Electric vehicle regulation power description and frequency support method considering user characteristics and willingness

Abstract

The invention relates to the technical field of electric vehicles, and discloses an electric vehicle regulation power description and frequency support method considering user characteristics and willingness, and the method comprises the steps: describing the related uncertainty factors of an electric vehicle based on a Gaussian mixture distribution method; wherein the uncertainty factors comprise a user behavior mode, a battery capacity, a charging rate and a discharging rate of the electric vehicle; predicting the schedulable willingness of the electric vehicle by adopting a Logit model; and depicting an electric vehicle adjustment boundary considering user characteristics and user intentions, determining an adjustment mode of the user by changing upper and lower limits of power fluctuation of the electric vehicle, and determining an electric vehicle frequency modulation strategy. The method provided by the invention solves the problem that the frequency modulation capability tends to be advanced due to the fact that the frequency modulation strategy of the electric vehicle (EV) rarely considers the uncertainty of the charging behavior.

Description

Technical Field

[0001] The present application relates to the technical field of electric vehicles, and in particular to a method for regulating power characterization and frequency support of electric vehicles taking into account user characteristics and intentions. Background Art

[0002] Electric vehicles (EVs) use electricity to replace traditional carbon-based fossil fuels, thereby significantly reducing fossil fuel consumption and greenhouse gas emissions. However, the charging and discharging behavior of EV users themselves is random and uncertain. When EVs are connected to the power grid at a high penetration rate, large-scale EV charging and discharging behavior will pose a huge threat to the power grid. Adopting a control strategy that takes EV uncertainty into account is the key to solving the problem.

[0003] With the popularization of electric vehicles, the real-time power supply and power conservation pressure of the power grid are also increasing. Accurate load forecasting helps balance the load of the power grid, promote the use of clean energy, and reduce dependence on fossil fuels. The literature "Zhu Lei, Huang He, Gao Song, et al. Research on optimal configuration of electric vehicle load considering wind power consumption [J]. Proceedings of the CSEE, 2021, 41(S1): 194-203." uses normal distribution to describe the disordered charging and discharging curves of EVs. The literature "Dou Xun, Wang Jun, Yang Zhihong, et al. Electric vehicle clustering control strategy for integrated energy system with AC / DC hybrid distribution network [J]. Proceedings of the CSEE, 2021, 41(14): 4829-4844." uses log-normal distribution to characterize the mileage of EVs. The literature "Zhang Shaoxing. Optimal operation of distribution network considering electricity price incentives after large-scale electric vehicle access [D]. North China Electric Power University, 2022." uses normal distribution to express EV travel and return time. In general, in terms of load forecasting for EVs participating in grid regulation, most of the user characteristics of EV users in relevant literature are approximately normally distributed, and fail to fully release the surplus energy of EVs to participate in grid regulation while ensuring user satisfaction by specifying reasonable control rules.

[0004] In recent years, the scale of EV clusters has continued to increase, and the grid connection of a large number of EV clusters will bring about the problem of sharp frequency fluctuations. To this end, the document "Wu Juai, Xue Yusheng, Xie Dongliang, et al. Joint risk dispatch of electric vehicles participating in the electricity market and the reserve market [J / OL]. Transactions of the Chinese Society of Electrical Engineering: 1-11 [2023-06-14]. "Based on the information interaction between EVA and EV users, the decoupling of EV normal charging behavior and the provision of reserve is realized, and a load frequency control model for aggregators to participate in the joint dispatch of the day-ahead energy market and the reserve market under multiple scenarios is constructed. The document "Huang Xiaoqing, Li Longyi, Xu Pengxin, et al. A multi-agent game-win-win method for sharing electric vehicle charging piles [J]. Transactions of the Chinese Society of Electrical Engineering, 2023, 38 (11): 2945-2961." proposes a method for electric vehicles to participate in load frequency control based on the master-slave game double-layer iteration of the Ford-Fulkerson maximum flow algorithm. The document "Zhang Qian, Deng Xiaosong, Yue Huanzhan, et al. Collaborative optimization strategy for electric vehicles participating in the energy-frequency regulation market taking into account battery life loss [J]. Transactions of the Chinese Society of Electrical Engineering, 2022, 37(1):72-81." proposes a load frequency control method for electric vehicles participating in the energy-frequency regulation market taking into account battery life loss. The document "Zhang Liang, Sun Chenglong, Cai Guowei, et al. Orderly charging and discharging two-stage optimization strategy for electric vehicles based on PSO algorithm [J]. Proceedings of the CSEE, 2022, 42(5):1837-1852." proposes a two-stage optimization strategy for orderly charging and discharging of electric vehicles based on particle swarm algorithm to analyze load frequency regulation. The above documents adopt different methods to optimize the grid connection of EV clusters, but when the number of EVs increases and the charging and discharging power fluctuates greatly, the load frequency regulation strategy will also cause new frequency spikes during the frequency regulation period or it will be difficult to effectively improve the frequency peak-to-average ratio. That is, the frequency regulation characteristics in the proposed control strategy are aggressive, there is no regulatory constraint on the upper and lower limits of frequency regulation, and no consideration is given to the satisfaction of individual users. Summary of the invention

[0005] The purpose of this application is to provide a method for characterizing the power regulation and frequency support of electric vehicles that takes into account user characteristics and intentions. Modeling and analysis are performed based on the travel behavior characteristics of EVs, including the time of entering and leaving the grid, the initial state of charge, and the user's intentions. Conventional EVs participate in the grid forecast load curve and the peak and valley charging price of the load curve generated by user characteristics, and then the initial state of charge, dispatch subsidy, and the time spent at the charging pile are normalized. A Logit model is established to analyze the user's behavioral characteristics, and the EV's adjustable power boundary is characterized according to the user's intention. A real-time power regulation strategy and frequency support strategy for EVs are proposed that takes into account the user's uncertainty characteristics and the user's dispatchable intentions while ensuring that the overall trend of the user's state of charge is rising.

