Pumped storage unit centralized bidding transaction strategy optimization method and device

Abstract

A centralized bidding transaction strategy optimization method for a pumped storage unit comprises the following steps: modeling a market subject: respectively establishing bidding decision targets of market members, taking the bidding decision targets as transaction objects in a real market, actively learning and adjusting a market strategy according to self and external environment information in a continuous transaction process, and establishing a bidding decision model; performing optimization decision according to an expected target; the upper and lower limits of bidding stipulated by the electricity market are inherent bidding intervals of the whole market main body, the upper limit of the bidding space for the pumped storage power station in the interval is the highest bidding value stipulated by the market, and the lower limit is marginal cost price; the upper limit of the bidding space for the power consumers is the historical average highest declaration price, and the lower limit is the market-specified lowest bidding value; the quotation decision optimization model outputs an optimal transaction strategy: the quotation decision optimization model performs iteration until the transaction round number t is greater than the preset T, and the dealer i stops learning; and outputting a quotation strategy obtained in the last round, and taking the quotation strategy as the optimal transaction strategy of the generator.

Description

Technical Field

[0001] The invention relates to an optimization method, and in particular to a method and a device for optimizing a centralized bidding transaction strategy for a pumped storage unit. Background Art

[0002] Centralized bidding transactions refer to a medium- and long-term market transaction method in which electricity buyers and sellers submit declaration information in the power trading center system, and then the power trading center matches the price difference and then conducts unified clearing. The final clearing price and transaction volume of centralized bidding transactions will be affected by a variety of numerical and non-numerical factors such as historical electricity prices, market supply and demand, and meteorological conditions. The final transaction price shows a certain degree of randomness and nonlinearity. This greatly affects the revenue of pumped storage power stations, so the quotation strategy for participating in centralized bidding transactions needs to be further improved.

[0003] At present, the quotation strategies for centralized bidding transactions are mainly divided into two categories: cost-plus pricing and futures price forecasting. The cost-plus pricing method determines the electricity price based on the cost of electricity production (such as fuel cost, environmental cost, etc.) and the supplier's markup. This pricing method is simple, transparent, and can ensure the coverage of producers' costs and stable profits, but it is more suitable for trading markets with obvious cost structures and little change in supply and demand, and cannot adapt well to the dynamic changes in electricity trading. The futures price forecasting method reflects market expectations, adapts to market changes, and provides real-time and highly liquid price signals, which helps to cope with market uncertainties and conduct risk management, but to a certain extent ignores the trading strategies of other competing entities, and cannot flexibly and quickly propose new quotation strategies based on the quotations of other entities.

[0004] Although the quotation basis and bidding strategy of the above methods are different, they all have certain limitations. The cost-plus pricing method is difficult to adapt to the increasingly volatile electricity market transactions, while the futures price forecasting method mainly focuses on factors such as market supply and demand and historical prices, and ignores the impact of quotations from other competing trading entities on the final clearing price. Summary of the invention

[0005] The present invention aims to propose a method for optimizing the centralized bidding transaction strategy of pumped storage units by integrating the idea of ​​portfolio optimization and reinforcement learning method, simulating the decision-making process of different market players in the power market to improve the market revenue of pumped storage power stations. The technical scheme is as follows:

[0006] In a first aspect, the present invention provides a method for optimizing a centralized bidding transaction strategy for a pumped storage unit, which is characterized by comprising:

[0007] Establish the bidding decision-making goals of market members and treat them as trading objects in the real market; in the continuous trading process, according to the members' own and external environment information, propose market quotation strategies within the trading space, and design quotation strategies according to the expected goals;

[0008] The future clearing price is predicted by time series analysis, and a price curve is drawn based on the prediction results. The clearing price curve is selected as the upper limit of the transaction, and the marginal cost price of pumped storage is selected as the lower limit, which is used as the trading space of pumped storage.

[0009] Based on the establishment of decision-making objectives of market players, a quotation strategy is selected within the pumped storage transaction space and input into the quotation decision optimization model for iteration until the number of transaction rounds t is greater than the preset T, trader i stops learning, and the optimal quotation strategy is obtained by outputting the last round.

[0010] Preferably, the establishment of the decision-making objectives of the market subject includes the following contents:

[0011] Power generation company transaction modeling: The revenue from electricity sales depends on the market clearing price and the amount of electricity cleared, both of which change with the bids of the power generation company and its competitors. The specific calculation method is shown in formula (1):

[0012]

[0013] In formula (1), F p The power sales revenue of power generators in the medium and long-term market includes two parts: the power sales revenue from annual bilateral negotiation transactions and the power sales revenue from monthly centralized bidding; s,l and Q s,l are the transaction price and transaction volume of the generator in the first transaction of the bilateral negotiation transaction; m,i and Q m,i are the clearing electricity price and winning electricity volume declared by the generator in the i-th segment in the centralized market;

[0014] The profit of a power generator is the difference between the revenue from electricity sales and the cost. For most power generators, the goal of participating in the centralized bidding market is to maximize profits. The bidding decision objective function is:

[0015]

[0016] In formula (2), F p-profit Profits obtained by power generators from participating in medium- and long-term market transactions; C g is the marginal generation cost of the generator;

[0017] Electricity user transaction modeling: Users who are eligible to directly participate in electricity market transactions participate in bidding transactions with the goal of obtaining lower electricity prices and effectively reducing corporate electricity costs. Their electricity purchase cost C in the centralized trading market is y for:

[0018]

[0019] In the formula, λ y,i and Q y,i are the clearing electricity price and winning bid volume declared by the electricity user in the i-th period in the medium and long term respectively;

[0020] In the user full electricity transaction mode, the deviation electricity generated is no longer settled according to the catalog electricity price, but needs to be processed through deviation assessment. The user needs to bear the deviation assessment cost. for:

[0021]

[0022] In the formula, λ r Penalize electricity price for deviation assessment; Q y The amount of electricity that electricity users need to purchase in the monthly centralized market; It is the electricity transaction volume of electricity users in the monthly centralized market, calculated as follows:

[0023]

[0024] Therefore, the market decision objective function of power users in the full power transaction mode is:

[0025]

[0026] Electricity user transaction modeling: Assuming that the remaining electricity demand of the power sales company needs to be traded through the monthly centralized bidding, the power purchase cost C of the power sales company in the monthly centralized market is s for:

[0027]

[0028] In the formula, λ s,i and Q s,i are the clearing electricity price and transaction volume declared by the power sales company in the i-th segment in the centralized bidding market;

[0029] Assuming that the power sales company bears all the deviation assessment for the user, the deviation assessment cost of the power sales company in the monthly centralized market is for:

[0030]

