A hierarchical charging strategy method for electric vehicles based on probability transfer matrix

Abstract

The present invention relates to the field of new energy electric vehicles, and provides a hierarchical charging strategy method for electric vehicles based on a probability transfer matrix, including: a main control center establishes an electric vehicle charging load prediction model based on a neural network, and takes the minimum peak-to-valley difference of the grid-side load as the objective function, optimizes the load curve and issues instructions to a secondary control center; the secondary control center formulates a charging plan by calculating a probability transfer matrix and broadcasts it to a group of controlled electric vehicles in the region; the electric vehicle updates the charging plan according to the probability transfer matrix and local constraints. The neural network load prediction model established by the present invention based on the historical charging load data of electric vehicles can accurately predict the charging load curve of electric vehicles for the next day, and the electric vehicle charging strategy based on the probability transfer matrix can effectively optimize the charging plan of electric vehicles for the next day. The present invention is an orderly charging control strategy combining hierarchical and distributed methods, which can effectively achieve the goal of peak shaving and valley filling.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy electric vehicles, and in particular to a hierarchical charging strategy method for electric vehicles based on a probability transfer matrix. Background Art

[0002] The large-scale development of electric vehicles has brought about significant changes to the transportation system, while also having profound impacts on the power system and the entire energy system. First, the large-scale, unorganized access of electric vehicles to the grid for charging will pose severe challenges to the operation and planning of the power system, primarily due to increased load and impacts on power quality. This will increase the difficulty of optimizing and controlling grid operations, and will place new demands on distribution network planning and operation. At the same time, the large-scale access of electric vehicles to the grid also presents significant opportunities for the future development of the power grid and even the Energy Internet. Through advanced, orderly charging control, the impact of charging loads can be reduced, grid efficiency can be improved, and the deep integration and interaction of large-scale electric vehicles and smart grids can be achieved. This will promote the rapid development of the Energy Internet, with electric vehicles as one of its pillars, and has become a key development trend.

[0003] Currently, electric vehicle (EV) and power grid interaction strategies can be categorized into three types: centralized control, distributed control, and hierarchical control. Centralized control involves a centralized control center that aggregates information from all EVs, making decisions about charging plans for each EV and distributing them to each EV. This approach offers advantages in terms of simpler control concepts and easy implementation. However, its disadvantages lie in the fact that information optimization and storage are performed at the control center, placing a significant computational burden on the control center when EVs are connected to the grid on a large scale. Distributed control involves local decisions about EV charging plans. Each EV develops a charging plan based on control signals. The control center then modifies control signals based on the charging and discharging plans provided by the EVs, guiding the EVs to achieve their desired control objectives. Distributed control is suitable for optimizing the control of large, decentralized EVs, distributing computational resources to each EV, reducing the burden on the control center and computation time. However, distributed control can pose a problem when EVs are connected to the grid on a large scale, leading to significant peak-to-valley load variations and significant degradation of power quality. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a tiered charging strategy for electric vehicles based on a probability transfer matrix, which mainly solves the problem that when electric vehicles are connected to the power grid on a large scale, the load peak-to-valley difference is too large, thereby greatly affecting the power quality.

[0005] Hierarchical control decomposes large-scale electric vehicles into multiple controlled electric vehicle groups. Each controlled electric vehicle group is controlled by a control center to achieve orderly charging. Top-level control focuses on the coordination and cooperation between multiple controlled electric vehicle groups. Through hierarchical control, the interactive control problem of large-scale electric vehicles is transformed into the interactive problem of small groups, reducing the scale and difficulty of solving the optimization problem.

