New energy annual transaction strategy optimization method and system based on supply and demand and game

Abstract

This application proposes a method and system for optimizing annual trading strategies for new energy based on supply and demand and game theory. The method includes: using a historical typical database; using the market supply and demand data of the target year, outputting the predicted spot price value through a spot price prediction model; using the predicted spot price value, outputting the predicted transaction ratio value through an annual centralized bidding transaction ratio prediction model; constructing and training a preliminary decision-making model for annual bilateral trading to generate a preliminary trading strategy set; inputting the preliminary trading strategy set, the predicted spot price value, and the predicted transaction ratio value into a comprehensive optimization model, and solving it using a differential evolution algorithm to obtain the annual trading strategy; verifying and optimizing the annual trading strategy using the historical typical database, and outputting the optimal annual trading strategy. The technical solution proposed in this application realizes the quantitative analysis of market supply and demand and game theory behavior, and can effectively improve the comprehensive returns and strategy robustness of new energy power plants in annual electricity market trading.

Description

Technical Field

[0001] This application relates to the fields of electricity market trading and new energy power generation technology, and in particular to a method and system for optimizing annual trading strategies for new energy based on supply and demand and game theory. Background Technology

[0002] With the introduction of dual-carbon targets, the scale of new energy in China has continued to expand. However, various provinces have insufficient capacity for new energy consumption and power supply, which has led to a gradual decline in the average settlement price of new energy in the spot market, making it difficult to meet the investment returns of new energy and adversely affecting the development of the new energy industry.

[0003] Under the unified power market system, the spot markets in various provinces are gradually advancing. With the large-scale commissioning of new energy, the inter-provincial trading market prices during the peak power generation period of new energy are also showing a gradual decline. Under the spot market model, the medium- and long-term trading of new energy, especially the annual trading, has strong risk resistance capabilities and has become an important decision-making stage to ensure the average settlement price of new energy and the revenue from power generation. However, the monthly and intra-month spot market clearing prices are affected by multiple factors such as resources, weather, and supply and demand, resulting in large fluctuations and high uncertainty. At present, the forecasting of intra-provincial spot prices is relatively mature, and the numerical analysis forecasting method has certain practicality, but it does not have practical application reference value at the annual and quarterly scales. There are still the following shortcomings in the optimization of strategies for new energy to participate in inter-provincial and intra-provincial annual trading: (1) There is relatively little research on the medium- and long-term quantity-price coupling mechanism of resources-weather intra-provincial and inter-provincial trading; (2) There is a lack of relevant assessment methods for quantitatively assessing the risks and returns of multi-provincial annual trading markets; (3) There is no targeted research that uses a combination of market supply and demand and game analysis to quantitatively analyze the annual trading decisions of new energy in a certain province. Therefore, there is an urgent need for an innovative optimization scheme for annual trading strategies of new energy that can comprehensively consider market supply and demand and game behavior, so as to overcome the above-mentioned defects of existing technologies, provide scientific and reliable annual trading decision support for new energy generators, and thus effectively improve the certainty and stability of their market returns. Summary of the Invention

[0004] This application provides a method and system for optimizing annual trading strategies for new energy based on supply and demand and game theory, in order to at least solve the technical problems of existing technologies failing to combine market supply and demand, market participants' game behavior and medium- and long-term price formation mechanisms, resulting in a lack of quantitative basis for annual trading decisions, uncontrollable risks, and difficulty in optimizing overall returns.

[0005] The first aspect of this application proposes a method for optimizing annual trading strategies for new energy based on supply and demand and game theory, the method comprising: Collect and process historical transaction data to build a database of typical historical data; The spot price forecasting model is trained based on the historical typical database, and the spot price forecast value is output through the spot price forecasting model using the market supply and demand data of the target year. The annual centralized bidding transaction ratio prediction model is trained based on the historical typical database, and the transaction ratio prediction value is output through the annual centralized bidding transaction ratio prediction model using the spot price prediction value. Using the historical typical database as a sample, and based on the transaction ratio prediction model and the spot price prediction model, an annual bilateral transaction preliminary decision model is constructed and trained to generate a preliminary transaction strategy set; The preliminary trading strategy set, the spot price forecast, and the transaction ratio forecast are input into a pre-built comprehensive optimization model and solved using the differential evolution algorithm to obtain the optimized annual trading strategy. The optimized annual trading strategy is validated and its parameters are tuned using the historical typical database, and the optimal annual trading strategy is output.