[0031] In formula (8), λ sis the deviation assessment electricity price, Q s To meet the power purchasing needs of power sales companies in the medium and long term market. The transaction volume of the power sales company in the medium and long term market, β is the deviation assessment coefficient of the power sales company, which depends on the specific deviation assessment rules;

[0032] The revenue that the power sales company obtains from the user side through the centralized trading market is related to the centralized trading electricity price signed by the user in the contract, namely:

[0033]

[0034] In formula (9), P s,k Q is the centralized trading electricity price agreed upon by the power sales company s and user k in the power purchase and sales contract; s,k The amount of electricity purchased by user k through centralized trading;

[0035] Therefore, the bidding objectives of the power sales company are:

[0036]

[0037] Preferably, the establishment of the pumped storage trading space includes the following contents:

[0038] Trading subject strategy: The trading space can provide the upper and lower limits of the trading price to the decision makers of power generators, ensuring the scientificity and rationality of the price adjustment process and the final bidding plan; the strategy set of trader i is X m ={x i1 ,x i2 ,L,x im}, initialize the Q value corresponding to each strategy; before the first round of bidding begins, assign the same Q to each strategy im (0) = 0;

[0039] Trading simulation parameter setting: The trading model needs to set parameters according to the regional trading rules to determine the trading space; set the number of experimental plates to O, and the maximum number of trading rounds per plate to T; usually the exploration factor ε can be taken as a smaller constant. The smaller ε is, the more times the agent tries new strategies. If the number of experimental trading rounds is set very large, the market will take longer to reach a convergence state, or even not converge. In order to speed up the convergence speed, ε can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (Formula 11):

[0040]

[0041] Among them, T is the maximum number of transaction rounds preset in the experiment, ε i0 The initial exploration probability pre-set for the experiment, ε it is the exploration probability of the tth round.

[0042] Preferably, the establishment of the optimal trading decision includes the following contents:

[0043] Bidding strategy selection: Starting from the first round, a random number γ in the interval (0,1) is generated in each round, and the size of γ is compared with the exploration factor ε. If γ < ε, a random number is selected from the strategy set X with uniform probability. i If γ>ε, select a strategy from the strategy set X i Select the strategy with the largest current Q value. If there are multiple strategies with the same Q value, randomly select one from them.

[0044] Simulated market clearing: After all traders submit their bids, the market is cleared according to the set clearing rules, and the clearing results of the winning electricity volume and winning electricity price are fed back to each trader;

[0045] Update Q value and profit value: Calculate the profit of trader i according to the market clearing information and update the Q value; if the strategy selected by trader i in the tth round of trading is a ik , according to the winning bid electricity price and winning bid electricity quantity calculation strategy a ik The benefits brought by R ik (t), update the Q value of each strategy in the next round:

[0046]

[0047] Among them, the profit function R i The calculation of is related to the decision-making objectives of specific market members. it is the reward discount coefficient of trader i in the tth round of trading, which is used to suppress previous experience. In order to speed up the market convergence, α can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (13):

[0048]

[0049] Update Q under the risk aversion model r Value: In the traditional decision-making model, Q value represents the cumulative average reward after the strategy is executed for a period of time. The calculation of Q value only considers the benefits brought by the strategy. On this basis, the improved risk aversion model quantifies the risk of the strategy into the decision-making process by calculating the standard deviation of the historical return data of each strategy. The larger the standard deviation, the more unstable the return of the strategy and the higher the risk. On the contrary, the smaller the standard deviation, the more stable the return of the strategy and the lower the risk. The specific formula is as follows:

[0050]

[0051] Trader i based on the above improved Q r The value is optimized and Q is selected from the strategy set X. r The strategy with the largest value Bidding, that is, a strategy with relatively stable returns and relatively low risks in terms of standard deviation:

[0052]

[0053] In formula (14), H im It means that trader i executes strategy x every time im The historical return data set, r im Yes H im Elements in Q im The calculation method is based on formula (12), r im is the cumulative reward considering risk; β is the decision maker’s personal risk preference coefficient, and its value range is [0,1]; the larger the value of β, the greater the risk aversion of the trader. When β=1, the trader is extremely risk-averse, and the size of the return has no effect on the decision. The only goal is to minimize the risk. As β decreases, the risk gradually increases, and the power generator pays more attention to the return.

[0054] In a second aspect, the present invention further discloses a device for optimizing a centralized bidding transaction strategy for a pumped storage unit, which is characterized by:

[0055] Establish a bidding decision target module for market members: treat it as a transaction object in the real market; in the continuous trading process, according to the members' own and external environment information, propose market quotation strategies within the transaction space, and design quotation strategies according to the expected goals;

[0056] Module for establishing pumped storage trading space: Module for establishing pumped storage trading space: predict the future clearing price through time series analysis method, draw the price curve according to the prediction results, select the clearing price curve as the upper limit of the transaction, and select the marginal cost price of pumped storage as the lower limit, so as to serve as the pumped storage trading space;

[0057] Module for establishing optimal trading decisions: Based on the establishment of decision-making objectives of market entities, a quotation strategy is selected within the pumped storage trading space and input into the quotation decision optimization model for iteration until the number of trading rounds t is greater than the preset T, trader i stops learning, and outputs the optimal quotation strategy for the last round.

[0058] Preferably, the module for establishing the decision-making objectives of the market subject includes the following contents:

[0059] Power producer transaction modeling submodule: The calculation method is shown in formula (1):

[0060]

[0061] In formula (1), F pThe power sales revenue of power generators in the medium and long-term market includes two parts: the power sales revenue from annual bilateral negotiation transactions and the power sales revenue from monthly centralized bidding; s,l and Q s,l are the transaction price and transaction volume of the generator in the first transaction of the bilateral negotiation transaction; m,i and Q m,i are the clearing electricity price and winning electricity volume declared by the generator in the i-th segment in the centralized market;

[0062] The profit of a power generator is the difference between the revenue from electricity sales and the cost. For most power generators, the goal of participating in the centralized bidding market is to maximize profits. The bidding decision objective function is:

[0063]

[0064] In formula (2), F p-profit Profits obtained by power generators from participating in medium- and long-term market transactions; C g is the marginal generation cost of the generator;

[0065] Electricity user transaction modeling submodule: Users who are eligible to directly participate in electricity market transactions participate in bidding transactions with the goal of obtaining lower electricity prices and effectively reducing corporate electricity costs. Their electricity purchase cost C in the centralized trading market is y for:

[0066]

[0067] In the formula, λ y,i and Q y,i are the clearing electricity price and winning bid volume declared by the electricity user in the i-th period in the medium and long term respectively;

