Power transaction auxiliary decision-making management method, device and equipment and storage medium

Abstract

The invention is applicable to the technical field of power transaction and energy management, and provides a power transaction aid decision management method, which comprises the following steps of: analyzing and predicting a power market environment through an energy large model according to power transaction environment information to obtain an energy prediction result, and performing decision management on the power market environment on the basis of the energy prediction result. The energy scheduling strategy is generated for the electricity market environment through the multi-target optimization model, and the electricity transaction in the electricity market environment is scheduled according to the energy scheduling strategy, so that the decision-making balance and comprehensiveness are improved, the method can better adapt to the complex scenes such as large-scale access of new energy and frequent change of electricity market rules, and the user experience is improved. And the market risk is reduced, so that the method has better adaptability and application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of power trading and energy management, and particularly relates to a power trading auxiliary decision-making management method, device, equipment and storage medium. Background Art

[0002] With the continuous development of the power system, power trading faces more and more challenges: First, the uncertainty of new energy is a major problem. New energy power generation has significant instability, which brings great obstacles to power trading. For example, solar power generation depends on light intensity and duration, and wind power generation depends on wind speed and direction. The uncontrollability of these natural factors makes it difficult to accurately predict the output of new energy power generation, resulting in many uncertainties in power trading in aspects such as power balance and price setting; Second, the complexity of the power market cannot be ignored. Currently, the participants in the power market are becoming increasingly diversified, covering power generation enterprises, grid enterprises, power sales companies, power users, and various emerging market players. At the same time, the trading rules are constantly updated and changed to adapt to the market development needs. This situation greatly increases the complexity of trading decisions, and it is difficult for market participants to quickly and accurately grasp the market dynamics and make reasonable decisions; Third, power trading has extremely high requirements for real-time performance. The operating characteristics of the power system determine that power trading needs to quickly respond to market changes. Whether it is the instantaneous change in power supply and demand or the real-time fluctuation of electricity prices, it requires the trading decision-making system to make timely responses, which poses unprecedented high requirements for the real-time performance of the trading system.

[0003] Traditional power trading decision-making methods mainly rely on expert experience, historical data analysis, and simple statistical models. However, these methods have obvious limitations. Their adaptability is poor and it is difficult to keep up with the rapid changes in the power market. Facing the large-scale access of new energy and the frequent adjustment of power market rules, traditional models are difficult to effectively respond and make accurate decisions. Moreover, the information utilization rate is low. Usually only part of the historical data is used, and the potential value hidden in the massive data cannot be fully explored, resulting in the lack of comprehensiveness and foresight in decision-making. Furthermore, the problem of decision-making lag is prominent. The traditional decision-making process often has a long time delay and cannot capture and respond to the rapid changes in the market in a timely manner, making market participants passive in trading. Although existing patented technologies (such as CN118333429A, CN118552058A, CN118367555A) introduce technologies such as machine learning and blockchain, there are still some deficiencies. First, the model is single: relying on a single prediction or optimization model, it is difficult to capture the complexity of the power system. Second, the optimization goal is single: usually only focusing on maximizing economic benefits and ignoring goals such as system stability and renewable energy utilization rate, which has obvious limitations in the context of sustainable development and ensuring the reliable operation of the power system. Third, the data is not fully utilized: multi-source heterogeneous data (such as meteorology, equipment status, market rules) is not fully integrated, and the huge value of massive data in improving the accuracy and scientific nature of trading decisions is not fully explored, and the rich data resources cannot be transformed into effective decision support. Summary of the Invention

[0004] The purpose of the present invention is to provide a power trading auxiliary decision-making management method, device, equipment, and storage medium, aiming to solve the problems of single power trading decision-making, poor balance, and difficulty in adapting to complex power trading markets caused by existing technologies.

[0005] On the one hand, the present invention provides a power trading auxiliary decision-making management method, and the method includes the following steps:

[0006] According to the power trading environment information, analyze and predict the power market environment through an energy large model to obtain an energy prediction result;

[0007] Based on the energy prediction result, generate an energy scheduling strategy for the power market environment through a multi-objective optimization model;

[0008] Schedule the power trading in the power market environment according to the energy scheduling strategy.

[0009] Preferably, after the step of scheduling the power trading in the power market environment according to the energy scheduling strategy, the method further includes:

[0010] Generate the operation curves of each power trading node in the power market environment under the energy scheduling strategy, and declare the operation curves to the power trading center.