[0006] In order to achieve the above purpose, the technical solution adopted in this application is as follows:

[0007] The present application provides a method for regulating power characterization and frequency support of an electric vehicle taking into account user characteristics and intentions, the method comprising:

[0008] The uncertainty factors related to electric vehicles are described based on the Gaussian mixture distribution method; wherein the uncertainty factors include the user behavior pattern, battery capacity, and charging and discharging rate of the electric vehicle;

[0009] The Logit model is used to predict the dispatchability willingness of electric vehicles;

[0010] Characterize the electric vehicle regulation boundary considering user characteristics and user willingness, determine the user's regulation method by changing the upper and lower limits of electric vehicle power fluctuation, and determine the electric vehicle frequency regulation strategy.

[0011] The beneficial effects of this application are:

[0012] 1) This application takes into account user characteristics and user wishes, characterizes the EV regulation boundary, constrains the EV real-time frequency regulation power, determines the user frequency regulation mode by changing the upper and lower limits of EV power fluctuation, and accurately characterizes the EV frequency regulation boundary.

[0013] 2) This application conducts a significance analysis of the factors affecting EV participation in grid frequency regulation based on SPSS. It is found that the initial SOC of EV users, the waiting time at the charging station, and the electricity price incentive are the three most important decision variables affecting EV's participation in grid frequency regulation. From an overall perspective, the impact of EV's participation in grid frequency regulation capacity boundaries is studied by perturbing electricity prices; from an individual perspective, combined with the correlation coefficient, it is found that the initial SOC is more correlated with the proportion of EV individual participation in frequency regulation time than the waiting time.

[0014] 3) The electric vehicle frequency regulation strategy proposed in this application ensures that the overall trend of the user's state of charge is on the rise and takes into account the user's uncertainty characteristics and the user's dispatchable willingness, which ensures the overall upward trend of the SOC of EV users during the regulation process. By considering the user's willingness, the fast frequency improvement control strategy is optimized, effectively reducing the fluctuation of load power, so that the frequency deviation is concentrated as much as possible in the ±0.033Hz range. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of a method for regulating power characterization and frequency support of an electric vehicle taking into account user characteristics and intentions provided in an embodiment of the present application;

[0016] Figure 2 An EV charging load curve diagram considering user characteristics provided in an embodiment of the present application;

[0017] Figure 3 A droop characteristic diagram when an EV participates in frequency modulation provided in an embodiment of the present application;

[0018] Figure 4 A graph depicting EV output power in different scenarios provided in an embodiment of the present application;

[0019] Figure 5 A diagram depicting the adjustable power capability of EVs under scenarios 2 and 3 provided in the embodiments of the present application;

[0020] Figure 6 Typical SOC curves of EVs in different scenarios provided by embodiments of the present application;

[0021] Figure 7 A comparison diagram of the effects of three different electricity prices on EV charging and discharging power boundaries under scenario 3 provided in an embodiment of the present application;

[0022] Figure 8 A relationship diagram of the initial soc and the proportion of EV frequency modulation time to the parking time provided in the embodiment of the present application;

[0023] Fig. 9 A relationship diagram of different stop times and the proportion of EV frequency modulation time relative to the stop time provided in the embodiment of the present application;

[0024] Fig.10 A variable scatter plot matrix diagram provided for an embodiment of the present application;

[0025] Fig.11 A matrix correlation heat map provided for an embodiment of the present application;

[0026] Fig.12 Schematic diagram of EV-assisted frequency modulation control analysis for scenario 2 and scenario 3 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0028] The specific implementation of the present application is further described in detail below in conjunction with the drawings and examples.

[0029] See also Figure 1, which is a flow chart of a method for adjusting power characterization and frequency support of an electric vehicle taking into account user characteristics and intentions, provided in an embodiment of the present application. An embodiment of the present application provides a method for adjusting power characterization and frequency support of an electric vehicle taking into account user characteristics and intentions, and the method includes the following steps S1 to S3.

[0030] S1: The uncertainty factors related to electric vehicles are described based on the Gaussian mixture distribution method; the uncertainty factors include the user behavior patterns, battery capacity, and charging and discharging rates of electric vehicles.

[0031] The charging power of individual electric vehicle users connected to the grid is determined by the charging pile, but the grid-connected power of electric vehicles (EVs) is often determined by the user characteristics of EVs. There are usually several methods to characterize EVs user characteristics: regression analysis, cluster analysis, decision trees, and machine learning algorithms. Current research mainly uses normal probability distribution fitting, and user travel distribution has morning and evening peaks, and the grid-connected distribution is also different according to different peak and valley electricity prices. Therefore, the normal distribution cannot accurately describe the multi-peak and peak-valley electricity price incentive phenomena, while the Gaussian mixture model (GMM) can accurately describe the multi-peak distribution. Therefore, this embodiment uses GMM to calculate the time of arrival of Ev (T _arival_i ) and the probability density function (P d ) is classified and fitted, then the specific probability distribution function P(T arrive_i |Θ) is:

[0032]

[0033] Where: N θ (T arrive_i ;μ θ ,D θ ) is θ th Gaussian component function, μ θ is the average value, D θ is the variance; J is the total number of Gaussian components; α θ is the weight of the Gaussian component; the expression of one-dimensional GMM: θ=(θ1,……,θ J ), by determining T arrive_i , μ θ , D θ To fit the Gaussian mixture distribution, J is set by the EV user before estimation, and J is adjusted according to the actual fitting effect after estimation. ;

[0034] Determine the probability density function of the initial energy of the electric vehicle It is expressed as:

[0035]

[0036] Where: E i0 and SOC i0 i th The initial energy and initial SOC of the EV; is the mean of the ith distribution; D θ is the variance of the probability density distribution function; τ i The number of days that an EV can be driven after charging; d iR For EV range;

[0037] The parameters of the initial energy probability density function of the electric vehicle are estimated through historical data, and the SOC i0 and T arrive_i The density function in is derived. During the simulation, random numbers are drawn using the Monte Carlo method according to each probability density. At the same time, the departure time T is obtained according to the stop time set by the user. _deprive (Total EV _Number ), to depict the complete user characteristics.