[0068] In the user full electricity transaction mode, the deviation electricity generated is no longer settled according to the catalog electricity price, but needs to be processed through deviation assessment. The user needs to bear the deviation assessment cost. for:

[0069]

[0070] In the formula, λ r Penalize electricity price for deviation assessment; Q y The amount of electricity that electricity users need to purchase in the monthly centralized market; It is the electricity transaction volume of electricity users in the monthly centralized market, calculated as follows:

[0071]

[0072] Therefore, the market decision objective function of power users in the full power transaction mode is:

[0073]

[0074] Electricity user transaction modeling submodule: Assuming that the remaining electricity demand of the power sales company needs to be traded through the monthly centralized bidding, the power purchase cost C of the power sales company in the monthly centralized market is s for:

[0075]

[0076] In the formula, λ s,i and Q s,i are the clearing electricity price and transaction volume declared by the power sales company in the i-th segment in the centralized bidding market;

[0077] Assuming that the power sales company bears all the deviation assessment for the user, the deviation assessment cost of the power sales company in the monthly centralized market is for:

[0078]

[0079] In formula (8), λ s is the deviation assessment electricity price, Q s To meet the power purchasing needs of power sales companies in the medium and long term market. The transaction volume of the power sales company in the medium and long term market, β is the deviation assessment coefficient of the power sales company, which depends on the specific deviation assessment rules;

[0080] The revenue that the power sales company obtains from the user side through the centralized trading market is related to the centralized trading electricity price signed by the user in the contract, namely:

[0081]

[0082] In formula (9), P s,k Q is the centralized trading electricity price agreed upon by the power sales company s and user k in the power purchase and sales contract; s,k The amount of electricity purchased by user k through centralized trading;

[0083] Therefore, the bidding objectives of the power sales company are:

[0084]

[0085] Preferably, the module for establishing the pumped storage trading space includes the following contents:

[0086] Trading subject strategy submodule: The trading space can provide the upper and lower limits of the trading price to the decision makers of power generators, ensuring the scientificity and rationality of the price adjustment process and the final bidding plan; the strategy set of trader i is X m ={x i1 ,x i2 ,L,xim}, initialize the Q value corresponding to each strategy; before the first round of bidding begins, assign the same Q to each strategy im (0) = 0;

[0087] Trading simulation parameter setting submodule: The trading model needs to set parameters according to the regional trading rules to determine the trading space; set the number of experimental plates to O, and the maximum number of trading rounds per plate experiment to T; usually the exploration factor ε can be taken as a smaller constant. The smaller ε is, the more times the agent tries new strategies. If the number of experimental trading rounds is set very large, the market will take longer to reach a convergence state, or even not converge. In order to speed up the convergence speed, ε can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (Formula 11):

[0088]

[0089] Among them, T is the maximum number of transaction rounds preset in the experiment, ε i0 The initial exploration probability pre-set for the experiment, ε it is the exploration probability of the tth round.

[0090] Preferably, the module for establishing the optimal trading decision includes the following contents:

[0091] Quotation strategy selection submodule: Starting from the first round, a random number γ in the interval (0,1) is generated in each round, and the size of γ is compared with the exploration factor ε. If γ < ε, a random number is selected from the strategy set X with uniform probability. i If γ>ε, select a strategy from the strategy set X i Select the strategy with the largest current Q value. If there are multiple strategies with the same Q value, randomly select one from them.

[0092] Simulated market clearing submodule: After all traders submit their bids, the market is cleared according to the set clearing rules, and the clearing results of the winning electricity volume and winning electricity price are fed back to each trader;

[0093] Update Q value and profit value submodule: Calculate the profit of trader i according to the market clearing information and update the Q value; if the strategy selected by trader i in the tth round of trading is a ik , according to the winning bid electricity price and winning bid electricity quantity calculation strategy a ik The benefits brought by R ik (t), update the Q value of each strategy in the next round:

[0094]

[0095] Among them, the profit function R i The calculation of is related to the decision-making objectives of specific market members. itis the reward discount coefficient of trader i in the tth round of trading, which is used to suppress previous experience. In order to speed up the market convergence, α can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (13):

[0096]

[0097] Update Q under the risk aversion model r Value submodule: In the traditional decision-making model, the Q value represents the cumulative average reward after the strategy is executed for a period of time. The calculation of the Q value only considers the benefits brought by the strategy. On this basis, the improved risk aversion model quantifies the risk of the strategy into the decision-making process by calculating the standard deviation of the historical return data of each strategy. The larger the standard deviation, the more unstable the return of the strategy and the higher the risk. On the contrary, the smaller the standard deviation, the more stable the return of the strategy and the lower the risk. The specific formula is as follows:

[0098]

[0099] Trader i based on the above improved Q r The value is optimized and Q in the strategy set X is selected. r The strategy with the largest value Bidding, that is, a strategy with relatively stable returns and relatively low risks in terms of standard deviation:

[0100]

[0101] In formula (14), H im It means that trader i executes strategy x every time im The historical return dataset, r im Yes H im Elements in Q im The calculation method is based on formula (12), r im is the cumulative reward considering risk; β is the decision maker’s personal risk preference coefficient, and its value range is [0,1]; the larger the value of β, the greater the risk aversion of the trader. When β=1, the trader is extremely risk-averse, and the size of the return has no effect on the decision. The only goal is to minimize the risk. As β decreases, the risk gradually increases, and the power generator pays more attention to the return.

[0102] Beneficial Effects

[0103] In a third aspect, the present invention further discloses a non-volatile storage medium, wherein the non-volatile storage medium includes a stored program, wherein when the program is run, the device where the non-volatile storage medium is located is controlled to execute any one of the above methods.

[0104] In a fourth aspect, the present invention further discloses an electronic device, comprising a processor and a memory; the memory stores computer-readable instructions, and the processor is used to execute the computer-readable instructions, wherein the computer-readable instructions execute any one of the methods described above when executed.

