Marketing analysis method and system based on medium and long term electricity transaction

Abstract

The invention discloses a marketing analysis method and system based on medium-and-long-term electricity transaction. The method comprises the following steps: acquiring historical electricity transaction related data of medium-and-long-term electricity transaction business; constructing a medium and long term electricity transaction marketing system; respectively extracting feature data of the historical power transaction related data, and training a medium and long term power transaction large model and a prediction model by using the corresponding feature data; inputting the historical power transaction related data into a medium and long term power transaction marketing system, outputting a prediction result of the medium and long term power transaction by using the trained prediction model, and inputting the prediction result and the historical power transaction related data into the trained medium and long term power transaction large model to output a decision transaction scheme; and optimizing the medium and long term power transaction large model according to the decision transaction scheme and the prediction result so as to optimize the decision transaction scheme based on the optimized model. According to the invention, the medium-and-long-term power transaction decision-making difficulty is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power market analysis, and in particular to a marketing analysis method and system based on medium- and long-term power transactions. Background Art

[0002] In the power market, medium- and long-term power transactions are of great significance to both power suppliers and users. However, traditional power trading marketing analysis methods often have problems such as low data processing efficiency, insufficient analysis depth, and difficulty in accurately predicting market demand. With the development of artificial intelligence technology, especially the advancement of large model technology, new ideas are provided to solve these problems.

[0003] 1. Data quality and accuracy issues: Traditional medium- and long-term power trading marketing analysis methods rely on a large amount of power trading data. If the data is wrong, missing or inaccurate, it may lead to deviations in the analysis results. For example, inaccurate power consumption data provided by some users will affect the prediction of the overall power demand in the region.

[0004] 2. Model limitations: The analytical models and algorithms used in traditional marketing analysis methods may have certain limitations and cannot fully and accurately reflect the complex and changing electricity market situation. For example, certain emergencies or sudden policy adjustments may cause the model prediction to fail. Changes in external factors such as the macroeconomic situation, energy policies, and climate change may have a significant impact on electricity trading, but the system may find it difficult to adapt to these changes quickly and comprehensively. For example, new environmental protection policies require a significant reduction in coal-fired power generation, and the system may fail to adjust the analysis and forecast of the power supply structure in a timely manner. Summary of the invention

[0005] The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

[0006] To this end, the present invention proposes a marketing analysis method based on medium- and long-term electricity transactions. The marketing analysis method of medium- and long-term electricity transactions is mainly based on the business characteristics of medium- and long-term electricity transactions, combined with large model technology, to construct a medium- and long-term electricity transaction analysis model with adaptive learning capabilities, and continuously analyzes and feedbacks the output results after use, and continuously optimizes the model to solve various data problems.

[0007] Another object of the present invention is to propose a marketing analysis system based on medium- and long-term electricity transactions.

[0008] To achieve the above-mentioned purpose, the present invention proposes a marketing analysis method based on medium- and long-term power transactions, comprising:

[0009] Obtain historical power trading related data for medium- and long-term power trading business;

[0010] Constructing a medium- and long-term power trading and marketing system; the medium- and long-term power trading and marketing system at least includes a medium- and long-term power trading large model and a prediction model;

[0011] Extracting characteristic data of historical power transaction related data respectively, and using the corresponding characteristic data to train the medium- and long-term power transaction large model and the prediction model, so as to obtain a trained medium- and long-term power transaction large model and a trained prediction model;

[0012] Inputting historical power transaction related data into the medium- and long-term power transaction marketing system, using the trained prediction model to output the prediction results of medium- and long-term power transactions, and inputting the prediction results and the historical power transaction related data into the trained medium- and long-term power transaction large model to output a decision-making transaction plan; wherein the prediction results include medium- and long-term load prediction results and electricity price prediction results;

[0013] The medium- and long-term electricity trading large model is optimized according to the decision-making trading plan and the prediction results to optimize the decision-making trading plan based on the optimized model.

