Electric vehicle cluster and unit combined robust frequency modulation method and device

Abstract

The invention discloses an electric vehicle cluster and unit joint robust frequency modulation method and device, and belongs to the technical field of power systems, and the method comprises the steps: generating a scene set of frequency modulation signals based on obtained historical frequency modulation indication signal data; a charge-discharge excitation mechanism is considered, a kernel density estimation interval is adopted to predict a result that an electric vehicle cluster participates in active regulation and control of capacity, and a distribution robust optimization model of conditional value-at-risk is established; inputting the scene set of the generated frequency modulation signal into the distribution robust optimization model for optimization to obtain an optimized distribution robust optimization model; acquiring real-time frequency modulation indication signal data, inputting the real-time frequency modulation indication signal data into the optimized distributed robust optimization model, and predicting to obtain an electric vehicle cluster and unit joint robust frequency modulation result; according to the method, economic benefits obtained by participating in the frequency modulation market can be minimized, and the reporting EV cluster frequency modulation capacity can be determined on the premise that an aggregator is helped to have the highest profit expectation, profit standard deviation and condition loss expectation.

Description

Technical Field

[0001] The invention relates to a method and a device for joint robust frequency regulation of an electric vehicle cluster and a generator set, belonging to the technical field of power systems. Background Art

[0002] In recent years, the large-scale development of renewable energy represented by wind power and photovoltaic power has brought severe challenges to its grid connection due to the random intermittent characteristics of new energy output power. The resulting system frequency instability problem has become increasingly prominent, and the system frequency problem has received more and more widespread attention in the new environment. At present, the auxiliary service of system frequency regulation in my country is mainly undertaken by thermal power units. In the new environment, the problems of thermal power units such as long response lag, low ramp rate, and inability to accurately track instructions are becoming increasingly prominent. An effective way is to configure a considerable scale of battery energy storage for the frequency regulation units. However, the current large-scale battery energy storage investment and operating costs are high; another effective way is to treat the electric vehicle (EV) cluster as a large-scale energy storage connected to the system node, and participate in frequency regulation together with the units to save energy storage investment costs.

[0003] To solve the problem of unstable system frequency, on the one hand, EV clusters have flexible battery energy storage systems that can provide or absorb electricity in a very short time, thereby achieving rapid regulation of grid frequency, allowing EV clusters to make up for the shortcomings of traditional units in terms of regulation delay; on the other hand, the benefits of traditional units participating in the frequency regulation market alone are becoming increasingly pessimistic. For this reason, various countries have formulated market rules oriented towards frequency regulation service effects, but the capacity and duration of differentiated EV users are difficult to meet the threshold of the frequency regulation market. The regulation characteristics and market rules have promoted the development of the frequency regulation model of cooperation between clusters and units.

[0004] Regarding the research on the joint frequency regulation of electric vehicle clusters and units, there are several technical problems:

[0005] 1. It mainly focuses on the basis that aggregators can control all EVs to participate in frequency regulation, without considering the uncertainty of EV users participating in charging and discharging control. Although some literature has begun to focus on the importance of economic and non-economic means such as charging cost incentives, compensation amounts, and free charging for EV participation in frequency regulation, it does not consider the differences in charging and discharging control of different EV response incentive mechanisms. The proposed incentive economic means fail to comprehensively consider the battery loss costs caused by discharge and the differences in incentives for each individual user due to additional compensation, so it is difficult to effectively reflect the incentive costs of EV users participating in frequency regulation.

[0006] 3. Existing inventions only consider the uncertainty on the EV cluster side or the market side in terms of uncertainty models, and fail to comprehensively consider the uncertainty of both parties in frequency regulation resources, making it difficult to effectively achieve the coordinated allocation of EV cluster and unit frequency regulation capacity. Although existing literature has analyzed the uncertainty of EV charging behavior in joint frequency regulation, it has ignored the analysis of the risk value of charging behavior on the results of joint frequency regulation. At the same time, on the market side, the correlation between signals issued by different frequency regulation markets has not been considered, resulting in a significant increase in the probability of continuous occurrence in extreme frequency regulation signal scenarios, making the results of joint frequency regulation appear too conservative.