[0011] Preferably, the multi-objective optimization model includes an encoding module and a decoding module. The step of generating an energy scheduling strategy for the power market environment through the multi-objective optimization model based on the energy prediction result includes:

[0012] Extract the multi-dimensional correlation features between the energy prediction result and the pre-acquired multi-source data through the encoding module;

[0013] Generate the energy scheduling strategy through the decoding module based on the multi-dimensional correlation features.

[0014] Preferably, before the step of generating an energy scheduling strategy for the power market environment through the multi-objective optimization model based on the energy prediction result, the method further includes:

[0015] Construct the multi-objective optimization model with the goals of maximizing economic benefits, maximizing system stability, and maximizing the utilization rate of renewable energy by adopting an architecture similar to Transformer;

[0016] Iteratively train the multi-objective optimization model by adopting a preset multi-stage training strategy.

[0017] Preferably, the step of iteratively training the multi-objective optimization model by adopting a preset multi-stage training strategy includes:

[0018] Perform pre-training of self-supervised learning on the multi-objective optimization model by using the pre-acquired training data to capture the deep features related to power trading;

[0019] Perform fine-tuning training of supervised learning on the pre-trained multi-objective optimization model by using specific scenario annotation data to optimize the prediction accuracy of the model for the target power market.

[0020] On the other hand, the present invention provides a power trading auxiliary decision-making management device, and the device includes:

[0021] A trading prediction unit, configured to analyze and predict the power market environment through an energy large model according to power trading environment information to obtain an energy prediction result;

[0022] A strategy generation unit, configured to generate an energy scheduling strategy for the power market environment through a multi-objective optimization model based on the energy prediction result;

[0023] An energy scheduling unit, configured to schedule the power trading in the power market environment according to the energy scheduling strategy.

[0024] Preferably, the device further includes:

[0025] An information reporting unit, configured to generate an operation curve of each power trading node in the power market environment under the energy scheduling strategy, and report the operation curve to the power trading center.

[0026] Preferably, the multi-objective optimization model includes an encoding module and a decoding module, and the strategy generation unit includes:

[0027] A feature extraction unit, configured to extract multi-dimensional correlation features between the energy prediction result and pre-acquired multi-source data through the encoding module;

[0028] A strategy generation subunit, configured to generate the energy scheduling strategy based on the multi-dimensional correlation features through the decoding module.

[0029] On the other hand, the present invention also provides a power trading management device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps described in the above-mentioned power trading auxiliary decision-making management method are implemented.

[0030] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps described in the above-mentioned power trading auxiliary decision-making management method are implemented.

[0031] According to the power trading environment information, the present invention analyzes and predicts the power market environment through an energy large model to obtain an energy prediction result. Based on the energy prediction result, a multi-objective optimization model generates an energy scheduling strategy for the power market environment, and schedules the power trading in the power market environment according to the energy scheduling strategy, thereby improving the decision-making balance and comprehensiveness, and being able to better adapt to complex scenarios such as large-scale access of new energy and frequent changes in power market rules, reducing market risks, and thus having better adaptability and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flowchart of the implementation of the power trading auxiliary decision-making management method provided in Embodiment 1 of the present invention;

[0033] Figure 2 is a schematic structural diagram of the power trading auxiliary decision-making management device provided in Embodiment 2 of the present invention;

[0034] Figure 3 is a schematic diagram of the preferred structure of the power trading auxiliary decision-making management device provided in Embodiment 2 of the present invention;

[0035] Figure 4 It is a schematic structural diagram of the power trading management device provided in the third embodiment of the present invention. Specific implementation manners

[0036] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0037] The following describes the specific implementation of the present invention in detail with reference to specific embodiments:

[0038] Example 1:

[0039] Figure 1 The implementation process of the power trading auxiliary decision-making management method provided in the first embodiment of the present invention is shown. For the sake of convenience of description, only the parts related to the embodiments of the present invention are shown and are described in detail as follows:

[0040] In step S101, according to the power trading environment information, the power market environment is analyzed and predicted through the energy large model to obtain an energy prediction result.