[0006] Preferably, the construction process of the historical typical database includes: Collect historical annual bilateral trading volume and price, centralized bidding trading volume and price, spot market time-of-use clearing price, actual power generation, and corresponding regional load, inter-provincial transmission plan and new energy output data for renewable energy power plants; Using formula Calculate the theoretical optimal comprehensive return for each historical period and form a sample of return scenarios, where, For at any time The overall revenue from bilateral transactions in the new energy sector in the current year. For at any time The total transaction volume For at any time The average transaction price For at any time The annual competitive bidding transaction can declare electricity volume. For at any time The predicted electricity price For at any time The transaction ratio on the power generation side For at any time The deviation in the actual annual power generation of new energy sources For at any time The average clearing price of the spot market for the entire year; A historical typical database was constructed based on the aforementioned revenue scenario samples.

[0007] Furthermore, the step of training a spot price forecasting model based on the historical typical database, and using the market supply and demand data for the target year to output a predicted spot price value through the spot price forecasting model, includes: Based on the historical regional load data, inter-provincial transmission data, and new energy output data in the aforementioned historical typical database, the historical time-sharing bidding space for thermal power is calculated. Using the bidding space as input features and the corresponding historical spot price as output label, the model is trained using a linear regression method to obtain the spot price prediction model. The target year's load forecast, transmission plan, and new energy output forecast data are obtained and input into the trained spot price forecast model to obtain the spot price forecast value.

[0008] Furthermore, the step of training an annual centralized auction transaction ratio prediction model based on the historical typical database, and using the spot price prediction value to output a transaction ratio prediction value through the annual centralized auction transaction ratio prediction model, includes: Based on the aforementioned spot price forecasts, the annual-scale price forecast series and the monthly-scale price forecast series for the target year are calculated respectively. Calculate the price difference vector between annual and monthly forecast prices; Using the historical price difference data stored in the historical typical database as input and the corresponding historical transaction ratio as output, the annual centralized bidding transaction ratio prediction model is obtained by training through the linear regression method. The calculated target annual price difference vector is input into the trained transaction ratio prediction model to obtain the transaction ratio prediction value.

[0009] Furthermore, using the historical typical database as a sample, and based on the transaction ratio prediction model and the spot price prediction model, a preliminary annual bilateral transaction decision model is constructed and trained to generate a preliminary transaction strategy set, including: The electricity volume and price characteristics of historical transactions are extracted from the historical typical database as input features, and the corresponding theoretical optimal comprehensive benefits are extracted as output labels. The transaction ratio prediction model and the spot price prediction model are integrated into the decision model framework as embedded modules to calculate the expected returns under different trading strategies. The decision model is trained using a random forest regression algorithm to obtain the preliminary decision model for the annual bilateral transactions; The current market inquiry information is input into the preliminary decision-making model for the annual bilateral transaction, and the output is the preliminary transaction strategy set containing multiple combinations of electricity volume and price.

[0010] Furthermore, the construction process of the comprehensive optimization model includes: To maximize the expected overall return, construct an objective function; The comprehensive optimization model is constructed based on the objective function and constraints, with constraints on power balance, declared power, upper and lower limits of price quotation, and risk-return constraints. The objective function is calculated as follows:

[0011] In the formula, For the annual trading time The overall revenue of the new energy bilateral trading market For the annual trading time The total transaction volume For the annual trading time The average transaction price in the market. For the annual trading time The annual centralized bidding transaction can declare electricity volume. For the annual trading time The predicted electricity price For the annual trading time The expected transaction ratio on the power generation side.

[0012] Furthermore, the step of using the historical typical database to verify and fine-tune the optimized annual trading strategy, and outputting the optimal annual trading strategy, includes: The optimized annual trading strategy was then tested by backtracking multiple historical year scenarios in the historical typical database to simulate and calculate its return and risk distribution under various heterogeneous market environments. Based on the return distribution results of the backtracking test, the grid search method is used to jointly optimize the key parameters of the preliminary decision model and the differential evolution algorithm. Output an executable optimal annual trading strategy file. The file is structured data, including: time-period electricity curves, price curves, expected returns and their confidence intervals, and strategy risk warning information based on return distribution.

[0013] The second aspect of this application proposes a new energy annual trading strategy optimization system based on supply and demand and game theory, including: The data acquisition module is used to collect and process historical transaction data and build a historical typical database. The first prediction module is used to train a spot price prediction model based on the historical typical database, and to output the spot price prediction value through the spot price prediction model using the market supply and demand basic data of the target year. The second prediction module is used to train an annual centralized bidding transaction ratio prediction model based on the historical typical database, and to output a transaction ratio prediction value through the annual centralized bidding transaction ratio prediction model using the spot price prediction value. The generation module is used to construct and train an annual bilateral transaction preliminary decision model based on the historical typical database as a sample and the transaction ratio prediction model and spot price prediction model, and generate a preliminary transaction strategy set. The optimization module is used to input the preliminary trading strategy set, the spot price prediction value and the transaction ratio prediction value into the pre-built comprehensive optimization model, and use the differential evolution algorithm to solve it to obtain the optimized annual trading strategy. The optimization module is used to verify and fine-tune the optimized annual trading strategy using the historical typical database, and output the optimal annual trading strategy.