[0007] Therefore, conducting research on joint frequency regulation of EV clusters and traditional units will help improve system frequency stability, increase the penetration rate of renewable energy, and optimize the allocation of frequency regulation resources. Summary of the invention

[0008] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a method and device for robust frequency regulation of electric vehicle clusters and units, which can minimize the economic benefits obtained from participating in the frequency regulation market and help aggregators determine the reportable EV cluster frequency regulation capacity under the premise of having the highest expected profit, standard deviation of profit and expected conditional loss.

[0009] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0010] In a first aspect, the present invention provides a method for joint robust frequency regulation of an electric vehicle cluster and a generator set, comprising:

[0011] Obtain real-time frequency modulation indication signal data, input it into the pre-built and optimized distributed robust optimization model, restructure the distributed robust optimization model into a second-order cone programming problem, and predict the result of the joint robust frequency modulation of the electric vehicle cluster and the unit;

[0012] The construction and optimization method of the distributed robust optimization model includes:

[0013] Generate a scene set of frequency modulation signals based on the acquired historical frequency modulation indication signal data;

[0014] Considering the charging and discharging incentive mechanism, the kernel density estimation interval is used to predict the results of the active regulation of electric vehicle clusters, and a distributed robust optimization model of conditional risk value is established;

[0015] The generated scene set of the frequency modulation signal is input into the distributionally robust optimization model for optimization to obtain an optimized distributionally robust optimization model.

[0016] Furthermore, the generating of a scene set of the frequency modulation signal based on the acquired historical frequency modulation indication signal data includes:

[0017] S11: Collect and normalize the historical frequency modulation indicator signal data of the PJM market. Randomly select from the historical FM indicator signal data as the initial cluster centers; S12: Calculate the Euclidean distance between all historical frequency modulation indicator signal data and each cluster center; and classify the data into the class represented by the cluster center closest to the data; S13: Calculate the mean of each category of historical frequency modulation indicator signal data and use it as the new cluster center of the category data, and calculate the average Euclidean distance of all historical frequency modulation indicator signal data to the cluster center; S14: Determine whether the cluster center and the average value of the Euclidean distance have changed. If so, continue with step S12; if not, end clustering and generate a scene set of the FM signal.

[0018] Furthermore, the charging and discharging incentive mechanism is considered, and the kernel density estimation interval is used to predict the results of the electric vehicle cluster participating in the active regulation of capacity, and a distributed robust optimization model of conditional risk value is established, including: Construct charging incentive mechanism and discharging incentive mechanism, and obtain charging incentive price and discharging incentive price; According to the charging incentive price and the discharging incentive price, the proportion of electric vehicles that respond to the charging incentive mechanism and the discharging incentive mechanism are obtained respectively; According to the obtained statistical data of basic state parameters of electric vehicles, the basic state parameter results of a single electric vehicle are sampled and obtained; According to the basic state parameter results of a single electric vehicle, the power constraint and energy operation constraint of the single electric vehicle are obtained by Monte Carlo simulation; Based on the power constraints and energy operation constraints of individual electric vehicles, multiple Monte Carlo simulations were performed according to the proportion of electric vehicles to obtain multiple preliminary results of the capacity of electric vehicle clusters participating in active regulation. The final electric vehicle cluster active regulation capacity result is obtained by using kernel density estimation interval prediction based on multiple preliminary electric vehicle cluster active regulation capacity results. The final electric vehicle cluster participates in active capacity regulation, and a distributed robust optimization model is established through the conditional risk value distributed robust optimization method.

[0019] Furthermore, the multiple groups of preliminary electric vehicle clusters participating in active regulation capacity results are used to obtain the final electric vehicle cluster participating in active regulation capacity results by using kernel density estimation interval prediction, and the calculation formula is as follows: ; ; Where: is the bandwidth parameter, which is used to control the width of the kernel function; is the number of data points in the test set; For a given bandwidth The estimated probability density function under ; For is the kernel function of the parameter; It is a sample composed of multiple preliminary groups of electric vehicle clusters participating in active capacity regulation results. is the cth observation in the sample; is the standardized kernel function; and They are the upper and lower limits of the prediction interval of the test set; The final result is the active regulation capacity of the electric vehicle cluster; is the confidence level; is the inverse function of the estimated probability density function.

