A method and terminal for formulating an electricity price for an electric vehicle based on multi-agent double-layer game

Abstract

The application discloses a kind of based on multi-agent double-layer game electric vehicle price formulation method and terminal, using Markov-Fourier correction prediction model to carry out load prediction to distribution network, obtain actual load demand, according to the historical electricity consumption data and the price data corresponding thereto collected to construct electricity-price elasticity model, and based on the model constructs two-stage electric vehicle demand response model, electric vehicle aggregator alliance is used as game optimization leader, distribution network operator is used as game optimization follower, constructs double-layer transaction model based on principal-agent game, based on actual load demand and two-stage electric vehicle demand response model to solve double-layer transaction model based on principal-agent game, obtain optimal price strategy, fully consider the interaction and benefit coordination between distribution network operator and electric vehicle aggregator, realize the multi-level optimization from macro-control to micro-incentive, to obtain effective, reliable price strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity price setting, and in particular to a method and terminal for setting electricity prices for electric vehicles based on multi-agent double-layer game. Background Art

[0002] With the global energy transition and the rapid growth of the electric vehicle (EV) market, the large-scale integration of EVs has brought unprecedented challenges and opportunities to distribution networks. On the one hand, irrational charging behavior can lead to increased peak-to-valley variations in the power grid, increasing operational pressure. On the other hand, by intelligently regulating the EV charging process, EVs are considered a flexible distributed energy storage resource, effectively assisting the grid in peak load shifting and valley filling, thereby improving grid efficiency and reliability. Therefore, how to rationally design time-of-use electricity pricing mechanisms to guide EV users to charge or discharge during different time periods has become a key issue in current research on power market and EV integration.

[0003] Conventional electricity pricing methods are often based on historical load data and simple demand forecasting models. These methods lack dynamic considerations for user behavior and are unable to accurately reflect market supply and demand fluctuations and user-side elastic demand. Furthermore, most research focuses on single-level game analysis, overlooking the complex two-tier decision-making structure between distribution network operators and electric vehicle aggregators. Furthermore, existing demand response models and algorithms often suffer from limited computational efficiency and optimization effectiveness when dealing with large user groups, making them incapable of meeting the demands of real-time response and refined management.

[0004] In summary, the existing electricity price setting methods cannot obtain effective and reliable electricity price strategies. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an electric vehicle electricity price setting method and terminal based on multi-agent double-layer game, which can obtain an effective and reliable electricity price strategy.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for setting electricity prices for electric vehicles based on a multi-agent two-layer game, comprising the following steps:

[0008] Use the Markov-Fourier modified forecasting model to forecast the load of the distribution network and obtain the actual load demand;

[0009] Building an electricity quantity-electricity price elasticity model based on the collected historical electricity consumption data and electricity price data corresponding to the historical electricity consumption data, and building a two-stage electric vehicle demand response model based on the electricity quantity-electricity price elasticity model;

[0010] An electric vehicle aggregator alliance is used as a game optimization leader, and a distribution network operator is used as a game optimization follower. A two-tier transaction model based on a master-slave game is constructed according to the game optimization leader and the game optimization follower.

[0011] The two-layer transaction model based on the master-slave game is solved based on the actual load demand and the two-stage electric vehicle demand response model to obtain the optimal electricity price strategy.

[0012] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0013] An electric vehicle electricity price setting terminal based on a multi-agent two-layer game includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented:

[0014] Use the Markov-Fourier modified forecasting model to forecast the load of the distribution network and obtain the actual load demand;

[0015] Building an electricity quantity-electricity price elasticity model based on the collected historical electricity consumption data and electricity price data corresponding to the historical electricity consumption data, and building a two-stage electric vehicle demand response model based on the electricity quantity-electricity price elasticity model;

[0016] An electric vehicle aggregator alliance is used as a game optimization leader, and a distribution network operator is used as a game optimization follower. A two-tier transaction model based on a master-slave game is constructed according to the game optimization leader and the game optimization follower.

[0017] The two-layer transaction model based on the master-slave game is solved based on the actual load demand and the two-stage electric vehicle demand response model to obtain the optimal electricity price strategy.

[0018] The beneficial effects of the present invention are: using the Markov-Fourier corrected prediction model to predict the load of the distribution network to obtain the actual load demand, constructing an electricity-price elasticity model based on the collected historical electricity consumption data and the corresponding electricity price data, and constructing a two-stage electric vehicle demand response model based on the model, taking the electric vehicle aggregator alliance as the game optimization leader and the distribution network operator as the game optimization follower, constructing a two-layer trading model based on the master-slave game, solving the two-layer trading model based on the master-slave game based on the actual load demand and the two-stage electric vehicle demand response model, and obtaining the optimal electricity price strategy. In this way, through accurate load forecasting, electricity-price elasticity model, two-stage electric vehicle demand response model and two-layer trading model based on the master-slave game, the interaction and interest coordination between the distribution network operator and the electric vehicle aggregator are fully considered, and multi-level optimization from macro-control to micro-incentive is achieved, thereby obtaining an effective and reliable electricity price strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flowchart of a method for formulating electricity prices for electric vehicles based on a multi-agent two-layer game according to an embodiment of the present invention;

[0020] Figure 2 The present invention is a schematic structural diagram of an electric vehicle electricity price setting terminal based on a multi-agent two-layer game according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0022] Please refer to Figure 1 , a method for setting electricity prices for electric vehicles based on multi-agent two-level game, comprising the steps of:

[0023] Use the Markov-Fourier modified forecasting model to forecast the load of the distribution network and obtain the actual load demand;

[0024] Building an electricity quantity-electricity price elasticity model based on the collected historical electricity consumption data and electricity price data corresponding to the historical electricity consumption data, and building a two-stage electric vehicle demand response model based on the electricity quantity-electricity price elasticity model;

[0025] An electric vehicle aggregator alliance is used as a game optimization leader, and a distribution network operator is used as a game optimization follower. A two-tier transaction model based on a master-slave game is constructed according to the game optimization leader and the game optimization follower.

[0026] The two-layer transaction model based on the master-slave game is solved based on the actual load demand and the two-stage electric vehicle demand response model to obtain the optimal electricity price strategy.

[0027] From the above description, it can be seen that the beneficial effects of the present invention are: using the Markov-Fourier corrected prediction model to predict the load of the distribution network to obtain the actual load demand, constructing an electricity-price elasticity model based on the collected historical electricity consumption data and the corresponding electricity price data, and constructing a two-stage electric vehicle demand response model based on the model, taking the electric vehicle aggregator alliance as the game optimization leader and the distribution network operator as the game optimization follower, constructing a two-layer trading model based on the master-slave game, solving the two-layer trading model based on the master-slave game based on the actual load demand and the two-stage electric vehicle demand response model, and obtaining the optimal electricity price strategy. In this way, through accurate load forecasting, electricity-price elasticity model, two-stage electric vehicle demand response model and two-layer trading model based on the master-slave game, the interaction and interest coordination between the distribution network operator and the electric vehicle aggregator are fully considered, and multi-level optimization from macro-control to micro-incentive is achieved, thereby obtaining an effective and reliable electricity price strategy.