[0006] The present invention provides a hierarchical charging strategy method for electric vehicles based on a probability transfer matrix, comprising the following steps:

[0007] S1. Establish an electric vehicle charging load prediction model based on LSTM neural network based on the historical charging load data of electric vehicles in the region;

[0008] The electric vehicle charging load prediction model uses the logic gate of the LSTM neural network: input gate f t , output gate o t 、Forget Gate t Control the discarding or adding of data information;

[0009] The logic gate: input gate f t , output gate o t 、Forget Gate t The functions are composed of their own sigmoid functions and dot multiplication operations;

[0010] The input gate i t Used to control the amount of data information input; input gate i t Combined with a tanh function, the tanh function generates a new candidate vector Input gate i t for Each item in generates a value in [0,1], which controls how much data information is added. The cell state of the updated memory cell is:

[0011]

[0012] In formula (1), σ is the input gate sigmoid function, W i and W c is the input gate weight coefficient, b i and b c is the input gate bias, C t is the updated memory unit state;

[0013] Forget Gate f t is the output h of the previous unit t-1 and the input x of this unit tThe sigmoid function is used as input to control how much data information of the historical state t-1 is allowed to flow into time t. The output is expressed as:

[0014] f t =σ(W f ·[h t-1 ,x t ]+b f ) (2)

[0015] In formula (2), σ is the forget gate sigmoid function, W f is the forget gate weight coefficient, x t h is the input feature of regional weather information, holiday conditions and travel purpose factors; t-1 is the output of the previous unit, b f Forget gate bias;

[0016] The output gate o t It is used to control the filtering of the state information at time t, generating a value in [0,1] for each item of the state information at time t, and controlling the degree to which the unit state is filtered; the output gate o t The expression is:

[0017]

[0018] In formula (3), σ is the output gate sigmoid function, W o is the output gate weight coefficient, b o is the output gate deviation, h t The output of the LSTM neural network is the prediction result of the electric vehicle charging load in the region for the next day, completing the electric vehicle charging load prediction model based on the LSTM neural network;

[0019] S2, the main control center optimizes the charging load curve of the electric vehicle charging load prediction model to obtain the charging power guidance curve H t , and sent to each secondary control center;

[0020] S3, each of the secondary control centers controls the charging power guidance curve H issued by the main control center. t Calculate the probability transfer matrix and broadcast it to the controlled electric vehicle group;

[0021] The method for issuing the charging power guidance curve includes: the objective function of the secondary control center following the charging power guidance curve is:

[0022]

[0023] In formula (4), T is the time series {1, 2, 3, ..., T}; P i,tis the electric vehicle charging load of the ith secondary control center at time t, H i,t The charging power guidance curve H that the main control center requires to follow t The component at time t; and minimizing the Euclidean distance between the total charging load curve of the electric vehicle and the charging power guidance curve by the objective function;

[0024] The method for calculating the probability transfer matrix includes: calculating the probability transfer matrix M:

[0025]

[0026] In formula (5), the element m in the i-th row and j-th column of the probability transfer matrix M is i,j represents the probability of transferring from state j to state i. Its physical meaning represents the probability of transferring the charging load of an electric vehicle at time j to time i. The peak period is defined as the period when the current load is greater than the charging guidance load, that is, P i,t ≥H i,t The valley period is defined as the period when the current load is less than the charging guidance load, that is, P i,t ≤H i,t The period of time, the charging load transfer period and the load non-transfer period are defined. The period to be transferred is the period when the electric vehicle charging load is greater than the charging guidance load curve, and the non-transfer period is the period when the electric vehicle charging load is less than the charging guidance load curve. The load surplus value Q of each period is calculated. t , as shown below:

[0027]

[0028] For peak period j, P i,t ≥H i,t , as shown below:

[0029]

[0030] In formula (7), i≠j and i∈T. For valley period j, P i,t ≤H i,t , as shown below:

[0031]

[0032] In formula (8), i≠j and i∈T.

[0033] S4. The electric vehicles in the controlled electric vehicle group update a charging plan according to the probability transfer matrix and local constraints.

[0034] Furthermore, the method for establishing the electric vehicle charging load prediction model of the LSTM neural network in step S1 includes: the main control center samples data of the historical charging load of electric vehicles in the area at 144 equally spaced time points within a day.

[0035] Furthermore, the method of optimizing the charging load curve of the electric vehicle charging load prediction model to obtain a charging power guidance curve and issuing the curve to each secondary control center in step S2 includes: the main control center optimizes the charging load curve to obtain a power guidance curve with the minimum peak-to-valley difference of the grid-side load as the objective function, and the objective function is:

[0036]

[0037] In formula (9), T is the time series {1, 2, 3, ..., T}, L t is the conventional power load in the area, and the charging load curve h t Optimize and obtain the power guidance curve H t ; The main control center will power guidance curve H t Passed down to each secondary control center.