[0014] The marketing analysis method based on medium- and long-term power transactions in the embodiment of the present invention may also have the following additional technical features:

[0015] In one embodiment of the present invention, feature data of historical power transaction related data is extracted, and the medium- and long-term power transaction large model is trained using the corresponding feature data, including:

[0016] Obtaining historical power transaction data from historical power transaction related data;

[0017] Preprocessing and feature extracting the historical power transaction data to obtain preprocessed transaction data;

[0018] Use the Transformer model to build a large model for medium- and long-term power trading, and set the Transformer model parameters and training hyperparameters;

[0019] Dividing the preprocessed transaction data into a training data set and a test data set according to a preset ratio;

[0020] The input dimension and output dimension of the Transformer model are set, the Transformer model is trained using the training data set, and the trained Transformer model is tested using the test data set to obtain a trained medium- and long-term power trading model, so as to perform data audit on the real historical power trading data and eliminate erroneous data in the real-time power trading data.

[0021] In one embodiment of the present invention, the prediction result of future electricity market price and the historical electricity transaction related data are input into the trained medium- and long-term electricity transaction large model to output a decision-making transaction plan, including:

[0022] Audit and revise historical power transaction-related data;

[0023] Input the audited and corrected historical power transaction related data into the trained medium- and long-term power transaction large model to output the historical best decision-making transaction plan;

[0024] Comparatively analyzing the historical best decision-making trading plan and the historical decision-making trading plan, and outputting a difference report according to the comparative analysis results;

[0025] A real-time decision-making transaction plan is output based on the difference report and the prediction result.

[0026] In one embodiment of the present invention, the historical electricity trading data includes multiple types of historical electricity consumption data, historical electricity supply data, historical market price data, historical load forecast data, historical weather data and historical economic indicators.

[0027] In one embodiment of the present invention, after obtaining the trained medium- and long-term power trading large model, the method further includes:

[0028] Obtain future load forecast data, power user declaration data, and historical declaration data after data audit;

[0029] The future load forecast data, power user declaration data and historical declaration data after data audit are input into the trained medium- and long-term power trading large model to predict future declared electricity volume and declared price.

[0030] In one embodiment of the present invention, the prediction model is trained using corresponding feature data to obtain a trained prediction model, including:

[0031] Obtain medium- and long-term historical electricity trading data and related macroeconomic data;

[0032] Cleaning the medium- and long-term power transaction history data and related macroeconomic data, processing missing values, outliers and erroneous data, and stabilizing the processed time series data to extract time series features;

[0033] Constructing a prediction model; the prediction model includes a deep learning network model based on a multi-layer perceptron;

[0034] The time series features are divided into a training set, a validation set and a test set, and the parameters of the deep learning network model are adjusted by a stochastic gradient descent optimization algorithm to train the deep learning network model to obtain a trained prediction model.

[0035] In one embodiment of the present invention, the trained prediction model is used to output the prediction results of medium- and long-term power transactions, including: obtaining historical power consumption data and external factors;

[0036] The historical electricity consumption data and external factors are input into a trained prediction model, and data analysis is performed on the historical electricity consumption data and external factors to predict future electricity demand and load changes.

[0037] In one embodiment of the present invention, the medium- and long-term power trading and marketing system further includes a large-screen display module; the method further includes:

[0038] The decision-making transaction plan and the forecast results of the future electricity market price are graphically displayed through a large-screen display module to obtain data visualization results.

[0039] In one embodiment of the present invention, the medium- and long-term power trading and marketing system further includes a file management module, a data management module and a settlement management module; the method further includes:

[0040] Use the archive management module to manage information on power users, power sales companies, and power sources;

[0041] Use the data management module to manage weather, holidays, power consumption of power users, reported power consumption, and power decomposition curve data;

[0042] Use the settlement management module to manage electricity confirmation sheets, electricity settlement sheets, settlement reports, and user billing functions; and,

[0043] The large-screen display module provides a graphical display of customer portraits and classifications, market trend analysis, load and demand forecasts, trading strategy effectiveness evaluation, and electricity consumption comparison data of electricity users.

[0044] To achieve the above object, the present invention further proposes a marketing analysis system based on medium- and long-term power transactions, comprising:

[0045] The power transaction data acquisition module is used to obtain historical power transaction related data of medium- and long-term power transaction business;

[0046] A trading and marketing system building module is used to build a medium- and long-term power trading and marketing system; the medium- and long-term power trading and marketing system at least includes a medium- and long-term power trading large model and a prediction model;

[0047] A system model training module, used to extract characteristic data of historical power transaction related data respectively, and use the corresponding characteristic data to train the medium- and long-term power transaction large model and the prediction model to obtain a trained medium- and long-term power transaction large model and a trained prediction model;

[0048] A trading decision plan output module is used to input historical power trading related data into the medium- and long-term power trading marketing system, output the prediction results of medium- and long-term power trading using the trained prediction model, and input the prediction results and the historical power trading related data into the trained medium- and long-term power trading large model to output a decision trading plan; wherein the prediction results include medium- and long-term load prediction results and electricity price prediction results;

[0049] The optimization analysis module is used to optimize the medium- and long-term power trading model according to the decision-making trading plan and the prediction results to optimize the decision-making trading plan based on the optimized model.

