Electric vehicle V2G power regulation and control method based on proxy model

Abstract

The invention discloses an electric vehicle V2G power regulation and control method based on an agent model. The method comprises the following steps: firstly, constructing a training sample set according to historical V2G data of the electric vehicle; thirdly, training the proxy model based on the neural network by using the training sample set to obtain an electric vehicle power flexibility prediction model; then, electric vehicle parameters and behavior information corresponding to each user are obtained and input into the prediction model, the model outputs the power flexibility of the electric vehicle corresponding to the user, and the power grid sends an electric vehicle actual regulation and control instruction to the corresponding charging pile according to the power flexibility of the electric vehicle, so that the charging and discharging power of the electric vehicle is controlled; and finally, the real-time battery SOC of the electric vehicle is obtained, rolling prediction of the power flexibility of the electric vehicle is carried out, and dynamic regulation and control of the V2G power of the electric vehicle are achieved. According to the method, the power flexibility of the electric vehicle is quickly solved, and the power grid adjusting capability can be effectively improved.

Description

Technical Field

[0001] The present invention relates to an electric vehicle regulation method in the technical field of electric vehicles, and particularly to an electric vehicle V2G power regulation method based on a surrogate model. Background Art

[0002] Existing electric vehicle regulation means are mainly divided into two categories: centralized control and distributed control. The centralized control method can achieve a global optimal solution, but there are also certain limitations. On the one hand, this method ignores the protection of user privacy; on the other hand, due to the high computational complexity, it is prone to bring a large computational burden and is difficult to meet the high-efficiency requirements in practical applications. In contrast, the distributed control method makes up for the deficiencies of the centralized control to a certain extent. The distributed control method usually integrates the feasible power boundaries provided by users through an electric vehicle aggregator and submits them to the distribution system operator. Subsequently, the distribution system operator issues a regulation instruction according to the real-time demand of the power grid, so as to achieve the efficient and reasonable regulation of electric vehicles.

[0003] However, both the existing centralized control method and the distributed control method do not consider and quantify the power flexibility of electric vehicles. Therefore, it is necessary to propose an electric vehicle V2G power regulation method that considers the power flexibility of electric vehicles. Summary of the Invention

[0004] In order to solve the problems and requirements in the background art, the present invention provides an electric vehicle V2G power regulation method based on a surrogate model.

[0005] The technical solution of the present invention is as follows: I. An electric vehicle V2G power regulation method based on a surrogate model S1: Calculate the power flexibility of the electric vehicle according to the historical V2G data of the electric vehicle, and then construct a training sample set; S2: Use the training sample set to train a surrogate model based on a neural network to obtain an electric vehicle power flexibility prediction model; S3: Obtain the electric vehicle parameters and behavior information corresponding to each user and input them into the electric vehicle power flexibility prediction model. The model outputs the power flexibility of the electric vehicle corresponding to the user. The power grid sends an actual regulation instruction for the electric vehicle to the corresponding charging pile according to the power flexibility of the electric vehicle, so as to control the charging and discharging power of the electric vehicle; S4: Obtain the real-time battery SOC of the electric vehicle and perform a rolling prediction of the power flexibility of the electric vehicle to achieve dynamic regulation of the electric vehicle V2G power.

[0006] In the above S1, the power flexibility of the electric vehicle satisfies the following formula: S = [P- i,1 ,P + i,1 ]×[P - i,2 ,P + i,2 ]×…×[P - i,t ,P + i,t ] Where S is the power flexibility of the electric vehicle; P - i,1 ,P + i,1 are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time 1; P - i,2 ,P + i,2 are the upper and lower limits of the charging and discharging power allowed by electric vehicle i at time 2; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t, respectively. The time interval between two adjacent moments is △t.

