storage scenario are collected and preprocessed. Based on the preprocessed power grid data, a generative AI model is used to make operational predictions and obtain prediction results for the source-grid-load-storage scenario in a future preset time period. Based on the prediction results, a multi-objective optimization model is used to generate multiple candidate scheduling strategies. The multiple candidate scheduling strategies are evaluated to obtain corresponding evaluation results. Based on the evaluation results, the target scheduling strategy is determined and the resource scheduling task is executed based on the target scheduling strategy.

[0079] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0080] This application embodiment accurately predicts the supply and demand of future preset source-grid-load-storage scenarios and provides optimized scheduling strategies to comprehensively coordinate and schedule complex energy systems, thereby effectively improving energy utilization efficiency and resource allocation flexibility, reducing grid load fluctuations, and enhancing grid stability and reliability. In addition, by continuously updating the model through adaptive learning, the real-time performance and flexibility of the scheduling strategy are effectively improved, effectively adapting to complex energy systems and responding to market uncertainties, and meeting high real-time requirements.

[0081] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0082] Therefore, embodiments of this application provide a storage medium storing multiple instructions, which can... The instruction can be loaded by the processor to execute the steps in any of the resource scheduling methods for a virtual power plant provided in the embodiments of this application. For example, the instruction can execute the following steps: Data on power grids, loads, and storage in a source-grid-load-storage scenario are collected and preprocessed. Based on the preprocessed power grid data, a generative AI model is used to make operational predictions and obtain prediction results for the source-grid-load-storage scenario in a future preset time period. Based on the prediction results, a multi-objective optimization model is used to generate multiple candidate scheduling strategies. The multiple candidate scheduling strategies are evaluated to obtain corresponding evaluation results. Based on the evaluation results, the target scheduling strategy is determined and the resource scheduling task is executed based on the target scheduling strategy.

[0083] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0084] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk, or optical disk, etc. Since the instructions stored in the storage medium can execute the steps of any of the virtual power plant resource scheduling methods provided in the embodiments of this application, the beneficial effects achievable by any of the virtual power plant resource scheduling methods provided in the embodiments of this application can be realized. See the preceding embodiments for details, which will not be repeated here.

[0085] The resource scheduling method, apparatus, electronic device, and storage medium of a virtual power plant provided in the embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A resource scheduling method of a virtual power plant, characterized by, The method comprises the following steps: collecting power grid data of a source grid load storage scene and preprocessing the power grid data; performing operation prediction based on the preprocessed power grid data through a generative AI large model to obtain a prediction result of the source grid load storage scene in a future preset time period; generating multiple candidate scheduling strategies based on the prediction result through a multi-objective optimization model; evaluating the multiple candidate scheduling strategies to obtain corresponding evaluation results; determining a target scheduling strategy based on the evaluation results and performing a resource scheduling task based on the target scheduling strategy.

2. The resource scheduling method of the virtual power plant according to claim 1, characterized in that, The collecting of the power grid data of the source grid load storage scene and the preprocessing of the power grid data comprise: collecting real-time power data, historical data and weather data of the source grid load storage scene; performing timestamp alignment processing and / or data preprocessing on the real-time power data, the historical data and the weather data.

3. The method of claim 1, wherein, The operation prediction based on the preprocessed power grid data through the generative AI large model to obtain the prediction result of the source grid load storage scene in the future preset time period comprises: performing standardization processing on the preprocessed power grid data; inputting the standardized processed power grid data into a pre-constructed generative AI large model to perform operation prediction, and outputting load prediction results, power generation prediction results and electricity price prediction results of the source grid load storage scene in the future preset time period.

4. The resource scheduling method of the virtual power plant according to claim 3, characterized in that, The construction method of the generative AI large model comprises: integrating the collected power grid data to obtain integrated power grid data; performing data set segmentation on the integrated power grid data according to a time sequence diagram analysis result to obtain a training data set and a test data set; constructing an initial generative AI large model through machine learning technology and deep learning technology; performing model training on the initial generative AI large model based on the training data set and a machine learning framework and a deep learning framework to obtain a trained generative AI large model; adjusting the trained generative AI large model using an optimization parameter adjustment technology to obtain an adjusted generative AI large model; evaluating the prediction accuracy of the adjusted generative AI large model based on the test data set.

5. The method of claim 1, wherein, The generation of multiple candidate scheduling strategies based on the prediction result through the multi-objective optimization model comprises: defining optimization objectives and constraint conditions of the multi-objective optimization model; analyzing a resource scheduling task to obtain a resource scheduling objective; analyzing adjustment parameters corresponding to the resource scheduling objective according to the prediction result, the optimization objectives and the constraint conditions through the multi-objective optimization model; generating multiple candidate scheduling strategies based on the adjustment parameters.

6. The resource scheduling method of a virtual power plant according to claim 5, wherein, The construction method of the multi-objective optimization model comprises: obtaining and labeling balance before energy data and balance after energy data corresponding to an energy system to construct training data; constructing an initial energy system balance model using a Transformer-like architecture, extracting and compressing key information and multi-dimensional relationships corresponding to the balance before energy data through an encoder, and converting the key information extracted by the encoder into a balanced energy system of a predetermined balance target through a decoder; Pre-training the initial energy system balance model according to the training data by using a self-supervised learning method to obtain deep features of the energy system; Supervised learning using training data of a specific project as labels to obtain an energy system balance model; Evaluating the energy system balance model and adjusting the energy system balance model according to the evaluation result.

7. The method of claim 1, wherein, The evaluation of the plurality of candidate scheduling strategies to obtain corresponding evaluation results comprises: Respectively evaluating each of the candidate scheduling strategies by using a simulation environment to evaluate machine learning evaluation indicators and multi-dimensional benefit indicators, to obtain machine learning evaluation results and multi-dimensional benefit evaluation results; Based on the machine learning evaluation results and the multi-dimensional benefit evaluation results, determine the comprehensive evaluation result of each of the candidate scheduling strategies.

8. A resource scheduling apparatus of a virtual power plant, characterized by, Comprise: A data acquisition module for acquiring power grid data of a source network load storage scene and preprocessing the power grid data; An operation prediction module for predicting the operation based on the preprocessed power grid data by using a generative AI large model to obtain a prediction result of the source network load storage scene in a future preset time period; A strategy generation module for generating a plurality of candidate scheduling strategies based on the prediction result by using a multi-objective optimization model; A strategy evaluation module for evaluating the plurality of candidate scheduling strategies to obtain corresponding evaluation results; A strategy execution module for determining a target scheduling strategy based on the evaluation results and executing a resource scheduling task based on the target scheduling strategy.

9. An electronic device, comprising: Comprise: A memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the resource scheduling method of the virtual power plant according to any one of claims 1-7.

10. A storage medium, characterized by A computer program capable of being loaded and executed by the processor to perform the resource scheduling method of the virtual power plant according to any one of claims 1-7 is stored.

Images