[0050] The marketing analysis method and system based on medium- and long-term power trading in the embodiments of the present invention can improve the quality of medium- and long-term power trading data through technical means of medium- and long-term power trading large model data cleaning, error correction and supplementation for data quality and accuracy issues, and can also continuously optimize and improve the model, overcoming the limitations of the model to a certain extent.

[0051] The beneficial effects of the present invention are:

[0052] 1. Collecting data through big model technology reduces the difficulty of data collection, improves data accuracy, and reduces the workload and data accuracy problems caused by manual collection. Data collection based on big models can automatically extract data from multiple data sources and automatically audit the data. For example, it can automatically extract weather, macro policy and other data, automatically identify erroneous reporting data submitted by electricity users and give correction suggestions.

[0053] 2. Through the large model technology, it can respond quickly to business changes. The traditional medium- and long-term power trading and marketing method requires programmers to modify the implementation of the system according to policy changes when certain emergencies or sudden policy adjustments occur, and it is impossible to respond quickly. The medium- and long-term power trading and marketing method based on the large model of the present invention only needs to train the model to adapt to such changes.

[0054] 3. This solution provides a large-screen display function to intuitively present data and analysis results, reducing the time decision makers spend on finding and integrating information, so that decisions can be made more quickly.

[0055] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0057] Figure 1 is a flow chart of a marketing analysis method based on medium- and long-term power transactions according to an embodiment of the present invention;

[0058] Figure 2 is an architecture diagram of a marketing analysis method based on medium- and long-term power transactions according to an embodiment of the present invention;

[0059] Figure 3 It is a structural diagram of a marketing analysis system based on medium- and long-term power transactions according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0061] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0062] The following describes a marketing analysis method and system based on medium- and long-term power transactions according to an embodiment of the present invention with reference to the accompanying drawings.

[0063] Figure 1 is a flow chart of a marketing analysis method based on medium- and long-term power transactions according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0064] S1, obtain historical power transaction related data of medium and long-term power transaction business;

[0065] S2, constructing a medium- and long-term power trading and marketing system; the medium- and long-term power trading and marketing system at least includes a medium- and long-term power trading large model and a prediction model;

[0066] S3, extracting characteristic data of historical power transaction related data respectively, and using the corresponding characteristic data to train the medium- and long-term power transaction large model and the prediction model to obtain a trained medium- and long-term power transaction large model and a trained prediction model;

[0067] S4, inputting historical power transaction related data into the medium- and long-term power transaction marketing system, using the trained prediction model to output the prediction results of medium- and long-term power transactions, and inputting the prediction results and the historical power transaction related data into the trained medium- and long-term power transaction large model to output a decision-making transaction plan; wherein the prediction results include medium- and long-term load prediction results and electricity price prediction results;

[0068] S5, optimizing the medium- and long-term power trading model according to the decision-making trading plan and the prediction results to optimize the decision-making trading plan based on the optimized model.

[0069] The marketing analysis method based on medium- and long-term power transactions in the embodiment of the present invention has an overall structure as follows: Figure 2 As shown, the overall architecture of the medium- and long-term electricity trading and marketing system of the present invention includes a data acquisition module and auxiliary decision-making module based on a large model, a prediction module based on artificial intelligence, a large-screen display module, etc.

[0070] In one embodiment of the present invention, it is first necessary to build a medium- and long-term electricity trading and marketing system to provide basic archive management, data management, electricity trading, settlement management, forecasting module, large-screen display and other modules. Among them, archive management is used to manage information of market entities such as electricity users, power sales companies, and power sources. The data management module is used to manage data such as weather, holidays, electricity consumption of electricity users, declared electricity, and electricity decomposition curves. The electricity trading module includes functions such as declaration plan, declaration management, and transaction result management. Settlement management includes functions such as electricity confirmation form, electricity settlement form, settlement report, and user bill. The forecasting module includes medium- and long-term load forecasting and electricity price forecasting functions. The large-screen display module provides graphic methods such as bar charts to display various medium- and long-term electricity trading data such as customer portraits and classifications, market trend analysis, load and demand forecasts, trading strategy effect evaluation, and electricity user electricity consumption comparison.