[0007] In S1, the power flexibility of the electric vehicle is calculated based on the historical V2G data of the electric vehicle, specifically: First, the electric vehicle V2G power optimization problem corresponding to each user is constructed, and the electric vehicle V2G power optimization problem satisfies the following formula: maxP gap -σ·k P gap =min(P + i,t -P - i,t ),t=t i a ,… ,t i ε t i ε = min(t i a +ε,t i d ) σ = ∑ abs(P + i,t -P + i,t+1 )+ abs(P - i,t -P- i,t+1 ) ,t=t i a ,…,t i d -△t P + i,t ≥P - i,t ,t=t i a ,..., t i d Among them, P gap is the minimum value of the flexibility of the electric vehicle power in each period within the target time period; σ is the power fluctuation coefficient of the upper and lower limits of the electric vehicle power flexibility; k is the power fluctuation penalty coefficient; △t is the time interval between two adjacent moments; t i a is the arrival time of the i-th electric car; t i d is the departure time of the i-th electric vehicle; ε is the duration of the target time period; t i ε P gap The end time of the target time period; abs() is the sign of the absolute value; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t+1, respectively; Then, based on the historical V2G data of electric vehicles, the capacity constraints of the battery itself, the initial energy state of the vehicle, the minimum required power constraints when the user leaves, the maximum charging and discharging power constraints, and the energy balance constraints, the optimization method is used to solve the V2G power optimization problem of the electric vehicle corresponding to each user to obtain the power flexibility of the electric vehicle corresponding to the user.

[0008] The capacity constraint of the battery itself satisfies the formula e i,min ≤e i,t ≤e i,max ,t i a ≤t≤t i d , e i,min is the minimum energy of the battery of the ith electric vehicle, e i,max is the maximum energy of the battery of the ith electric vehicle, ei,t is the energy state of the i-th electric car at time t.

[0009] The initial energy state of the vehicle and the minimum required power constraint when the user leaves satisfy the formula e i need ≤e i init +∑P i,t △t,t=t i a ,…,t i d , P i,t is the power of the i-th electric car at time t, e i need is the minimum energy required by the electric vehicle battery when the i-th electric vehicle leaves; e i ini is the energy of the electric vehicle battery when the i-th electric vehicle arrives.

[0010] The energy balance constraint satisfies the formula e i,t+1 = e i,t + P i,t △t,e i,t is the energy state of the ith electric car at time t, e i,t+1 is the energy state of the i-th electric car at time t+1, P i,t is the power of the i-th electric car at time t.

[0011] 2. An electric vehicle V2G power control device based on agent model A training sample set generating unit, used for calculating the power flexibility of the electric vehicle according to the historical V2G data of the electric vehicle, and then constructing the training sample set; A model storage unit, used to store a neural network-based proxy model and corresponding network parameters; A training unit is used to train a proxy model based on a neural network using a training sample set, obtain trained network parameters and send them to a prediction unit; A prediction unit is used to obtain the electric vehicle parameters and behavior information corresponding to each user and input them into a trained neural network-based agent model, and the model outputs the power flexibility of the electric vehicle corresponding to the user; The power grid control unit is used to send actual control instructions of the electric vehicle to the corresponding charging pile according to the power flexibility of the electric vehicle, so as to control the charging and discharging power of the electric vehicle.

[0012] 3. A computer device The device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the electric vehicle V2G power control method based on an agent model when executing the computer program.

[0013] 4. A computer-readable storage medium The medium stores a computer program, and when the computer program is executed by a processor, the steps of the electric vehicle V2G power control method based on an agent model are implemented.

[0014] 5. A computer program product The product includes a computer program / instruction, which, when executed by a processor, implements the steps of the electric vehicle V2G power control method based on an agent model.

[0015] The beneficial effects of the present invention are: The method proposed in the present invention is based on electric vehicle parameters and user behavior information, and adopts an agent model to quantitatively evaluate the adjustable power flexibility in the future period. It can effectively aggregate electric vehicle resources for power grid operators, improve the power grid regulation capability, and help ensure the operation and maintenance management of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the method flow provided by an embodiment of the present invention.

