Charging station bidding method based on order classification prediction

Abstract

The invention discloses a charging station bidding method based on order classification prediction, and the method comprises the following steps: S1, obtaining charging station order data, and carrying out the standardization processing of the order data; s2, constructing a charging load classification prediction model of the charging station based on a long short-term memory network; s3, making a charging station invitation type bidding strategy according to the invitation type response; and S4, formulating charging station market type bidding strategies in different scenes according to the market type response. According to the method, optimization modeling is carried out on real factors such as graded subsidy and baseline difference in different response modes, so that the interactive benefits of the charging station in different scenes are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-grid interaction technology, and in particular to a charging station bidding method based on order classification prediction. Background Art

[0002] With the access of large-scale renewable energy generation, the adjustable capacity of the grid side is insufficient, and the system safety issues are becoming increasingly prominent. Electric vehicles, due to their controllable load characteristics, have gradually become the largest distributed flexibility resource. As a key node for electric vehicle energy replenishment, charging stations can flexibly participate in vehicle-grid integration and interaction based on demand response through deep integration with the electricity market, which not only brings economic benefits to themselves, but also provides support for the stable operation of the power grid. In this context, studying the operation strategy of charging stations in corresponding scenarios can not only provide practical decision-making support for charging station operators, but also provide theoretical support for more comprehensive vehicle-grid integration and interaction efficiency improvement, which has important engineering value and practical significance. In recent years, a lot of research has been carried out on the operation strategy of charging stations participating in vehicle-grid interaction, but there are still the following defects:

[0003] Charging load prediction is the prerequisite for charging stations to participate in vehicle-grid interaction. Accurately predicting future charging loads helps charging stations optimize energy scheduling and reduce operating costs. However, most of the current quantitative analysis of charging load demand for charging stations or a specific area does not take into account the charging power preferences of different electric vehicles, making it difficult for charging stations to guide and manage different charging users in a refined manner.

[0004] The bidding strategy is the core means for charging stations to participate in vehicle-grid interaction, which is directly related to the economic benefits and market competitiveness of charging stations. However, the current bidding optimization model is more inclined to theoretical research, lacking modeling and analysis of factors such as graded subsidies and baseline differences in the invitation-based and market-based interactive responses in the actual engineering context, making it difficult to apply to actual engineering. Summary of the invention

[0005] In order to solve the problems that it is difficult for current charging stations to guide and manage different charging users in a refined manner, and there is a lack of modeling and analysis of invitation-based and market-based interactive responses in actual projects, the present invention proposes a charging station bidding method based on order classification prediction, which can assist charging stations in screening stable loads to participate in interactive responses and enhance risk resistance. It also optimizes and models factors such as tiered subsidies and load baseline differences that exist in actual engineering applications, and proposes a bidding strategy for invitation-based and market-based interactive responses that is more in line with the actual project, so as to improve the demand response capability of charging stations and solve the above problems.

[0006] The present application discloses a charging station bidding method based on order classification prediction, comprising the following steps:

[0007] S1. Obtain charging station order data and standardize the order data;

[0008] S2. Construct a charging load classification prediction model for charging stations based on long short-term memory network;

[0009] S3. Formulate a charging station solicitation bidding strategy for solicitation responses;

[0010] S4. Develop market-based bidding strategies for charging stations in different scenarios based on market responses.

[0011] Preferably, the charging station order data is as follows:

[0012]

[0013] Where D represents the charging station order dataset, P t represents the total charging load of the charging station at time t, P t，i represents the charging power of electric vehicles of type i at time t, i = 1, 2, ..., n, and n is the total number of categories of electric vehicles obtained according to the charging power;

[0014] The formula for normalizing order data is as follows:

[0015]

[0016] Among them, μ i is the mean value of load data of category i, σ i is the standard deviation of load data for category i.

[0017] Preferably, the input of the charging load classification prediction model is the electric vehicle load data of the previous 24 hours. in The output of the charging load classification prediction model includes the charging power of n types of electric vehicles, and the output of the charging load classification prediction model is the predicted value of the charging power of the same n types of electric vehicles within 24 hours of the next day. in

[0018] The loss function of the charging load classification prediction model is:

[0019]

[0020] Where N is the number of samples, is the charging power predicted by the model, is the actual charging power.