[0014] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in the first aspect embodiment.

[0015] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0016] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects: This application proposes a method and system for optimizing annual trading strategies for new energy based on supply and demand and game theory. The method includes: collecting and processing historical trading data to construct a historical typical database; training a spot price prediction model based on the historical typical database, and using the market supply and demand data for the target year, outputting a predicted spot price value through the spot price prediction model; training an annual centralized bidding transaction ratio prediction model based on the historical typical database, and using the predicted spot price value, outputting a predicted transaction ratio value through the annual centralized bidding transaction ratio prediction model; using the historical typical database as a sample, constructing and training an annual bilateral trading preliminary decision model based on the transaction ratio prediction model and the spot price prediction model to generate a preliminary trading strategy set; inputting the preliminary trading strategy set, the predicted spot price value, and the predicted transaction ratio value into a pre-constructed comprehensive optimization model, and solving it using a differential evolution algorithm to obtain the optimized annual trading strategy; verifying and fine-tuning the optimized annual trading strategy using the historical typical database, and outputting the optimal annual trading strategy. The technical solution proposed in this application enables quantitative analysis of market supply and demand and game behavior, which can effectively improve the overall returns and strategic robustness of new energy power plants in annual electricity market transactions.

[0017] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0018] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for optimizing an annual trading strategy for new energy based on supply and demand and game theory, according to an embodiment of this application. Figure 2 A detailed flowchart of a new energy annual trading strategy optimization method based on supply and demand and game theory, according to an embodiment of this application; Figure 3 This is a structural diagram of a new energy annual trading strategy optimization system based on supply and demand and game theory, according to an embodiment of this application. Detailed Implementation

[0019] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0020] This application proposes a method and system for optimizing annual trading strategies for new energy based on supply and demand and game theory. The method includes: collecting and processing historical trading data to construct a historical typical database; training a spot price prediction model based on the historical typical database, and using the market supply and demand data for the target year, outputting a predicted spot price value through the spot price prediction model; training an annual centralized bidding transaction ratio prediction model based on the historical typical database, and using the predicted spot price value, outputting a predicted transaction ratio value through the annual centralized bidding transaction ratio prediction model; using the historical typical database as a sample, constructing and training an annual bilateral trading preliminary decision model based on the transaction ratio prediction model and the spot price prediction model to generate a preliminary trading strategy set; inputting the preliminary trading strategy set, the predicted spot price value, and the predicted transaction ratio value into a pre-constructed comprehensive optimization model, and solving it using a differential evolution algorithm to obtain an optimized annual trading strategy; verifying and fine-tuning the optimized annual trading strategy using the historical typical database, and outputting the optimal annual trading strategy. The technical solution proposed in this application enables quantitative analysis of market supply and demand and game behavior, which can effectively improve the overall returns and strategic robustness of new energy power plants in annual electricity market transactions.

[0021] The following describes, with reference to the accompanying drawings, a method and system for optimizing annual trading strategies for new energy based on supply and demand and game theory, according to embodiments of this application.

[0022] Example 1 Figure 1 The flowchart below shows a method for optimizing an annual trading strategy for new energy based on supply and demand and game theory, according to an embodiment of this application. Figure 1 As shown, the method includes: Step 1: Collect and process historical transaction data to build a historical typical database; In this embodiment of the disclosure, the process of constructing the historical typical database includes: Collect historical annual bilateral trading volume and price, centralized bidding trading volume and price, spot market time-of-use clearing price, actual power generation, and corresponding regional load, inter-provincial transmission plan and new energy output data for renewable energy power plants; Using formula Calculate the theoretical optimal comprehensive return for each historical period and form a sample of return scenarios, where, For at any time The overall revenue from bilateral transactions in the new energy sector in the current year. For at any time The total transaction volume For at any time The average transaction price For at any time The annual competitive bidding transaction can declare electricity volume. For at any time The predicted electricity price For at any time The transaction ratio on the power generation side For at any time The deviation in the actual annual power generation of new energy sources For at any time The average clearing price of the spot market for the entire year; A historical typical database was constructed based on the aforementioned revenue scenario samples.

[0023] It should be noted that the analysis of the impact of historical transaction volumes and prices of new energy power plants on the annual comprehensive returns of new energy power plants through bilateral and centralized bidding at different time periods establishes a typical database of theoretically optimal comprehensive returns under typical historical annual transaction volumes and prices. The calculation expression is as follows:

[0024] In the formula: , , and It indicates at time The total revenue, rated maximum capacity, total transaction volume, and average transaction price in the annual bilateral transactions of new energy sources. , , It indicates at time The annual competitive bidding transaction can include the declared electricity volume, predicted electricity price, and the transaction ratio on the generation side. , , It indicates at time The deviation of the actual annual power generation of new energy sources, the actual power generation, and the average clearing price of the day-ahead spot market throughout the year. This represents the overall return on new energy. A typical database of optimal comprehensive returns under discrete volume and price conditions in bilateral market transactions is established based on historical monthly trading and spot market transaction data.