[0105] The present invention aims to establish an intelligent model that can continuously learn and adjust the strategy market in the actual transaction process by simulating the bidding decisions of different market players (such as power generators, users, power sales companies, etc.) in the power market. The model can construct a corresponding strategy set based on the private information and bidding rules of market members, and simulate the intelligent behavior of market members to continuously learn new knowledge and adjust quotations in transactions. Finally, by introducing the goal of risk aversion, a risk control model suitable for medium- and long-term market transactions of pumped storage modules is constructed, which can better reflect the decision-making process and risk attitude of each subject in the power market, ensure that the quotation strategy of the pumped storage power station fully considers the quotation situation of other market players, and dynamically adjusts the pricing strategy through real-time price signals to adapt to the fluctuating power trading market. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 This is a flow chart of the centralized bidding transaction strategy optimization method for pumped storage units of the present invention. DETAILED DESCRIPTION

[0107] Example 1

[0108] See also Figure 1 As shown, the present invention proposes a method for optimizing the centralized bidding transaction strategy of pumped storage units by integrating the idea of ​​portfolio optimization and reinforcement learning method. The method mainly includes: market subject modeling, transaction space construction, quotation decision optimization model, and output of optimal transaction strategy. The specific contents are as follows:

[0109] Market subject modeling: Establish the bidding decision-making goals of market members and treat them as trading objects in the real market; in the continuous trading process, according to the members' own and external environment information, propose market quotation strategies within the trading space, and design quotation strategies according to the expected goals;

[0110] In the actual operation of centralized bidding transactions in the power market, bidding decisions are made by different market entities, including power generators, users, and power sales companies. This decision-making behavior belongs to the social and economic activities of human intelligence. Therefore, it is necessary to establish bidding decision goals for market members respectively, and treat them as trading objects in the real market. In the continuous trading process, according to their own and external environmental information, they actively learn and adjust market strategies, and make optimal decisions according to the expected goals.

[0111] Modeling of power generation company transactions. Generally speaking, the purpose of power generation companies participating in market bidding is to obtain maximum profits through strategic bidding. In the medium and long-term market, the profit of power generation companies is determined by the cost of power generation and the revenue from power sales. The value of this part has been studied in detail in Chapter 4, so the research results above are continued here. Among them, the revenue from power sales depends on the market clearing price and the clearing quantity, which change with the bids of power generation companies and their opponents. The specific calculation method is shown in formula (1).

[0112]

[0113] In formula (1), F p The power sales revenue of power generators in the medium and long-term market includes two parts: the power sales revenue from annual bilateral negotiation transactions and the power sales revenue from monthly centralized bidding; s,l and Q s,l are the transaction price and transaction volume of the generator in the first transaction of the bilateral negotiation transaction; m,i and Q m,i They are respectively the clearing electricity price and winning bid electricity quantity declared by the generator in the i-th section in the centralized market.

[0114] The profit of a power generator is the difference between the revenue from electricity sales and the cost. For most power generators, their goal in participating in the centralized bidding market is to maximize profits. Their bidding decision objective function is:

[0115]

[0116] In formula (2), F p-profit Profits obtained by power generators from participating in medium- and long-term market transactions; C g is the marginal generation cost of the generator.

[0117] Electricity user transaction modeling. Users who are eligible to directly participate in electricity market transactions are usually large industrial enterprises with high voltage levels, electricity consumption capacity exceeding a certain scale, and relatively stable electricity load. They have sufficient electricity demand and market participation capabilities, and can trade in the market and obtain economic benefits. The electricity cost of such users accounts for a high proportion of the total cost of the enterprise, which has a greater impact on the final profit level of the enterprise. Therefore, their goal in participating in bidding transactions is to strive for lower electricity prices and effectively reduce the electricity costs of enterprises. Their electricity purchase cost C in the centralized trading market y for:

[0118]

[0119] In the formula, λ y,i and Q y,i They are respectively the clearing electricity price and winning bid electricity quantity declared by the electricity user in the i-th period in the medium and long term.

[0120] In the user full electricity transaction mode, the deviation electricity generated is no longer settled according to the catalog electricity price, but needs to be processed through deviation assessment. The user needs to bear the deviation assessment cost. for:

[0121]

[0122] In the formula, λ r Penalize electricity price for deviation assessment; Q y The amount of electricity that electricity users need to purchase in the monthly centralized market; It is the electricity transaction volume of electricity users in the monthly centralized market, calculated as follows:

[0123]

[0124] Therefore, the market decision objective function of power users in the full power transaction mode is:

[0125]

[0126] Electricity user transaction modeling. Usually, the profit model of power sales companies is divided into two categories: earning price difference and value-added services. Here we mainly discuss power sales companies that specialize in providing power purchase services to earn price difference. Taking into account the uncertainty of centralized market electricity prices, power sales companies will gradually enter the full electricity mode in the future. In order to reduce the deviation of direct transactions, power sales companies need to make supplementary transactions in the monthly centralized market to adjust the difference based on the changes in the previous month's demand of the users they represent on the basis of bilateral transactions. In this invention, it is assumed that the remaining demand electricity of the power sales company needs to be traded through the monthly centralized bidding. Then the power purchase cost C of the power sales company in the monthly centralized market is s for:

[0127]

[0128] In the formula, λ s,i and Q s,i They are respectively the clearing electricity price and transaction volume declared by the power sales company in the i-th section in the centralized bidding market.

[0129] In this case, if the power sales company's quotation is inappropriate, the monthly centralized bidding volume cannot meet the actual demand, and this part of the deviation volume needs to be assessed. In the actual market, the power sales company and the user will sign a power purchase and sales contract in advance to clarify the deviation assessment sharing ratio between the two parties, the user's intended purchase volume, and the bilateral and centralized market electricity prices. This article assumes that the power sales company bears all deviation assessments for the user. The deviation assessment cost of the power sales company in the monthly centralized market is for:

[0130]

[0131] In formula (8), λ s is the deviation assessment electricity price, Q s To meet the power purchasing needs of power sales companies in the medium and long term market. The transaction volume of the power sales company in the medium and long term market, β is the deviation assessment coefficient of the power sales company, which depends on the specific deviation assessment rules.

[0132] The revenue that the power sales company obtains from the user side through the centralized trading market is related to the centralized trading electricity price signed by the user in the contract, namely:

[0133]

[0134] In formula (9), P s,k Q is the centralized trading electricity price agreed upon by the power sales company s and user k in the power purchase and sales contract; s,k The amount of electricity purchased by user k through centralized transactions.

[0135] From this, we can infer that the bidding target of the power sales company is:

[0136]

[0137] Construction of trading space: The future clearing price is predicted by time series analysis, and the price curve is drawn according to the prediction results. The clearing price curve is selected as the upper limit of the transaction, and the marginal cost price of pumped storage is selected as the lower limit, which is used as the trading space of pumped storage;

[0138] In general, the transaction results of the electricity market are considered to be the consequences of the decision-making behaviors of various market players, and the generation of these decision-making behaviors is closely related to the transaction space of the electricity market. The upper and lower limits of bidding stipulated by the electricity market are the inherent bidding ranges of all market players. In this range, the upper limit of the bidding space available for pumped storage power stations is the highest bidding value stipulated by the market, and the lower limit is the marginal cost price. The upper limit of the bidding space available for power users is the historical average highest bid price, and the lower limit is the lowest bidding value stipulated by the market. In a given market environment, market players formulate their own strategic behaviors based on the market system, and the common strategic behaviors of all market players jointly form the transaction results of the market.