[0021] Preferably, S3 comprises the following steps:

[0022] S31. Taking the hierarchical subsidy income and load regulation cost as the optimization objectives, a charging station demand response bidding model and model constraints are constructed;

[0023] S32. Relax the high-order terms in the charging station demand response bidding model constraints to obtain a final charging station demand response bidding model.

[0024] Preferably, the S31 comprises the following steps:

[0025] For the invitation-based response users, graded subsidies are provided. When the subsidy is settled, the response P t The determination of the load baseline L t related:

[0026]

[0027] Among them, L res，t represents the charging load after the charging station participates in the response, γ is a binary variable, when the system regulation demand is the peak shaving demand, γ = 0, when the system regulation demand is the valley filling demand, γ = 1, L t It is mainly based on the average load calculation of the same time period in the past few days. D represents the total number of days used to calculate the load baseline.

[0028] The charging station demand response bidding model is constructed as follows:

[0029]

[0030] P 1，min ≤P 1，t ≤P 1，max , P 2，min ≤P 2，t ≤P 2，max ; (12)

[0031] α 1，t , α 2，t ∈{0,1}; (13)

[0032] Among them, p1 is the short-term response subsidy price when the continuous response time does not exceed 2 hours, p2 is the long-term response subsidy price when the continuous response time exceeds 2 hours, and p s，t is the charging station to unit load adjustment fee at time t, P 1，t is the charging load after participating in short-term demand response at the corresponding time, P 2，t is the charging load after participating in long-term demand response at the corresponding time, P 1，min P 1，t Adjustment lower limit, P 1，max P 1，t Upper limit of regulation, P 2，min P 2，t Adjustment lower limit, P2，max P 2，t Upper limit of adjustment, L pre，t is the predicted value of charging load at the charging station, α 1，t , α 2，t is a binary variable, α 1，t =1 means participating in short-term demand response at time t, α 2，t =1 means participating in long-term response demand at time t;

[0033] α 1，t , α 2，t The following timing constraints are met:

[0034] α 1，t +α 1，t+1 +α 1，t+2 ≤2t=0,1,…,21; (14)

[0035] 2·(α 2，t -α 2，t-1 )≤α 2，t+1 +α 2，t+2 t=1,…,21; (15)

[0036] Among them, formula (14) means that when participating in short-term response, the continuous response time shall not exceed 2 hours, and formula (15) means that once participating in long-term response, the continuous response time must be greater than or equal to 3 hours;

[0037] For special points, the following constraints are satisfied:

[0038] 2α 2，0 ≤α 2，1 +α 2，2 ; (16)

[0039] 2α 2，21 ≥α 2，22 +α 2，23 ; (17)

[0040] α 1，t , α 2，t The following response mutual exclusion constraints are satisfied:

[0041] α 1，t +α 2，t =α t ; (18)

[0042] 2-α 1，t-1 -α 2，t+1 ≥α t t=1,2,…,22; (19)

[0043] 2-α 1，t+1 -α 2，t-1 ≥α tt=1,2,…,22; (20)

[0044] η 1，t =α 1，t ·(P 1，t -L t ), η 2，t =α 2，t ·(P 2，t -L t ); (twenty one)

[0045] η 1，t +η 2，t =η t ; (twenty two)

[0046] α t α t+1 ·η t =α t α t+1 ·η t+1 t=0,1,…,22; (23)

[0047] Among them, α t is a newly introduced binary variable. t =1, it means that the charging station participates in short-term or long-term demand response at time t. t =0 means that the charging station does not participate in the response at time t, η 1，t represents the bidding amount for short-term response at time t, η 2，t represents the bid amount for long-term response at time t, η t is the total number of bids participating in the response at time t.

[0048] Preferably, the S32 comprises the following steps:

[0049] The relaxation of the quadratic term in equation (21) is as follows:

[0050] η 1，t ≥P 1，t -L t -(1-α 1，t )·M; (24)

[0051] η 1，t ≤P 1，t -L t ; (25)

[0052] η 1，t ≤α 1，t ·M; (26)

[0053] η 1，t ≥-α 1，t ·M; (27)

[0054] η2，t ≥P 2，t -L t -(1-α 2，t )·M; (28)

[0055] η 2，t ≤P 2，t -L t ; (29)

[0056] η 2，t ≤α 2，t ·M; (30)