[0025] Step 2: Train a spot price forecasting model based on the historical typical database, and use the market supply and demand data for the target year to output the spot price forecast value through the spot price forecasting model; In this embodiment of the disclosure, step 2 specifically includes: Based on the historical regional load data, inter-provincial transmission data, and new energy output data in the aforementioned historical typical database, the historical time-sharing bidding space for thermal power is calculated. Using the bidding space as input features and the corresponding historical spot price as output label, the model is trained using a linear regression method to obtain the spot price prediction model. The target year's load forecast, transmission plan, and new energy output forecast data are obtained and input into the trained spot price forecast model to obtain the spot price forecast value.

[0026] It should be noted that the historical time-of-use bidding space for thermal power, formed by market supply and demand, is used to establish an annual time-of-use bilateral and centralized trading volume and price correlation analysis index with the time-of-use load factor of thermal power. The expression is as follows: (1) Considering the spatial distribution characteristics of historical thermal power time-of-use bidding formed by market supply and demand, the linear regression spot clearing electricity price prediction model is expressed as follows:

[0027]

[0028] In the formula: Time-sharing average spot price forecast Lower limit price , Parameters of linear regression Upper limit price, , , , This refers to the competitive pricing space in market supply and demand, regional market load, regional power transmission or reception, and renewable energy output. Regression-predicted prices and These are the lower and upper limits of the spot market price. This refers to the system's boot capacity. (2) Establish correlation analysis indicators between annual time-segmented bilateral and centralized trading volume and price and thermal power time-segmented bidding space:

[0029] In the formula, Annual bilateral transaction revenue, Annual comprehensive return, number of years in multiple consecutive samples, , These represent the annual bilateral transaction revenues of new energy sources at times t=1, 2, 3, ..., 24, respectively. and They are respectively the corresponding The annual comprehensive income of new energy at any time; This is the correlation coefficient between the annual bilateral transaction volume and price of new energy and the overall revenue. In different... and Given the value, The value is between -1 and +1. >0 indicates that and The two variables are positively correlated under the following values; if <0 indicates that and The two variables are negatively correlated when the values ​​are given. The larger the absolute value, the stronger the correlation, and the greater the probability of profit or loss; if =0 indicates that there is no linear relationship between the two.

[0030] Step 3: Train an annual centralized bidding transaction ratio prediction model based on the historical typical database, and use the spot price prediction value to output the transaction ratio prediction value through the annual centralized bidding transaction ratio prediction model; In this embodiment of the disclosure, step 3 specifically includes: Based on the aforementioned spot price forecasts, the annual-scale price forecast series and the monthly-scale price forecast series for the target year are calculated respectively. Calculate the price difference vector between annual and monthly forecast prices; Using the historical price difference data stored in the historical typical database as input and the corresponding historical transaction ratio as output, the annual centralized bidding transaction ratio prediction model is obtained by training through the linear regression method. The calculated target annual price difference vector is input into the trained transaction ratio prediction model to obtain the transaction ratio prediction value.

[0031] It should be noted that, based on Euclidean distance and market game theory, the annual centralized auction transaction price difference is compared with the monthly transaction price difference using similarity and linear regression calculations to form the predicted value of the annual centralized auction transaction volume ratio, the expression of which is as follows:

[0032]

[0033]

[0034]

[0035] In the formula: This indicates the time frame of annual auction transactions and monthly transactions. The average spot price spread and This represents the price difference vector. with vector Between The Euclidean distance between elements, when the distances in two vectors are the same. The value is zero. This represents the number of points with a distance of zero. This is the linear regression function expression for the transaction ratio and price difference. Indicates the time of centralized bidding. The transaction ratio is calculated as follows: when the transaction is sold, it is the transaction ratio on the power generation side; when the transaction is repurchased, it is the transaction ratio on the land use side.

[0036] Step 4: Using the historical typical database as a sample, and based on the transaction ratio prediction model and the spot price prediction model, construct and train the annual bilateral transaction preliminary decision model to generate a preliminary transaction strategy set; In this embodiment of the disclosure, step 4 specifically includes: The electricity volume and price characteristics of historical transactions are extracted from the historical typical database as input features, and the corresponding theoretical optimal comprehensive benefits are extracted as output labels. The transaction ratio prediction model and the spot price prediction model are integrated into the decision model framework as embedded modules to calculate the expected returns under different trading strategies. The decision model is trained using a random forest regression algorithm to obtain the preliminary decision model for the annual bilateral transactions; The current market inquiry information is input into the preliminary decision-making model for the annual bilateral transaction, and the output is the preliminary transaction strategy set containing multiple combinations of electricity volume and price.