[0017] Figure 2 1 is a schematic diagram of the quantification results of the V2G power flexibility of an electric vehicle provided in an embodiment of the present invention, wherein (a) is a schematic diagram of the quantification results of the V2G power flexibility of an electric vehicle provided in Example 2, and (b) is a schematic diagram of the quantification results of the V2G power flexibility of an electric vehicle provided in Example 1.

[0018] Figure 3 It is a schematic diagram of the proxy model replacement process provided by an embodiment of the present invention.

[0019] Figure 4 It is a schematic diagram of the quantification result of V2G power flexibility provided by an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of the rolling prediction results when the target period for maximizing the flexibility of electric vehicles provided in Example 1 is 1 hour.

[0021] Figure 6 This is a schematic diagram of the rolling prediction results provided in Example 2 when the target period for maximizing the flexibility of electric vehicles is the total stay time. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] In the description of the present invention, it is to be understood that the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0024] Example 1

[0025] like Figure 1 As shown, the present invention proposes an electric vehicle V2G power control method based on an agent model, which specifically includes the following steps: S1: Calculate the power flexibility of the electric vehicle based on the historical V2G data of the electric vehicle, and then construct a training sample set; the historical V2G data includes the user's electric vehicle parameters and behavior information, and each sample in the training sample set includes the user's electric vehicle parameters and behavior information and power flexibility.

[0026] The power flexibility of electric vehicles satisfies the following formula: S=[P - i,1 ,P + i,1 ]×[P - i,2 ,P + i,2 ]×…×[P - i,t ,P + i,t ] Where S is the power flexibility of the electric vehicle; P - i,1 ,P + i,1 are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time 1; P - i,2 ,P + i,2 are the upper and lower limits of the charging and discharging power allowed by electric vehicle i at time 2; P - i,t ,P +i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t, respectively. The time interval between two adjacent moments is △t, that is, the power state at each moment will last for △t time, and △t is 15 minutes.

[0027] The present invention quantifies the power flexibility S of electric vehicles in different time periods based on the inner box approximation method, and can achieve decoupling of the power feasible area in the time dimension.

[0028] The power flexibility of electric vehicles is calculated based on the historical V2G data of electric vehicles, specifically: First, construct the electric vehicle V2G power optimization problem corresponding to each user. The electric vehicle V2G power optimization problem satisfies the following formula: maxP gap -σ·k P gap =min(P + i,t -P - i,t ),t=t i a ,… ,t i ε t i ε = min(t i a +ε,t i d ) σ = ∑ abs(P + i,t -P + i,t+1 )+ abs(P - i,t -P - i,t+1 ) ,t=t i a ,…,t i d -△t P + i,t ≥P - i,t ,t=t i a ,..., t i d Among them, P gap is the minimum value of the flexibility of the electric vehicle power in each period within the target time period; σ is the power fluctuation coefficient of the upper and lower limits of the electric vehicle power flexibility; k is the power fluctuation penalty coefficient; t ia is the arrival time of the i-th electric car; t i d is the departure time of the i-th electric vehicle; ε is the duration of the target time period in the objective function, which is 1h here; t i ε P gap The end time of the target time period; abs() is the sign of the absolute value; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t+1, respectively; like Figure 2 As shown in (b), when P gap When the target period is selected as the nearest target period of the electric vehicle, the power flexibility of the nearest period is greatly improved. Although the power flexibility in the following periods is limited to a very small area, the power flexibility can be continuously updated through the continuous prediction and evaluation of rolling optimization, so that the power flexibility at the nearest moment is always increasing.

[0029] Then, based on the historical V2G data of electric vehicles, the capacity constraints of the battery itself, the initial energy state of the vehicle, the minimum required power constraints when the user leaves, the maximum charging and discharging power constraints, and the energy balance constraints, the optimization method is used to solve the V2G power optimization problem of the electric vehicle corresponding to each user to obtain the power flexibility of the electric vehicle corresponding to the user.