[0038] Furthermore, the method for constructing the objective function of the secondary control center following the charging power guidance curve in step S3 includes:

[0039] In terms of model constraints, the power boundary constraints and energy boundary constraints need to be considered as follows:

[0040] The power boundary constraints and energy boundary constraints are:

[0041]

[0042] In formula (10), P i ,t E is the upper and lower limits of the total charging power of the electric vehicle at time t; i,t is the total energy that the electric vehicle has charged at time t; E i,t The upper and lower limits of the total charging energy of the electric vehicle at time t; the total charging energy of the electric vehicle E i，t The expression is:

[0043] E i,t =ΣP i,t ·ΔT (11)

[0044] In formula (11), ΔT is the time interval. The lower limit of the total charging power of electric vehicles in the distributed control model of secondary control center i is zero, and the upper limit of the total charging power of electric vehicles in the distributed control model of secondary control center i is:

[0045]

[0046] In formula (12), T i,n is the charging period of the i-th electric vehicle, is the rated charging power of the i-th electric vehicle; P is the sum of the rated charging power of electric vehicles in all charging periods including time t, limit,t is the upper limit of total power, and the upper limit of total charging power of electric vehicles at time t is the smaller value between the two;

[0047] Energy upper and lower bounds of the distributed control model for secondary control center i E i,t Satisfy the following formula:

[0048]

[0049] In formula (13), represents the energy required by the nth electric vehicle in the secondary control center i. The lower limit of the total charging energy of electric vehicles at time t is the total energy required by all electric vehicles that have completed charging at time t; the upper limit of the total charging energy of electric vehicles at time t is the total energy required by all electric vehicles that have started charging or completed charging at time t;

[0050] The charging guidance power curve is tracked by calculating the probability transfer matrix:

[0051] H i,t =M×P i,t (14).

[0052] Furthermore, the method of updating the charging plan of the electric vehicle according to the probability transfer matrix in combination with local constraints in step S4 includes:

[0053] The electric vehicle receives the updated charging plan for the next day issued by the secondary control center, and updates the charging plan in combination with its own local constraints. In general, the received matrix M is not a 0-1 matrix and cannot be directly used as the charging plan for the electric vehicle and may not meet the local constraints of each electric vehicle. Therefore, after receiving the probability transfer matrix M, the electric vehicle needs to make local corrections to M according to its own constraints before further updating the charging plan. According to the charging period constraints of the electric vehicle, M is corrected. In the non-rechargeable period, There is no charging load. The probability of the charging load in the non-charging period being transferred to other periods is 0, and the probability of the charging load in other periods being transferred to the non-charging period is also 0. The correction rule is as follows:

[0054]

[0055] Then, M is modified according to the current charging plan of the electric vehicle. If the electric vehicle does not charge in period j, the probability of transferring the charging load in period j to other periods is 0. If the nth electric vehicle does not charge at time j, the modification rule is as follows:

[0056] m t,j =0 (16)

[0057] If the nth electric vehicle is charged in period j, the probability that the charging load in other periods is transferred to period j is 0. The correction rule is as follows:

[0058] m j,t =0 (17)

[0059] The method of updating the charging plan of the electric vehicle according to the probability transfer matrix combined with local constraints in step S4 further includes:

[0060] Modify M to a 0-1 matrix. According to the roulette wheel selection principle, the probability of an element in the i-th row being selected is:

[0061]

[0062] If it falls into the i-th interval, then m i,j is selected and corrected to 1, and the other elements in row i and column j are corrected to 0. The correction rules are as follows:

[0063]

[0064] After the above corrections, a 0-1 matrix M′ is obtained, which ensures that the charging energy of the electric vehicle remains unchanged before and after the charging plan is updated.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] The present invention establishes a neural network load forecasting model based on the historical charging load data of electric vehicles, which can accurately predict the charging load curve of electric vehicles for the next day. The electric vehicle charging strategy based on the probability transfer matrix can effectively optimize the charging plan of electric vehicles for the next day. The present invention is an orderly charging control strategy that combines hierarchical and distributed methods, which can effectively achieve the goal of peak shaving and valley filling. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0068] In the attached figure:

[0069] Figure 1 This is a flow chart of a method for hierarchical charging strategy of electric vehicles based on a probability transfer matrix according to the present invention;

[0070] Figure 2 This is a flow chart of a hierarchical control strategy for orderly charging of electric vehicles based on a probability transfer matrix according to an embodiment of the present invention;

[0071] Figure 3 This is a schematic diagram of an LSTM neural network prediction model according to an embodiment of the present invention;

[0072] Figure 4 It is the charging load prediction result of the electric vehicle charging load prediction model based on the LSTM neural network model in an embodiment of the present invention;

[0073] Figure 5 It is the total power load prediction result of the electric vehicle charging load prediction model based on the LSTM neural network model in an embodiment of the present invention;

[0074] Figure 6 It is the charging power guidance curve of the electric vehicle issued by the main control center in area 1 of the embodiment of the present invention;

[0075] Figure 7 It is the charging power guidance curve for electric vehicles issued by the main control center in area 2 of the embodiment of the present invention;

[0076] Figure 8 It is the charging power guidance curve for electric vehicles issued by the main control center in area three of the embodiment of the present invention;

[0077] Figure 9 This is the load curve after optimization of the electric vehicle orderly charging strategy based on the probability transfer matrix in region 1 of the embodiment of the present invention;

[0078] Figure 10 This is the load curve after optimization of the electric vehicle orderly charging strategy based on the probability transfer matrix in region 2 of the embodiment of the present invention;

[0079] Figure 11 This is the load curve after optimization of the electric vehicle orderly charging strategy based on the probability transfer matrix in region three of the embodiment of the present invention. DETAILED DESCRIPTION

[0080] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of devices and products consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0081] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. As used in this disclosure and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0082] It should be understood that although the terms first, second, third, etc. may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining."

[0083] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. The embodiments of the present invention include but are not limited to the following embodiments.

[0084] Example

[0085] The embodiment of the present invention provides a hierarchical charging strategy method for electric vehicles based on a probability transfer matrix. Figure 1 As shown, the following steps are included:

[0086] S1. Establish an electric vehicle charging load prediction model based on LSTM neural network based on the historical charging load data of electric vehicles in the region;

[0087] S2, the main control center optimizes the charging load curve of the electric vehicle charging load prediction model to obtain the charging power guidance curve G i，t , and sent to each secondary control center;

[0088] S3, the secondary control center calculates the probability transfer matrix according to the charging power guidance curve issued by the main control center and broadcasts it to the controlled electric vehicle group;

[0089] S4. The electric vehicles in the controlled electric vehicle group update a charging plan according to the probability transfer matrix and local constraints.

[0090] In order to realize the orderly charging control strategy of large-scale electric vehicles, it is necessary to solve the problem of coordinated operation of multiple control centers and electric vehicles, and integrate distributed and hierarchical control modes, such as Figure 2As shown in Figure 1, this model architecture consists of three layers: a primary control center, secondary control centers, and a controlled electric vehicle cluster. The primary control center coordinates the secondary control centers. The secondary control centers employ a centralized control strategy to control the relatively concentrated distribution of electric vehicles within the grid, while a distributed control strategy provides orderly charging control for distributed electric vehicles. A next-day charging load forecasting model for electric vehicles is established within the primary control center. The primary control center samples historical charging load data for electric vehicles in the region at 144 equal intervals throughout the day, resulting in a relatively long time series of collected historical data.