[0028] Furthermore, the use of the Markov-Fourier modified prediction model to perform load forecasting on the distribution network to obtain the actual load demand includes:

[0029] Accumulate the original power load sequences of the nodes in the distribution network to obtain the cumulative load sequence;

[0030] Constructing a first-order differential equation based on the cumulative load sequence, and performing mean calculation on the cumulative load sequence to obtain a mean load sequence;

[0031] Constructing a grey prediction equation based on the cumulative load sequence, the first-order differential equation and the mean load sequence;

[0032] Obtaining a cumulative load forecast sequence based on the grey forecast equation, and obtaining a power load forecast sequence according to the cumulative load forecast sequence;

[0033] Calculate the load residual sequence of the power load forecast sequence using a residual formula;

[0034] Performing residual correction on the load residual sequence using Fourier series to obtain a corrected load residual sequence;

[0035] Based on the original power load sequence, the modified load residual sequence is divided into states of a Markov chain to obtain a plurality of states, and the transition probabilities between the plurality of states are calculated according to a state transition probability formula to obtain a state transition probability matrix;

[0036] determining a fitting accuracy index value based on the state transition probability matrix and the plurality of states;

[0037] A future power supply load value is generated based on the fitting accuracy index value and the power load forecast sequence, and the future power supply load value is used as the actual load demand.

[0038] From the above description, it can be seen that the Markov-Fourier modified forecasting model takes into account the time series characteristics of historical load data, effectively improves the accuracy of load forecasting, and provides a more solid data foundation for subsequent electricity price design.

[0039] Furthermore, the performing residual correction on the load residual sequence by using Fourier series to obtain a corrected load residual sequence includes:

[0040] Converting the load residual sequence into a Fourier series sequence;

[0041] Calculating parameter estimates of the Fourier series sequence;

[0042] Perform residual correction on the load residual sequence based on the parameter estimation value to obtain a corrected load residual sequence.

[0043] From the above description, it can be seen that using Fourier series to correct residuals can improve the accuracy of the final prediction value.

[0044] Furthermore, the constructing of the electricity-price elasticity model based on the collected historical electricity consumption data and the electricity price data corresponding to the historical electricity consumption data includes:

[0045] Performing data cleaning and standardization on the collected historical electricity consumption data and the electricity price data corresponding to the historical electricity consumption data to obtain processed historical electricity consumption data and electricity price data;

[0046] The elasticity of electricity quantity to electricity price is estimated based on the processed historical electricity consumption data and electricity price data using statistical methods or econometric methods to obtain an electricity quantity-electricity price elasticity model.

[0047] From the above description, it can be seen that using statistical methods or econometric methods to estimate the elasticity of electricity consumption to electricity price based on processed historical electricity consumption data and electricity price data, and obtaining an electricity-price elasticity model, can more accurately quantify the changes in user electricity consumption behavior under different electricity price levels, thereby providing a theoretical basis for formulating more targeted electricity price strategies and ensuring the optimal electricity price strategy.

[0048] Furthermore, the constructing of a two-stage electric vehicle demand response model based on the electricity quantity-electricity price elasticity model includes:

[0049] In the first phase, the charging demand characteristics of EVs in the EV aggregator’s jurisdiction are obtained;

[0050] Calculating the change of electric vehicle load when declaring different electricity prices based on the electricity quantity-electricity price elasticity model according to the charging demand characteristics of the electric vehicle;

[0051] In the second stage, a user combination benefit model and green electricity proportion constraints, charge and discharge capacity constraints and distribution response power constraints corresponding to the user combination benefit model are constructed based on the changes in electric vehicle load when different electricity prices are declared.

[0052] From the above description, it can be seen that, unlike the traditional single-form demand response mechanism, the two-stage electric vehicle demand response model realizes the demand response of electric vehicle load in stages, which is conducive to electric vehicle aggregators to integrate a higher level of response power resources and reduce the electricity costs of electric vehicle users.

[0053] Furthermore, the user combination benefit model is:

[0054] I all =I c +I DR ;

[0055]

[0056]

[0057] Where, I all represents the user portfolio benefit, I c represents the additional carbon trading income, I DR represents the electric vehicle charging and discharging reward, α u represents the user profit coefficient, T DR represents the demand response period, P t represents the electricity price during period t, Q′ DR,t represents the second stage charge and discharge load of the user group during period t, T DR,v2g represents the peak period of demand response discharge, T DR,g2v represents the valley period of demand response charging, L ev Indicates the mileage per unit of electricity of electric vehicles, D c represents the carbon emissions per unit mileage of fuel vehicles, ρ represents the carbon dioxide density, P c represents the carbon emission quota market clearing price, Q end,t represents the power demand at the end of the response phase, β t represents the proportion of green electricity in period t;

[0058] The green electricity ratio constraint is:

[0059]

[0060] Where Q gp,trepresents the green electricity generation in the upper power generation market during period t, Q all,t represents the total power generation in the upper power generation market during period t;

[0061] The charge and discharge capacity constraints are:

[0062] -Q' DR,i,t ≤0.5Q initial,i,t ,i=1,2,…,n;

[0063] Where Q' DR,i,t represents the discharge amount of electric vehicles of the user group during period t, Q initial,i,t represents the initial load, n represents the number of users;

[0064] The allocation response power constraint includes an allocation response power balance constraint and an allocation response power limit constraint;

[0065] The distribution response power balance constraint is:

[0066]

[0067] Where Q DR,i,t It represents the demand response power allocated by the distribution network operator;

[0068] The allocation response power limit constraint is:

[0069] 0≤|Q initial,i,t -Q PR,i,t |≤|Q DR,i,t |≤|Q DR,DSO,t |,i=1,2,…,n;

[0070] Where Q PR,i,t Indicates the remaining power of the electric vehicle after discharge, Q DR,DSO,t It represents the total response power preset by the distribution network operator during period t.

[0071] From the above description, it can be seen that in the case that the electric vehicle aggregator realizes load transfer in the first phase, by constructing a user portfolio profit model and obtaining the remaining load required to complete the demand response according to the demand response power expected to be allocated by the distribution network operator, a combination incentive package containing a higher level of electric vehicle charging and discharging rewards and additional carbon trading income is set to maintain user stickiness, and electric vehicle users participating in the demand response are arranged to perform peak shaving and valley filling tasks in designated time periods, so as to realize the demand response of deep scheduling incentives in the second phase, thereby continuously attracting users to participate in the demand response work.

[0072] Furthermore, the constructing of a two-tier transaction model based on a master-slave game according to the game optimization leader and the game optimization follower includes:

[0073] Determining a bidding strategy set of the electric vehicle aggregator alliance and a power allocation strategy set of the distribution network operator;

[0074] determining a slightly altruistic utility function of the EV aggregator alliance and a utility maximization function of the distribution network operator;

[0075] generating an upper-layer transaction model based on a master-slave game according to the power allocation strategy set of the distribution network operator and the utility maximization function of the distribution network operator;

[0076] generating a lower-level transaction model based on a master-slave game according to the bidding strategy set of the electric vehicle aggregator alliance and the slightly altruistic utility function of the electric vehicle aggregator alliance;

[0077] A double-layer transaction model based on the master-slave game is obtained according to the upper-layer transaction model based on the master-slave game and the lower-layer transaction model based on the master-slave game.