[0057] η 2，t ≥-α 2，t ·M; (31)

[0058] Wherein, M represents a sufficiently large positive number;

[0059] Formula (23) contains the cubic product of variables. A two-stage relaxation method is used. First, the auxiliary variable A is introduced. t =α t α t+1 , χ 1，t =α t α t+1 ·η t , χ 2，t =α t α t+1 ·η t+1 , formula (23) can be expressed as:

[0060] χ 1，t = x 2，t ; (32)

[0061] χ 1，t =A t ·η t ; (33)

[0062] χ 2，t =A t ·η t+1 ; (34)

[0063] In the first stage, equations (33)-(34) are relaxed:

[0064] χ 1，t ≥η t -(1-A t )·M; (35)

[0065] χ 1，t ≤η t ; (36)

[0066] χ 1，t ≤At ·M; (37)

[0067] χ 1，t ≥-A t ·M; (38)

[0068] χ 2，t ≥η t+1 -(1-A t )·M; (39)

[0069] χ 2，t ≤η t+1 ; (40)

[0070] χ 2，t ≤A t ·M; (41)

[0071] χ 2，t ≥-A t ·M; (42)

[0072] In the second stage, the introduced A t =α t α t+1 To perform relaxation:

[0073] A t ≤α t ; (43)

[0074] A t ≤α t+1 ; (44)

[0075] A t ≥α t +α t+1 -1; (45)

[0076] Preferably, the final charging station demand response bidding model is:

[0077]

[0078] The final charging station demand response bidding model can be obtained based on a solver solution.

[0079] Preferably, for a market-based response, given a bid price combination (P t，k , p t，k ), the charging station revenue function is expressed as:

[0080]

[0081] Among them, P t，k represents the response power of the kth group of quotations, p t，krepresents the corresponding price, k = 1, 2, ..., K, K is the number of bid combinations submitted, δ t，k It is a binary variable, indicating whether the kth group of quotes at time t is accepted.

[0082] Preferably, the charging station market bidding strategies in different scenarios in S4 include:

[0083] The system publishes load reduction requirements and charging load forecast value L pre，t <L t The first segment bidding strategy when

[0084] The system publishes load reduction requirements and charging load forecast value L pre，t >L t The second segment bidding strategy when

[0085] The system publishes the load filling demand and the charging load forecast value L pre，t >L t The third segment bidding strategy when

[0086] The system publishes the load filling demand and the charging load forecast value L pre，t <L t The fourth segment bidding strategy at the time.

[0087] Preferably, the first segment bidding strategy is as follows:

[0088]

[0089] Among them, p min，down represents the lowest price that a charging station can accept for participating in peak load response, p max，down represents the upper limit of the peak shaving subsidy price stipulated by local policies, L min，t The lower limit of load reduction for charging stations;

[0090] The second segment bidding strategy is as follows:

[0091]

[0092] P t，1 represents the lowest price that the charging station wants to make a profit, so when P t，1 >p max，down When the charging station is not involved in this round of bidding;

[0093] The third segment bidding strategy is as follows:

[0094]

[0095] Among them, p min，up represents the lowest price that a charging station can accept for participating in valley filling response, p max，upThe price cap for the grain-filling subsidy is set by local policy;

[0096] The fourth segment bidding strategy is as follows:

[0097]

[0098] When p t，1 >p max，up The charging station does not participate in this round of bidding.

[0099] Beneficial effects of the present invention:

[0100] (1) This paper proposes a charging load classification prediction method based on order data to assist in the effective load screening of charging stations under random access of large-scale electric vehicles. On this basis, considering the difficulty of adapting bidding decisions to existing policy models, this paper optimizes and models realistic factors such as graded subsidies and baseline differences under different response models to ensure the interactive benefits of charging stations in different scenarios.

[0101] (2) The technology proposed in the present invention can achieve a 71% improvement in load forecasting accuracy before and after load screening and a maximum 55% response load increment in a single period, which not only improves the charging station's ability to resist load fluctuation risks, but also effectively enhances the potential for vehicle-grid interactive regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 This is a flow chart of a charging station bidding method based on order classification prediction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0103] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and examples.

[0104] The embodiment of the present application discloses a charging station bidding method based on order classification prediction, and the process is as follows: Figure 1 As shown, the following steps are included:

[0105] S1. Obtain the charging station order data and standardize the order data.