[0071] In one embodiment of the present invention, the medium- and long-term power trading big model includes obtaining the open source big model source code from the open source project or community and building a big model service. By associating multiple data sources, data such as weather, holidays, and macro policies can be obtained and synchronized to the data management module of the marketing system. The model can be trained based on historical data to automatically audit the declared data provided by power users and give correction suggestions, and automatically generate the best decision-making plan based on historical power consumption, forecast and other data.

[0072] The present invention obtains open source big model source code from open source projects or communities, builds a big model service, and obtains a big model based on medium- and long-term electricity trading business through training. A large amount of historical trading data, market information, user behavior data, etc. are used to train a big model that can understand and process medium- and long-term electricity trading business. The big model can process and analyze basic data, such as the user's electricity usage history, contract terms, market supply and demand conditions, etc., to provide data support for subsequent trading decisions. Based on the analysis results of the model, trading strategies and decision recommendations are automatically generated to help traders make more scientific and reasonable decisions.

[0073] Specifically, the training process of the medium- and long-term power trading model is as follows:

[0074] 1) Obtain open source big models from open source projects or communities, such as fastGPT, and build medium- and long-term power trading big model services. Conduct training in the medium- and long-term power trading business field based on the model to output more accurate information.

[0075] 2) Associate multiple data sources to collect a large amount of data related to the medium- and long-term electricity market, including but not limited to historical transaction data, electricity supply and demand data, weather data, policies and regulations, holidays, etc.

[0076] 3) Extract valuable features from the sorted data. These features should be able to reflect the key factors and potential laws of the medium- and long-term power market. For example, extract the power supply and demand ratio in different seasons, the impact of weather on power demand, etc.

[0077] 4) According to the characteristics of the medium- and long-term power trading model and the characteristics of the medium- and long-term power trading data, training is performed based on ChatGPT or similar large models. During the training process, a large amount of normal sample data is used to allow the model to learn the patterns and rules of medium- and long-term power trading data for subsequent applications.

[0078] For example, the model can be trained with a large amount of electricity consumption data from electricity users so that the model can identify erroneous electricity consumption data submitted by electricity users and automatically give correction suggestions. The model can be trained with a large amount of historical transaction data, electricity consumption, electricity prices and other data so that the model can output the declaration plan with the highest cost performance, and can also be compared with historical declaration plans to output a difference report. Finally, the decision-making transaction plan is optimized based on forecast and other data, providing a declaration plan for subsequent medium- and long-term electricity transactions.

[0079] 5) During the large model training process, the Adagrad optimization algorithm and the cross entropy loss function are used, and batch training is adopted to gradually adjust the model parameters to minimize the loss function and improve the accuracy of the model.

[0080] Preferably, the present invention may include the following steps for training a medium- and long-term power trading model based on power trading data:

[0081] 1) Collect electricity trading data for the past five years, including electricity consumption data, electricity supply data, market price data, load forecast data, weather data, economic indicators, etc.

[0082] 2) Preprocess and extract features from these data. For example: convert date and time into periodic features (e.g., hours in a day, days in a week, etc.), calculate the mean, maximum, and minimum values ​​of weather data, extract the rate of change of economic indicators, calculate the historical trend of electricity consumption and supply, etc.

[0083] 3) For medium and long-term electricity trading, the Transformer model is used.

[0084] 4) Divide the data collected in step 1, for example, use 95% of the data as the training set and the remaining 5% of the data as the test set.

[0085] 5) Set input dimensions, output dimensions, etc. For example, use historical declared electricity, load forecast data, weather, policies, etc. as features. Use future declaration plans as output dimensions.

[0086] 6) Set Transformer model parameters, training hyperparameters, etc.

[0087] Furthermore, after the training of the medium- and long-term power trading model is completed, it can be used to generate new declaration schemes. For example, given tomorrow's load forecast data, historical declaration data, power user declaration data, etc., they can be input into the model to obtain tomorrow's declared electricity and declared price.

[0088] In one embodiment of the present invention, a medium- to long-term electricity trading prediction model is trained.