[0030] The capacity constraint of the battery itself satisfies the formula e i,min ≤e i,t ≤e i,max ,t i a ≤t≤t i d , e i,min is the minimum energy of the battery of the ith electric vehicle, e i,max is the maximum energy of the battery of the ith electric vehicle, e i,t is the energy state of the i-th electric vehicle at time t. The initial energy state of the vehicle and the minimum required power constraint when the user leaves satisfy the formula e i need ≤e i init +∑P i,t △t,t=t i a,…,t i d , P i,t is the power of the i-th electric car at time t, e i need is the minimum energy required by the electric vehicle battery when user i leaves, e i ini is the energy of the electric vehicle battery when user i arrives, and the maximum charge and discharge power constraint satisfies the formula P i,min ≤P i,t ≤P i,max , P i,max is the maximum charging power allowed for the i-th electric vehicle, P i,min is the maximum discharge power allowed by the i-th electric vehicle. The energy balance constraint satisfies the formula e i,t+1 = e i,t + P i,t △t,e i,t+1 is the energy state of the i-th electric car at time t+1.

[0031] S2: Use the training sample set to train the neural network-based proxy model so that it can effectively learn the mapping process of the original problem and obtain the electric vehicle power flexibility prediction model; the electric vehicle power flexibility prediction model replaces the original optimization method, and the process is as follows Figure 3 shown.

[0032] S3: Each user uploads his or her electric vehicle parameters and behavior information to the cloud. Electric vehicle parameters include electric vehicle battery capacity, maximum charge and discharge power, minimum battery SOC required for travel, and current battery SOC. Behavior information includes arrival time and departure time, where the current battery SOC is collected through the charging pile. Obtain the electric vehicle parameters and behavior information corresponding to each user and input them into the electric vehicle power flexibility prediction model. The model outputs the power flexibility of the electric vehicle corresponding to the user. The power grid sends the actual control instructions of the electric vehicle to the corresponding charging pile based on the power flexibility of the electric vehicle, thereby controlling the charge and discharge power of the electric vehicle; S4: Obtain the real-time battery SOC of the electric vehicle and make a rolling prediction of the power flexibility of the electric vehicle, that is, input the real-time battery SOC of the electric vehicle together with other electric vehicle parameters and behavior information into the electric vehicle power flexibility prediction model. The model continuously predicts the power flexibility of the electric vehicle. The power grid sends the actual control instructions of the electric vehicle to the corresponding charging pile according to the power flexibility of the electric vehicle, thereby controlling the charging and discharging power of the electric vehicle, thereby realizing dynamic control of the V2G power of the electric vehicle.

[0033] In this embodiment, 1000 electric vehicle charging events are selected as historical V2G data. Table 1 shows the performance of solving the power flexibility of electric vehicles using a proxy model (specifically, MLP multi-layer perceptron) and a mathematical optimization method (specifically, yalmip programming and solving with gurobi solver), where the accuracy of the proxy model is evaluated using the mean square error index. As can be seen from the table, the accuracy of the proxy model has achieved good results. Compared with the mathematical optimization method, the proxy model has achieved hundreds of times reduction in computing time, which fully proves the efficiency and practical application potential of the method proposed in the present invention.

[0034] Table 1 shows the performance of the proxy model and optimization method. from Figure 4 It can be seen that the electric vehicle V2G power control method based on the proxy model proposed in the present invention is almost completely consistent with the results obtained by the mathematical optimization model. This result verifies the high accuracy of the method proposed in the present invention.

[0035] Figure 5 The feasible domain of rolling evaluation of power flexibility of a typical electric vehicle is shown. The range of the target time period ε is set to 1h. The power feasible domain is re-evaluated after every 1h using the electric vehicle V2G power control method proposed in the present invention. It can be clearly observed that the power flexibility of electric vehicles, which are highly coupled with time, is closely related to the current battery energy state and historical power trajectory. The use of rolling prediction can significantly expand the range of the power feasible domain in the future period.