[0106] Load forecasting of electric vehicle charging stations is an important part of optimizing operations. Accurately predicting future charging loads helps charging stations optimize energy scheduling and reduce operating costs. However, with the rapid development of charging infrastructure, different types of electric vehicles often have different charging power requirements. In order to quantify the differences in charging power and charging pile selection for different types of electric vehicles, unlike traditional methods that directly predict the total charging load, the embodiment of this application classifies and manages charging users based on order data:

[0107]

[0108] Where D represents the charging station order dataset, P t represents the total charging load of the charging station at time t, P t，i represents the charging power of electric vehicles of type i at time t, i = 1, 2, ..., n, and n is the total number of categories of electric vehicles obtained according to the charging power;

[0109] In order to unify the scales of different types of loads, further data standardization is performed, and the formula is as follows:

[0110]

[0111] Among them, μ i is the mean value of load data of category i, σ i is the standard deviation of load data for category i.

[0112] S2. Based on the Long Short-Term Memory (LSTM) network, a charging load classification prediction model for charging stations is further constructed.

[0113] The input of the charging load classification prediction model is the electric vehicle load data of the previous 24 hours in The output of the charging load classification prediction model includes the charging power of n types of electric vehicles, and the output of the charging load classification prediction model is the predicted value of the charging power of the same n types of electric vehicles within 24 hours of the next day. in

[0114]

[0115] The basic LSTM model is as follows:

[0116] f t =σ(W f ·[h t-1 , x t ]+b f ); (3)

[0117] i t =σ(W i ·[h t-1 , x t ]+b i ); (4)

[0118]

[0119] o t =σ(W o ·[h t-1 , x t ]+b o ); (7)

[0120] h t =o t tanh(C t ); (8)

[0121] Among them, f t is the activation value of the forget gate, i t is the input gate activation value, C t-1 is the cell state at the previous time step, C t is the cell state at the current time step, is the new candidate cell state at the current time step, o t is the output gate activation value, h t is the hidden state of the current time step, W f is the weight matrix of the forget gate, W i is the weight matrix of the input gate, W c is the weight matrix of candidate memory content, W o is the weight matrix of the output gate, b f is the bias term of the forget gate, b i is the bias term of the input gate, b C is the bias term of the candidate memory content, b o is the bias term of the output gate, h t-1 is the hidden state at the previous moment, x t is the current input data, and σ is the Sigmoid activation function.

[0122] The loss function of the charging load classification prediction model is:

[0123]

[0124] Where N is the number of samples, is the charging power predicted by the model, is the actual charging power.

[0125] Different electric vehicle users have different demands for time efficiency and will choose charging piles of different power for charging. In addition, due to differences in vehicle types and models, the actual distribution of charging power connected to the network is highly random. Classifying electric vehicles based on order data is more conducive to considering geographic information and user travel characteristics, and is also more conducive to charging stations screening out electric vehicle users with greater adjustment potential to participate in vehicle-grid interaction.

[0126] In the process of charging stations participating in the grid interaction, according to the real-time demand of the grid, charging stations may face different interaction scenarios, including invitation-based response and market-based response. Based on the charging demand forecast, the charging station invitation bidding strategy and market bidding bidding strategy are further proposed.

[0127] S3. Develop a charging station invitation bidding strategy for invitation responses.

[0128] S31. Taking the tiered subsidy income and load regulation cost as optimization objectives, a charging station demand response bidding model and model constraints are constructed.

[0129] Demand response refers to the behavior of users reducing electricity consumption during peak hours and increasing electricity consumption during low hours through price signal guidance, thereby obtaining economic subsidies. For users who respond by invitation, tiered subsidies are provided. When the subsidy is settled, the responding P t The determination of the load baseline L t related:

[0130]

[0131] Among them, L res，t represents the charging load after the charging station participates in the response, γ is a binary variable, when the system regulation demand is the peak shaving demand, γ = 0, when the system regulation demand is the valley filling demand, γ = 1, L t It is mainly based on the average load calculation of the same time period in the past few days. D represents the total number of days used to calculate the load baseline.