[0038] S2: Use the Logit model to predict the dispatchability willingness of electric vehicles.

[0039] The Logit model is mainly used to analyze and predict the selection behavior of EV users when facing different charging strategies and grid demands in the electricity market environment. It determines whether the users are willing to accept grid dispatch based on their characteristics, and determines the mode of participation in grid regulation based on the size of their willingness.

[0040] The Logit model is suitable for representing scenarios with binary results (acceptance, rejection), and can effectively present the impact of multiple variables such as charging status, scheduling subsidy incentives, and control flexibility on the decision-making process. This allows a clear understanding of how different factors affect decision-making. Using the Logit probability model to represent EV dispatchability willingness mainly includes the following steps:

[0041] S21: Considering the EV charge state, the dispatch subsidy of the power grid in different time periods, and the EV user’s stay time in the charging area as multiple factors affecting user willingness, the hidden variable y in the probability model of user dispatch willingness is i It can be expressed as:

[0042] y i =α1+β1SOC i0 +β2K EV_i +β3T on_i +u i (4)

[0043] Where: y i for i th The decision intention hidden variable; α i is the coefficient of the benchmark probability distribution function; β1, β2, β3 are the coefficients of each explanatory variable; SOC i , K Ev_i 、T on_i Represents electric vehicle i th The initial state of charge, scheduling subsidy incentives, and charging pile residence time input by the user; u i is the random error variable of decision intention;

[0044] In practical applications, the coefficients of each explanatory variable are usually obtained by performing Logit regression analysis on historical data using the mnrfit function. This function is specifically used to fit multinomial logistic regression models and is suitable for parameter estimation with binary outcomes.

[0045] S22: Use the Logit function to transform the latent variable y i Transformed into the user's choice probability, the probability functions of EV accepting grid dispatch and not accepting dispatch are expressed by equations (5) and (6) respectively, where the hidden variable y in the dispatch willingness probability model is i The value range of is (-∞, +∞), and the scheduling willingness probability F(y i ) has a value range of (0, 1).

[0046]

[0047] Where: Y i for i th The decision intention variable, Y i =1 means the user chooses to accept the grid dispatch, Y i =0 means the user chooses to be idle, P i (Y i =1) represents the probability of the user making the decision Y, and its value range is (0,1), P i (Y i =0) represents the probability that the i-th electric vehicle user chooses not to accept grid dispatch, F(Y i ) represents the cumulative distribution function of decision willingness, and e represents the base of the natural logarithm.

[0048] This decision variable is affected by factors such as the state of charge, the dispatch subsidy of the power grid at different time periods, and the time EV users stay in the charging area. Among them, the dispatch subsidy has the highest weight on the willingness to choose, which affects the willingness of users to participate in the regulation in the logit model according to the different peak and valley electricity prices of each vehicle in each hour during their stay. Therefore, this embodiment is first based on Figure 2Considering the daily EV charging load curve of user characteristics, the peak and valley membership is divided using the trapezoidal membership function calculation method. p is the valley load threshold, T v is the peak load threshold, and the daily load curve is a set of data L = {l1,l2,...,l n}, where each l i Represents the grid load at a specific time point. The trapezoidal valley membership function is usually defined as four parameters, T pa ,T p ,T p ' and T pd , where [T p ,T p '] represents the core area with a membership degree of 1, and [T pa ,T p ] and [T p ',T pd ] is the rising edge and falling edge, and the membership increases from 0 to 1, or decreases from 1 to 0. The four parameters of the trapezoidal peak membership function are T va ,T v ,T v ' and T vd For peak-valley division, the peak load and valley load membership functions are shown in equations (7) and (8) respectively.

[0049]

[0050] Where M v (l) is the valley membership function representing the load curve at a certain moment. It is used to measure the closeness between the current load and the valley load. p (l i ) represents a certain moment l in the load curve i The peak membership function is used to measure the current load l i The degree of proximity to the peak load, l is the load value at a specific moment in the power grid load curve.

[0051] The peak-valley membership of the load in each period of the day is calculated using the trapezoidal peak-valley membership function as shown in Table 1.

[0052] Table 1 Peak-valley membership distribution based on EVs daily load curve

[0053] time Peak membership Valley membership time Peak membership Valley membership 1 0.64 0.14 13 0.77 0.96 2 0.54 0.24 14 0.76 0.95 3 0.48 0.32 15 0.76 0.96 4 0.44 0.39 16 0.85 0.93 5 0.40 0.45 17 0.94 0.89 6 0.41 0.52 18 0.98 0.79 7 0.44 0.55 19 0.98 0.66 8 0.51 0.63 20 0.98 0.54 9 0.55 0.71 21 0.93 0.42 10 0.61 0.77 22 0.91 0.32 11 0.69 0.87 23 0.84 0.23 12 0.76 0.96 24 0.72 0.14

[0054] S3: Characterize the electric vehicle regulation boundary considering user characteristics and user willingness, determine the user's regulation method by changing the upper and lower limits of electric vehicle power fluctuation, and determine the electric vehicle frequency regulation strategy.