[0041] In the embodiments of the present invention, it is necessary to analyze and predict the power market environment with the help of the energy large model according to the power trading environment information, and then obtain an energy prediction result. Among them, the power trading environment information includes multi-source data such as power market data, meteorological data and policy and regulation information. Specifically, first, collect the power trading environment information: collect the current real-time power market data (such as power price) in the power market environment, the historical power market data within a preset time period (such as the recent one month), and the latest policy and regulation information related to the power market. At the same time, obtain the meteorological data in the same period (such as temperature, humidity, wind speed, etc.) for analyzing its influence on the power generation side and the power consumption side. Then, process the power trading environment information: including but not limited to data cleaning to remove outliers in the data, filling in missing values in the data, and data format conversion (such as unified text format processing for policy and regulation information). Finally, input the processed power trading environment information into the trained energy large model, and use the energy large model to predict the supply and demand situation of the power market within a preset future time (such as the power price and its trend within the next 24 hours) to obtain an energy prediction result, so as to provide a basis for formulating an energy scheduling strategy in the future. For example, the energy scheduling strategy will specifically determine how much power to purchase from the power market according to the predicted power price, and adjust the load situation of the rest to ensure that the specific energy consumption requirements can be met.

[0042] In a feasible embodiment, the energy large model adopts a machine learning or deep learning architecture, and through learning from multi-source data such as massive power market data and meteorological data, it realizes accurate prediction of energy trends (such as electricity prices).

[0043] In yet another feasible embodiment, the energy large model outputs energy prediction results within a preset future time period in a sliding window recursive manner.

[0044] In step S102, based on the energy prediction results, an energy scheduling strategy is generated for the power market environment through a multi-objective optimization model.

[0045] In the embodiment of the present invention, based on the energy prediction results, combined with the real-time status data of the edge devices, an energy scheduling strategy is generated for the power market environment through a multi-objective optimization model, where the multi-objective optimization model includes an encoding module and a decoding module.

[0046] In a feasible embodiment, before generating an energy scheduling strategy for the power market environment through the multi-objective optimization model, the training of the multi-objective optimization model is achieved through the following steps:

[0047] (1) A multi-objective optimization model with the goals of maximizing economic benefits, maximizing system stability, and maximizing the utilization rate of renewable energy is constructed using a Transformer-like architecture;

[0048] In the embodiment of the present invention, an encoder-decoder structure using a Transformer-like architecture is adopted to construct a multi-objective optimization model. At the same time, the objective function of the multi-objective optimization model covers maximizing economic benefits, maximizing system stability, and maximizing the utilization rate of renewable energy. Compared with traditional single-objective optimization methods, it realizes a more comprehensive and balanced decision-making, significantly improves the quality of decision-making, and ensures the sustainable development of the power system.

[0049] (2) The multi-objective optimization model is iteratively trained using a preset multi-stage training strategy.

[0050] In the embodiment of the present invention, using the PyTorch framework and combined with distributed training technology, the multi-objective optimization model is iteratively trained through a multi-stage training strategy that combines self-supervised learning and supervised learning to achieve the dynamic balance and strategy generation of the energy system.

[0051] When iteratively training the multi-objective optimization model using a preset multi-stage training strategy, preferably, the multi-objective optimization model is first pre-trained through self-supervised learning using pre-acquired training data to capture deep features related to power trading, and then the pre-trained multi-objective optimization model is fine-tuned through supervised learning using specific scenario annotation data to optimize the prediction accuracy of the model for the target power market.

[0052] In an embodiment of the present invention, a large amount of previously labeled energy data (including energy data before balancing (such as energy consumption, new energy power generation, power trading, etc.) and energy data after balancing corrected by the power trading team) is used as the training data of the model. Using this training data, a multi-objective optimization model is pre-trained by a self-supervised learning method to capture deep features related to power trading. Then, supervised learning fine-tuning training is performed on the pre-trained multi-objective optimization model using specific scenario labeled data to optimize the prediction accuracy of the model for the target power market, so that it can adapt to the special requirements for the balance of the energy system in a specific scenario.