[0036] Example 2

[0037] like Figure 1 As shown, the present invention proposes an electric vehicle V2G power control method based on an agent model, which specifically includes the following steps: S1: Calculate the power flexibility of the electric vehicle based on the historical V2G data of the electric vehicle, that is, the power flexibility of the electric vehicle during all the stay periods, and then construct a training sample set; the historical V2G data includes the user's electric vehicle parameters and behavior information, and each sample in the training sample set includes the user's electric vehicle parameters and behavior information and power flexibility.

[0038] The power flexibility of electric vehicles satisfies the following formula: S=[P - i,1 ,P + i,1 ]×[P - i,2 ,P + i,2 ]×…×[P- i,t ,P + i,t ] Where S is the power flexibility of the electric vehicle; P - i,1 ,P + i,1 are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time 1; P - i,2 ,P + i,2 are the upper and lower limits of the charging and discharging power allowed by electric vehicle i at time 2; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t, respectively. The time interval between two adjacent moments is △t, that is, the power state at each moment will last for △t time, and △t is 15 minutes.

[0039] The present invention quantifies the power flexibility S of electric vehicles in different time periods based on the inner box approximation method, and can achieve decoupling of the power feasible area in the time dimension.

[0040] The power flexibility of electric vehicles is calculated based on the historical V2G data of electric vehicles, specifically: First, construct the electric vehicle V2G power optimization problem corresponding to each user. The electric vehicle V2G power optimization problem satisfies the following formula: maxP gap -σ·k P gap =min(P + i,t -P - i,t ),t=t i a ,… , t i d σ = ∑ abs(P + i,t -P + i,t+1 )+ abs(P - i,t -P - i,t+1 ) ,t=t i a ,…,t i d -△t P +i,t ≥P - i,t ,t=t i a ,..., t i d Among them, P gap is the minimum value of the flexibility of the electric vehicle power in each period within the target time period; σ is the power fluctuation coefficient of the upper and lower limits of the electric vehicle power flexibility; k is the power fluctuation penalty coefficient; △t is the time interval between two adjacent moments; t i a is the arrival time of the i-th electric car; t i d is the departure time of the ith electric car; abs( ) is the sign of the absolute value; t i d is the departure time of the i-th electric car; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t+1, respectively; Figure 2 (a) is P gap The power flexibility diagram obtained when the target period is selected as the entire stay period of the electric vehicle. At this time, the power of the electric vehicle only needs to change within its power flexibility range to meet the operation constraints of the electric vehicle. The energy change range is given by Figure 2 As shown in the figure above (a), the upper and lower boundaries correspond to the control results of the electric vehicle power at the upper and lower limits of power flexibility.

[0041] Then, based on the historical V2G data of electric vehicles, the capacity constraints of the battery itself, the initial energy state of the vehicle, the minimum required power constraints when the user leaves, the maximum charging and discharging power constraints, and the energy balance constraints, the optimization method is used to solve the V2G power optimization problem of the electric vehicle corresponding to each user to obtain the power flexibility of the electric vehicle corresponding to the user.

[0042] The capacity constraint of the battery itself satisfies the formula e i,min ≤e i,t ≤e i,max ,t i a ≤t≤t i d , e i,minis the minimum energy of the battery of the ith electric vehicle, e i,max is the maximum energy of the battery of the ith electric vehicle, e i,t is the energy state of the i-th electric vehicle at time t. The initial energy state of the vehicle and the minimum required power constraint when the user leaves satisfy the formula e i need ≤e i init +∑P i,t △t,t=t i a ,…,t i d , P i,t is the power of the i-th electric car at time t, e i need is the minimum energy required by the electric vehicle battery when user i leaves, e i ini is the energy of the electric vehicle battery when user i arrives, and the maximum charge and discharge power constraint satisfies the formula P i,min ≤P i,t ≤P i,max , P i,max is the maximum charging power allowed for the i-th electric vehicle, P i,min is the maximum discharge power allowed by the i-th electric vehicle. The energy balance constraint satisfies the formula e i,t+1 = e i,t + P i,t △t,e i,t+1 is the energy state of the i-th electric car at time t+1.