[0139] Trading subject strategy. The trading space can provide the upper and lower limits of the trading price to the decision makers of power generators, ensuring the scientificity and rationality of the price adjustment process and the final bidding plan. The strategy set of trader i is X m ={x i1 ,x i2 ,L,x im}, initialize the Q value corresponding to each strategy. Before the first round of bidding begins, assign the same Q to each strategy. im (0)=0.

[0140] Trading simulation parameter setting. The trading model needs to set parameters according to the regional trading rules to determine the trading space. Set the number of experimental plates to O, and the maximum number of trading rounds per plate to T. Usually the exploration factor ε can be taken as a small constant, such as 0.01. The smaller ε is, the more times the agent tries new strategies. If the number of experimental trading rounds is set very large, it will take longer for the market to reach a convergence state, or even not converge. In order to speed up the convergence, ε can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (Formula 11):

[0141]

[0142] Among them, T is the maximum number of transaction rounds preset in the experiment, ε i0 The initial exploration probability pre-set for the experiment, ε it is the exploration probability of the tth round.

[0143] Quotation decision optimization model: Based on the establishment of decision-making objectives of market entities, the quotation strategy is selected within the scope of the pumped storage transaction space and input into the quotation decision optimization model for iteration until the number of transaction rounds t is greater than the preset T, trader i stops learning, and the optimal quotation strategy is obtained by outputting the last round;

[0144] Bidding strategy selection: Starting from the first round, a random number γ in the interval (0,1) is generated in each round, and the size of γ is compared with the exploration factor ε. If γ < ε, a random number is selected from the strategy set X with uniform probability. i If γ>ε, select a strategy from the strategy set X i Select the strategy with the largest current Q value. If there are multiple strategies with the same Q value, one is randomly selected.

[0145] Simulated market clearing: After all traders submit their bids, the market is cleared according to the set clearing rules, and the clearing results such as the winning electricity volume and winning electricity price are fed back to each trader.

[0146] Update Q value and profit value: Calculate the profit of trader i based on the market clearing information and update the Q value. If the strategy selected by trader i in the tth round of trading is a ik , according to the winning bid electricity price and winning bid electricity quantity calculation strategy a ik The benefits brought by R ik (t), update the Q value of each strategy in the next round.

[0147]

[0148] Among them, the profit function R i The calculation of is related to the decision-making objectives of specific market members. itis the reward discount coefficient of trader i in the tth round of trading, which serves to suppress previous experience. In order to speed up the market convergence, α can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (13):

[0149]

[0150] Update Q under the risk aversion model r Value: In traditional decision-making models, the Q value represents the cumulative average reward after a strategy is executed for a period of time. The calculation of the Q value only considers the benefits brought by the strategy. On this basis, the improved risk aversion model quantifies the risk of the strategy into the decision-making process by calculating the standard deviation of the historical return data of each strategy. The larger the standard deviation, the more unstable the strategy's return and the higher the risk; on the contrary, the smaller the standard deviation, the more stable the strategy's return and the lower the risk. The specific formula is as follows:

[0151]

[0152] Trader i based on the above improved Q r The value is optimized and Q in the strategy set X is selected. r The strategy with the largest value Bidding, that is, a strategy with relatively stable returns and relatively low risks in terms of standard deviation:

[0153]

[0154] In formula (14), H im It means that trader i executes strategy x every time im The historical return dataset, r im Yes H im Elements in Q im The calculation method is based on formula (12), r im is the cumulative reward considering risk; β is the personal risk preference coefficient of the decision maker, and its value range is [0,1]. The larger the value of β, the greater the risk aversion of the trader. When β = 1, the trader is extremely risk-averse, and the size of the return has no effect on the decision. The only goal is to minimize the risk. As β decreases, the risk gradually increases, and the power generator pays more attention to the return.

[0155] Output the optimal trading strategy. The bidding decision optimization model is iterated until the number of trading rounds t is greater than the preset T, and trader i stops learning. Output the bidding strategy obtained in the last round as the optimal trading strategy for the generator.

[0156] Example 2

[0157] The present invention also discloses a pumped storage unit centralized bidding transaction strategy optimization device, comprising the following contents:

[0158] Establish a bidding decision target module for market members: treat it as a transaction object in the real market; in the continuous trading process, according to the members' own and external environment information, propose market quotation strategies within the transaction space, and design quotation strategies according to the expected goals;

[0159] The module for establishing the bidding decision target of market members includes the following contents:

[0160] Power producer transaction modeling submodule: The specific calculation method is shown in formula (1):

[0161]

[0162] In formula (1), F p The power sales revenue of power generators in the medium and long-term market includes two parts: the power sales revenue from annual bilateral negotiation transactions and the power sales revenue from monthly centralized bidding; s,l and Q s,l are the transaction price and transaction volume of the generator in the first transaction of the bilateral negotiation transaction; m,i and Q m,i are the clearing electricity price and winning electricity volume declared by the generator in the i-th segment in the centralized market;

[0163] The profit of a power generator is the difference between the revenue from electricity sales and the cost. For most power generators, the goal of participating in the centralized bidding market is to maximize profits. The bidding decision objective function is:

[0164]

[0165] In formula (2), F p-profit Profits obtained by power generators from participating in medium- and long-term market transactions; C g is the marginal generation cost of the generator.

[0166] Electricity user transaction modeling submodule: Users who are eligible to directly participate in electricity market transactions participate in bidding transactions with the goal of obtaining lower electricity prices and effectively reducing corporate electricity costs. Their electricity purchase cost C in the centralized trading market is y for:

[0167]

[0168] In the formula, λ y,i and Q y,i are the clearing electricity price and winning bid volume declared by the electricity user in the i-th period in the medium and long term respectively;

[0169] In the user full electricity transaction mode, the deviation electricity generated is no longer settled according to the catalog electricity price, but needs to be processed through deviation assessment. The user needs to bear the deviation assessment cost. for:

[0170]

[0171] In the formula, λ r Penalize electricity price for deviation assessment; Q y The amount of electricity that electricity users need to purchase in the monthly centralized market; It is the electricity transaction volume of electricity users in the monthly centralized market, calculated as follows:

[0172]

[0173] Therefore, the market decision objective function of power users in the full power transaction mode is:

[0174]

[0175] Electricity user transaction modeling submodule: Assuming that the remaining electricity demand of the power sales company needs to be traded through the monthly centralized bidding, the power purchase cost C of the power sales company in the monthly centralized market is s for:

[0176]

[0177] In the formula, λ s,i and Q s,i are the clearing electricity price and transaction volume declared by the power sales company in the i-th segment in the centralized bidding market;

[0178] Assuming that the power sales company bears all the deviation assessment for the user, the deviation assessment cost of the power sales company in the monthly centralized market is for:

[0179]

[0180] In formula (8), λ s is the deviation assessment electricity price, Q s To meet the power purchasing needs of power sales companies in the medium and long term market. The transaction volume of the power sales company in the medium and long term market, β is the deviation assessment coefficient of the power sales company, which depends on the specific deviation assessment rules;

[0181] The revenue that the power sales company obtains from the user side through the centralized trading market is related to the centralized trading electricity price signed by the user in the contract, namely:

[0182]

[0183] In formula (9), Ps,k Q is the centralized trading electricity price agreed upon by the power sales company s and user k in the power purchase and sales contract; s,k The amount of electricity purchased by user k through centralized trading;

[0184] Therefore, the bidding objectives of the power sales company are:

[0185]

[0186] Pumped storage trading space establishment module: The future trading quotation curve is inferred through time series analysis, and the highest lower limit and the lowest upper limit are selected as the pumped storage trading space;

[0187] The module for establishing the pumped storage trading space includes the following contents:

[0188] Trading subject strategy submodule: The trading space can provide the upper and lower limits of the trading price to the decision makers of power generators, ensuring the scientificity and rationality of the price adjustment process and the final bidding plan; the strategy set of trader i is X m ={x i1 ,x i2 ,L,x im}, initialize the Q value corresponding to each strategy; before the first round of bidding begins, assign the same Q to each strategy im (0) = 0;

[0189] Trading simulation parameter setting submodule: The trading model needs to set parameters according to the regional trading rules to determine the trading space; set the number of experimental plates to O, and the maximum number of trading rounds per plate experiment to T; usually the exploration factor ε can be taken as a smaller constant. The smaller ε is, the more times the agent tries new strategies. If the number of experimental trading rounds is set very large, the market will take longer to reach a convergence state, or even not converge. In order to speed up the convergence speed, ε can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (Formula 11):

[0190]

[0191] Among them, T is the maximum number of transaction rounds preset in the experiment, ε i0 The initial exploration probability pre-set for the experiment, ε it is the exploration probability of the tth round.

[0192] Module for establishing optimal trading decisions: Based on the establishment of decision-making objectives of market entities, a quotation strategy is selected within the scope of the pumped storage trading space and input into the quotation decision optimization model for iteration until the number of trading rounds t is greater than the preset T, trader i stops learning, and the optimal quotation strategy is obtained by outputting the last round:

[0193] The module for establishing the optimal trading decision includes the following contents:

[0194] Quotation strategy selection submodule: Starting from the first round, a random number γ in the interval (0,1) is generated in each round, and the size of γ is compared with the exploration factor ε. If γ < ε, a random number is selected from the strategy set X with uniform probability. i If γ>ε, select a strategy from the strategy set X i Select the strategy with the largest current Q value. If there are multiple strategies with the same Q value, randomly select one from them.

[0195] c2: Simulated market clearing submodule: After all traders submit their bids, the market is cleared according to the set clearing rules, and the clearing results of the winning electricity volume and winning electricity price are fed back to each trader;

[0196] Update Q value and profit value submodule: Calculate the profit of trader i according to the market clearing information and update the Q value; if the strategy selected by trader i in the tth round of trading is a ik , according to the winning bid electricity price and winning bid electricity quantity calculation strategy a ik The benefits brought by R ik (t), update the Q value of each strategy in the next round:

[0197]

[0198] Among them, the profit function R i The calculation of is related to the decision-making objectives of specific market members. it is the reward discount coefficient of trader i in the tth round of trading, which is used to suppress previous experience. In order to speed up the market convergence, α can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (13):

[0199]

[0200] Update Q under the risk aversion model r Value submodule: In the traditional decision-making model, the Q value represents the cumulative average reward after the strategy is executed for a period of time. The calculation of the Q value only considers the benefits brought by the strategy. On this basis, the improved risk aversion model quantifies the risk of the strategy into the decision-making process by calculating the standard deviation of the historical return data of each strategy. The larger the standard deviation, the more unstable the return of the strategy and the higher the risk. On the contrary, the smaller the standard deviation, the more stable the return of the strategy and the lower the risk. The specific formula is as follows:

[0201]

[0202] Trader i based on the above improved Q r The value is optimized and Q in the strategy set X is selected. r The strategy with the largest value Bidding, that is, a strategy with relatively stable returns and relatively low risks in terms of standard deviation:

[0203]

[0204] In formula (14), H im It means that trader i executes strategy x every time im The historical return dataset, r im Yes H im Elements in Q im The calculation method is based on formula (12), r im is the cumulative reward considering risk; β is the decision maker’s personal risk preference coefficient, and its value range is [0,1]; the larger the value of β, the greater the risk aversion of the trader. When β=1, the trader is extremely risk-averse, and the size of the return has no effect on the decision. The only goal is to minimize the risk. As β decreases, the risk gradually increases, and the power generator pays more attention to the return.

[0205] The technical solution of the present invention constructs a transaction model for different market entities by analyzing the transaction characteristics of power generators, power users and power sales companies. Furthermore, by simulating the bidding decisions of different market entities in the power market, a trading strategy optimization model is constructed to establish a market that can continuously learn and adjust strategies during actual transactions. The model can construct a corresponding strategy set based on the private information and bidding rules of market members, simulating the intelligent behavior of market members to continuously learn new knowledge and adjust quotations during transactions. Finally, by introducing the goal of risk aversion, a risk control model suitable for medium- and long-term market transactions of pumped storage modules is constructed, thereby better reflecting the decision-making process and risk attitude of various entities in the power market. It helps pumped storage to reasonably arrange output strategies for different time periods, helps pumped storage power stations optimize market participation decisions, and formulates reasonable centralized bidding transaction strategies, so as to obtain better economic benefits in the market.

[0206] Embodiment 3:

[0207] Based on the same inventive concept, the present invention also provides an electronic device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in a computer storage medium to implement a corresponding method flow or a corresponding function, so as to implement the steps of a method in the above embodiment.

[0208] Embodiment 4:

[0209] Based on the same inventive concept, the present invention also provides a non-volatile storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in electronic devices and, of course, extended storage media supported by electronic devices. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of any method in the above embodiment.

[0210] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0211] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0212] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0213] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0214] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

[0215] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for optimizing centralized bidding transaction strategy for pumped storage units, characterized by: include: Establish the bidding decision-making objectives of market members and treat them as transaction objects in the real market; In the continuous trading process, according to the members' own and external environment information, market quotation strategies are proposed within the trading space, and quotation strategies are designed according to the expected goals; The future clearing price is predicted by time series analysis, and a price curve is drawn based on the prediction results. The clearing price curve is selected as the upper limit of the transaction, and the marginal cost price of pumped storage is selected as the lower limit, which is used as the trading space of pumped storage. Based on the establishment of decision-making objectives of market players, a quotation strategy is selected within the pumped storage transaction space and input into the quotation decision optimization model for iteration until the number of transaction rounds t is greater than the preset T, trader i stops learning, and the optimal quotation strategy is obtained by outputting the last round.