[0078] From the above description, it can be seen that the upper-level transaction model based on the master-slave game is generated according to the power allocation strategy set of the distribution network operator and the utility maximization function of the distribution network operator. The lower-level transaction model based on the master-slave game is generated according to the quotation strategy set of the electric vehicle aggregator alliance and the slightly altruistic utility function of the electric vehicle aggregator alliance. A two-level transaction model based on the master-slave game is established using game theory to solve the problem of interest distribution between users and power grids. It not only ensures that the power grid can be efficiently adjusted according to the load conditions, but also protects the economic interests of residential users and encourages users to actively participate in the demand response plan, thereby achieving mutual benefit and win-win results for both supply and demand sides.

[0079] Furthermore, the slightly altruistic utility function F of the electric vehicle aggregator alliance is league for:

[0080]

[0081] Where, F EVA,i represents the objective function of each electric vehicle aggregator, α r Indicates mild altruism factor, F DSO represents the utility maximization function of the distribution network operator;

[0082] The utility maximization function F of the distribution network operator DSO for:

[0083]

[0084] Where, P i,trepresents the electricity price in the t period of the day in the jurisdiction of the ith electric vehicle aggregator, γ1 represents the first fluctuation economic loss conversion coefficient of the load in the jurisdiction of the electric vehicle aggregator and the load of the regional distribution network, γ2 represents the second fluctuation economic loss conversion coefficient of the load in the jurisdiction of the electric vehicle aggregator and the load of the regional distribution network, Q i,ave represents the average daily load in the jurisdiction of the i-th electric vehicle aggregator, Q DSO,t represents the load of the distribution network in the distribution network operator’s area during period t, Q DSO,ave It represents the average daily load of the distribution network in the distribution network operator's area.

[0085] From the above description, it can be seen that since the electric vehicle aggregator alliance, which is in the position of optimization leader, is actually a follower of the distribution network operator's strategy based on the demand response effect, that is, there is both antagonistic competition and collusion to make concessions between the two, and they want to obtain more benefits in the long-term goal, then the leader must regard the follower's interests as part of the leader's interests. Therefore, in order to avoid the optimization causing excessive greed of the leader alliance and the reduction of the demand response effect of the follower distribution network operators, the objective function of the electric vehicle aggregator alliance participating in the master-slave game is set to a slightly altruistic utility function; when the demand response period electricity price declared by each electric vehicle aggregator in the management area is obtained, the distribution network operator will allocate the electricity obtained by each electric vehicle aggregator in each demand response period based on the utility maximization function composed of electricity sales revenue, load fluctuation loss in the electric vehicle aggregator's jurisdiction, and load fluctuation loss of the regional distribution network.

[0086] Furthermore, the two-layer transaction model based on the master-slave game is solved based on the actual load demand and the two-stage electric vehicle demand response model to obtain the optimal electricity price strategy, which includes:

[0087] Using a Cplex solver to solve the upper-layer transaction model based on the master-slave game, obtaining a solution result;

[0088] Based on the solution result, an improved genetic algorithm is used to solve the lower-level transaction model based on the master-slave game to obtain the optimal electricity price strategy.

[0089] As can be seen from the above description, combining the Cplex solver, which is easy to program and efficient for solving quadratic programming problems, with an improved genetic algorithm with strong scalability and strong search capabilities to solve the two-level trading model based on the master-slave game can efficiently solve the master-slave game problem between distribution network operators and electric vehicle aggregator alliances.

[0090] Please refer to Figure 2Another embodiment of the present invention provides an electric vehicle electricity price setting terminal based on multi-agent two-layer game, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the above-mentioned electric vehicle electricity price setting method based on multi-agent two-layer game is implemented.

[0091] The electric vehicle electricity price setting method and terminal based on multi-agent two-layer game of the present invention can be applied to electricity price setting scenarios, and the following is an explanation through specific implementation methods:

[0092] Please refer to Figure 1 , embodiment 1 of the present invention is:

[0093] A method for setting electricity prices for electric vehicles based on a multi-agent two-layer game, comprising the following steps:

[0094] S1. Use the Markov-Fourier modified prediction model to predict the load of the distribution network and obtain the actual load demand, which specifically includes S11-S19:

[0095] S11. Accumulate the original power load sequences of the nodes in the distribution network to obtain a cumulative load sequence.

[0096] Among them, the original power load sequence of the node is: (0) =(r (0) (1),r (0) (2),...,r (0) (n)), where r (0) (n) represents the original power load of the nth node;

[0097] The cumulative load sequence is: (1) =(r (1) (1),r (1) (2),…,r (1) (n)), Where r (1) (n) represents the cumulative load of the nth node, and k represents the kth node.

[0098] S12. Constructing a first-order differential equation based on the cumulative load sequence, and performing mean calculation on the cumulative load sequence to obtain a mean load sequence.

[0099] Since the cumulative load series has a gray finger distribution, a first-order differential equation is constructed as follows: In the formula, a represents the first estimated parameter, b represents the second estimated parameter, R (1) (t) represents the load, which is an unknown function;

[0100] The mean load sequence is:(1) =(z (1) (1),z (1) (2),...,z (1) (k)), z (1) (k)=0.5r (1) (k)+0.5r (1) (k-1), where z (1) (k) represents the kth mean load; therefore, the first-order differential equation is transformed into: (0) (k)+az (1) (k)=b, k=2,3,4,...,n.

[0101] S13. Constructing a grey prediction equation based on the cumulative load sequence, the first-order differential equation and the mean load sequence.

[0102] Specifically, the parameter vector is estimated using the least squares method with reference to the grey prediction method to obtain the third estimated parameter θ. In this way, the first estimated parameter a and the second estimated parameter b are obtained, and then the grey prediction equation is constructed: Represents load forecast parameters.

[0103] S14. Obtain a cumulative load prediction sequence based on the grey prediction equation, and obtain a power load prediction sequence according to the cumulative load prediction sequence.

[0104] Specifically, a cumulative load forecast sequence is obtained based on the grey forecast equation, and the cumulative load forecast sequence is subtracted to obtain a power load forecast sequence.

[0105] The cumulative load forecast sequence is: The power load forecast sequence is: Where, Indicates the nth power load forecast value.

[0106] S15. Calculate the load residual sequence of the power load forecast sequence using the residual formula, which is: x (0) (k) represents the actual load value at time k, Represents the load forecast value at time k.

[0107] S16, performing residual correction on the load residual sequence using Fourier series to obtain a corrected load residual sequence, specifically including S161-S163:

[0108] S161. Convert the load residual sequence into a Fourier series sequence, which is: In the formula, a0 represents the first constant, a i represents the second constant, b i represents the third constant, k a represents a finite Fourier series, and T represents the period.