[0089] It is understandable that artificial intelligence technologies such as machine learning and deep learning are used to develop modules that can predict market changes, such as load forecasting and electricity price forecasting. By analyzing historical electricity consumption data and external factors (such as weather, seasons, holidays, etc.), future electricity demand and load changes are predicted. Through the influence of factors such as market supply and demand, policy changes, and fuel costs, future electricity market prices are predicted. Combining the prediction results with trading strategies provides traders with real-time decision support, helping them avoid risks and seize market opportunities.

[0090] Specifically, based on the collected data, including weather and historical load, medium- and long-term electricity consumption forecasts and future electricity price forecasts are conducted for power users to provide decision support for medium- and long-term power transactions. The specific process of training the medium- and long-term power transaction prediction model is as follows:

[0091] a) Data collection and integration: Collect long-term and medium-term historical data on electricity transactions over the years, including transaction volume, transaction price, transaction subject information, electricity supply and demand conditions, etc. Integrate relevant macroeconomic data, such as GDP growth, industrial production index, inflation rate, etc. Incorporate information on changes in energy policies and regulations, and development plans for the power industry.

[0092] b) Data preprocessing and feature engineering: Clean the collected data and deal with missing values, outliers and erroneous data. Stabilize the time series data, for example, by differential operations. Extract valuable features from the original data, such as generating seasonal features based on time, generating classification features based on transaction subject attributes, etc.

[0093] c) Choose a suitable prediction model: Consider using deep learning models such as multi-layer perceptron (MLP), long short-term memory network (LSTM) combined with attention mechanism to handle the long-term dependencies of time series data.

[0094] d) Model training: Divide the preprocessed data into training set, validation set and test set. Use the training set to train the selected model and adjust the model parameters through optimization algorithms such as stochastic gradient descent. Use the validation set to search and tune hyperparameters, such as learning rate, number of network layers, number of neurons, etc.

[0095] In one embodiment of the present invention, the large-screen display module in the medium- and long-term electricity trading and marketing system of the present invention displays the characteristics and classifications of different customer groups through bar charts, pie charts and other graphical methods, helping marketers to better understand customer needs and behaviors. Use charts to display the overall trend of the market, including price fluctuations, changes in supply and demand, etc., to provide a macro perspective for trading decisions. Display the predicted load and demand data in a graphical way so that traders can intuitively understand the future direction of the market. By comparing the actual trading results with the forecasting strategy, the effectiveness and accuracy of the trading strategy are evaluated to provide a basis for strategy adjustment. Display the comparison of electricity consumption of different electricity users to help analyze users' electricity consumption behavior and potential optimization space.

[0096] In summary, the functions realized by the main models in the medium- and long-term power trading and marketing system of the present invention include:

[0097] Models are used to understand and process medium- and long-term power trading business. Forecast models are used to predict market changes, such as load and electricity prices. Large-screen display modules are used to graphically display data and analysis results.

[0098] The large model requires a large amount of historical transaction data, market information, user behavior data, etc. to train the model. The prediction model requires many years of historical data on power transactions, including transaction volume, price, supply and demand conditions, etc., as well as macroeconomic data and policy change information. Large screen display module: The data and analysis results obtained from the large model and prediction model are required for visual display.

[0099] The large model cleans the collected data, handles missing values, outliers, and erroneous data, and extracts valuable features. The prediction model stabilizes the time series data, extracts seasonal features and classification features, etc. The large screen display module does not need to directly participate in preprocessing, but will use preprocessed data for display.

[0100] The big model uses the open source big model source code and obtains the big model based on the medium- and long-term power trading business through training. The prediction model selects a suitable deep learning model, such as Transformer or LSTM, for training and parameter tuning. The large screen display module does not directly participate in model training, but will display the training results.

[0101] The big model provides basic data analysis and decision support, such as automatically auditing the declared data provided by power users and giving correction suggestions. The forecasting model predicts future power demand and load changes, as well as market prices based on historical data and external factors. The large-screen display module displays forecast results and market trends, providing real-time decision support for traders.

[0102] The large model will display the analysis results and decision suggestions graphically through the large screen display module. The forecasting model will display the predicted load and demand data through the large screen display module. The large screen display module is directly responsible for displaying the output of the large model and the forecasting model to the user in a graphical manner, such as bar charts, pie charts, etc.