[0020] Furthermore, the objective function of the distributed robust optimization model is to minimize the joint frequency regulation cost of the electric vehicle cluster and the unit under the regulation of the electric vehicle aggregator, wherein the joint frequency regulation cost of the electric vehicle cluster and the unit includes the startup cost, shutdown cost, no-load cost, coal consumption cost, frequency regulation cost and frequency regulation benefit of the unit; the frequency regulation cost, frequency regulation benefit, charging incentive cost and discharge incentive cost of the electric vehicle.

[0021] Furthermore, the objective function of the distributed robust optimization model is calculated as follows:

[0022] ;

[0023] In the formula, is the objective function, S is the total number of FM signal scenes after clustering; s is the scene; is the probability of the scenario occurring; and The frequency regulation capacity matrix of the unit and the frequency regulation capacity matrix of the EV cluster obtained in advance, To consider the worst-case EV cluster frequency regulation benefit.

[0024] In a second aspect, the present invention provides a robust frequency modulation device for electric vehicle clusters and generator sets, comprising:

[0025] The frequency modulation unit is used to obtain real-time frequency modulation indication signal data, input it into a pre-built and optimized distributed robust optimization model, reconstruct the distributed robust optimization model into a second-order cone programming problem, and predict the result of the joint robust frequency modulation of the electric vehicle cluster and the unit;

[0026] The construction and optimization method of the distributed robust optimization model includes:

[0027] Generate a scene set of frequency modulation signals based on the acquired historical frequency modulation indication signal data;

[0028] Considering the charging and discharging incentive mechanism, the kernel density estimation interval is used to predict the results of the active regulation of electric vehicle clusters, and a distributed robust optimization model of conditional risk value is established;

[0029] The generated scene set of the frequency modulation signal is input into the distributionally robust optimization model for optimization to obtain an optimized distributionally robust optimization model.

[0030] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the aforementioned methods.

[0031] In a fourth aspect, the present invention provides a computer device, comprising:

[0032] Memory, for storing computer programs / instructions;

[0033] A processor is used to execute the computer program / instructions to implement the steps of any of the aforementioned methods.

[0034] In a fifth aspect, the present invention provides a computer program product, comprising a computer program / instruction, which implements the steps of any one of the aforementioned methods when executed by a processor.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] (1) The present invention provides a method and device for robust frequency regulation of electric vehicle clusters and units, which fully considers the robust optimization scheduling model of EV cluster and unit joint frequency regulation with dual uncertainty, can minimize the economic benefits obtained from participating in the frequency regulation market, and help aggregators determine the reportable EV cluster frequency regulation capacity under the premise of having the highest expected profit, standard deviation of profit and expected conditional loss;

[0037] (2) The stochastic optimization method and the CVaR-based distributed robust optimization method in the present invention can handle the uncertainty of the frequency modulation signal and EV charging behavior in the proposed model. The evaluation indicators PICP and CWC of KDE interval prediction reach 0.334 and 213.04, respectively.

[0038] (3) The calculation results of the present invention show that when only traditional units are considered to participate in market frequency regulation, the frequent start and stop of the generator sets brings the worst economic effect; when only EV clusters are considered to participate in market frequency regulation, a large number of EVs are required to meet the frequency regulation demand; when EV clusters and units are considered to participate in market frequency regulation together, the rapid regulation of EV clusters compensates for the delay of traditional unit regulation, and the cost of frequency regulation is reduced by 22.45% compared with unit frequency regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the operating area of ​​a single EV in the present invention;

[0040] Figure 2 It is a schematic diagram of survey data of EV user participation and response in a specific embodiment;

[0041] Figure 3 is a frequency modulation signal scenario in a specific embodiment (the scenario number is 10 schematic diagrams);

[0042] Figure 4 Schematic diagram of root mean error and loss value of the training set in a specific embodiment;

[0043] Figure 5 Schematic diagram of probability distribution of a training set in a specific embodiment;

[0044] Figure 6 A schematic diagram of a cumulative probability density function in a specific embodiment;

[0045] Figure 7 It is a schematic diagram of KDE interval prediction results in a specific embodiment;

[0046] Figure 8 It is a schematic diagram of cluster frequency regulation capacity under different confidence levels and discharge incentive prices in a specific embodiment;

[0047] Fig. 9 It is a schematic diagram of the mean and variance of charging demand of EV cluster in each time period in a specific embodiment;

[0048] Fig.10 It is a schematic diagram of the reporting capacity of the frequency modulation of the EV cluster in each period in a specific embodiment;