[0109] S162: Calculate parameter estimates of the Fourier series sequence.

[0110] Specifically, using formula C a =(P T P) -1 P T E r Compute parameter estimates for a Fourier series sequence in, T=n-1,k a =(n-1) / 2-1.

[0111] S163: Perform residual correction on the load residual sequence based on the parameter estimation value to obtain a corrected load residual sequence.

[0112] S17. Based on the original power load sequence, the modified load residual sequence is divided into states of a Markov chain to obtain multiple states, and the transition probabilities between the multiple states are calculated according to a state transition probability formula to obtain a state transition probability matrix.

[0113] Among them, the multiple states are m represents the number of states, m, A i and B i According to the original data, represents the dot product of the vector, Y(t) represents the residual, A i Indicates the lower limit of the load value range covered by each state, B i Indicates the upper limit of the load value range covered by each state.

[0114] The transition probability P between the multiple states ij (k) State transition probability matrix Where, P ij (k) indicates the relevant parameters from After k steps, the state is transferred to The probability value of the state, M ij (k) represents the relevant parameters of the data. After k steps, the state is transferred to The number of original samples of the state, M i Indicates that The raw number of samples for the state.

[0115] S18. Determine a fitting accuracy index value based on the state transition probability matrix and the multiple states.

[0116] Specifically, let the state corresponding to the power load forecast at a certain moment in the past be the initial state By examining the row vector of the corresponding state transition probability matrix, the state transition probability vector of Y(t) at this moment is obtained as: P i (k)=(P i1 (k),P i2 (k),...P im (k)); by the m-order P i The matrix formed by (k) is called the m-order weighted state probability transfer matrix, and the weighted transition probabilities under the same state are used as the transition probability P of Y(t). i : Take the maximum P i The corresponding state is the weighted Markov state of Y(t) at that moment. After determining the state of the gray fitting accuracy index, the linear interpolation method is used to determine the fitting accuracy index value:

[0117] S19. Generate a future power supply load value based on the fitting accuracy index value and the power load forecast sequence, and use the future power supply load value as the actual load demand.

[0118] Among them, the future power supply load value at time n+1 is for:

[0119] S2. Build an electricity-price elasticity model based on the collected historical electricity consumption data and the electricity price data corresponding to the historical electricity consumption data, and build a two-stage electric vehicle demand response model based on the electricity-price elasticity model, specifically including S21-S25:

[0120] S21 . Perform data cleaning and standardization on the collected historical electricity consumption data and the electricity price data corresponding to the historical electricity consumption data to obtain processed historical electricity consumption data and electricity price data.

[0121] In an optional implementation, the historical electricity consumption data includes electricity consumption pattern data at different electricity price levels. For electric vehicles, charging records of charging stations also need to be considered.

[0122] S22. Use statistical methods or econometric methods to estimate the elasticity of electricity volume to electricity price based on the processed historical electricity consumption data and electricity price data, and obtain an electricity volume-electricity price elasticity model.

[0123] The statistical method or econometric method includes linear regression, nonlinear regression, piecewise linear model, etc. In an optional embodiment, after S22, the validity and robustness of the model can be tested by cross-validation, residual analysis, etc.

[0124] Compared to distribution network operators and electric vehicle aggregators, which rely on electricity purchase and sales transactions for profit, electric vehicle users do not place the same level of importance on charging costs. Furthermore, electric vehicle users are subject to a multi-period response model. This refers to users not simply reducing their electricity consumption but rather shifting their load from high-price periods to low-price periods. Therefore, multi-period response is not only dependent on the current electricity price, but also on prices at other times. Therefore, elasticity theory is more suitable for analyzing the relationship between charging and discharging volume and electricity prices for a specific user group, reducing the interference of the highly random nature of individual electric vehicle charging behavior and better revealing the collective patterns of response behavior among electric vehicle users within the same jurisdiction. Electric vehicle users participating in regional demand response receive a dynamic electricity price set by the electric vehicle aggregator (i.e., the price reported by the electric vehicle aggregator to the distribution network operator) on top of a fixed time-of-use electricity price during the demand response period. This price signal guides users to change their charging behavior.

[0125] The electricity-price elasticity model is defined as the ratio of the percentage change in electricity volume to the percentage change in electricity price. The electricity-price elasticity model is an important means to understand the sensitivity of electricity demand to changes in electricity prices, especially when designing demand response strategies. The model helps to quantitatively analyze how changes in electricity prices affect users' electricity consumption behavior, and then guides how to effectively use price mechanisms to regulate grid load, ensure supply and demand balance, reduce pressure during peak hours, and promote the absorption of renewable energy.

[0126] Assuming 1 hour as 1 period, the electricity consumption-price elasticity model E of 24 periods a day is:

[0127]

[0128] ε represents the electricity-price elasticity coefficient. The larger the coefficient, the greater the degree of change in electricity demand due to electricity price changes during the period. Electric vehicle users are more sensitive to price signals. In addition, the ε of price-sensitive user groups will be smaller, while the ε of price-oriented user groups will be larger overall. Among them, the electricity-price elasticity coefficient ε is further divided into the self-elasticity coefficient ε i，i and the cross elastic coefficient ε i，j Composition, i=1,2,…,24, j=1,2,…,24, self-elasticity coefficient ε i，i The change in charging cost in time period i affects the electricity demand in time period i; the cross elasticity matrix ε i，j Refers to the response of the change in electricity consumption in time period i to the change in charging cost in time period j.

[0129] The final electricity demand is calculated as follows:

[0130]

[0131] Where Q PR,t Indicates the changed power demand within time period t, Q initial,t Indicates the original power demand in time period t, ΔP t Indicates the change in electricity price within time period t, P initial,t P represents the original electricity price in time period t, which is subject to the quotation range: t =P initial,t +ΔP t , minP BR ≤P t ≤maxP BR , t∈T DR , P BR Indicates the quotation range, minP BR Indicates the lower limit of the quotation range, maxP BR Indicates the upper limit of the quotation range.

[0132] S23. In the first stage, the charging demand characteristics of electric vehicles in the jurisdiction of the electric vehicle aggregator are obtained.

[0133] S24. Calculate the change in electric vehicle load when declaring different electricity prices based on the electricity quantity-electricity price elasticity model according to the charging demand characteristics of the electric vehicle.

[0134] Electric vehicle aggregators use the electricity quantity-price elasticity model to reflect the sensitivity of electricity demand of each electric vehicle group to price changes based on the charging demand characteristics of electric vehicle users in their jurisdiction. On the basis of time-of-use electricity prices, they calculate the changes in electric vehicle load when declaring different electricity prices, thereby guiding the first stage of price signal demand response.

[0135] Under the condition of load transfer achieved in the first phase, the electric vehicle aggregator obtains the remaining load required to complete the demand response based on the demand response electricity expected to be allocated by the distribution network operator, and sets a combined incentive package containing a higher level of electric vehicle charging and discharging rewards and additional carbon trading income to maintain user stickiness, and arranges electric vehicle users participating in the demand response to perform peak shaving and valley filling tasks in designated time periods to achieve the demand response of deep scheduling incentives in the second phase, as described in S25.