[0091] A next-day charging load forecasting model for electric vehicles in a region is established based on the historical data of electric vehicle charging loads in the region. When the time series of historical data is too long, the use of recurrent neural networks (RNNs) will have the problem of reduced fitting accuracy. The long short-term memory neural network (LSTM) model controls the discarding or adding of information through its unique input gate, output gate, and forget gate. The logic gate function consists of a sigmoid function and a dot product operation. When processing long time series data, the three logic control units will selectively memorize or forget the information, effectively overcoming the shortcomings of recurrent neural networks that are prone to gradient vanishing and gradient exploding. The schematic diagram of the LSTM neural network is shown below. Figure 3 As shown, the present invention selects weather information, holiday conditions and travel purpose in three regions as input features x t , forget gate f t is the output h of the previous unit t-1 and the input x of this unit t The sigmoid function is used as input to control how much data information of the historical state t-1 is allowed to flow into time t. The output is expressed as:

[0092] f t =σ(W f ·[h t-1 ,x t ]+b f ) (2)

[0093] In formula (2), σ is the forget gate sigmoid function, W f is the forget gate weight coefficient, b f The input gate and a tanh function cooperate to control the addition of new information, and the tanh function generates a new candidate vector The input gate is Each item in produces a value in [0,1], which controls how much new information is added. The cell state of the updated memory cell is:

[0094]

[0095] In formula (1), σ is the input gate sigmoid function, W i and W c is the input gate weight coefficient, b i and b c is the input gate bias, C t is the updated memory cell state. The output gate is used to control the filtering of the current cell state information. First, the cell state is activated. The output gate generates a value in [0,1] for each item, controlling the degree to which the cell state is filtered. Its expression is:

[0096]

[0097] In formula (3), σ is the output gate sigmoid function, W o is the output gate weight coefficient, b o is the output gate deviation, h t is the output of the LSTM neural network, that is, the electric vehicle charging load information in the area for the next day, thereby completing the electric vehicle charging load prediction model based on the LSTM neural network model.

[0098] Based on the established electric vehicle charging load prediction model based on the LSTM neural network model, the main control center obtains the electric vehicle charging load curve in the region through the electric vehicle charging load prediction model based on the LSTM neural network model, such as Figure 4 As shown in Figure 1, this is the result of the electric vehicle charging load prediction model based on the LSTM neural network model for predicting the electric vehicle charging load in the three areas. It can be seen that the charging load of electric vehicles in the three areas is concentrated between 17:00 and 21:00; Figure 5 The figure below shows the predicted electric vehicle charging load curve combined with the conventional residential electricity load to generate the total load forecast curve for the three regions. The main control center optimizes the charging load curve using the minimum peak-to-valley difference in grid-side load as the objective function to generate the charging power guidance curve.

[0099] The main control center obtains the electric vehicle charging load curve h in the area through the electric vehicle charging load prediction model based on the LSTM neural network model. t The main control center optimizes the charging load curve with the minimum peak-to-valley difference of the grid side load as the objective function to obtain the power guidance curve. The objective function is:

[0100]

[0101] In formula (9), T is the time series {1, 2, 3, ..., T}, L t is the conventional power load in the area, and the charging load curve h t Optimize and obtain the power guidance curve H t ; The main control center will power guidance curve H t The main control center sends the charging power guidance curve to each secondary control center.

[0102] Based on the charging power guidance curve issued by the main control center, the secondary control center t By controlling the charging load of electric vehicles to avoid peak load periods and transfer charging to low load periods, the variance is reduced and the peak-valley difference is minimized. The essence of the peak shaving and valley filling control goal is to make the load curve as smooth as possible. In order to follow the charging power guidance curve, the objective function of the secondary control center is to minimize the Euclidean distance between the total charging load curve of electric vehicles and the charging power guidance curve. That is, the objective function is shown as follows:

[0103]

[0104] In formula (4), T is the time series {1, 2, 3, ..., T}; P i,t is the electric vehicle charging load of the ith secondary control center at time t, H i,t The charging power guidance curve H that the main control center requires to follow t The component at time t; and minimizing the Euclidean distance between the total charging load curve of the electric vehicle and the charging power guidance curve by the objective function;

[0105] In terms of model constraints, power boundary constraints and energy boundary constraints need to be considered, as shown in the following formula:

[0106]

[0107] In formula (10), P i ,t E is the upper and lower limits of the total charging power of the electric vehicle at time t; i,t is the total energy that the electric vehicle has charged at time t; E i,t The upper and lower limits of the total charging energy of the electric vehicle at time t; the total charging energy of the electric vehicle E i,t The expression is:

[0108] E i,t =∑P i,t·ΔT (11)

[0109] In formula (11), ΔT is the time interval. The lower limit of the total charging power of electric vehicles in the distributed control model of secondary control center i is zero, and the upper limit of the total charging power of electric vehicles in the distributed control model of secondary control center i is:

[0110]

[0111] In formula (12), T i,n is the charging period of the i-th electric vehicle, is the rated charging power of the i-th electric vehicle; P is the sum of the rated charging power of electric vehicles in all charging periods including time t, limit,t is the upper limit of total power, and the upper limit of total charging power of electric vehicles at time t is the smaller value between the two;

[0112] Energy upper and lower bounds of the distributed control model for secondary control center i E i,t Satisfy the following formula:

[0113]

[0114] In formula (13), Represents the energy required by the nth electric vehicle in the secondary control center i. The lower limit of the total charging energy of electric vehicles at time t is the total energy required by all electric vehicles that have completed charging at time t; the upper limit of the total charging energy of electric vehicles at time t is the total energy required by all electric vehicles that have started charging or completed charging at time t. Figure 6 , Figure 7 , Figure 8 As shown in the figure, it is the charging power guidance curve of electric vehicles issued by the main control center in the three areas.

[0115] The purpose of tracking the charging guidance power curve is achieved by calculating the probability transfer matrix, as shown in the following formula:

[0116] H i,t =M×P i,t (14)

[0117] The secondary control center calculates the probability transfer matrix M, as shown below:

[0118]

[0119] In formula (5), the element m in the i-th row and j-th column of the probability transfer matrix M is i,jrepresents the probability of transferring from state j to state i. Its physical meaning represents the probability of transferring the charging load of an electric vehicle at time j to time i. The peak period is defined as the period when the current load is greater than the charging guidance load, that is, P i,t ≥H i,t The valley period is defined as the period when the current load is less than the charging guidance load, that is, P i,t ≤H i,t The period of time, the charging load transfer period and the load non-transfer period are defined. The period to be transferred is the period when the electric vehicle charging load is greater than the charging guidance load curve, and the non-transfer period is the period when the electric vehicle charging load is less than the charging guidance load curve. The load surplus value Q of each period is calculated. t , as shown below:

[0120]

[0121] For peak period j, P i,t ≥H i,t , as shown below:

[0122]

[0123] In formula (7), i≠j and i∈T. For valley period j, P i,t ≤H i,t , as shown below:

[0124]

[0125] In formula (8), i≠j and i∈T.

[0126] The secondary control center updates the next day's electric vehicle charging plan based on the calculated probability transfer matrix and broadcasts the updated next day's electric vehicle charging plan to the controlled electric vehicle group.

[0127] The secondary control center issues the next day's electric vehicle charging plan. Electric vehicles in the region receive the updated next day's electric vehicle charging plan issued by the secondary control center and complete the electric vehicle charging plan update in combination with the constraints of each electric vehicle. In most cases, the charging plan received by the electric vehicle cannot meet the local constraints of each electric vehicle, so it cannot be directly executed according to the charging plan. Each electric vehicle should optimize the probability transfer matrix in combination with the constraints. There should be no charging load during the non-charging period, that is, the charging load of electric vehicles during the non-charging period cannot be transferred to other periods, and the charging load of any other period cannot be transferred to this period. The specific optimization rules are shown in the following formula:

[0128]

[0129] Then, M is modified according to the current charging plan of the electric vehicle. If the electric vehicle does not charge in period j, the probability of transferring the charging load in period j to other periods is 0. If the nth electric vehicle does not charge at time j, the modification rule is as follows:

[0130] m t,j =0 (16)

[0131] If the nth electric vehicle is charged in period j, the probability that the charging load in other periods is transferred to period j is 0. The correction rule is as follows:

[0132] m j,t =0 (17)

[0133] Further modify m to be a 0-1 matrix. According to the roulette wheel selection principle, the probability of an element in the i-th row being selected is:

[0134]

[0135] If it falls into the i-th interval, then m i,j is selected and corrected to 1, and the other elements in row i and column j are corrected to 0. The correction rules are as follows:

[0136]

[0137] After the above correction, a 0-1 matrix M′ is obtained. This can ensure that the charging energy of the electric vehicle remains unchanged before and after the charging plan is updated. After the electric vehicle updates the charging plan, it uploads the updated charging plan to the local secondary control center. Figure 9 , Figure 10 , Figure 11 As shown in Figure 1, this is the load curve in the region after the electric vehicle orderly charging control strategy is optimized based on the probability transfer matrix.