[0053] In another feasible embodiment, the generation of the energy scheduling strategy is realized through the following steps:

[0054] ① Extract multi-dimensional correlation features between the energy prediction results and the pre-acquired multi-source data through an encoding module;

[0055] In an embodiment of the present invention, the pre-acquired multi-source data includes but is not limited to end-side real-time status data (such as energy storage system SOC, adjustable load demand, new energy power generation power), meteorological data (temperature, wind speed, light intensity), historical transaction data, and market rule information, etc. Specifically, through devices such as intelligent gateway meters and secondary anti-backflow meters, the input and output loads of the power system are tracked in real time to ensure the real-time and accuracy of the data, and the real-time data of distributed energy facilities are collected, such as the output power of photovoltaic panels, the rotation speed and power output of wind turbines, the charging status and capacity of energy storage batteries, etc. Sensors and monitoring systems are used to monitor the health status and efficiency of the equipment in real time to ensure the accuracy and reliability of the data. Third-party meteorological services are accessed to obtain key meteorological parameters such as light, wind speed, and temperature in real time. Here, data preprocessing is first performed on the energy prediction results (such as future 24-hour electricity price, new energy power generation prediction) and multi-source data: standardize and normalize different types of data, eliminate dimension differences, and fill in missing values. Then, the preprocessed energy prediction results and multi-source data are input into the encoding module. After that, through the time series encoding layer of the encoding module, long-term dependencies of electricity price fluctuations and load cycle changes are captured. Through the graph neural network (GNN) of the encoding module, the topological relationship between distributed energy devices (such as photovoltaic, energy storage, load) is modeled to analyze the impact of energy flow between devices. Through the attention mechanism of the encoding module, the relevance of meteorological data to the power generation side (such as light affecting photovoltaic output) and the power consumption side (such as temperature affecting air conditioning load) is dynamically weighted to generate fused multi-dimensional correlation features.

[0056] ② Generate an energy scheduling strategy based on the multi-dimensional correlation features through a decoding module.

[0057] In the embodiments of the present invention, multi-dimensional associated features are input into a decoding module. Through the fully connected layer of the decoding module, they are mapped to a potential policy space (such as charge and discharge plans, trading volume intervals, load regulation ranges). An reinforcement learning framework is adopted, with economic benefits, system stability, and renewable energy utilization rate as reward functions, to guide the model to generate multiple sets of candidate policies that meet multi-objective optimization (such as energy storage charge and discharge schemes and power trading volume allocations at different time periods), and screen out candidate policies that achieve a balance among economic benefits, stability, and environmental friendliness. Here, the screened candidate policies can be input into a simulation environment to simulate the system state after execution (such as grid frequency, energy storage capacity, market revenue), verify their feasibility, correct the policies that violate the constraints (such as the energy storage charge and discharge rate exceeding the limit), and finally output executable candidate energy scheduling policies. Finally, the executable candidate energy scheduling policies are presented to the power trading decision-makers, and the decision-makers select the optimal energy scheduling policy according to the actual market situation and experience as the final energy scheduling policy for power trading scheduling.

[0058] The above steps ①② are implemented by an encoding module to fuse multi-source heterogeneous data and capture the complex dynamic associations of the power system, avoiding policy deviations caused by traditional models ignoring cross-dimensional relationships. In the decoding module, by combining generative AI and multi-objective optimization, the diversity and balance of policies are achieved, significantly enhancing the adaptability of the virtual power plant in a volatile market.

[0059] In step S103, the power transactions in the power market environment are scheduled according to the energy scheduling policy.

[0060] In the embodiments of the present invention, the energy scheduling policy is converted into corresponding power trading instructions, energy storage charge and discharge instructions, and load regulation instructions, which are respectively sent to the power trading side, the energy storage system side, and the adjustable load side. The power trading side conducts power trading according to the optimal power trading volume and market trading rules in the power trading instructions. The energy storage system side performs corresponding charge and discharge operations according to the charge and discharge scheme in the energy storage charge and discharge instructions. For example, it charges when the electricity price is low and discharges during peak load to adjust the power supply and demand balance. The adjustable load side executes the adjustment amount based on the load regulation instructions and conducts load regulation according to various control algorithms in the end-side control system, such as adjusting the electricity consumption time of the production equipment of industrial users, and feeds back the adjustment results to the system for the next round of balance optimization.

[0061] In a feasible embodiment, after scheduling the power transactions in the power market environment according to the energy scheduling policy, an operating curve of each power trading node in the power market environment under the energy scheduling policy is generated and reported to the power trading center.