[0136] S25. In the second stage, a user combination benefit model and green electricity proportion constraints, charge and discharge capacity constraints, and distribution response power constraints corresponding to the user combination benefit model are constructed based on the changes in electric vehicle load when different electricity prices are declared.

[0137] Among them, the second phase of deep scheduling incentive demand response will give users a combination of incentive packages with higher levels of electric vehicle charging and discharging rewards and additional carbon trading income, continuously attracting users to participate in demand response work. The carbon trading market follows the current domestic carbon market's "one-day-one-price" carbon emission price. The user combination income model is:

[0138] I all =I c +I DR ;

[0139]

[0140]

[0141] Where, I all represents the user portfolio benefit, I c represents the additional carbon trading income, I DR represents the electric vehicle charging and discharging reward, α u Indicates the user profit coefficient, which is 0.7, T DR represents the demand response period, P t represents the electricity price during period t, Q′ DR,t It represents the second stage charging and discharging load of the user group during period t. It is negative when discharging and positive when charging. It is 0 during the non-demand response period. DR,v2g represents the peak period of demand response discharge, T DR,g2v represents the valley period of demand response charging, L ev Indicates the unit power mileage of electric vehicles, which is 8.9km / kWh, D c represents the carbon emissions per unit mileage of fuel vehicles, which is 1.6L / km, ρ represents the carbon dioxide density, which is 1.96g / L, and P c represents the carbon emission quota market clearing price, which is 55rmb / t, Q end,t represents the power demand at the end of the response phase, β t represents the proportion of green electricity in period t;

[0142] The carbon emission quota obtained during the charging process of electric vehicles only comes from the green electricity portion of the power generation in the upper power generation market. The green electricity ratio constraint is:

[0143]

[0144] Where Q gp,t represents the green electricity generation in the upper power generation market during period t, Q all,t represents the total power generation in the upper power generation market during period t;

[0145] When electric vehicles within the jurisdiction of each electric vehicle aggregator obtain discharge revenue, the electric vehicle group must follow the electric vehicle aggregator's demand response resource integration arrangement to avoid excessive discharge causing new load troughs and achieve the required discharge capacity in the jurisdiction. Under this condition, the discharge capacity can be expressed as:

[0146] Q' DR,i,t =Q initial,i,t +Q DR,i,t -Q PR,i,t ,i=1,2,…,n,t∈T DR ;

[0147] Due to the user's vehicle usage requirements and the charging device's discharge power limit, it is stipulated that the maximum discharge of the user group's electric vehicles in period t shall not exceed half of the initial load. The charge and discharge capacity constraint is:

[0148] -Q' DR,i,t ≤0.5Q initial,i,t ,i=1,2,…,n;

[0149] Where Q' DR,i,t represents the discharge amount of electric vehicles of the user group during period t, Q initial,i,t represents the initial load, n represents the number of users;

[0150] The allocation response power constraint includes an allocation response power balance constraint and an allocation response power limit constraint;

[0151] Considering the consistency of the overall charging demand of the electric vehicle group within the jurisdiction of each electric vehicle aggregator, the scalar sum of the demand response power allocated by the distribution network operator is 0, so the balance constraint of the allocated response power is:

[0152]

[0153] Where Q DR,i,t It represents the demand response power allocated by the distribution network operator;

[0154] When allocating response power, the distribution network operator will ensure the first-stage demand response results of users within the electric vehicle aggregator guided by price signals. The response power obtained by a single electric vehicle aggregator in period t is not less than the power change caused by price signal guidance, and the response power obtained by a single electric vehicle aggregator in period t cannot be greater than the total response power preset by the distribution network operator in that period. Therefore, the allocation response power limit constraint is:

[0155] 0≤|Q initial,i,t -Q PR,i,t |≤|Q DR,i,t |≤|Q DR,DSO,t |,i=1,2,…,n;

[0156] Where QPR,i,t Indicates the remaining power of the electric vehicle after discharge, Q DR,DSO,t It represents the total response power preset by the distribution network operator during period t.

[0157] Finally, after the guidance of the first stage price signal demand response and the second stage deep scheduling incentive demand response, the load power of each electric vehicle aggregator can be expressed as: Q end,i,t =Q PR,i,t +Q' DR,i,t,i =1,2,…,n.

[0158] S3. The electric vehicle aggregator alliance is used as a game optimization leader, and the distribution network operator is used as a game optimization follower. A two-tier transaction model based on a master-slave game is constructed based on the game optimization leader and the game optimization follower.

[0159] The DNO publishes the DR power consumption for each DR period, a process that occurs only once per day. EV aggregators determine their bids for each DR period based on the DR power consumption for each period. Once the action (distribution operator power allocation) is taken, the DNO delivers it to the EV aggregator, who then acts as the leader in a master-slave game, updating the bid price based on their own benefits. The DNO then acts as the follower in the subsequent power allocation action, adjusting the amount of power each EV aggregator receives or needs to replenish during the DR period. This adjustment has the following benefits: first, the adjusted game order is more suitable for the specified DR transaction scenario, and the published DR power serves as a floor for DR adjustments, reducing the risk of poor DR effectiveness. This demonstrates the advantage of the EV aggregator's bid price, as the leader, over the dynamic pricing set by the DNO following the traditional game order. Second, while guaranteeing a guaranteed DR benefit, the DNO cedes power to the EV aggregator, leveraging the EV aggregator's strengths as an independent entity. This allows the EV aggregator to improve its own profitability while also enhancing DR benefits for the DNO's management area.

[0160] The Shapley method is widely used in the profit distribution activities of cooperative games. After the formation of the electric vehicle aggregator alliance, the Shapley method was introduced to distribute the profits obtained from the alliance's cooperative game according to contribution. The essence of the Shapley method is that each individual obtains corresponding benefits in the cooperative alliance based on its contribution, which can make the distribution of benefits among the electric vehicle aggregators in the alliance more fair.

[0161] The revenue of the electric vehicle aggregator alliance V(S), which is the charging and discharging revenue cost compensation part in the slightly altruistic utility function and the slightly altruistic part considering the utility target of the distribution network operator, can be expressed as:

[0162]

[0163] V(i), i∈I represents the benefits that the i-th electric vehicle aggregator can share in the decision-making alliance cooperation game composed of n electric vehicle aggregators. The Shapley benefit distribution is performed according to the following formula:

[0164]

[0165] Where n represents the number of EV aggregators, S represents the subset of decision-making entities participating in the EV aggregator alliance, |S| represents the number of EV aggregators in subset S, v(S) represents the revenue earned from cooperation among all EV aggregators in the EV aggregator alliance, and v(S / i) represents the revenue earned from the EV aggregator alliance’s cooperative operation after excluding entity i. Specifically, it includes S31-S36:

[0166] S31. The electric vehicle aggregator alliance is regarded as the game optimization leader, and the distribution network operator is regarded as the game optimization follower.