[0055] In this embodiment, step S3 specifically includes:

[0056] S31: Determine an improved control strategy for EVs participating in rapid frequency regulation.

[0057] S32: Determine rules (or strategies) for EVs to participate in grid frequency regulation based on the improved control strategy for EVs to participate in rapid frequency regulation.

[0058] Wherein, step S31 can be specifically implemented in the following manner:

[0059] Considering that the response time of bidirectional converters in real work is usually in milliseconds, the electromagnetic transient process of electric vehicles is often ignored. The EV frequency modulation power expression can be expressed as:

[0060] ΔP EV =Δf·K EV (9)

[0061] Where: ΔP EVi is the change in charging and discharging power of the i-th electric vehicle, Δf is the frequency deviation of the system, and K EVi is the frequency response coefficient of the i-th electric vehicle;

[0062] According to formula (9), there is a linear relationship between the output power fluctuation of EV grid connection and the system frequency. The power-frequency droop characteristic curve of electric vehicle is as follows: Figure 3 shown.

[0063] Figure 3 In the figure, the black line is the droop characteristic of EV grid-connected frequency modulation in bidirectional V2G (BC) mode. In this mode, EV can both charge and discharge, and its power fluctuation limit is shown in equations (10) and (11). The red line represents the droop characteristic of EV grid-connected frequency modulation in unidirectional V2G (UC) mode. Both of them are K when the system frequency fluctuates. EV The frequency response coefficient of the EV system is connected to the grid for frequency regulation. However, the regulation power required for grid frequency fluctuations is often greater than the EV's own power fluctuation range. Therefore, the characterization of the regulation power constraint directly affects the EV's ability to participate in grid regulation. The upper and lower limits of the EV regulation power are:

[0064]

[0065]

[0066] Where: and The power fluctuation limits in UC mode and BC mode are respectively determined by the maximum charge and discharge power The charge and discharge power P of the previous sampling time BC , P UC The difference between the BC mode and the UC mode and the upper limit of the FM power The minimum value between the two determines is the lower limit of power fluctuation in BC mode and UC mode, which is calculated by the charge and discharge power P at the previous sampling time. BC , P UC Minimum charge and discharge power The difference between the BC mode and the UC mode and the lower limit of the FM power The maximum value between the two determines;

[0067] In BC mode and Also in UC mode The absolute value of is the maximum value of the charging and discharging power of the electric vehicle. The absolute value of is half of the maximum value of the electric vehicle charging and discharging power, and is 0;

[0068] According to the upper and lower limits of the electric vehicle power regulation, the expression of the change in the charging and discharging power of the electric vehicle is obtained:

[0069]

[0070] Where: ΔP EV K is the change in charging and discharging power of electric vehicles; EV is the response coefficient of the system frequency response of electric vehicles;

[0071] Taking into account user characteristics and user willingness, the EV regulation boundary is characterized, the real-time frequency regulation power of EV is constrained, and the user's frequency regulation method is determined by changing the upper and lower limits of EV power fluctuation, thereby accurately characterizing the EV frequency regulation characteristics.

[0072] Step S32 can be specifically implemented in the following manner:

[0073] In order to improve EV frequency regulation participation and its frequency regulation capability, this embodiment proposes a frequency regulation strategy that takes into account different user characteristics and intentions while ensuring that the overall trend of user charge status is on the rise.

[0074] EV participates in grid frequency regulation and reaches the end charging amount set by the user (SOC e )The required charging time is shown in formula (14).

[0075]

[0076] Where: T UC_i for i th The set SOC is achieved in UC mode eRequired charging time; ΔT UC_i for i th The charging time margin in UC mode (minutes) is designed to reduce the error caused by using open circuit voltage (OCV) instead of EV battery terminal voltage, and to compensate for the fluctuation of the mean shift of actual grid load fluctuations, thereby ensuring that the calculated charging time is not underestimated. Find the right ΔT UC_i The value needs to be analyzed and calculated by the power grid dispatching center based on long-term operation data;

[0077] In order to accurately assess the charging willingness of EV users, the initial SOC (SOC i0 ), EV battery capacity, termination SOC (SOC ie ), and the EV battery pack type and other information to calculate the battery energy required to be replenished. The calculation formula is as follows:

[0078]

[0079] Where: E i Representative th Energy required to be supplemented (kWh), S i for i th Number of power batteries connected in series, OCV i for i th The open circuit voltage of a single cell. SOC i With OCV i The functional relationship between them is expressed as follows, taking lithium manganese oxide battery as an example:

[0080]

[0081] According to the charging time required for EV to replenish energy and the actual grid-connected start time and set end time obtained by formula (14), the rules for the vehicle to participate in grid frequency regulation are determined. The rules are designed to establish a model to improve user satisfaction, which can not only provide EV users with more humane and reliable services, but also provide grid operators with more accurate frequency regulation resources.

[0082] Table 2 Rules for EVs to participate in grid frequency regulation

[0083]

[0084] In Table 2, T on_i For electric vehicles th Frequency modulation control time, that is, the end time t set by the EV user e_i The grid-connected start time t b_i When mode=1, it works in unidirectional V2G mode; when mode=2, it works in bidirectional V2G mode.