[0043] S2: Use the training sample set to train the neural network-based proxy model so that it can effectively learn the mapping process of the original problem and obtain the electric vehicle power flexibility prediction model; the electric vehicle power flexibility prediction model replaces the original optimization method, and the process is as follows Figure 3 shown.

[0044] S3: Each user uploads his or her electric vehicle parameters and behavior information to the cloud. Electric vehicle parameters include electric vehicle battery capacity, maximum charge and discharge power, minimum battery SOC required for travel, and current battery SOC. Behavior information includes arrival time and departure time, where the current battery SOC is collected through the charging pile. Obtain the electric vehicle parameters and behavior information corresponding to each user and input them into the electric vehicle power flexibility prediction model. The model outputs the power flexibility of the electric vehicle corresponding to the user. The power grid sends the actual control instructions of the electric vehicle to the corresponding charging pile based on the power flexibility of the electric vehicle, thereby controlling the charge and discharge power of the electric vehicle; S4: Obtain the real-time battery SOC of the electric vehicle and make a rolling prediction of the power flexibility of the electric vehicle, that is, input the real-time battery SOC of the electric vehicle together with other electric vehicle parameters and behavior information into the electric vehicle power flexibility prediction model. The model continuously predicts the power flexibility of the electric vehicle. The power grid sends the actual control instructions of the electric vehicle to the corresponding charging pile according to the power flexibility of the electric vehicle, thereby controlling the charging and discharging power of the electric vehicle, thereby realizing dynamic control of the V2G power of the electric vehicle.

[0045] Figure 6 A rolling evaluation feasible domain of power flexibility of a typical electric vehicle is demonstrated. The range of the target time period ε is set to the total stay time of the electric vehicle. The power feasible domain is re-evaluated after every 1 hour using the electric vehicle V2G power control method proposed in the present invention.

[0046] In the above two embodiments, since the range ε ​​of the target time period is set differently, the obtained flexibility intervals are also different. Figure 5 and Figure 6 It can be seen that Figure 5 The power flexibility benefits achieved are more significant, and the flexibility range obtained is larger. However, the rolling strategy has high performance requirements in terms of computing time. Unlike traditional mathematical methods that only need to calculate power flexibility once, the rolling method needs to continuously evaluate new flexibility areas as time goes on, which poses a higher challenge to the calculation speed of the model. To this end, the present invention introduces a proxy model, which can significantly reduce the calculation time in the process of solving the power flexibility optimization problem, thereby providing important help for practical engineering applications.

[0047] The above are preferred embodiments of the present invention. It should be noted that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for controlling V2G power of electric vehicles based on an agent model, characterized in that: The following steps are involved: S1: Calculate the power flexibility of electric vehicles based on the historical V2G data of electric vehicles, and then construct a training sample set; S2: Using the training sample set to train the neural network-based agent model to obtain an electric vehicle power flexibility prediction model; S3: Obtain the electric vehicle parameters and behavior information corresponding to each user and input them into the electric vehicle power flexibility prediction model. The model outputs the power flexibility of the electric vehicle corresponding to the user. The power grid sends the actual control instructions of the electric vehicle to the corresponding charging pile according to the power flexibility of the electric vehicle, thereby controlling the charging and discharging power of the electric vehicle; S4: Obtain the real-time battery SOC of the electric vehicle and make a rolling prediction of the power flexibility of the electric vehicle to achieve dynamic regulation of the V2G power of the electric vehicle.

2. The method for controlling the power of an electric vehicle V2G based on an agent model according to claim 1, characterized in that: In S1, the power flexibility of the electric vehicle satisfies the following formula: S=[P - i,1 ,P + i,1 ]×[P - i,2 ,P + i,2 ]×…×[P - i,t ,P + i,t ] Where S is the power flexibility of electric vehicle i; P - i,1 ,P + i,1 are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time 1; P - i,2 ,P + i,2 are the upper and lower limits of the charging and discharging power allowed by electric vehicle i at time 2; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t, respectively. The time interval between two adjacent moments is △t.