[0167] S32. Determine the quotation strategy set θp = {θ1, θ2, ..., θn} of the electric vehicle aggregator alliance and the power allocation strategy set λq = {λ1, λ2, ..., λn} of the distribution network operator, where θn represents the quotation made by the n-th electric vehicle aggregator in all demand response periods, and λn represents the demand response power allocated to the n-th electric vehicle aggregator by the distribution network operator in all demand response periods.

[0168] S33. Determine the slightly altruistic utility function of the electric vehicle aggregator alliance and the utility maximization function of the distribution network operator.

[0169] Among them, since the electric vehicle aggregator alliance in the optimization leader position is actually a follower of the strategy of the distribution network operator with the demand response effect as the main purpose, that is, there is both antagonistic competition and collusion to make concessions between the two, and to obtain more benefits in the long-term goal, then the leader must regard the follower's interests as part of the leader's interests. Therefore, in order to avoid the excessive greed of the leader alliance caused by optimization, which will lead to a decrease in the demand response effect of the follower distribution network operator, the slightly altruistic utility function F of the electric vehicle aggregator alliance is league for:

[0170]

[0171] Where, F EVA,i represents the objective function of each electric vehicle aggregator, α r Indicates a slight altruism factor of -0.1, F DSOrepresents the utility maximization function of the distribution network operator;

[0172] Considering that electric vehicle aggregators often take charging electric vehicle users an annual fixed service fee as their main profit model, the utility maximization function F of the distribution network operator is: DSO for:

[0173]

[0174]

[0175]

[0176]

[0177] Where, P i,t represents the electricity price in the t period of the day in the jurisdiction of the ith electric vehicle aggregator, γ1 represents the first fluctuation economic loss conversion coefficient of the load in the jurisdiction of the electric vehicle aggregator and the load of the regional distribution network, which is 0.03, γ2 represents the second fluctuation economic loss conversion coefficient of the load in the jurisdiction of the electric vehicle aggregator and the load of the regional distribution network, which is 0.05, Q i,ave represents the average daily load in the jurisdiction of the i-th electric vehicle aggregator, Q DSO,t represents the load of the distribution network in the distribution network operator’s area during period t, Q DSO,ave It represents the average daily load of the distribution network in the distribution network operator’s area, Q DSO,initial,t It represents the conventional load in the distribution network operator's management area during period t, excluding the load in the electric vehicle aggregator's jurisdiction.

[0178] In the solution process, there are Theorem 1 and Theorem 2. Theorem 1: If the bidding strategy sets of the members in the electric vehicle aggregator alliance are all compact convex sets, the slightly altruistic utility function of the electric vehicle aggregator alliance is continuous and non-concave on the domain of definition, and the original master-slave game problem has a Nash equilibrium solution, then When , the master-slave game considering a slightly altruistic equilibrium also has at least one equilibrium solution. Therefore, we must first prove the sufficient condition of Theorem 1: if the bidding strategy sets of members in the electric vehicle aggregator alliance are all compact convex sets, the slightly altruistic utility function of the electric vehicle aggregator alliance is continuous and non-convex in the domain of definition. The proof is as follows:

[0179] The bidding strategy sets of each electric vehicle aggregator follow the electricity price declaration range published by the distribution network operator. Therefore, the bidding strategy sets of the members of the electric vehicle aggregator alliance are all compact convex sets, and the slightly altruistic utility function is continuous on the definition domain formed by the bidding range. Prove that the slightly altruistic utility function is a non-convex function on the definition domain: Every time the game returns to a decision of an electric vehicle aggregator, the distribution network operator’s decision on Q DR The decision has been given, Q DRIt can be regarded as a constant. At this time, the slightly altruistic utility function is calculated with respect to P t =(P1,P2,...,P 24 ), we get:

[0180]

[0181]

[0182] Then conduct a mild altruistic utility function about P t 、P j The second-order partial derivative of (j≠t) gives:

[0183]

[0184]

[0185] It can be seen that the slightly altruistic utility function is about P t 、P j The second-order partial derivative of (j≠t) is not related to any variable, Q initial 、P initial , ε are constants. In the subsequent examples, the “det()” command of Matlab is used to quickly solve the slightly altruistic utility function about P t 、P j The value of the determinant corresponding to the Hessian matrix H1 composed of the second-order partial derivatives of (j≠t) can be verified to be less than or equal to 0 to prove that the slightly altruistic utility function is a non-convex function in the domain of definition.

[0186] Theorem 2: When the master-slave game model satisfies the following conditions, there is a unique equilibrium solution.

[0187] (1) The utility function of the game participants is a non-empty, continuous function of the game strategy set.

[0188] (2) The utility function of the game participants is a non-empty, continuous function of the game strategy set.

[0189] Proof: The following proves the existence and uniqueness of the equilibrium solution of the master-slave game problem composed of the electric vehicle aggregator alliance and the distribution network operator: (1) First, the strategy set constraints that the leader electric vehicle aggregator alliance and the follower distribution network operator need to follow, and the variables involved have no discrete variables; secondly, the utility function of the leader electric vehicle aggregator alliance, the utility function of the internal electric vehicle aggregator, and the utility function of the follower distribution network operator investment operator in the master-slave game. It can be seen that the utility functions of the game participants are all non-empty and continuous functions of the game strategy set. The utility functions of the game participants are non-empty and continuous functions of the game strategy set. (2) Every time the game returns to a distribution network operator's decision, the electric vehicle aggregator alliance's decision on P i,t The decision has been given, P i,t Can be regarded as a constant. In order to solve the utility maximization function of the distribution network operator with respect to Q DR,i,t According to the derivative rule of composite function, in the same derivable domain t∈T DR First, solve the utility maximization function about Q end,i,t The partial derivative of Q end,i,t About Q DR,i,t Substituting the derivative of back into the partial derivative result, the utility maximization function with respect to Q end,i,t The partial derivative of is:

[0190]

[0191] Substitute the above formula into The second-order partial derivatives of the utility maximization function of the distribution network operator are solved for four cases:

[0192] (1) With Q end,i,t Q with the same i and t end,i,t Second-order partial derivatives:

[0193]

[0194] (2) With Q end,i,t Q with the same i but different t end,i,j Second-order partial derivatives:

[0195]

[0196] (3) With Q end,i,t Q with the same i but different t end,k,t Second-order partial derivatives:

[0197]

[0198] (4) With Q end,i,t Q for different i and t end,k,j Second-order partial derivatives:

[0199]

[0200] The calculation results of A1, A2, A3, and A4 are all constants, so the utility maximization function is about Q end The second-order partial derivative of is equal to the utility maximization function with respect to Q DR The second-order partial derivative of . The Hessian matrix H2 of the utility maximization function is arranged as follows:

[0201]

[0202] Among them, B1 on the diagonal of H2 matrix represents the Q from the same electric vehicle aggregator in the same period t. DR The second-order derivative of , there are n B1; B2 in the rest of the H2 matrix represents the Q from different electric vehicle aggregators DR The second-order derivative of . The internal components of B1 and B2 are as follows:

[0203]

[0204]

[0205] Among them, A1 on the diagonal of B1 matrix represents the data from the same electric vehicle aggregator at the same time period t about Q DR The second-order derivative of the matrix B1 is the same as the number of demand response periods. The A2 in the rest of the matrix B1 represents the relationship between period t and period j from the same electric vehicle aggregator with respect to Q. DR The second derivative of ; A3 on the diagonal of B2 matrix represents the data from the i-th electric vehicle aggregator and the k-th electric vehicle aggregator about Q in the same period t DR The number of A3 is the same as the number of demand response periods; the A4 in the rest of the B1 matrix represents the relationship between the i-th electric vehicle aggregator period t and the k-th electric vehicle aggregator period j on Q DR The second derivative of .