[0085] In order to verify the reliability of EV participation in grid regulation rules in providing services to individual users, the incentive effect of the Logit model on user willingness, the accuracy of the GMM model in describing user characteristics, and the effect of the regulation strategy in controlling load frequency fluctuations, this embodiment uses matlab / simulink to perform simulation verification using a single-area system as an example. Predicting the total system load peak value P LOAD_max Based on the actual data, the GMM method is used to simulate 3,000 electric vehicles. The EV vehicle type for grid-connected frequency regulation control is a pure electric vehicle, and the power battery type is a lithium manganese oxide battery. The relevant parameters of the remaining EVs participating in grid frequency control are shown in Table 3, where the electric vehicle real-time regulation power characterization parameters are typical values ​​of a single-region power system (please refer to the document "Prabha K. Power System Stability and Control [M]. Beijing: China Electric Power Press, 2002. "), and the electric vehicle battery parameters are provided by the document "Bao Yan, Jiang Jiuchun, Zhang Weiguo, et al. Research on Model and Control Strategy of Electric Vehicle Mobile Energy Storage System [J]. Automation of Electric Power Systems. 2012, 36 (22): 3643. ". The electricity prices during peak hours (11:00-24:00), valley hours (3:00-9:00), and normal hours (24:00-3:00, 9:00-11:00) are: peak hours (1.0044), valley hours (0.6950), and normal hours (0.4532).

[0086] Table 3 Parameters related to EV participating in grid frequency control

[0087] parameter parameter <![CDATA[System power generation P G / MW]]> 300 <![CDATA[Reference power P base / MW]]> 300 <![CDATA[Reference frequency f base / Hz]]> 50 <![CDATA[Governor time constant T g / s]]> 0.2 <![CDATA[Prime mover time constant T ch / s]]> 0.3 Inertia time constant M / s 10 Load-damping constant D 1 Unit adjustment rate R / (%Hz / MW(pu)) 5 Regional frequency response coefficient B / (MW / Hz(pu)) 21 EV battery capacity Q / A·h 80 Number of power batteries connected in series S 80 Battery cell internal resistance R / mΩ 0.8 <![CDATA[SOC maximum value SOC max (%)]]> 100 <![CDATA[SOC minimum SOC min (%)]]> 20 <![CDATA[Absolute value of the maximum charge and discharge power of the EV, P max / MW (pu)]]> 3×10－5 <![CDATA[Upper limit of power grid frequency fluctuation value Δf max / Hz]]> 0.1 <![CDATA[Lower limit of power grid frequency fluctuation value Δf min / Hz]]> －0.1 <![CDATA[EV frequency response coefficient K EV / (MW / Hz(pu))]]> －3×10－3 <![CDATA[EV time constant T EV / ms]]> 0.3

[0088] EV real-time power regulation is mainly divided into three scenarios:

[0089] Scenario 1: EV charging power characterization considering user characteristics;

[0090] Scenario 2: EV charging and discharging power characterization considering charging and discharging control strategy;

[0091] Scenario 3: EV charging and discharging power characterization considering user willingness and charging and discharging control strategy.

[0092] Figure 4 This is the real-time power output diagram of EV in different scenarios. Figure 1 The load curve of normal power output is Figure 4 Compared with the EV charging and discharging power curve considering the charging and discharging control strategy in (a), the load power generated by the EV cluster participating in the grid frequency regulation is significantly reduced, reducing the load power intensity by 41% and increasing the load factor by 5.23%. Figure 4(b) is the EV charging and discharging power curve that takes into account user willingness and charging and discharging control strategy. The data results show that the charging and discharging control method of scenario 3 effectively reduces the load power generated by the EV cluster. At the same time, based on the electricity price incentive provided by the peak and valley charging of EVs daily load curve, scenario 3 provides reverse power when the grid demand is peak.

[0093] Figure 5 Schematic diagram of the adjustable power capability of EVs in scenarios 2 and 3. Figure 5 (a) and Figure 5 (c) shows that the EV regulation power in scenario 2 can provide upward adjustable power during the period of excess renewable energy production from about 11 noon to 12 midnight, absorbing the excess electricity generated when the wind and solar radiation are strong. By characterizing the upward regulation power reserve, the electricity during the off-peak period can be absorbed in a planned manner and released during the peak period. Therefore, this is also a characterization of the frequency regulation capability of EVs. It helps to reduce the high-priced peak electricity that needs to be purchased, thereby reducing the overall operating cost.

[0094] Figure 5 (b) and Figure 5 (d) shows that its downward adjustment power curve has two peaks, namely the noon peak and the evening peak, which can increase the EV cluster to quickly release the stored power to the grid during the period when the grid load demand is high and the fluctuation is large. Therefore, the EV cluster can be regarded as a distributed load buffer device, providing power support within the difference between the real-time power and the upper and lower boundaries, balancing the supply and demand relationship, and thus improving the stability of the grid.

[0095] Figure 5 (c) and Figure 5 (d) and Figure 5 (a) and Figure 5 The difference in (d) lies mainly in the different electricity price incentives. According to the simulation results of the Logit model, users are more willing to participate in grid regulation during periods with higher peak-valley membership, which not only has a good effect on real-time power regulation, but also improves the reserve reserve for upward and downward power regulation.

[0096] Figure 6 The SOC change curve of a typical EV during the stay time in scenario 2 and scenario 3. Figure 6 As can be seen in (a), the control strategy proposed in this embodiment can effectively utilize the surplus EV residence time to flexibly increase or decrease the EV charging power to participate in grid frequency regulation without affecting user travel. Figure 6(b) is the SOC change curve of EV within the dwelling time under the scenario of taking into account the control strategy and user's dispatchable willingness. Under the premise of meeting the basic vehicle demand of users, the logit model is used to enable users to participate in the reverse charging of the power grid to obtain relevant subsidies as much as possible. After participating in the grid regulation, users can not only reduce their expenses, but also obtain relevant economic benefits from it. It not only meets the frequency regulation needs of the power grid, but also gives back to users, getting rid of the game between the subject and the individual in the traditional method and achieving a win-win situation.