[0132] Taking into account the feasibility of the actual project, taking the hierarchical subsidy income and load regulation cost as the optimization target, the charging station demand response bidding model is constructed as follows:

[0133]

[0134] P 1，min ≤P 1，t ≤P 1，max , P 2，min ≤P 2，t ≤P 2，max ; (12)

[0135] α 1，t , α 2，t ∈{0,1}; (13)

[0136] Among them, p1 is the short-term response subsidy price when the continuous response time does not exceed 2 hours, p2 is the long-term response subsidy price when the continuous response time exceeds 2 hours, and p s，t is the charging station to unit load adjustment fee at time t, P 1，t is the charging load after participating in short-term demand response at the corresponding time, P 2，t is the charging load after participating in long-term demand response at the corresponding time, P 1，min P 1，t Adjustment lower limit, P 1，max P 1，t Upper limit of regulation, P 2，min P 2，tAdjustment lower limit, P 2，max P 2，t Upper limit of adjustment, L pre，t is the predicted value of charging load at the charging station, α 1，t , α 2，t is a binary variable, α 1，t =1 means participating in short-term demand response at time t, α 2，t =1 means participating in long-term response demand at time t.

[0137] α 1，t , α 2，t The following timing constraints are met:

[0138] α 1，t +α 1，t+1 +α 1，t+2 ≤2t=0,1,…,21; (14)

[0139] 2·(α 2，t -α 2，t-1 )≤α 2，t+1 +α 2，t+2 t=1,…,21; (15)

[0140] Among them, formula (14) means that when participating in short-term response, the continuous response time shall not exceed 2 hours, and formula (15) means that once participating in long-term response, the continuous response time must be greater than or equal to 3 hours;

[0141] For special points, the following constraints are satisfied:

[0142] 2α 2，0 ≤α 2，1 +α 2，2 ; (16)

[0143] 2α 2，21 ≥α 2，22 +α 2，23 ; (17)

[0144] Since it is impossible to participate in both short-term and long-term responses at the same time, α 1，t , α 2，t The following response mutual exclusion constraints should also be satisfied:

[0145] α 1，t +α 2，t =α t ; (18)

[0146] 2-α 1，t-1 -α 2，t+1 ≥α t t=1,2,…,22; (19)

[0147] 2-α 1，t+1 -α2，t-1 ≥α t t=1,2,…,22; (20)

[0148] η 1，t =α 1，t ·(P 1，t -L t ), η 2，t =α 2，t ·(P 2，t -L t ); (twenty one)

[0149] η 1，t +η 2，t =η t ; (twenty two)

[0150] α t α t+1 ·η t =α t α t+1 ·η t+1 t=0,1,…,22; (23)

[0151] Among them, α t is a newly introduced binary variable. t =1, it means that the charging station participates in short-term or long-term demand response at time t. t = 0 means that the charging station does not participate in the response η at time t 1，t represents the bidding amount for short-term response at time t, η 2，t represents the bid amount for long-term response at time t, η t is the total bid amount participating in the response at time t. Formulas (18)-(20) ensure that the short-term response period and the long-term response period are not adjacent (if adjacent, both periods are attributed to the long-term response period). Formulas (21)-(23) ensure that the power bid amount at each moment in the same response period remains consistent.

[0152] S32. Relax the high-order terms in the charging station demand response bidding model constraints to obtain a final charging station demand response bidding model.

[0153] Constraints (21) and (23) both contain high-order terms of variable multiplication. In order to facilitate solver optimization, the high-order terms need to be relaxed.

[0154] The relaxation of the quadratic term in equation (21) is as follows:

[0155] η 1，t ≥P 1，t -L t -(1-α 1，t )·M; (24)

[0156] η 1，t ≤P 1，t -L t ; (25)

[0157] η 1，t ≤α 1，t ·M; (26)

[0158] η 1，t ≥-α 1，t ·M; (27)

[0159] η 2，t ≥P 2，t -L t -(1-α 2，t )·M; (28)

[0160] η 2，t ≤P 2，t -L t ; (29)

[0161] η 2，t ≤α 2，t ·M; (30)

[0162] η 2，t ≥-α 2，t ·M; (31)

[0163] Wherein, M represents a sufficiently large positive number, and in this embodiment, the value is 50000. For the high-order terms included in the objective function, linear processing can also be performed through the above relaxation.