[0138] The technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is clearly not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0139] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A hierarchical charging strategy method for electric vehicles based on probability transfer matrix, characterized in that: The following steps are involved: S1. Combine the historical charging load data of electric vehicles in the region to establish an electric vehicle charging load prediction model based on LSTM neural network; The electric vehicle charging load prediction model is implemented through the logic gate of the LSTM neural network: input gate i t , output gate o t 、Forget gate t Control the discarding or adding of data information; The logic gate: input gate i t , output gate o t 、Forget gate t The functions are composed of their own sigmoid function and dot product operation; The input gate i t Used to control the amount of data information input; input gate i t and a tanh function, which generates a new candidate vector Input Gate i t for Each item in generates a value in [0,1], controlling how much data information is added. The cell state of the updated memory cell is: In formula (1), σ is the input gate sigmoid function, W i and W c is the input gate weight coefficient, b i and b c is the input gate deviation, C t is the updated memory unit state; Forget Gate t is the output of the previous unit h t-1 and the input x of this unit t The sigmoid function is used as the input to control how much data information of the historical state t-1 is allowed to flow into the t moment. The output is expressed as: f t =σ(W f ·[h t-1 ,x t ]+b f ) (2) In formula (2), σ is the forget gate sigmoid function, W f is the weight coefficient of the forget gate, x t Input features for regional weather information, holiday conditions, and travel purpose factors; h t-1 is the output of the previous unit, b f is the forget gate bias; The output gate o t It is used to control the filtering of the state information at time t, generating a value in [0,1] for each item of the state information at time t, and controlling the degree to which the unit state is filtered; the output gate o t The expression is: In formula (3), σ is the output gate sigmoid function, W o is the output gate weight coefficient, b o is the output gate deviation, h t The output of the LSTM neural network is the prediction result of the electric vehicle charging load in the region for the next day, and the electric vehicle charging load prediction model based on the LSTM neural network is completed; S2. The main control center optimizes the charging load curve of the electric vehicle charging load prediction model to obtain a charging power guidance curve H t , and sent to each secondary control center; S3, each of the secondary control centers is based on the charging power guidance curve H issued by the main control center t Calculate the probability transfer matrix and broadcast it to the controlled electric vehicle group; The method for issuing the charging power guidance curve includes: the objective function of the secondary control center following the charging power guidance curve is: In formula (4), T is the time series {1, 2, 3, ..., T}; P i,t is the electric vehicle charging load of the ith secondary control center at time t, H i,t The charging power guidance curve H given by the main control center is required to follow t The component at time t; through the objective function to minimize the Euclidean distance between the total charging load curve of the electric vehicle and the charging power guidance curve; The method for calculating the probability transfer matrix includes: calculating the probability transfer matrix M: In formula (5), the element m in the i-th row and j-th column of the probability transfer matrix M is i,j represents the probability of transferring from state j to state i. Its physical meaning represents the probability of transferring the charging load of an electric vehicle at time j to time i. The peak period is defined as the period when the current load is greater than the charging guidance load, that is, P i,t ≥H i,t The valley period is defined as the period when the current load is less than the charging guidance load, that is, P i,t ≤H i,t The period of time, the charging load transfer period and the load non-transfer period are defined. The waiting transfer period is the period when the electric vehicle charging load is greater than the charging guidance load curve, and the non-transfer period is the period when the electric vehicle charging load is less than the charging guidance load curve. The load surplus value Q of each period is calculated. t , as shown below: For peak period j, P i,t ≥H i,t , as shown below: In formula (7), i≠j and i∈T; for valley period j, P i,t ≤H i,t , as shown below: In formula (8), i≠j and i∈T; S4. The electric vehicles in the controlled electric vehicle group update the charging plan according to the probability transfer matrix combined with local constraints.