[0103] In practical applications, electricity trading-related data are collected in a unified manner; data are cleaned and features are extracted to prepare for model training; pre-processed data are used to train large models to provide basic analysis and decision support; pre-processed data are used to train prediction models to predict market changes; the output results of the large model and the prediction model are graphically displayed through the large-screen display module to assist decision-making.

[0104] According to the marketing analysis method based on medium- and long-term power transactions in accordance with the embodiment of the present invention, the difficulty of medium- and long-term power transaction decision-making is greatly reduced. At the same time, the machine learning algorithm is used to identify and correct outliers and errors in the data, and the missing data is supplemented by associating other reliable data sources. The model can also be better adapted to new power market conditions and emergencies by continuously updating data and retraining. In view of the lack of adaptability to changes in external factors, the large model can enhance the perception and analysis capabilities of external factors by integrating multi-source data, including macroeconomic indicators, policy and regulatory information, climate change data, etc., thereby improving the adaptability of the system.

[0105] In order to implement the above embodiment, Figure 3 As shown, this embodiment also provides a marketing analysis system 10 based on medium- and long-term power transactions, including:

[0106] The power transaction data acquisition module 100 is used to acquire historical power transaction related data of medium- and long-term power transaction business;

[0107] The trading and marketing system construction module 200 is used to construct a medium- and long-term power trading and marketing system; the medium- and long-term power trading and marketing system at least includes a medium- and long-term power trading large model and a prediction model;

[0108] The system model training module 300 is used to extract the characteristic data of the historical power transaction related data respectively, and use the corresponding characteristic data to train the medium- and long-term power transaction large model and the prediction model to obtain the trained medium- and long-term power transaction large model and the trained prediction model;

[0109] The trading decision plan output module 400 is used to input the historical power trading related data into the medium- and long-term power trading marketing system, output the prediction results of the medium- and long-term power trading using the trained prediction model, and input the prediction results and the historical power trading related data into the trained medium- and long-term power trading large model to output the decision trading plan; wherein the prediction results include the medium- and long-term load prediction results and the electricity price prediction results;

[0110] The optimization analysis module 500 is used to optimize the medium- and long-term power trading model according to the decision-making trading plan and the prediction results to optimize the decision-making trading plan based on the optimized model.

[0111] According to the marketing analysis system based on medium- and long-term power transactions according to the embodiment of the present invention, the difficulty of medium- and long-term power transaction decision-making is greatly reduced. At the same time, machine learning algorithms are used to identify and correct outliers and errors in the data. At the same time, missing data is supplemented by associating other reliable data sources. The model can also be better adapted to new power market conditions and emergencies by continuously updating data and retraining. In view of the lack of adaptability to changes in external factors, the large model can enhance the perception and analysis capabilities of external factors by integrating multi-source data, including macroeconomic indicators, policy and regulatory information, climate change data, etc., thereby improving the adaptability of the system.

[0112] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0113] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

Claims

1. A marketing analysis method based on medium- and long-term power transactions, characterized in that: include: Obtain historical power trading related data for medium- and long-term power trading business; Constructing a medium- and long-term power trading and marketing system; the medium- and long-term power trading and marketing system at least includes a medium- and long-term power trading large model and a prediction model; Extracting characteristic data of historical power transaction related data respectively, and using the corresponding characteristic data to train the medium- and long-term power transaction large model and the prediction model, so as to obtain a trained medium- and long-term power transaction large model and a trained prediction model; Inputting historical power transaction related data into the medium- and long-term power transaction marketing system, using the trained prediction model to output the prediction results of medium- and long-term power transactions, and inputting the prediction results and the historical power transaction related data into the trained medium- and long-term power transaction large model to output a decision-making transaction plan; wherein the prediction results include medium- and long-term load prediction results and electricity price prediction results; The medium- and long-term electricity trading large model is optimized according to the decision-making trading plan and the prediction results to optimize the decision-making trading plan based on the optimized model.

2. The method according to claim 1, characterized in that Extracting characteristic data of historical power transaction related data, and using the corresponding characteristic data to train the medium- and long-term power transaction large model, including: Obtaining historical power transaction data from historical power transaction related data; Preprocessing and feature extracting the historical power transaction data to obtain preprocessed transaction data; Use the Transformer model to build a large model for medium- and long-term power trading, and set the Transformer model parameters and training hyperparameters; Dividing the preprocessed transaction data into a training data set and a test data set according to a preset ratio; The input dimension and output dimension of the Transformer model are set, the Transformer model is trained using the training data set, and the trained Transformer model is tested using the test data set to obtain a trained medium- and long-term power trading model, so as to perform data audit on the real historical power trading data and eliminate erroneous data in the real-time power trading data.