[0164] Formula (23) contains the cubic product of variables. A two-stage relaxation method is used. First, the auxiliary variable A is introduced. t =α t α t+1 , χ 1，t =α t α t+1 ·η t , χ 2，t =α t α t+1 ·η t+1 , formula (23) can be expressed as:

[0165] χ 1，t = x 2，t ; (32)

[0166] χ1,t=A t ·η t ; (33)

[0167] χ 2，t =A t ·ηt+1 ; (34)

[0168] In the first stage, equations (33)-(34) are relaxed:

[0169] χ 1，t ≥η t -(1-A t )·M; (35)

[0170] χ 1，t ≤η t ; (36)

[0171] χ 1，t ≤A t ·M; (37)

[0172] χ 1，t ≥-A t ·M; (38)

[0173] χ 2，t ≥η t+1 -(1-A t )·M; (39)

[0174] χ 2，t ≤η t+1 ; (40)

[0175] χ 2，t ≤A t ·M; (41)

[0176] χ 2，t ≥-A t ·M; (42)

[0177] In the second stage, the introduced A t =α t α t+1 To perform relaxation:

[0178] A t ≤α t ; (43)

[0179] A t ≤α t+1 ; (44)

[0180] A t ≥α t +α t+1 -1; (45)

[0181] The final charging station demand response bidding model is:

[0182]

[0183] At this point, the final charging station demand response bidding model can be obtained based on the solver solution.

[0184] S4. Develop market-based bidding strategies for charging stations in different scenarios based on market responses.

[0185] Different from the invitation-based response, the subsidy price of the market-based response is determined by the market clearing and cannot be predicted in advance, and the cost price changes corresponding to different response quantities are also difficult to accurately represent. Therefore, this type of interactive scenario has a high degree of uncertainty and it is difficult to build an accurate mathematical optimization model. In this regard, a segmented quotation strategy is proposed, that is, each period contains multiple sets of quantity and price combinations to ensure the benefits under different clearing conditions.

[0186] Given a bid price combination (P t，k , p t，k ), the charging station revenue function is expressed as:

[0187]

[0188] Among them, P t，k represents the response power of the kth group of quotations, p t，k represents the corresponding price, k = 1, 2, ..., K, K is the number of bid combinations submitted, δ t，k It is a binary variable, indicating whether the kth group of quotes at time t is accepted.

[0189] Considering the two types of peak shaving and valley filling, the charging station market bidding strategies in different scenarios include:

[0190] The system publishes load reduction requirements and charging load forecast value Lpre ，t <L t The first segment bidding strategy at this time. t -L pre，t It will bring considerable additional subsidy income, and charging stations have a natural advantage in participating in peak load response. Considering the upper and lower limits of the adjustment power and a certain expected rate of return, the first segment bidding strategy can be expressed as follows:

[0191]

[0192] Among them, p min，down represents the lowest price that a charging station can accept for participating in peak load response, p max，down represents the upper limit of the peak shaving subsidy price stipulated by local policies, L min，t It is the lower limit of load reduction of the charging station.

[0193] The system publishes load reduction requirements and charging load forecast value Lpre ，t >L tThe second segment bidding strategy at this time. Since the load baseline is lower than the charging load of the next day, the power difference Lpre ，t -L t This will bring additional adjustment costs. If charging stations want to ensure their own profits, they need to further increase their bids. The second segment bidding strategy can be expressed as follows:

[0194]

[0195] P t，1 represents the lowest price that the charging station wants to make a profit, so when P t，1 >p max，down The charging station does not participate in this round of bidding.

[0196] The system publishes the load filling demand and the charging load forecast value L pre，t >L t The third segment bidding strategy when . Similar to the first segment bidding strategy, the power difference L pre，t -L t It will bring considerable additional subsidy income. At this time, the charging station has a natural advantage in participating in valley filling response. The third segment bidding strategy can be expressed as follows:

[0197]

[0198] Among them, p min，up represents the lowest price that a charging station can accept for participating in valley filling response, p max，up This is the upper limit of the valley filling subsidy price stipulated by local policies.

[0199] The system publishes the load filling demand and the charging load forecast value L pre，t <L t The fourth segment bidding strategy when . Similar to the situation in the second segment bidding strategy, the power difference L t -L pre，t It will bring additional adjustment costs. At this time, the bidding strategy for the fourth segment of the charging station participating in the valley filling corresponding segment is as follows:

[0200]

[0201] When p t，1 >p max，up The charging station does not participate in this round of bidding.