2. The electric vehicle tiered charging strategy method according to claim 1, characterized in that: The method for establishing an LSTM neural network electric vehicle charging load prediction model in step S1 includes: a main control center samples data of the historical charging load of electric vehicles in a region at 144 equally spaced time points within a day.

3. The electric vehicle tiered charging strategy method according to claim 1, characterized in that: The method of optimizing the charging load curve of the electric vehicle charging load prediction model to obtain a charging power guidance curve and issuing it to each secondary control center in step S2 includes: the main control center obtains the electric vehicle charging load curve h in the region through the electric vehicle charging load prediction model based on the LSTM neural network model t The main control center optimizes the charging load curve to obtain the power guidance curve with the minimum peak-to-valley difference of the grid load as the objective function. The objective function is: In formula (9), T is the time series {1, 2, 3, ..., T}, L t is the conventional power load in the area, and the charging load curve h t Optimize to get the power guidance curve H t ; The main control center will power guidance curve H t Passed down to each secondary control center.

4. The electric vehicle tiered charging strategy method according to claim 1, characterized in that: The method for constructing the objective function of the secondary control center following the charging power guidance curve in step S3 includes: the power boundary constraint and the energy boundary constraint are: In formula (10), P i ,t E is the upper and lower limits of the total charging power of electric vehicles at time t; i,t is the total energy that has been charged into the electric vehicle at time t; E i,t is the upper and lower limits of the total charging energy of the electric vehicle at time t; the total charging energy of the electric vehicle E i，t The expression is: AND i,t =∑P i,t ·ΔT (11) In formula (11), ΔT is the time interval. The lower limit of the total charging power of electric vehicles in the distributed control model of secondary control center i is zero, and the upper limit of the total charging power of electric vehicles in the distributed control model of secondary control center i is: In formula (12), T i,n is the charging period of the i-th electric vehicle, is the charging rated power of the i-th electric vehicle; is the sum of the rated charging power of electric vehicles in all charging periods including time t, P limit,t is the upper limit of total power, and the upper limit of total charging power of electric vehicles at time t is the smaller value between the two; Energy upper and lower limits of distributed control model for secondary control center i E i,t Satisfy the following formula: In formula (13), represents the energy required by the nth electric vehicle in the secondary control center i. The lower limit of the total charging energy of electric vehicles at time t is the total energy required by all electric vehicles that have completed charging at time t; the upper limit of the total charging energy of electric vehicles at time t is the total energy required by all electric vehicles that have started charging or completed charging at time t; The charging guidance power curve is tracked by calculating the probability transfer matrix: H i,t =M×P i,t (14)。 5. The electric vehicle tiered charging strategy method according to claim 1, characterized in that: The method for the electric vehicle in step S4 to update the charging plan according to the probability transfer matrix combined with local constraints includes: According to the charging period constraint of the electric vehicle, M is modified to There is no charging load. The probability of the charging load in the non-charging period transferring to other periods is 0, and the probability of the charging load in other periods transferring to the non-charging period is also 0. The correction rule is as follows: Then, M is corrected according to the current charging plan of the electric vehicle. If the electric vehicle does not charge in the j period, the probability of transferring the charging load in the j period to other periods is 0; if the nth electric vehicle does not charge at time j, the correction rule is as follows: m t,j ＝0 (16) If the nth electric vehicle is charged in period j, the probability of the charging load in other periods being transferred to period j is 0. The correction rule is as follows: m j,t ＝0 (17)。 6. The electric vehicle tiered charging strategy method according to claim 1, characterized in that: The method of updating the charging plan of the electric vehicle according to the probability transfer matrix combined with local constraints in step S4 also includes: Modify M to a 0-1 matrix. According to the roulette selection principle, the probability of being selected from the i-th row is: If it falls into the i-th interval, then m i,j is selected and corrected to 1, and other elements in the i-th row and j-th column are corrected to 0. The correction rules are as follows: After the above corrections, a 0-1 matrix M′ is obtained.

Images