2. The method for optimizing the centralized bidding transaction strategy of pumped storage units according to claim 1 is characterized by: The bidding decision objectives of establishing market members include the following: Power producer transaction modeling: The calculation method is shown in formula (1): In formula (1), F p The power sales revenue of power generators in the medium and long-term market includes two parts: the power sales revenue from annual bilateral negotiation transactions and the power sales revenue from monthly centralized bidding; s,l and Q s,l are the transaction price and transaction volume of the generator in the first transaction of the bilateral negotiation transaction; m,i and Q m,i are the clearing electricity price and winning electricity volume declared by the generator in the i-th segment in the centralized market; The profit of the power generator is the difference between the power sales revenue and the cost, and its bidding decision objective function is: In formula (2), F p-profit Profits obtained by power generators from participating in medium- and long-term market transactions; C g is the marginal generation cost of the generator; Electricity user transaction modeling: the electricity purchase cost C in the centralized trading market y for: In the formula, λ y,i and Q y,i are the clearing electricity price and winning bid volume declared by the electricity user in the i-th period in the medium and long term respectively; In the user full power transaction mode, the deviation assessment cost C that the user needs to bear is y * for: In the formula, λ r Penalize electricity price for deviation assessment; Q y The amount of electricity that electricity users need to purchase in the monthly centralized market; It is the electricity transaction volume of electricity users in the monthly centralized market, calculated as follows: Therefore, the market decision objective function of power users in the full power transaction mode is: Electricity user transaction modeling: Assuming that the remaining electricity demand of the power sales company needs to be traded through the monthly centralized bidding, the power purchase cost C of the power sales company in the monthly centralized market is s for: In the formula, λ s,i and Q s,i are the clearing electricity price and transaction volume declared by the power sales company in the i-th segment in the centralized bidding market; Assuming that the power sales company bears all the deviation assessment for the user, the deviation assessment cost of the power sales company in the monthly centralized market is for: In formula (8), λ s is the deviation assessment electricity price, Q s To meet the power purchasing needs of power sales companies in the medium and long term market. The transaction volume of the power sales company in the medium and long term market, β is the deviation assessment coefficient of the power sales company, which depends on the specific deviation assessment rules; The revenue that the power sales company obtains from the user side through the centralized trading market is related to the centralized trading electricity price signed by the user in the contract, namely: In formula (9), P s,k Q is the centralized trading electricity price agreed upon by the power sales company s and user k in the power purchase and sales contract; s,k The amount of electricity purchased by user k through centralized trading; Therefore, the bidding objectives of the power sales company are:

3. The method for optimizing the centralized bidding transaction strategy of pumped storage units according to claim 1 is characterized by: The process of establishing the pumped storage trading space includes the following contents: Trading subject strategy: The strategy set of trader i is X m ={x i1 ,x i2 ,L,x im }, initialize the Q value corresponding to each strategy; before the first round of bidding begins, assign the same Q to each strategy im (0) = 0; Trading simulation parameter setting: Set the number of experimental plates to O, and the maximum number of trading rounds per plate to T; usually the exploration factor ε can be taken as a small constant. The smaller ε is, the more times the agent tries new strategies. If the number of experimental trading rounds is set very large, the market will take longer to reach a convergence state, or even fail to converge. In order to speed up the convergence speed, let ε gradually decrease with the increase of the number of trading rounds t. The calculation method is as follows (Formula 11): Among them, T is the maximum number of transaction rounds preset in the experiment, ε i0 The initial exploration probability pre-set for the experiment, ε it is the exploration probability of the tth round.

4. The method for optimizing the centralized bidding transaction strategy of pumped storage units according to claim 1 is characterized by: The process of establishing the optimal trading decision includes the following contents: Bidding strategy selection: Starting from the first round, a random number γ in the interval (0,1) is generated in each round, and the size of γ is compared with the exploration factor ε. If γ < ε, a random number is selected from the strategy set X with uniform probability. i If γ>ε, select a strategy from the strategy set X i Select the strategy with the largest current Q value. If there are multiple strategies with the same Q value, randomly select one from them. Simulated market clearing: After all traders submit their bids, the market is cleared according to the set clearing rules, and the clearing results of the winning electricity volume and winning electricity price are fed back to each trader; Update Q value and profit value: Calculate the profit of trader i based on market clearing information and update Q value; If the strategy chosen by trader i in round t is a ik , according to the winning bid electricity price and winning bid electricity quantity calculation strategy a ik The benefits brought by R ik (t), update the Q value of each strategy in the next round: Among them, the profit function R i The calculation of is related to the decision-making objectives of specific market members. it is the reward discount coefficient of trader i in the tth round of trading, which is used to suppress previous experience. In order to speed up the market convergence, α can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (13): c4: Update Q under the risk aversion model r Value: Q value represents the cumulative average reward after the strategy is executed for a period of time. The calculation of Q value only considers the benefits brought by the strategy. The risk of the strategy is quantified into the decision-making process by calculating the standard deviation of the historical benefit data of each strategy. The larger the standard deviation, the more unstable the strategy's benefits and the higher the risk. On the contrary, the smaller the standard deviation, the more stable the strategy's benefits and the lower the risk. The specific formula is as follows: Trader i based on the above improved Q r The value is optimized and Q in the strategy set X is selected. r The strategy with the largest value Bidding, that is, a strategy with relatively stable returns and relatively low risks in terms of standard deviation: In formula (14), H im It means that trader i executes strategy x every time im The historical return data set, r im Yes H im Elements in Q im The calculation method is based on formula (12), r im is the cumulative reward considering risk; β is the decision maker’s personal risk preference coefficient, and its value range is [0,1]; the larger the value of β, the greater the risk aversion of the trader. When β=1, the trader is extremely risk-averse, and the size of the return has no effect on the decision. The only goal is to minimize the risk. As β decreases, the risk gradually increases, and the power generator pays more attention to the return.

5. A device for optimizing centralized bidding transaction strategy for pumped storage units, characterized in that: Establish a bidding decision target module for market members: treat it as a transaction object in the real market; in the continuous trading process, according to the members' own and external environment information, propose market quotation strategies within the transaction space, and design quotation strategies according to the expected goals; Module for establishing pumped storage trading space: predict the future clearing price through time series analysis, draw a price curve based on the prediction results, select the clearing price curve as the upper limit of the transaction, and select the pumped storage marginal cost price as the lower limit, which is used as the pumped storage trading space; Module for establishing optimal trading decisions: Select a quotation strategy within the pumped storage trading space and input it into the quotation decision optimization model for iteration until the number of trading rounds t is greater than the preset T, trader i stops learning, and outputs the optimal quotation strategy for the last round.