[0206] It can be seen from this that the components of the Hessian matrix H2 of the utility maximization function are not related to any variable and are all constants. In the subsequent examples, the "det()" command of Matlab is also used to quickly solve the value of the determinant corresponding to the Hessian matrix H2. Verifying that the value is not equal to 0 proves that the utility maximization function is a continuous concave / convex function with respect to the respective game strategy sets.

[0207] S34. Generate an upper-layer transaction model based on a master-slave game according to the power allocation strategy set of the distribution network operator and the utility maximization function of the distribution network operator.

[0208] S35. Generate a lower-level transaction model based on the master-slave game according to the quotation strategy set of the electric vehicle aggregator alliance and the slightly altruistic utility function of the electric vehicle aggregator alliance.

[0209] S36 . Obtain a two-layer transaction model based on the master-slave game according to the upper-layer transaction model based on the master-slave game and the lower-layer transaction model based on the master-slave game.

[0210] S4. Solving the master-slave game-based two-layer transaction model based on the actual load demand and the two-stage electric vehicle demand response model to obtain an optimal electricity price strategy, specifically including S41-S42:

[0211] S41. Use a Cplex solver to solve the upper-layer transaction model based on the master-slave game to obtain a solution result.

[0212] Specifically, Yalmip is used to convert the master problem of the upper-level trading model based on the master-slave game into a form that can be processed by the Cplex solver. This problem is then passed to the Cplex solver for solution, resulting in a solution. The master problem involves distribution network operators setting electricity pricing strategies, taking into account the effectiveness of demand response, and optimizing overall grid operating costs and revenues. This typically involves a quadratic programming problem that considers the nonlinear relationship between cost, revenue, and load balance.

[0213] Yalmip provides a high-level syntax for defining complex optimization problems, including nonlinear ones. The Cplex solver is known for its powerful linear, mixed integer, and quadratic programming capabilities, enabling it to quickly find global or near-optimal solutions.

[0214] S42. Based on the solution result, an improved genetic algorithm is used to solve the lower-layer transaction model based on the master-slave game to obtain an optimal electricity price strategy.

[0215] Among them, the improved genetic algorithm includes elite retention strategy, adaptive crossover mutation probability or local search, etc., and the slave problem of the lower-level transaction model based on the master-slave game is that the electric vehicle aggregator optimizes the charging scheduling of its users according to the published electricity price to minimize the charging cost or maximize the user welfare.

[0216] The use of an improved genetic algorithm provides powerful search capabilities and good global optimization performance. The upper and lower layer transaction models are solved in an iterative or interactive manner to ensure that the solution is globally optimal or near-optimal.

[0217] The specific solution process is as follows:

[0218] Step 1: Generate the initial demand response period quotation population of the electric vehicle aggregator alliance;

[0219] Step 2: Calculate the charging plan of electric vehicle users adjusted based on the initial demand response bid of the electric vehicle aggregator in the jurisdiction;

[0220] Step 3: Calculate and allocate the demand response power of each EV aggregator in each demand response period according to the utility maximization function of the distribution network operator;

[0221] Step 4: Calculate the slightly altruistic utility function of the EV aggregator alliance for individuals in the population;

[0222] Step 5: Retain the individuals with the best slightly altruistic utility function of the electric vehicle aggregator alliance within the population;

[0223] Step 6: The remaining individuals except the elite individuals undergo possible crossover and mutation to form the offspring bidding strategy;

[0224] Step 7: Repeat steps 3-4 to select the top 50% individuals in the offspring with the slightly altruistic utility function of the electric vehicle aggregator alliance;

[0225] Step 8: Repeat steps 5 to 7 until the iteration converges and meets the preset iteration termination conditions, and output the optimal bidding individual and the distribution network operator's demand response power allocation strategy.

[0226] Among them, the preset iteration termination condition is: |F league (g+1)-F league (g)|≤η, g represents the number of iterations, and η represents a sufficiently small positive number.

[0227] This invention not only improves the flexibility and efficiency of electric vehicles in participating in demand response, but also promotes the harmonious coexistence of the power grid and electric vehicles through accurate load forecasting and sophisticated electricity price design, providing a powerful tool for achieving efficient operation of the power grid and sustainable development of energy.

[0228] Please refer to Figure 2 , the second embodiment of the present invention is:

[0229] A terminal for setting electricity prices for electric vehicles based on a multi-agent two-layer game comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the method for setting electricity prices for electric vehicles based on a multi-agent two-layer game in Example 1 is implemented.

[0230] In summary, the present invention provides an electric vehicle electricity price setting method and terminal based on multi-agent two-layer game, which uses the Markov-Fourier modified prediction model to predict the load of the distribution network to obtain the actual load demand, and constructs an electricity-price elasticity model based on the collected historical electricity consumption data and the corresponding electricity price data. A two-stage electric vehicle demand response model is constructed based on the model, and the electric vehicle aggregator alliance is used as the game optimization leader and the distribution network operator is used as the game optimization follower to construct a two-layer transaction model based on the master-slave game. The two-layer transaction model based on the master-slave game is solved based on the actual load demand and the two-stage electric vehicle demand response model to obtain the optimal electricity price strategy, thereby achieving the optimal electricity price strategy through accurate load forecasting, electricity-price elasticity model, and two-stage electric vehicle demand. The response model and the two-tier transaction model based on the master-slave game fully consider the interaction and interest coordination between distribution network operators and electric vehicle aggregators, realize multi-level optimization from macro-control to micro-incentives, and thus obtain an effective and reliable electricity price strategy; and, in the first stage, when the electric vehicle aggregator realizes load transfer, by constructing a user combination benefit model, the remaining load required to complete the demand response is obtained according to the demand response power expected to be allocated by the distribution network operator, and a combination incentive package with a higher level of electric vehicle charging and discharging rewards and additional carbon trading income is set to maintain user stickiness, and arrange electric vehicle users participating in demand response to perform peak shaving and valley filling tasks in designated time periods, so as to realize the demand response of deep scheduling incentives in the second stage, thereby continuously attracting users to participate in demand response work.