3. The method according to claim 1, characterized in that The prediction results of future electricity market prices and the historical electricity transaction related data are input into the trained medium- and long-term electricity transaction model to output a decision-making transaction plan, including: Audit and revise historical power transaction-related data; Input the audited and corrected historical power transaction related data into the trained medium- and long-term power transaction large model to output the historical best decision-making transaction plan; Comparatively analyzing the historical best decision-making trading plan and the historical decision-making trading plan, and outputting a difference report according to the comparative analysis results; A real-time decision-making transaction plan is output based on the difference report and the prediction result.

4. The method according to claim 2, characterized in that: The historical electricity trading data includes multiple ones of historical electricity consumption data, historical electricity supply data, historical market price data, historical load forecast data, historical weather data and historical economic indicators.

5. The method according to claim 2, characterized in that: After obtaining the trained medium- and long-term power trading model, the method further includes: Obtain future load forecast data, power user declaration data, and historical declaration data after data audit; The future load forecast data, power user declaration data and historical declaration data after data audit are input into the trained medium- and long-term power trading large model to predict future declared electricity volume and declared price.

6. The method according to claim 1, characterized in that The prediction model is trained using the corresponding feature data to obtain a trained prediction model, including: Obtain medium- and long-term electricity transaction historical data and related macroeconomic data; Cleaning the medium- and long-term power transaction history data and related macroeconomic data, processing missing values, outliers and erroneous data, and stabilizing the processed time series data to extract time series features; Constructing a prediction model; the prediction model includes a deep learning network model based on a multi-layer perceptron; The time series features are divided into a training set, a validation set and a test set, and the parameters of the deep learning network model are adjusted by a stochastic gradient descent optimization algorithm to train the deep learning network model to obtain a trained prediction model.

7. The method according to claim 6, characterized in that Use the trained prediction model to output the prediction results of medium- and long-term power transactions, including: Obtain historical electricity consumption data and external factors; The historical electricity consumption data and external factors are input into a trained prediction model, and data analysis is performed on the historical electricity consumption data and external factors to predict future electricity demand and load changes.

8. The method according to claim 1, characterized in that: The medium- and long-term power trading and marketing system further includes a large-screen display module; the method further includes: The decision-making transaction plan and the forecast results of the future electricity market price are graphically displayed through a large-screen display module to obtain data visualization results.

9. The method according to claim 8, characterized in that The medium- and long-term power trading and marketing system further includes a file management module, a data management module and a settlement management module; the method further includes: Use the archive management module to manage information on power users, power sales companies, and power sources; Use the data management module to manage weather, holidays, power consumption of power users, reported power consumption, and power decomposition curve data; Use the settlement management module to manage electricity confirmation sheets, electricity settlement sheets, settlement reports, and user billing functions; and, The large-screen display module provides a graphical display of customer portraits and classifications, market trend analysis, load and demand forecasts, trading strategy effectiveness evaluation, and electricity consumption comparison data of electricity users.

10. A marketing analysis system based on medium and long-term power transactions, characterized in that: include: The power transaction data acquisition module is used to obtain historical power transaction related data of medium- and long-term power transaction business; A trading and marketing system building module is used to build a medium- and long-term power trading and marketing system; the medium- and long-term power trading and marketing system at least includes a medium- and long-term power trading large model and a prediction model; A system model training module, used to extract characteristic data of historical power transaction related data respectively, and use the corresponding characteristic data to train the medium- and long-term power transaction large model and the prediction model to obtain a trained medium- and long-term power transaction large model and a trained prediction model; A trading decision plan output module is used to input historical power trading related data into the medium- and long-term power trading marketing system, output the prediction results of medium- and long-term power trading using the trained prediction model, and input the prediction results and the historical power trading related data into the trained medium- and long-term power trading large model to output a decision trading plan; wherein the prediction results include medium- and long-term load prediction results and electricity price prediction results; The optimization analysis module is used to optimize the medium- and long-term power trading model according to the decision-making trading plan and the prediction results to optimize the decision-making trading plan based on the optimized model.

Images