[0097] This embodiment also conducts a visual analysis of the main decision-making factors that affect electric vehicles' participation in grid frequency regulation. First, from an overall perspective, the impact of different electricity prices on the boundary of EV's participation in grid frequency regulation is analyzed; second, from an individual perspective, the correlation between EV users' differentiated initial SOC and waiting time and the proportion of their participation in grid frequency regulation is studied; specifically, the following three aspects are included.

[0098] First, the impact of different electricity price incentives on EV participation in frequency regulation.

[0099] Figure 7 The figure is a comparison of the effects of three different electricity prices on the EV charging and discharging power boundary under scenario 3. It can be seen from the lines in the figure that the weighted decision variable electricity price incentive of the Logit model proposed in this embodiment is of great significance for describing the EV charging and discharging power boundary. When the electricity price is low, the charging demand is strong, the upper power boundary is higher, and the lower power boundary is lower. When the electricity price is high, the charging demand is weak, and the EV discharges to the grid in the opposite direction. The lower power boundary is higher and the upper power boundary is lower. The time-of-use electricity price obtained by solving the peak-valley membership takes into account the advantages of both high and low electricity prices. While reducing the power fluctuation, the economy of the power supply of the grid is guaranteed.

[0100] Secondly, the impact of EV initial SOC on EV participation in grid frequency regulation.

[0101] like Figure 8 As shown, the horizontal axis is the initial SOC percentage, and the vertical axis is the weighted average of the ratio of the time when all EV individuals participate in frequency modulation within the percentage interval to the waiting time of the individual at the charging pile. Figure 8 There are fewer EVs with an initial SOC of 100%, and the proportion of EVs participating in frequency regulation is obvious. The trends of other data also reflect that the initial SOC size is positively correlated with the proportion of EVs participating in grid frequency regulation, that is, as the initial SOC of electric vehicles increases, the ability of EVs to participate in grid frequency regulation increases significantly.

[0102] Third, the impact of EV residence time on EV participation in frequency modulation.

[0103] like Fig. 9As shown in Figure 2, EVs with longer residence times are more likely to be included in frequency regulation services. This provides a larger time window to participate in the grid’s demand response and frequency regulation services. This means that the battery has ample time to reach an ideal state of charge, thereby maximizing its effectiveness in responding to frequency regulation signals.

[0104] This embodiment also performs correlation analysis.

[0105] The correlation analysis results obtained based on SPSS are as follows Fig.10 and Fig.11 shown. Fig.10 It is a variable scatter plot matrix, whose main diagonal is a bar chart of the number distribution of 3,000 EV individuals in their respective decision variables, and the non-diagonal elements are the respective cross-correlation scatter plots. Through the first two figures in the third column, it can be intuitively analyzed that the initial SOC and waiting time are positively correlated with the proportion of participating in frequency modulation. Fig.11 The matrix correlation heat map shows that, through color scale analysis, although both initial SOC and waiting time are positively correlated with the proportion of participants in frequency modulation, the correlation coefficient of initial SOC is 0.6024, which is higher than that of waiting time (0.3937), showing a stronger correlation.

[0106] Combining the above examples and corresponding analysis, it can be seen that when the EV-assisted frequency regulation control strategy is adopted, although the system frequency and active output fluctuations are improved to a certain extent compared with the conventional charging scenario, the frequency regulation effect is poor compared with the control strategy that considers user willingness. The reason is that the simple EV-assisted control scenario does not take into account the impact of electricity price incentives and does not take into account that the benefits obtained by users during charging will greatly increase user willingness, thus restricting the frequency regulation potential of the EV cluster. In the design of the EV frequency regulation control strategy of this embodiment, through real-time state feedback of user characteristics and user willingness and compensation of external disturbances, the EV cluster can respond to changes in grid frequency more flexibly, reduce the grid frequency deviation range, and smooth the fluctuations in the active output of the prime mover. Fig.12 (a) shows the comparison curve of the regulation power provided by scenario 2 and scenario 3. It can be seen that the difference in regulation capacity between scenario 3 and scenario 2 during the peak period from 11:00 am to 11:00 pm is significantly higher than that during the valley period and normal period. Fig.12 (b) shows the impact of active power output on the system frequency change characteristics in three scenarios: conventional charging, considering EV-assisted frequency regulation, and considering user willingness model.

[0107] Combined with the simulation index calculation results, it can be seen that scenario 3 shows excellent frequency modulation performance in all frequency indicators. When the EV cluster adopts the improved method, Δf max With Δf min Compared with scenario 2 and scenario 1, the reduction is 13.91% and 29.27% ​​respectively. In addition, from the system Δf RMSFrom the perspective of Δf, the improved method reduces 11.13% and 29.59% respectively compared with the traditional EV auxiliary control and non-frequency modulation control; dtime From the time scale of [±0.033], the improved method showed the shortest duration and shortened the frequency deviation time by 18.21% and 42.69% respectively compared with the other two control methods; from the COVAR (f sys ,P EV ), the system frequency fluctuation degree caused by the active output fluctuation of the EV cluster under the improved method was reduced by 17.76% and 41.30% respectively; in addition, from Fig.12 From the frequency deviation PDF shown in (c), it can be seen that the probability that the frequency deviation of the improved method is concentrated in the ±0.033Hz range is greater than that of the other two control methods, which effectively prevents the generation of new frequency spikes during the frequency modulation period.