6. A pumped storage unit centralized bidding transaction strategy optimization device according to claim 5, characterized in that: The module for establishing the decision-making objectives of the market subject includes the following contents: Power producer transaction modeling submodule: The calculation method is shown in formula (1): In formula (1), F p The revenue from electricity sales of power generators in the medium and long-term market includes two parts: the revenue from electricity sales from annual bilateral negotiated transactions and the revenue from electricity sales from monthly centralized bidding. λ s,l and Q s,l are the transaction price and transaction volume of the generator in the first transaction of the bilateral negotiation transaction; m,i and Q m,i are the clearing electricity price and winning electricity volume declared by the generator in the i-th segment in the centralized market; The profit of a power generator is the difference between the revenue from electricity sales and the cost. For most power generators, the goal of participating in the centralized bidding market is to maximize profits. The bidding decision objective function is: In formula (2), F p-profit Profits obtained by power generators from participating in medium- and long-term market transactions; C g is the marginal generation cost of the generator; Electricity user transaction modeling submodule: The electricity purchase cost C in the centralized trading market y for: In the formula, λ y,i and Q y,i are the clearing electricity price and winning bid volume declared by the electricity user in the i-th period in the medium and long term respectively; In the user's full power transaction mode, the deviation assessment method is used to handle the deviation assessment cost that the user needs to bear. for: In the formula, λ r Penalize electricity price for deviation assessment; Q y The amount of electricity that electricity users need to purchase in the monthly centralized market; It is the electricity transaction volume of electricity users in the monthly centralized market, calculated as follows: Therefore, the market decision objective function of power users in the full power transaction mode is: Electricity user transaction modeling submodule: Assuming that the remaining electricity demand of the power sales company needs to be traded through the monthly centralized bidding, the power purchase cost C of the power sales company in the monthly centralized market is s for: In the formula, λ s,i and Q s,i are the clearing electricity price and transaction volume declared by the power sales company in the i-th segment in the centralized bidding market; Assuming that the power sales company bears all the deviation assessment for the user, the deviation assessment cost of the power sales company in the monthly centralized market is for: In formula (8), λ s is the deviation assessment electricity price, Q s To meet the power purchasing needs of power sales companies in the medium and long term market. The transaction volume of the power sales company in the medium and long term market, β is the deviation assessment coefficient of the power sales company, which depends on the specific deviation assessment rules; The revenue that the power sales company obtains from the user side through the centralized trading market is related to the centralized trading electricity price signed by the user in the contract, namely: In formula (9), P s,k Q is the centralized trading electricity price agreed upon by the power sales company s and user k in the power purchase and sales contract; s,k The amount of electricity purchased by user k through centralized trading; Therefore, the bidding objectives of the power sales company are:

7. The device for optimizing centralized bidding transaction strategy for pumped storage units according to claim 5 is characterized by: The module for establishing the pumped storage trading space includes the following contents: Trading subject strategy submodule: The strategy set of trader i is X m ={x i1 ,x i2 ,L,x im }, initialize the Q value corresponding to each strategy; before the first round of bidding begins, assign the same Q to each strategy im (0) = 0; Trading simulation parameter setting submodule: Set the number of experimental plates to O, and the maximum number of trading rounds per plate experiment to T; usually the exploration factor ε can be taken as a small constant. The smaller ε is, the more times the agent tries new strategies. If the number of experimental trading rounds is set very large, the market will take longer to reach a convergence state, or even not converge. In order to speed up the convergence speed, ε can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (Formula 11): Among them, T is the maximum number of transaction rounds preset in the experiment, ε i0 The initial exploration probability pre-set for the experiment, ε it is the exploration probability of the tth round.

8. The device for optimizing centralized bidding transaction strategy for pumped storage units according to claim 1 is characterized by: The module for establishing the optimal trading decision includes the following contents: Quotation strategy selection submodule: Starting from the first round, a random number γ in the interval (0,1) is generated in each round, and the size of γ is compared with the exploration factor ε. If γ < ε, a random number is selected from the strategy set X with uniform probability. i If γ>ε, select a strategy from the strategy set X i Select the strategy with the largest current Q value. If there are multiple strategies with the same Q value, randomly select one from them. Simulated market clearing submodule: After all traders submit their bids, the market is cleared according to the set clearing rules, and the clearing results of the winning electricity volume and winning electricity price are fed back to each trader; Update Q value and profit value submodule: calculate the profit of trader i according to the market clearing information and update the Q value; If the strategy chosen by trader i in round t is a ik , according to the winning bid electricity price and winning bid electricity quantity calculation strategy a ik The benefits brought by R ik (t), update the Q value of each strategy in the next round: Among them, the profit function R i The calculation of is related to the decision-making objectives of specific market members. it is the reward discount coefficient of trader i in the tth round of trading, which is used to suppress previous experience. In order to speed up the market convergence, α can be gradually reduced as the number of trading rounds t increases. The calculation method is as follows (13): Update Q under the risk aversion model r Value submodule: By calculating the standard deviation of each strategy's historical return data, the risk of the strategy is quantified into the decision-making process. The larger the standard deviation, the more unstable the strategy's return and the higher the risk. On the contrary, the smaller the standard deviation, the more stable the strategy's return and the lower the risk. The specific formula is as follows: Trader i based on the above improved Q r The value is optimized and Q in the strategy set X is selected. r The strategy with the largest value x k * Bidding, that is, a strategy with relatively stable returns and relatively low risks in terms of standard deviation: In formula (14), H im It means that trader i executes strategy x every time im The historical return dataset, r im Yes H im Elements in Q im The calculation method is based on formula (12), r im is the cumulative reward considering risk; β is the decision maker’s personal risk preference coefficient, and its value range is [0,1]; the larger the value of β, the greater the risk aversion of the trader. When β=1, the trader is extremely risk-averse, and the size of the return has no effect on the decision. The only goal is to minimize the risk. As β decreases, the risk gradually increases, and the power generator pays more attention to the return.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored program, wherein the program controls the device where the non-volatile storage medium is located to execute the method according to any one of claims 1 to 4 when the program is executed.

10. An electronic device, characterized in that: It comprises a processor and a memory; the memory stores computer-readable instructions, and the processor is used to execute the computer-readable instructions, wherein the computer-readable instructions execute the method described in any one of claims 1 to 4 when executed.

Images