[0231] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for setting electric vehicle electricity prices based on multi-agent two-layer game, characterized in that: Including steps: Use the Markov-Fourier modified forecasting model to forecast the load of the distribution network and obtain the actual load demand; Building an electricity quantity-electricity price elasticity model based on the collected historical electricity consumption data and electricity price data corresponding to the historical electricity consumption data, and building a two-stage electric vehicle demand response model based on the electricity quantity-electricity price elasticity model; An electric vehicle aggregator alliance is used as a game optimization leader, and a distribution network operator is used as a game optimization follower. A two-tier transaction model based on a master-slave game is constructed according to the game optimization leader and the game optimization follower. Solving the master-slave game-based two-layer transaction model based on the actual load demand and the two-stage electric vehicle demand response model to obtain an optimal electricity price strategy; The use of the Markov-Fourier modified prediction model to perform load forecasting on the distribution network to obtain the actual load demand includes: Accumulate the original power load sequences of the nodes in the distribution network to obtain the cumulative load sequence; Constructing a first-order differential equation based on the cumulative load sequence, and performing mean calculation on the cumulative load sequence to obtain a mean load sequence; Constructing a grey prediction equation based on the cumulative load sequence, the first-order differential equation and the mean load sequence; Obtaining a cumulative load forecast sequence based on the grey forecast equation, and obtaining a power load forecast sequence according to the cumulative load forecast sequence; Calculate the load residual sequence of the power load forecast sequence using a residual formula; Performing residual correction on the load residual sequence using Fourier series to obtain a corrected load residual sequence; Based on the original power load sequence, the modified load residual sequence is divided into states of a Markov chain to obtain a plurality of states, and the transition probabilities between the plurality of states are calculated according to a state transition probability formula to obtain a state transition probability matrix; determining a fitting accuracy index value based on the state transition probability matrix and the plurality of states; generating a future power supply load value based on the fitting accuracy index value and the power load forecast sequence, and using the future power supply load value as the actual load demand; The constructing of a two-stage electric vehicle demand response model based on the electricity quantity-electricity price elasticity model includes: In the first phase, the charging demand characteristics of EVs in the EV aggregator’s jurisdiction are obtained; Calculating the change of electric vehicle load when declaring different electricity prices based on the electricity quantity-electricity price elasticity model according to the charging demand characteristics of the electric vehicle; In the second stage, a user combination benefit model and green electricity proportion constraints, charge and discharge capacity constraints and distribution response power constraints corresponding to the user combination benefit model are constructed based on the changes in electric vehicle load when different electricity prices are declared.

2. The electric vehicle electricity price setting method based on multi-agent double-layer game according to claim 1 is characterized in that: The method of performing residual correction on the load residual sequence by using Fourier series to obtain a corrected load residual sequence includes: Converting the load residual sequence into a Fourier series sequence; Calculating parameter estimates of the Fourier series sequence; Perform residual correction on the load residual sequence based on the parameter estimation value to obtain a corrected load residual sequence.

3. The electric vehicle electricity price setting method based on multi-agent double-layer game according to claim 1 is characterized in that: The constructing of the electricity-price elasticity model based on the collected historical electricity consumption data and the electricity price data corresponding to the historical electricity consumption data includes: Performing data cleaning and standardization on the collected historical electricity consumption data and the electricity price data corresponding to the historical electricity consumption data to obtain processed historical electricity consumption data and electricity price data; The elasticity of electricity quantity to electricity price is estimated based on the processed historical electricity consumption data and electricity price data using statistical methods or econometric methods to obtain an electricity quantity-electricity price elasticity model.

4. The electric vehicle electricity price setting method based on multi-agent double-layer game according to claim 1 is characterized in that: The user portfolio benefit model is: ； ； ； Where, I all represents the user portfolio benefit, I c represents the additional carbon trading income, I DR represents the electric vehicle charging and discharging reward, α u represents the user profit coefficient, T DR represents the demand response period, P t represents the electricity price during period t, represents the second stage charge and discharge load of the user group during period t, represents the peak period of demand response discharge, represents the valley period of demand response charging, L ev Indicates the mileage per unit of electricity of electric vehicles, D c represents the carbon emissions per unit mileage of fuel vehicles, ρ represents the carbon dioxide density, P c represents the carbon emission quota market clearing price, Indicates the power demand at the end of the response phase. represents the proportion of green electricity in period t; The green electricity ratio constraint is: ； Where, represents the green electricity generation in the upper power generation market during period t, represents the total power generation in the upper power generation market during period t; The charge and discharge capacity constraints are: ； Where, represents the discharge amount of electric vehicles of the user group during period t, represents the initial load, n represents the number of users; The allocation response power constraint includes an allocation response power balance constraint and an allocation response power limit constraint; The distribution response power balance constraint is: ； Where, It represents the demand response power allocated by the distribution network operator; The allocation response power limit constraint is: ； Where, Indicates the remaining power of the electric vehicle after discharge. It represents the total response power preset by the distribution network operator during period t.

5. The electric vehicle electricity price setting method based on multi-agent double-layer game according to claim 1 is characterized in that: The two-layer transaction model based on the master-slave game is constructed according to the game optimization leader and the game optimization follower, including: Determining a bidding strategy set of the electric vehicle aggregator alliance and a power allocation strategy set of the distribution network operator; determining a slightly altruistic utility function of the EV aggregator alliance and a utility maximization function of the distribution network operator; generating an upper-layer transaction model based on a master-slave game according to the power allocation strategy set of the distribution network operator and the utility maximization function of the distribution network operator; generating a lower-level transaction model based on a master-slave game according to the bidding strategy set of the electric vehicle aggregator alliance and the slightly altruistic utility function of the electric vehicle aggregator alliance; A double-layer transaction model based on the master-slave game is obtained according to the upper-layer transaction model based on the master-slave game and the lower-layer transaction model based on the master-slave game.

6. The electric vehicle electricity price setting method based on multi-agent double-layer game according to claim 5 is characterized in that: The slightly altruistic utility function F of the electric vehicle aggregator alliance league for: ； Where, represents the objective function of each electric vehicle aggregator, Indicates mild altruism factor, F DSO represents the utility maximization function of the distribution network operator; The utility maximization function F of the distribution network operator DSO for: ； Where, represents the electricity price during period t within the jurisdiction of the i-th electric vehicle aggregator, The first fluctuation economic loss conversion coefficient representing the load in the jurisdiction of the electric vehicle aggregator and the load in the regional distribution network, The second fluctuation economic loss conversion coefficient representing the load in the jurisdiction of the electric vehicle aggregator and the load in the regional distribution network, represents the average daily load in the jurisdiction of the i-th electric vehicle aggregator, represents the load of the distribution network in the distribution network operator’s area during period t, It represents the average daily load of the distribution network in the distribution network operator's area.

7. The electric vehicle electricity price setting method based on multi-agent double-layer game according to claim 5 is characterized in that: Solving the master-slave game-based two-layer transaction model based on the actual load demand and the two-stage electric vehicle demand response model to obtain the optimal electricity price strategy includes: Using a Cplex solver to solve the upper-layer transaction model based on the master-slave game, obtaining a solution result; Based on the solution result, an improved genetic algorithm is used to solve the lower-level transaction model based on the master-slave game to obtain the optimal electricity price strategy.

8. An electric vehicle electricity price setting terminal based on a multi-agent two-layer game, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the electric vehicle electricity price setting method based on multi-agent double-layer game according to any one of claims 1 to 7 is implemented.

Images