[0202] In summary, based on the above analysis, a segmented pricing strategy for charging stations that can adapt to different scenarios can be obtained.

[0203] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A charging station bidding method based on order classification prediction, characterized in that: The following steps are involved: S1. Obtain charging station order data and standardize the order data; S2. Construct a charging load classification prediction model for charging stations based on long short-term memory network; S3. Formulate a charging station solicitation bidding strategy for solicitation responses; S4. Develop market-based bidding strategies for charging stations in different scenarios based on market responses.

2. The charging station bidding method based on order classification prediction according to claim 1 is characterized in that: The charging station order data is as follows: Where D represents the charging station order dataset, P t represents the total charging load of the charging station at time t, P t,i represents the charging power of electric vehicles of type i at time t, i = 1, 2, ..., n, and n is the total number of categories of electric vehicles according to the charging power; The formula for normalizing order data is as follows: Among them, μ i is the mean value of load data of category i, σ i is the standard deviation of load data for category i.

3. The charging station bidding method based on order classification prediction according to claim 2 is characterized in that: The input of the charging load classification prediction model is the electric vehicle load data of the previous 24 hours. in The output of the charging load classification prediction model includes the charging power of n types of electric vehicles, and the output is the predicted value of the charging power of the same n types of electric vehicles within 24 hours of the next day. in The loss function of the charging load classification prediction model is: Where N is the number of samples, is the charging power predicted by the model, is the actual charging power.

4. The charging station bidding method based on order classification prediction according to claim 3 is characterized in that: The S3 comprises the following steps: S31. Taking the hierarchical subsidy income and load regulation cost as the optimization objectives, a charging station demand response bidding model and model constraints are constructed; S32. Relax the high-order terms in the charging station demand response bidding model constraints to obtain a final charging station demand response bidding model.

5. The charging station bidding method based on order classification prediction according to claim 4 is characterized in that: The S31 comprises the following steps: For the invitation-based response users, graded subsidies are provided. When the subsidy is settled, the response P t The determination of the load baseline L t related: Among them, L res,t represents the charging load after the charging station participates in the response, γ is a binary variable, when the system regulation demand is the peak shaving demand, γ = 0, when the system regulation demand is the valley filling demand, γ = 1, D represents the total number of days used to calculate the load baseline; The charging station demand response bidding model is constructed as follows: P 1,min ≤P 1,t ≤P 1,max ,P 2,min ≤P 2,t ≤P 2,max ;(12) α 1,t ,α 2,t ∈{0,1}；(13) Among them, p1 is the short-term response subsidy price when the continuous response time does not exceed 2 hours, p2 is the long-term response subsidy price when the continuous response time exceeds 2 hours, and p s,t is the charging station to unit load adjustment fee at time t, P 1,t is the charging load after participating in short-term demand response at the corresponding time, P 2,t is the charging load after participating in long-term demand response at the corresponding time, P 1,min P 1,t Adjustment lower limit, P 1,max P 1,t Upper limit of regulation, P 2,min P 2,t Adjustment lower limit, P 2,max P 2,t Upper limit of adjustment, L pre,t is the predicted value of charging load at the charging station, α 1,t , α 2,t is a binary variable, α 1,t =1 means participating in short-term demand response at time t, α 2,t =1 means participating in long-term response demand at time t; α 1,t , α 2,t The following timing constraints are met: α 1,t +α 1,t+1 +α 1,t+2 ≤2t＝0,1,…,21； (14) 2·(α 2,t -α 2,t-1 )≤α 2,t+1 +α 2,t+2 t＝1,…,21； (15) Among them, formula (14) means that when participating in short-term response, the continuous response time shall not exceed 2 hours, and formula (15) means that once participating in long-term response, the continuous response time must be greater than or equal to 3 hours; For special points, the following constraints are satisfied: 2α 2,0 ≤α 2,1 +α 2,2 ； (16) 2α 2,21 ≥α 2,22 +α 2,23 ； (17) α 1,t , α 2,t The following response mutual exclusion constraints are satisfied: α 1,t +α 2,t ＝α t ； (18) 2-α 1,t-1 -α 2,t+1 ≥α t t＝1,2,…,22； (19) 2-α 1,t+1 -α 2,t-1 ≥α t t＝1,2,…,22； (20) or 1,t =a 1,t ·(P 1,t -P t ),η2, t =a 2,t ·(P 2,t -L t ); (21) or 1,t +n 2,t =the t (22) a t ·a t+1 ·or t =a t ·a t+1 ·or t+1 t=0,1,…,22; (23) Among them, α t is a newly introduced binary variable. t =1, it means that the charging station participates in short-term or long-term demand response at time t. t =0 means that the charging station does not participate in the response at time t, η 1,t represents the bidding amount for short-term response at time t, η 2,t represents the bid amount for long-term response at time t, η t is the total number of bids participating in the response at time t.