[0049] Fig.11 It is a schematic diagram of comparing the frequency modulation capacity of Example 1 in multiple scenarios in a specific embodiment;

[0050] Fig.12 It is a schematic diagram of comparing the frequency modulation capacity of Example 3 in multiple scenarios in a specific embodiment;

[0051] Fig.13 This is a flow chart of the electric vehicle cluster and unit joint robust frequency modulation method provided by the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0053] Embodiment 1, as Fig.13 As shown, this embodiment introduces a method for joint robust frequency regulation of electric vehicle clusters and generator sets, including:

[0054] Obtain real-time frequency modulation indication signal data, input it into the pre-built and optimized distributed robust optimization model, restructure the distributed robust optimization model into a second-order cone programming problem, and predict the result of the joint robust frequency modulation of the electric vehicle cluster and the unit;

[0055] The construction and optimization method of the distributed robust optimization model includes:

[0056] Generate a scene set of frequency modulation signals based on the acquired historical frequency modulation indication signal data;

[0057] Considering the charging and discharging incentive mechanism, the kernel density estimation interval is used to predict the results of the active regulation of electric vehicle clusters, and a distributed robust optimization model of conditional risk value is established;

[0058] The generated scene set of the frequency modulation signal is input into the distributionally robust optimization model for optimization to obtain an optimized distributionally robust optimization model.

[0059] The application process of the electric vehicle cluster and unit joint robust frequency modulation method provided in this embodiment specifically involves the following steps:

[0060] Step 1: Cluster the historical frequency modulation data using the random optimization method to construct the frequency modulation signal scenario set required for the joint frequency modulation analysis of EV clusters and units;

[0061] S11: Collect and normalize the historical frequency modulation indicator signal data of the PJM market. Randomly select from the historical FM indicator signal data as the initial cluster centers;

[0062] S12: Calculate the Euclidean distance between all historical frequency modulation indicator signal data and each cluster center; and classify the data into the class represented by the cluster center closest to the data;

[0063] S13: Calculate the mean of each type of frequency modulation signal curve and use it as the new cluster center of the curve of this type, and calculate the average value of the Euclidean distance from all frequency modulation signal curves to their cluster center;

[0064] S14: Determine whether the cluster center and the average value of the Euclidean distance have changed. If so, continue with step S12; if not, end clustering and generate a scene set of the FM signal.

[0065] Step 2: Taking full account of the charging and discharging incentive mechanism and using the Kernel Density Estimation (KDE) interval to predict the probability of EV users participating in active regulation, the uncertainty problem of EV charging behavior is analyzed based on the Conditional Value at Risk (CVaR) distributed robust optimization method;

[0066] S21: Construct a charging and discharging incentive mechanism, which includes a charging incentive mechanism and a discharging incentive mechanism:

[0067] S211: Charging incentive mechanism, the difference between the power generation price in the real-time energy market and the power price announced by the aggregator is proposed as the charging incentive price. When the EV needs to be charged, the user can learn about the current charging incentive price. To decide whether to respond to CIM. In the EV cluster, the proportion of EVs that respond to CIM It can be expressed as:

[0068] (1);

[0069] Where: Function Used to calculate the proportion of EVs that respond to CIM in the EV cluster.

[0070] Based on the attached Figure 1 The operating area of ​​a single EV in The charging cycle of an EV is , then considering the EV response CIM, When an EV is charging, the charging time period that the aggregator can adjust is:

[0071] (2);

[0072] Where: is the battery capacity of the EV; For the The minimum time it takes for an EV to complete charging needs; and Respectively The starting SOC of the EV when it starts charging and the required SOC when it ends charging; For the The adjustable end time of the EV response to CIM; The upper limit of the charging and discharging power of the EV; Charging efficiency for EVs; is the time interval; is a 0-1 variable, and its value is 1, indicating EV responds to CIM, and a value of 0 indicates The EV did not respond to the CIM.

[0073] S212: Discharge incentive mechanism. Under normal temperature conditions, the battery loss of lithium battery EV is mainly affected by the discharge depth and the number of cycles. During the battery life cycle, when the discharge depth is When the total charge and discharge of EV It can be expressed as shown in formula (3): the additional battery loss cost caused by unit discharge power is As shown in formula (4).