[0037] It should be noted that, based on the probability distribution characteristics of the annual centralized bidding transaction ratio and clearing price prediction, a random forest regression algorithm is used to establish a bilateral time-sharing trading decision model for new energy annually. Due to the high uncertainty of the average price of annual bilateral trading and annual bidding transactions across different time periods, it is difficult to use a clear single price prediction scheme to compare and analyze bilateral market volume and price and make trading decisions. It is necessary to quantify and compare the comprehensive trading returns of bilateral market trading prices and corresponding trading volumes based on the probability distribution of historical spot time-sharing average price prediction errors. With the goal of optimizing comprehensive returns, the model solves for the bilateral volume and price transaction thresholds and ranges under different prediction error probability ranges. The relevant expressions are as follows: (1) The probability distribution of the annual auction clearing price prediction error follows a normal distribution. The sample mean of its independently distributed random error variable converges to a normal distribution, and its normal probability density function is:

[0038] in, The clearing average of the t-period competition period, The average clearing value of the spot market over different time periods (t) recently. Let be the expectation of the sample sequence to be estimated; Let be the standard deviation of the sample sequence to be estimated; For time period The mean error between the average price of the ten-day auction and the average price of the spot market clearing.

[0039] (2) With the goal of maximizing overall returns, it solves for the bilateral volume and price transaction thresholds and ranges under different prediction error probability ranges. The relevant expressions are as follows:

[0040] In the formula: , and This refers to the annual trading time. The overall revenue, total transaction volume, and average transaction price of the new energy bilateral trading market. , , It indicates at time The annual centralized bidding transactions can include declared electricity volume, predicted electricity price, and expected transaction ratio on the generation side. , , It indicates at time The forecast includes the deviation of the estimated annual power generation from new energy sources, the estimated annual power generation, and the average clearing price of the day-ahead spot market for the whole year. This represents the expected total revenue from new energy sources.

[0041] (3) A new energy annual bilateral time-sharing trading decision model was established using the random forest regression algorithm. Samples were randomly drawn using the Bootstrap sampling method, and the optimal parameters of the random forest regression model were obtained by using oob_score as the standard for model parameter tuning. The evaluation result is represented by oob_score, where oob_score is... Its expression is as follows:

[0042]

[0043]

[0044] In the formula: It is the sum of squared residuals; The total sum of squares; The number of samples; The sample number; These are the regression values ​​for the model; The true numerical labels for the sample points; This is the average value of the actual numerical labels.

[0045] Step 5: Input the preliminary trading strategy set, the spot price forecast, and the transaction ratio forecast into the pre-built comprehensive optimization model, and solve it using the differential evolution algorithm to obtain the optimized annual trading strategy; In this embodiment of the disclosure, the process of constructing the comprehensive optimization model includes: To maximize the expected overall return, construct an objective function; The comprehensive optimization model is constructed based on the objective function and constraints, with constraints on power balance, declared power, upper and lower limits of price quotation, and risk-return constraints. The objective function is calculated as follows:

[0046] In the formula, For the annual trading time The overall revenue of the new energy bilateral trading market For the annual trading time The total transaction volume For the annual trading time The average transaction price in the market. For the annual trading time The annual centralized bidding transaction can declare electricity volume. For the annual trading time The predicted electricity price For the annual trading time The expected transaction ratio on the power generation side.

[0047] It should be noted that the comprehensive optimization model is the annual bilateral and centralized bidding trading strategy integrated decision optimization model based on market supply and demand and market game theory. (1) Optimization objective:

[0048] (2) Model optimization is achieved using a difference algorithm. Mutation is performed by superimposing difference vectors on parent individuals to generate mutated individuals. Then, parent individuals are compared with experimental individuals with a certain probability, and the superior individuals enter the next generation population. Let the population size be NP and the dimension of the individual decision variables be n. Three operators are included: difference mutation, crossover, and selection, with the following expressions: Differential mutation operator. For each individual in the g-th generation of the population. Three distinct parent individuals were randomly selected. , , (r1, r2, r3 ∈ [1, NP] and r1, r2, r3 ≠ i), perform differential mutation operation according to equation (8) to generate mutated individuals. for: , where F is the difference scaling factor, which controls the magnitude of the difference variation.

[0049] Crossover operator. For parent individuals and mutated individuals Crossover is performed to generate experimental individuals. for , where: rand() is a uniform random number between [0, 1]; CR∈[0, 1] is the crossover probability, and rand(1, n) is a random integer between [1, n] to ensure that at least one of the experimental individuals comes from the mutant individual.