6. The charging station bidding method based on order classification prediction according to claim 5 is characterized in that: The S32 comprises the following steps: The relaxation of the quadratic term in equation (21) is as follows: or 1,t ≥P 1,t -L t -(1-a 1,t )·M; (24) or 1,t ≤P 1,t -L t (25) or 1,t ≤α 1,t ·M; (26) or 1,t ≥-a 1,t ·M; (27) or 2,t ≥P 2,t -L t -(1-a 2,t )·M; (28) or 2,t ≤P 2,t -L t (29) or 2,t ≤α 2,t ·M; (30) or 2,t ≥-a 2,t ·M; (31) Wherein, M represents a sufficiently large positive number; Formula (23) contains the cubic product of variables. A two-stage relaxation method is used. First, the auxiliary variable A is introduced. t =·α t α t+1 , χ 1,t =α t α t+1 ·η t , χ 2,t =α t α t+1 ·η t+1 , formula (23) can be expressed as: x 1,t =x 2,t (32) x 1,t =A t ·or t (33) x 2,t =A t ·or t+1 (34) In the first stage, equations (33)-(34) are relaxed: x 1,t ≥η tt -(1-A t )·M; (35) x 1,t ≤η t (36) x 1,t ≤A t ·M; (37) x 1,t ≥-A t ·M; (38) x 2,t ≥η t+1 -(1-A t )·M; (39) x 2,t ≤η t+1 (40) x 2,t ≤A t ·M; (41) x 2,t ≥-A t ·M; (42) In the second stage, the introduced A t =α t α t+1 To perform relaxation: A t ≤α t ； (43) A t ≤α t+1 ； (44) A t ≥α t +α t+1 -1； (45) At this point, the two-stage relaxation method is used to complete the relaxation of equation (23).

7. The charging station bidding method based on order classification prediction according to claim 6 is characterized in that: The final charging station demand response bidding model is: The final charging station demand response bidding model can be obtained based on a solver solution.

8. The charging station bidding method based on order classification prediction according to claim 7 is characterized in that: For market-type responses, given a bid-offer combination (P t,k ,p t,k ), the charging station revenue function is expressed as: Among them, P t,k represents the response power of the kth group of quotations, p t,k represents the corresponding price, k = 1, 2, ..., K, K is the number of bid combinations submitted, δ t,k It is a binary variable, indicating whether the kth group of quotes at time t is accepted.

9. The charging station bidding method based on order classification prediction according to claim 8 is characterized in that: The charging station market bidding strategies under different scenarios in S4 include: The system publishes load reduction requirements and charging load forecast value L pre,t <L t The first segment bidding strategy when The system publishes load reduction requirements and charging load forecast value L pre,t >L t The second segment bidding strategy when The system publishes the load filling demand and the charging load forecast value L pre,t >L t The third segment bidding strategy when The system publishes the load filling demand and the charging load forecast value L pre,t <L t The fourth segment bidding strategy at the time.

10. The charging station bidding method based on order classification prediction according to claim 9 is characterized in that: The first segment bidding strategy is as follows: Among them, p min,down represents the lowest price that a charging station can accept for participating in peak load response, p max,down represents the upper limit of the peak shaving subsidy price stipulated by local policies, L min,t The lower limit of load reduction for charging stations; The second segment bidding strategy is as follows: P t,1 represents the lowest price that the charging station wants to make a profit, so when P t,1 >p max,down When the charging station is not involved in this round of bidding; The third segment bidding strategy is as follows: Among them, p min,up represents the lowest price that a charging station can accept for participating in valley filling response, p max,up The price cap for the grain-filling subsidy is set by local policy; The fourth segment bidding strategy is as follows: When p t,1 >p max,up The charging station does not participate in this round of bidding.

Images