[0074] (3);

[0075] (4);

[0076] Where: For the A specific depth of discharge value; is the number of discharge depth values ​​taken; Purchase cost for batteries.

[0077] Calculate the discharge incentive price of EV aggregators according to formula (5): ,EV users can choose whether to respond to DIM according to their preferences.

[0078] (5);

[0079] Where: Compensate for battery discharge loss of EV during the entire regulation cycle; Discharge subsidy for EV aggregators.

[0080] In view of the differences in the degree of response of EV users to DIM, based on market research data, the proportion of EVs that respond to DIM is estimated according to formula (6): :

[0081] (6);

[0082] Where: Function Used to calculate the proportion of EVs in the cluster that respond to DIM.

[0083] EV users' response to CIM helps save charging costs, while their response to DIM can bring them additional benefits. EV users' response to DIM is mainly to obtain more economic benefits, so they have a stronger motivation to participate. Considering EV's response to DIM, the discharge time period that the aggregator can adjust is:

[0084] (7);

[0085] Where: ti,2 respectively represent the controllable end time of the response of the i-th single EV to DIM; is a 0-1 variable, and its value is 1, indicating The EV responds to DIM, and a value of 0 indicates that The EV does not respond to DIM; is a 0-1 variable, and its value is 1, indicating The EV responds to CIM or DIM, and the value is 0, indicating that the The EV does not respond to the CIM and DIM.

[0086] S22: Use KDE intervals to predict the probability of EV users participating in active regulation, and establish the uncertainty set of frequency regulation capacity at each moment according to formula (8): , in order to solve the frequency modulation capacity that aggregators can adjust at different time periods The unknown nature of the distribution.

[0087] (8);

[0088] Where: is the uncertain set of frequency modulation capacity at each moment; is the distribution of frequency modulation capacity at each moment; is the actual frequency regulation capacity matrix of the EV cluster at each moment; and are the mean and variance matrices of the uncertainty frequency modulation capacity distribution, respectively; and are the upper and lower limit matrices of the uncertain frequency regulation demand capacity respectively.

[0089] S221: Monte Carlo sampling of basic state parameters of the i-th EV;

[0090] S222: Determine the EV proportions in response to CIM and DIM using formulas (1-6) , ;

[0091] S223: Constructing random numbers and proportions , Compare and confirm , , ;

[0092] S224: The upper and lower bounds of the frequency regulation demand capacity of the i-th EV, including the upper and lower bounds of the power constraint of the single EV and the upper and lower bounds of the energy operation constraint of the single EV, as shown in the following formula:

[0093] (1) Single EV power constraints:

[0094] (9);

[0095] Where: , They are Time period The charging and discharging power of an EV; , They are Time period The upper and lower limits of the charging and discharging power of an EV; It is a 0-1 variable, and its value is 1. Time period The EV is in charging state, and its value is 0. Time period The EV is in a discharging state, which is used to restrict the EV from being charged and discharged at the same time.

[0096] (2) Single EV energy operation constraints:

[0097] (10);

[0098] Where: , They are Moment The upper and lower limits of the SOC of the EV; for Moment The SOC of an EV;

[0099] S225: Determine if i ≤ total number of vehicles: if not, continue with S221; if yes, start S226;

[0100] S226: Determine the frequency regulation demand capacity of the jth EV cluster based on time series superposition ;

[0101] S227: putting the data of the jth EV cluster into the experience playback area;

[0102] S228: Determine whether j ≤ the total number of clusters: if not, continue with S226; if yes, start S229;

[0103] S229: Use LSTM regression to train the experience playback area data;

[0104] S229: Probability interval prediction frequency regulation demand capacity upper limit matrix and the lower bound matrix ; Calculate the mean and variance matrix of the frequency modulation capacity distribution under different confidence intervals and The interval prediction method is evaluated by combining the Prediction Interval Coverage Probability (PICP) and Coverage Width Criterion (CWC) evaluation indicators.

[0105] The present invention uses KDE interval prediction, which is a non-parametric statistical method used to estimate the probability density function of a random variable. In KDE interval prediction, it is assumed that the data is extracted from an unknown probability distribution; then a kernel function is placed at each data point, and the kernel function can be expressed as shown in formula (11); then these kernel functions are superimposed to estimate the probability density function ; Finally, the confidence intervals of the prediction results are calculated as shown in formula (12).