3. The method for controlling electric vehicle V2G power based on agent model according to claim 1, characterized in that: In S1, the power flexibility of the electric vehicle is calculated based on the historical V2G data of the electric vehicle, specifically: First, the electric vehicle V2G power optimization problem corresponding to each user is constructed, and the electric vehicle V2G power optimization problem satisfies the following formula: maxP gap -s·k P gap =min(P + i,t -P - i,t ),t=t i a ,… ,t i ε t i ε = min(t i a +ε,t i d ) σ=∑abs(P + i,t -P + i,t+1 )+ abs(P - i,t -P - i,t+1 ) ,t=t i a ,…,t i d -△t P + i,t ≥P - i,t ,t=t i a ,..., t i d Among them, P gap is the minimum value of the flexibility of the electric vehicle power in each period within the target time period; σ is the power fluctuation coefficient of the upper and lower limits of the electric vehicle power flexibility; k is the power fluctuation penalty coefficient; △t is the time interval between two adjacent moments; t i a is the arrival time of the i-th electric car; t i d is the departure time of the i-th electric vehicle; ε is the duration of the target time period; t i ε P gap The end time of the target time period; abs() is the sign of the absolute value; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t; P - i,t ,P + i,t are the upper and lower limits of the charging and discharging power allowed for electric vehicle i at time t+1, respectively; Then, based on the historical V2G data of electric vehicles, the capacity constraints of the battery itself, the initial energy state of the vehicle, the minimum required power constraints when the user leaves, the maximum charging and discharging power constraints, and the energy balance constraints, the optimization method is used to solve the V2G power optimization problem of the electric vehicle corresponding to each user to obtain the power flexibility of the electric vehicle corresponding to the user.

4. The method for controlling V2G power of an electric vehicle based on an agent model according to claim 3, characterized in that: The capacity constraint of the battery itself satisfies the formula e i,min ≤e i,t ≤e i,max ,t i a ≤t≤t i d , e i,min is the minimum energy of the battery of the ith electric vehicle, e i,max is the maximum energy of the battery of the ith electric vehicle, e i,t is the energy state of the i-th electric car at time t.

5. The method for controlling electric vehicle V2G power based on agent model according to claim 3 is characterized in that: The initial energy state of the vehicle and the minimum required power constraint when the user leaves satisfy the formula e i need ≤e i init +∑P i,t △t,t=t i a ,…,t i d , P i,t is the power of the i-th electric car at time t, e i need is the minimum energy required by the electric vehicle battery when the i-th electric vehicle leaves; e i ini is the energy of the electric vehicle battery when the i-th electric vehicle arrives.

6. The method for controlling electric vehicle V2G power based on agent model according to claim 3, characterized in that: The energy balance constraint satisfies the formula e i,t+1 = e i,t + P i,t △t,e i,t is the energy state of the ith electric car at time t, e i,t+1 is the energy state of the i-th electric car at time t+1, P i,t is the power of the i-th electric car at time t.

7. An electric vehicle V2G power control device based on an agent model, characterized in that: include: A training sample set generating unit, used for calculating the power flexibility of the electric vehicle according to the historical V2G data of the electric vehicle, and then constructing the training sample set; A model storage unit, used to store a neural network-based proxy model and corresponding network parameters; A training unit is used to train a proxy model based on a neural network using a training sample set, obtain trained network parameters and send them to a prediction unit; A prediction unit is used to obtain the electric vehicle parameters and behavior information corresponding to each user and input them into a trained neural network-based agent model, and the model outputs the power flexibility of the electric vehicle corresponding to the user; The power grid control unit is used to send actual control instructions of the electric vehicle to the corresponding charging pile according to the power flexibility of the electric vehicle, so as to control the charging and discharging power of the electric vehicle.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the electric vehicle V2G power control method based on the agent model as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the electric vehicle V2G power control method based on an agent model as described in any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for V2G power control of an electric vehicle based on an agent model as described in any one of claims 1 to 6 are implemented.

Images