[0062] In the embodiment of the present invention, an operation curve is generated and optimized according to the energy scheduling strategy to ensure compliance with system constraints and market rules. The optimized operation curve is converted into the format specified by the power trading center and the declaration is completed. Specifically, first, all relevant node data of each power trading node in the power market environment under the scheduling of the energy scheduling strategy is collected, including energy storage status, load demand, power generation capacity, etc. An initial operation curve for the corresponding power trading node is generated based on the collected node data. Then, the initial operation curve is smoothed to remove noise and abnormal fluctuations. At the same time, it is checked whether the smoothed curve meets the technical constraints such as voltage and frequency of the system, as well as market rules and trading requirements. If not, the curve is adjusted to meet the market rules and trading requirements during the trading period. In addition, the user can view the operation curve through the provided human-computer interaction interface and adjust the curve according to load, energy storage system or other actual business requirements. For example, the load curve is modified according to the temporary adjustment of the production plan. After the user confirms the adjusted curve, the operation curve is prepared for declaration. For this, first, the curve format required by the power trading center is identified, such as a specific data structure and file format. The optimized operation curve is converted according to the identified curve format to make it meet the format required by the power trading center and ensure the accuracy and integrity of the data. Then, in accordance with the trading rules of the power trading center, it is checked whether the converted curve is compliant (such as checking whether the data such as electricity quantity and price in the curve conforms to the trading rules). If any non-compliant places are found, necessary corrections are made. Finally, the finally compliant operation curve is declared to the power trading center to complete the declaration process of power trading.

[0063] In the embodiment of the present invention, according to the power trading environment information, the power market environment is analyzed and predicted through an energy large model to obtain an energy prediction result. Based on the energy prediction result, a multi-objective optimization model is used to generate an energy scheduling strategy for the power market environment, and the power trading in the power market environment is scheduled according to the energy scheduling strategy, thereby improving the decision-making balance and comprehensiveness, and being able to better adapt to complex scenarios such as the large-scale access of new energy and the frequent changes of power market rules, reducing market risks, and thus having better adaptability and application prospects.

[0064] Example 2:

[0065] Figure 2 The structure of the power trading auxiliary decision-making management device provided in the second embodiment of the present invention is shown. For the sake of convenience of description, only the parts related to the embodiment of the present invention are shown, including:

[0066] A trading prediction unit 21, configured to analyze and predict the power market environment through an energy large model according to the power trading environment information to obtain an energy prediction result;

[0067] A strategy generation unit 22, configured to generate an energy scheduling strategy for a power market environment through a multi-objective optimization model based on an energy prediction result;

[0068] An energy scheduling unit 23, configured to schedule power transactions in a power market environment according to the energy scheduling strategy.

[0069] Preferably, as Figure 3 shown, the power transaction auxiliary decision-making management device according to an embodiment of the present invention further includes:

[0070] An information reporting unit 24, configured to generate an operation curve of each power transaction node in a power market environment under an energy scheduling strategy, and report the operation curve to a power trading center.

[0071] Preferably, the multi-objective optimization model includes an encoding module and a decoding module.

[0072] Preferably, the strategy generation unit 22 includes:

[0073] A feature extraction unit 221, configured to extract multi-dimensional correlation features between an energy prediction result and pre-acquired multi-source data through an encoding module;

[0074] A strategy generation subunit 222, configured to generate an energy scheduling strategy based on the multi-dimensional correlation features through a decoding module.

[0075] In an embodiment of the present invention, each unit of the power transaction auxiliary decision-making management device may be implemented by a corresponding hardware or software unit. Each unit may be an independent software or hardware unit, or may be integrated into a software or hardware unit, which is not used to limit the present invention here. Specifically, the implementation manners of each unit may refer to the description of the foregoing Embodiment 1, and will not be elaborated here.

[0076] Example 3:

[0077] Figure 4 The structure of a power transaction management device provided in Embodiment 3 of the present invention is shown. For the sake of convenience of description, only the parts related to the embodiment of the present invention are shown.

[0078] The power transaction management device 4 according to an embodiment of the present invention includes a processor 40, a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. When the processor 40 executes the computer program 42, the steps in the foregoing embodiment of a power transaction auxiliary decision-making management method are implemented, such as Figure 1 the steps S101 to S103 shown. Alternatively, when the processor 40 executes the computer program 42, the functions of each unit in the foregoing device embodiments are implemented, such as Figure 2 the functions of the units shown.

[0079] In the embodiments of the present invention, according to the power trading environment information, the power market environment is analyzed and predicted through an energy large model to obtain an energy prediction result. Based on the energy prediction result, a multi-objective optimization model is used to generate an energy scheduling strategy for the power market environment, and the power trading in the power market environment is scheduled according to the energy scheduling strategy, thereby improving the decision-making balance and comprehensiveness, and being able to better adapt to complex scenarios such as the large-scale access of new energy and the frequent changes of power market rules, reducing market risks, and thus having better adaptability and application prospects.