[0050] Selection operator. Compare the fitness values ​​of the parent individuals and the experimental individuals; the one with the better fitness value is selected for the next generation. .

[0051] Step 6: Validate and fine-tune the optimized annual trading strategy using the historical typical database, and output the optimal annual trading strategy.

[0052] In this embodiment of the disclosure, step 6 specifically includes: The optimized annual trading strategy was then tested by backtracking multiple historical year scenarios in the historical typical database to simulate and calculate its return and risk distribution under various heterogeneous market environments. Based on the return distribution results of the backtracking test, the grid search method is used to jointly optimize the key parameters of the preliminary decision model and the differential evolution algorithm. Output an executable optimal annual trading strategy file. The file is structured data, including: time-period electricity curves, price curves, expected returns and their confidence intervals, and strategy risk warning information based on return distribution.

[0053] It should be noted that when training the model considering different volume-price combinations in bilateral new energy transactions and historical training sets of different months and quarters, the error probability distribution and central value are different. Therefore, it is necessary to adjust and optimize the parameters for future prediction time and volume-price combination decisions. The grid search method is used to calculate the cross-validation performance of each volume-price combination on the new sample training set and then adjust and optimize the parameters.

[0054] The model is based on the Python language and uses a grid search method to calculate the cross-validation performance of each quantity-price combination on a new sample training set for parameter adjustment and optimization.

[0055] Specifically, such as Figure 2The diagram shows a detailed flowchart of a new energy annual trading strategy optimization method based on supply and demand and game theory proposed in this embodiment. Using multi-year bilateral, centralized bidding, and spot market time-segmented trading prices and trading volumes as the basic database, the correlation between the magnitude of comprehensive returns and bilateral trading volume and price is analyzed, and features are extracted. A random forest model is used to generate a comprehensive return prediction model 1. Simultaneously, predictions are made for future spot market time-segmented prices, price differences, and transaction ratios in centralized bidding. The prediction results and the bilateral trading volume and price market inquiry results are substituted into model 1, and the comprehensive return distribution is evaluated to see if it meets the expected threshold. If it is lower than expected, a higher and better volume and price curve is awaited; if the return expectation is met, the bilateral transaction is completed.

[0056] In summary, the proposed method for optimizing annual trading strategies for new energy based on supply and demand and game theory in this embodiment achieves quantitative analysis of market supply and demand and game theory behavior, and can effectively improve the overall returns and strategy robustness of new energy power plants in annual electricity market transactions.

[0057] Example 2 Figure 3 This is a structural diagram of a new energy annual trading strategy optimization system based on supply and demand and game theory, according to an embodiment of this application. Figure 3 As shown, the system includes: The data acquisition module 100 is used to collect and process historical transaction data and build a historical typical database. The first prediction module 200 is used to train a spot price prediction model based on the historical typical database, and to output a spot price prediction value through the spot price prediction model using the market supply and demand basic data of the target year. The second prediction module 300 is used to train an annual centralized bidding transaction ratio prediction model based on the historical typical database, and to output a transaction ratio prediction value through the annual centralized bidding transaction ratio prediction model using the spot price prediction value. The generation module 400 is used to construct and train an annual bilateral transaction preliminary decision model based on the historical typical database as a sample and the transaction ratio prediction model and spot price prediction model, and generate a preliminary transaction strategy set. The optimization module 500 is used to input the preliminary trading strategy set, the spot price prediction value and the transaction ratio prediction value into the pre-built comprehensive optimization model, and use the differential evolution algorithm to solve it to obtain the optimized annual trading strategy. It should be noted that the construction process of the comprehensive optimization model includes: To maximize the expected overall return, construct an objective function; The comprehensive optimization model is constructed based on the objective function and constraints, with constraints on power balance, declared power, upper and lower limits of price quotation, and risk-return constraints. The objective function is calculated as follows:

[0058] In the formula, For the annual trading time The overall revenue of the new energy bilateral trading market For the annual trading time The total transaction volume For the annual trading time The average transaction price in the market. For the annual trading time The annual centralized bidding transaction can declare electricity volume. For the annual trading time The predicted electricity price For the annual trading time The expected transaction ratio on the power generation side.

[0059] The optimization module 600 is used to verify and fine-tune the optimized annual trading strategy using the historical typical database, and output the optimal annual trading strategy.