[0108] In summary, this application proposes a method for EV real-time power regulation characterization and frequency support based on the premise of considering different user characteristics and user needs, taking into account user uncertainty characteristics and user dispatchable willingness. Through Gaussian mixture distribution and Logit model, a charging scheme that takes into account auxiliary frequency control and user satisfaction is formulated, and the following conclusions are drawn:

[0109] 1) Taking into account user characteristics and user willingness, the EV regulation boundary is characterized, the real-time frequency regulation power of EV is constrained, the user frequency regulation mode is determined by changing the upper and lower limits of EV power fluctuation, and the EV frequency regulation boundary is accurately characterized.

[0110] 2) Based on SPSS, a significance analysis of the factors affecting EV's participation in grid frequency regulation was conducted. It was found that the initial SOC of EV users, the waiting time at the charging station, and the electricity price incentive are the three most important decision variables affecting EV's participation in grid frequency regulation. From an overall perspective, this paper studies the impact of EV's participation in the grid frequency regulation capability boundary by perturbing the electricity price; from an individual perspective, combined with the correlation coefficient, it is found that the initial SOC is more correlated with the proportion of EV individual participation in frequency regulation than the waiting time.

[0111] 3) The proposed electric vehicle frequency regulation strategy ensures the overall upward trend of the user's state of charge and takes into account the user's uncertainty characteristics and user's dispatchable willingness, which ensures the overall upward trend of the SOC of EV users during the regulation process. By considering the user's willingness, the fast frequency improvement control strategy is optimized, effectively reducing the fluctuation of load power, and making the frequency deviation as concentrated as possible in the ±0.033Hz range.

[0112] The above implementation modes are only used to illustrate the present application, and are not intended to limit the present application. Ordinary technicians in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present application. Therefore, all equivalent technical solutions also belong to the scope of the present application, and the scope of patent protection of the present application shall be limited by the claims.

Claims

1. A method for regulating power characterization and frequency support of electric vehicles considering user characteristics and intentions, characterized in that: The method comprises: The uncertainty factors related to electric vehicles are described based on the Gaussian mixture distribution method; wherein the uncertainty factors include the user behavior pattern, battery capacity, and charging and discharging rate of the electric vehicle; The Logit model is used to predict the dispatchability willingness of electric vehicles; Characterize the electric vehicle regulation boundary considering user characteristics and user willingness, determine the user's regulation method by changing the upper and lower limits of electric vehicle power fluctuation, and determine the electric vehicle frequency regulation strategy.

2. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 1 is characterized in that: The uncertainty factors related to electric vehicles are described based on the Gaussian mixture distribution method, including: Based on the Gaussian mixture model, according to the arrival time T of the electric vehicle _arival_i The probability density function established by the actual data of the initial state of charge is classified and fitted, and the specific probability distribution function P(T arrive_i |Θ) is: Where: N θ (T arrive_i ;μ θ ,D θ ) is θ th Gaussian component function, μ θ is the average value, D θ is the variance; J is the total number of Gaussian components; α θ is the weight of the Gaussian component; the expression of one-dimensional GMM: θ=(θ1,……,θ J ), by determining T arrive_i , μ θ , D θ To fit the Gaussian mixture distribution, J is set by the EV user before estimation, and J is adjusted according to the actual fitting effect after estimation; Determine the probability density function of the initial energy of the electric vehicle It is expressed as: Where: E i0 and SOC i0 i th The initial energy and initial SOC of the EV; is the mean of the ith distribution; D θ is the variance of the probability density distribution function; τ i The number of days that an EV can be driven after charging; d iR For EV range; The parameters of the initial energy probability density function of the electric vehicle are estimated through historical data, and the SOC i0 and T arrive_i The density function in is derived. During the simulation, random numbers are drawn using the Monte Carlo method according to each probability density. At the same time, the departure time T is obtained according to the stop time set by the user. _deprive , depicting the complete user characteristics.

3. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 2 is characterized in that: The Logit model is used to predict the dispatchability of electric vehicles, including: Based on the state of charge of electric vehicles, dispatch subsidies of the power grid in different time periods, and the time that electric vehicle users stay in the charging area, the hidden variables in the probability model of user dispatch willingness are determined; The Logit function is used to convert the latent variable into the user's choice probability; wherein the user's choice probability is represented by the probability function of the electric vehicle accepting the grid dispatch and not accepting the dispatch.

4. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 3 is characterized in that: Based on the state of charge of electric vehicles, the dispatch subsidy of the power grid in different time periods, and the time that electric vehicle users stay in the charging area, the hidden variables in the probability model of user dispatch willingness are determined by the following formula: y i =α1+β1SOC i0 +β2K EV_i +β3T on_i +u i (4) Where: y i for i th The decision intention hidden variable, i th For electric vehicles; i is the coefficient of the benchmark probability distribution function; β1, β2, β3 are the coefficients of each explanatory variable; SOC i , K Ev_i 、T on_i Represents electric vehicle i th Initial state of charge, dispatch subsidy incentives, and user-entered charging station dwell time; u i is the random error variable of decision intention.

5. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 4 is characterized in that: The user's selection probability is expressed by the probability function of the electric vehicle accepting grid dispatch and not accepting dispatch as follows: Where: Y i for i th The decision intention variable, Y i =1 means the user chooses to accept the grid dispatch, Y i =0 means the user chooses to be idle, P i (Y i =1) represents the probability of the user making the decision Y, and its value range is (0,1), P i (Y i =0) represents the probability that the i-th electric vehicle user chooses not to accept grid dispatch, F(Y i ) represents the cumulative distribution function of decision willingness, and e represents the base of the natural logarithm; The decision-making willingness variable is affected by the state of charge, the dispatch subsidy of the power grid in different time periods, and the stay time of electric vehicle users in the charging area; among them, the dispatch subsidy has the highest weight on the willingness choice, which affects the user's willingness to participate in the regulation in the Logit model according to the different peak and valley electricity prices of each vehicle in the electric vehicle cluster in each hour during its stay.

6. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 5 is characterized in that: The willingness of users to participate in the regulation in the Logit model is affected by the different peak and valley electricity prices of each vehicle in the electric vehicle cluster during their respective stay periods, including: Firstly, based on the daily load curve of electric vehicle charging taking into account user characteristics, the peak and valley memberships are divided using the trapezoidal membership function calculation method. T p is the valley load threshold, T v is the peak load threshold, then the daily load curve is a set of data L = {l1,l2,...,l i ...,l n }, where each l i Represents the grid load at a set point in time; The trapezoidal valley membership function is defined as four parameters, namely, T pa ,T p ,T p ' and T pd ; where [T p ,T p '] represents the core area with a membership degree of 1, [T pa ,T p ] and [T p ',T pd ] is the rising edge and falling edge, and the membership increases from 0 to 1, or decreases from 1 to 0; The four parameters of the trapezoidal peak membership function are set as T va ,T v ,T v ' and T vd ; For peak-valley division, the peak load and valley load membership functions are shown in equations (7) and (8) respectively: Where M v (l) is the valley membership function representing the load curve at a certain moment, which is used to measure the closeness between the current load and the valley load. p (l i ) represents a certain moment l in the load curve i The peak membership function is used to measure the current load l i The degree of proximity to the peak load, l is the load value at a specific moment in the power grid load curve.

7. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 3 is characterized in that: Characterize the electric vehicle regulation boundary considering user characteristics and user willingness, determine the user's regulation method by changing the upper and lower limits of electric vehicle power fluctuation, and determine the electric vehicle frequency regulation strategy, including: The electric vehicle frequency modulation power expression is established as follows: ΔP EV =Δf·K EV (9) Where: ΔP EVi is the change in charging and discharging power of the i-th electric vehicle, Δf is the frequency deviation of the system, and K EVi is the frequency response coefficient of the i-th electric vehicle; Establish upper and lower limits of electric vehicle power regulation constraints, expressed as: Where: ΔP U 1 C and ΔP B 1 C The power fluctuation limits in UC mode and BC mode are respectively determined by the maximum charge and discharge power The charge and discharge power P of the previous sampling time BC , P UC The difference between the BC mode and the UC mode and the upper limit of the FM power The minimum value between the two determines is the lower limit of power fluctuation in BC mode and UC mode, which is calculated by the charge and discharge power P at the previous sampling time. BC , P UC Minimum charge and discharge power The difference between the BC mode and the UC mode and the lower limit of the FM power The maximum value between the two determines; In BC mode and Also in UC mode The absolute value of is the maximum value of the charging and discharging power of the electric vehicle. The absolute value of is half of the maximum value of the electric vehicle charging and discharging power, and is 0; According to the upper and lower limits of the electric vehicle power regulation, the expression of the change in the charging and discharging power of the electric vehicle is obtained: Where: ΔP EV K is the change in charging and discharging power of electric vehicles; EV is the response coefficient of the system frequency response of electric vehicles; Taking into account user characteristics and user willingness, the electric vehicle regulation boundary is characterized, the real-time frequency modulation power of the electric vehicle is constrained, and the user's frequency modulation method is determined by changing the upper and lower limits of the electric vehicle power fluctuation to characterize the frequency modulation characteristics of the electric vehicle; Based on the frequency regulation characteristics of electric vehicles, the frequency regulation strategy of electric vehicles is determined.

8. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 7 is characterized in that: Based on the frequency regulation characteristics of electric vehicles, determine the frequency regulation strategy of electric vehicles, including: The charging time required for electric vehicles to participate in grid frequency regulation and reach the set termination charging amount is calculated using the following formula: Where: T UC_i For electric vehicles th In UC mode, the set end charge SOC is reached ie Required charging time; ΔT UC_i represents the charging time margin of the i-th electric vehicle in UC mode; E i for i th The amount of energy required to replenish; Taking into account the initial state of charge SOC i0 , electric vehicle battery capacity, end state of charge SOC ie , electric vehicle battery pack type, calculate the battery energy required to be replenished using the following formula: Where: S i for i th Number of power batteries connected in series, OCV i for i th The open circuit voltage of a single cell; The strategy for the electric vehicle to participate in the grid frequency regulation is determined based on the charging time required for the electric vehicle to replenish energy, the actual grid-connected start time, and the set end time obtained from formula (14).

9. The electric vehicle power regulation characterization and frequency support method considering user characteristics and intentions as claimed in claim 8, characterized in that: The strategy for electric vehicles to participate in grid frequency regulation is: If T UC_i ≥T on_i , then mode=2 until t e_i time; among them, T on_i For electric vehicles th Frequency modulation control time, mode = 2 means working in BC mode, t e_i is the end time; If SOC0<50% and T UC_i <T on_i , then mode = 1 until SOC = 50%, then mode = 2 for T on_i -T UC_i , then mode=1 until t e_i moment; where SOC represents the state of charge of the electric vehicle battery; where mode = 1, it means working in UC mode; If T UC_i <T on_i If SOC0≥50%, then check whether the time from SOC=50% to SOC0 is greater than the charging power P max / 2 The time from SOC=50% to SOC0, if T≥T pmax / 2 , then mode=2; where T represents the actual time required for the electric vehicle to change from the current SOC to the target SOC, T pmax / 2 To represent 50% of the peak charging power - P max / 2 When charging, the theoretical time required to go from SOC = 50% to the target SOC; If SOC0 ≥ 50%, T ≥ T pmax / 2 And T UC_i <T on_i , then mode=2 for T on_i -T u , then mode=1 until t e time; where t e Indicates the termination time of electric vehicles participating in grid frequency regulation or charging / discharging tasks.

Images