[0080] The power trading management device in the embodiments of the present invention can be a personal computer or a server. When the processor 40 in the power trading management device 4 executes the computer program 42 to implement a power trading auxiliary decision-making management method, the steps implemented can refer to the description of the foregoing method embodiments and will not be elaborated here.

[0081] Example 4:

[0082] In the embodiments of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps in the foregoing embodiments of a power trading auxiliary decision-making management method. For example, Figure 1 the steps S101 to S103 shown. Or, when the computer program is executed by a processor, it implements the functions of each unit in the foregoing device embodiments. For example Figure 2 the functions of the units shown.

[0083] In the embodiments of the present invention, according to the power trading environment information, the power market environment is analyzed and predicted through an energy large model to obtain an energy prediction result. Based on the energy prediction result, a multi-objective optimization model is used to generate an energy scheduling strategy for the power market environment, and the power trading in the power market environment is scheduled according to the energy scheduling strategy, thereby improving the decision-making balance and comprehensiveness, and being able to better adapt to complex scenarios such as the large-scale access of new energy and the frequent changes of power market rules, reducing market risks, and thus having better adaptability and application prospects.

[0084] The computer-readable storage medium in the embodiments of the present invention may include any entity or device, recording medium that can carry computer program code, such as memories such as ROM / RAM, disks, optical discs, flash memories, etc.

[0085] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A power trading auxiliary decision management method, characterized in that: The method comprises the following steps: According to the power trading environment information, the power market environment is analyzed and predicted through the energy big model to obtain energy forecast results; Based on the energy forecast results, generating an energy dispatch strategy for the power market environment through a multi-objective optimization model; Electricity transactions in the electricity market environment are scheduled according to the energy scheduling strategy.

2. The method according to claim 1, characterized in that After the step of scheduling power transactions in the power market environment according to the energy scheduling strategy, the method further includes: Generate an operation curve of each power trading node in the power market environment under the energy scheduling strategy, and report the operation curve to the power trading center.

3. The method according to claim 1, characterized in that The multi-objective optimization model includes an encoding module and a decoding module. Based on the energy forecast result, the step of generating an energy dispatch strategy for the power market environment through the multi-objective optimization model includes: Extracting multi-dimensional correlation features between the energy prediction results and pre-acquired multi-source data through the encoding module; The energy scheduling strategy is generated based on the multi-dimensional correlation features through the decoding module.

4. The method according to claim 3, characterized in that Before the step of generating an energy dispatch strategy for the power market environment through a multi-objective optimization model based on the energy forecast result, the method further includes: A Transformer-like architecture is used to construct the multi-objective optimization model with the goals of maximizing economic benefits, maximizing system stability, and maximizing renewable energy utilization; The multi-objective optimization model is iteratively trained using a preset multi-stage training strategy.

5. The method according to claim 4, characterized in that The step of iteratively training the multi-objective optimization model using a preset multi-stage training strategy comprises: Pre-training the multi-objective optimization model by self-supervised learning using pre-acquired training data to capture deep features related to power trading; The pre-trained multi-objective optimization model is fine-tuned by supervised learning using specific scenario labeled data to optimize the model's prediction accuracy for the target power market.

6. An auxiliary decision-making management device for power trading, characterized in that: The device comprises: The transaction prediction unit is used to analyze and predict the power market environment through the energy big model according to the power transaction environment information to obtain the energy prediction result; A strategy generation unit, configured to generate an energy dispatch strategy for the power market environment through a multi-objective optimization model based on the energy forecast result; An energy scheduling unit is used to schedule power transactions in the power market environment according to the energy scheduling strategy.

7. The device according to claim 6, characterized in that The device also includes: The information reporting unit is used to generate an operation curve of each power trading node in the power market environment under the energy scheduling strategy, and report the operation curve to the power trading center.

8. The device according to claim 6, characterized in that The multi-objective optimization model includes an encoding module and a decoding module, and the strategy generation unit includes: A feature extraction unit, configured to extract multi-dimensional correlation features between the energy prediction result and pre-acquired multi-source data through the encoding module; A strategy generation subunit is used to generate the energy scheduling strategy based on the multi-dimensional correlation features through the decoding module.

9. An electric power trading management device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images