[0060] In this embodiment of the disclosure, the acquisition module 100 is further configured to: Collect historical annual bilateral trading volume and price, centralized bidding trading volume and price, spot market time-of-use clearing price, actual power generation, and corresponding regional load, inter-provincial transmission plan and new energy output data for renewable energy power plants; Using formula Calculate the theoretical optimal comprehensive return for each historical period and form a sample of return scenarios, where, For at any time The overall revenue from bilateral transactions in the new energy sector in the current year. For at any time The total transaction volume For at any time The average transaction price For at any time The annual competitive bidding transaction can declare electricity volume. For at any time The predicted electricity price For at any time The transaction ratio on the power generation side For at any time The deviation in the actual annual power generation of new energy sources For at any time The average clearing price of the spot market for the entire year; A historical typical database was constructed based on the aforementioned revenue scenario samples.

[0061] In this embodiment of the disclosure, the first prediction module 200 is further configured to: Based on the historical regional load data, inter-provincial transmission data, and new energy output data in the aforementioned historical typical database, the historical time-sharing bidding space for thermal power is calculated. Using the bidding space as input features and the corresponding historical spot price as output label, the model is trained using a linear regression method to obtain the spot price prediction model. The target year's load forecast, transmission plan, and new energy output forecast data are obtained and input into the trained spot price forecast model to obtain the spot price forecast value.

[0062] In this embodiment of the disclosure, the second prediction module 300 is further configured to: Based on the aforementioned spot price forecasts, the annual-scale price forecast series and the monthly-scale price forecast series for the target year are calculated respectively. Calculate the price difference vector between annual and monthly forecast prices; Using the historical price difference data stored in the historical typical database as input and the corresponding historical transaction ratio as output, the annual centralized bidding transaction ratio prediction model is obtained by training through the linear regression method. The calculated target annual price difference vector is input into the trained transaction ratio prediction model to obtain the transaction ratio prediction value.

[0063] In this embodiment of the disclosure, the generation module 400 is further configured to: The electricity volume and price characteristics of historical transactions are extracted from the historical typical database as input features, and the corresponding theoretical optimal comprehensive benefits are extracted as output labels. The transaction ratio prediction model and the spot price prediction model are integrated into the decision model framework as embedded modules to calculate the expected returns under different trading strategies. The decision model is trained using a random forest regression algorithm to obtain the preliminary decision model for the annual bilateral transactions; The current market inquiry information is input into the preliminary decision-making model for the annual bilateral transaction, and the output is the preliminary transaction strategy set containing multiple combinations of electricity volume and price.

[0064] In this embodiment of the disclosure, the tuning module 600 is further configured to: The optimized annual trading strategy was then tested by backtracking multiple historical year scenarios in the historical typical database to simulate and calculate its return and risk distribution under various heterogeneous market environments. Based on the return distribution results of the backtracking test, the grid search method is used to jointly optimize the key parameters of the preliminary decision model and the differential evolution algorithm. Output an executable optimal annual trading strategy file. The file is structured data, including: time-period electricity curves, price curves, expected returns and their confidence intervals, and strategy risk warning information based on return distribution.

[0065] In summary, the new energy annual trading strategy optimization system proposed in this embodiment, based on supply and demand and game theory, realizes quantitative analysis of market supply and demand and game behavior, and can effectively improve the overall returns and strategy robustness of new energy power plants in the annual electricity market trading.

[0066] Example 3 To implement the above embodiments, this disclosure also proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in Embodiment 1.

[0067] Example 4 To implement the above embodiments, this disclosure also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.

[0068] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0069] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0070] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for optimizing annual trading strategies for new energy based on supply and demand and game theory, characterized in that, The method includes: Collect and process historical transaction data to build a database of typical historical data; The spot price forecasting model is trained based on the historical typical database, and the spot price forecast value is output through the spot price forecasting model using the market supply and demand data of the target year. The annual centralized bidding transaction ratio prediction model is trained based on the historical typical database, and the transaction ratio prediction value is output through the annual centralized bidding transaction ratio prediction model using the spot price prediction value. Using the historical typical database as a sample, and based on the transaction ratio prediction model and the spot price prediction model, an annual bilateral transaction preliminary decision model is constructed and trained to generate a preliminary transaction strategy set; The preliminary trading strategy set, the spot price forecast, and the transaction ratio forecast are input into a pre-built comprehensive optimization model and solved using the differential evolution algorithm to obtain the optimized annual trading strategy. The optimized annual trading strategy is validated and its parameters are tuned using the historical typical database, and the optimal annual trading strategy is output.

2. The method as described in claim 1, characterized in that, The construction process of the historical typical database includes: Collect historical annual bilateral trading volume and price, centralized bidding trading volume and price, spot market time-of-use clearing price, actual power generation, and corresponding regional load, inter-provincial transmission plan and new energy output data for renewable energy power plants; Using formula Calculate the theoretical optimal comprehensive return for each historical period and form a sample of return scenarios, where, For at any time The overall revenue from bilateral transactions in the new energy sector in the current year. For at any time The total transaction volume, For at any time The average transaction price For at any time The annual competitive bidding transaction can declare electricity volume. For at any time The predicted electricity price For at any time The transaction ratio on the power generation side For at any time The deviation in the actual annual power generation of new energy sources For at any time The average clearing price of the spot market for the entire year; A historical typical database was constructed based on the aforementioned revenue scenario samples.

3. The method as described in claim 2, characterized in that, The step of training a spot price forecasting model based on the historical typical database and using the market supply and demand data for the target year to output a predicted spot price value through the spot price forecasting model includes: Based on the historical regional load data, inter-provincial transmission data, and new energy output data in the aforementioned historical typical database, the historical time-sharing bidding space for thermal power is calculated. Using the bidding space as input features and the corresponding historical spot price as output label, the model is trained using a linear regression method to obtain the spot price prediction model. The target year's load forecast, transmission plan, and new energy output forecast data are obtained and input into the trained spot price forecast model to obtain the spot price forecast value.

4. The method as described in claim 3, characterized in that, The step of training an annual centralized auction transaction ratio prediction model based on the historical typical database, and using the spot price prediction value, outputting a transaction ratio prediction value through the annual centralized auction transaction ratio prediction model, includes: Based on the aforementioned spot price forecasts, the annual-scale price forecast series and the monthly-scale price forecast series for the target year are calculated respectively. Calculate the price difference vector between annual and monthly forecast prices; Using the historical price difference data stored in the historical typical database as input and the corresponding historical transaction ratio as output, the annual centralized bidding transaction ratio prediction model is obtained by training through the linear regression method. The calculated target annual price difference vector is input into the trained transaction ratio prediction model to obtain the transaction ratio prediction value.

5. The method as described in claim 4, characterized in that, The process involves using the historical typical database as a sample, and constructing and training an annual bilateral transaction preliminary decision-making model based on the transaction ratio prediction model and the spot price prediction model to generate a preliminary transaction strategy set, including: The electricity volume and price characteristics of historical transactions are extracted from the historical typical database as input features, and the corresponding theoretical optimal comprehensive benefits are extracted as output labels. The transaction ratio prediction model and the spot price prediction model are integrated into the decision model framework as embedded modules to calculate the expected returns under different trading strategies. The decision model is trained using a random forest regression algorithm to obtain the preliminary decision model for the annual bilateral transactions; The current market inquiry information is input into the preliminary decision-making model for the annual bilateral transaction, and the output is the preliminary transaction strategy set containing multiple combinations of electricity volume and price.

6. The method as described in claim 5, characterized in that, The process of constructing the comprehensive optimization model includes: To maximize the expected overall return, construct an objective function; The comprehensive optimization model is constructed based on the objective function and constraints, with constraints on power balance, declared power, upper and lower limits of price quotation, and risk-return constraints. The objective function is calculated as follows: In the formula, For the annual trading time The overall revenue of the bilateral trading market for new energy sources For the annual trading time The total transaction volume, For the annual trading time The average transaction price in the market. For the annual trading time The annual centralized bidding transaction can declare electricity volume. For the annual trading time The predicted electricity price For the annual trading time The expected transaction ratio on the power generation side.

7. The method as described in claim 6, characterized in that, The process of validating and fine-tuning the optimized annual trading strategy using the historical typical database, and outputting the optimal annual trading strategy, includes: The optimized annual trading strategy was then tested by backtracking multiple historical year scenarios in the historical typical database to simulate and calculate its return and risk distribution under various heterogeneous market environments. Based on the return distribution results of the backtracking test, the grid search method is used to jointly optimize the key parameters of the preliminary decision model and the differential evolution algorithm. Output an executable optimal annual trading strategy file. The file is structured data, including: time-period electricity curves, price curves, expected returns and their confidence intervals, and strategy risk warning information based on return distribution.

8. A new energy annual trading strategy optimization system based on supply and demand and game theory, characterized in that, The system includes: The data acquisition module is used to collect and process historical transaction data and build a historical typical database. The first prediction module is used to train a spot price prediction model based on the historical typical database, and to output the spot price prediction value through the spot price prediction model using the market supply and demand basic data of the target year. The second prediction module is used to train an annual centralized bidding transaction ratio prediction model based on the historical typical database, and to output a transaction ratio prediction value through the annual centralized bidding transaction ratio prediction model using the spot price prediction value. The generation module is used to construct and train an annual bilateral transaction preliminary decision model based on the historical typical database as a sample and the transaction ratio prediction model and spot price prediction model, and generate a preliminary transaction strategy set. The optimization module is used to input the preliminary trading strategy set, the spot price prediction value and the transaction ratio prediction value into the pre-built comprehensive optimization model, and use the differential evolution algorithm to solve it to obtain the optimized annual trading strategy. The optimization module is used to verify and fine-tune the optimized annual trading strategy using the historical typical database, and output the optimal annual trading strategy.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images