[0106] (11);

[0107] (12);

[0108] Where: ℎ is the bandwidth parameter, which is used to control the width of the kernel function; is the number of data points in the test set; is the estimated probability density function under a given bandwidth ℎ; is the kernel function with ℎ as parameter; is the cth observation in the sample; is the standardized kernel function; and They are the upper and lower limits of the prediction interval of the test set; Predict results for the test set; is the confidence level; is the inverse function of the estimated probability density function.

[0109] PICP indicates the probability that the actual observation value falls within the prediction interval, and it is hoped that the prediction interval of the model can cover the actual observation value as much as possible, so a higher PICP indicates that the prediction interval of the model is more credible; and the smaller the CWC index, the smaller the width of the prediction interval, while the coverage probability is still high enough.

[0110] Step 3: Construct a distributed robust optimization model for the joint frequency regulation of EV clusters and units with the goal of minimizing the frequency regulation cost, and reconstruct the model into a second-order cone programming problem.

[0111] S31: Establish a distributed robust deterministic model for the joint frequency regulation of EV and units. The objective function of the model is to minimize the cost of the joint frequency regulation of EV clusters and units under the control of EV aggregators. Among them, the cost of the joint frequency regulation of EV clusters and units includes the startup cost, shutdown cost, no-load cost, coal consumption cost, frequency regulation cost and frequency regulation benefit of the unit; the frequency regulation cost, frequency regulation benefit, charging incentive cost and discharge incentive cost of EV are as follows:

[0112] (1) Unit startup cost:

[0113] (13);

[0114] In the formula: t is the time period; g is the unit number; is the startup cost of unit g in period t; is the startup unit price of unit g; is a binary variable, if If unit g is in operation during period t, the unit can be shut down but not started, otherwise it is 0.

[0115] (2) Shutdown cost of the unit:

[0116] (14);

[0117] Where: is the shutdown cost of unit g in period t; is the shutdown unit price of unit g; is a binary variable, if If the generator set g is out of operation during period t, the generator set can be started but not shut down; otherwise it is 0.

[0118] (3) No-load cost of the unit:

[0119] (15);

[0120] Where: is the no-load cost of unit g during period t; is the no-load unit price of unit g; is a binary variable, if The unit g is in the startup state during period t, otherwise it is 0.

[0121] (4) Coal consumption cost of the unit:

[0122] (16);

[0123] Where: is the coal consumption cost of unit g in period t; is the coal consumption unit price of unit g; is the output power of unit g during period t.

[0124] (5) Frequency regulation cost and frequency regulation benefit of the unit:

[0125] The unit's participation in the frequency regulation market will incur costs such as the cost of responding to frequency regulation instructions and the reduced efficiency caused by frequency changes. At the same time, the unit will also obtain benefits from frequency regulation services, including capacity benefits and mileage benefits. Therefore, the unit's frequency regulation cost is and FM gain It can be expressed as:

[0126] (17);

[0127] Where: is the frequency regulation cost of unit g in period t; and are the upward and downward adjustment capacities of unit g in period t respectively; and They are the upward and downward price adjustments for the units respectively; is the frequency regulation benefit of unit g in period t; It is the frequency regulation performance index of the unit; is the frequency regulation capacity electricity price of the unit in period t; is the frequency regulation mileage electricity price of the unit in period t; is the frequency regulation output mileage of the unit in period t.

[0128] (6) Frequency regulation costs and benefits of EVs:

[0129] EV clusters can use their large-scale EV battery energy storage systems as adjustable loads to participate in the frequency regulation market, thereby obtaining frequency regulation benefits, but they need to generate frequency regulation costs when the EV cluster participates in the day-ahead energy output plan. Therefore, the frequency regulation cost of the EV cluster is and FM gain It can be expressed as:

[0130] (18);

[0131] Where: is the frequency regulation cost of the EV cluster in period t; is the charging electricity price of the EV cluster during period t; and are the upward and downward adjustment capacities of the EV cluster in period t, respectively; It is the frequency regulation performance index of the EV cluster; is the frequency regulation benefit of the EV cluster in period t; It is the frequency regulation performance index of the unit; is the frequency regulation capacity electricity price of the EV cluster in period t; is the frequency regulation mileage electricity price of the EV cluster in period t; is the frequency modulation output mileage of the EV cluster in period t.

[0132] (7) EV charging and discharging incentive costs

[0133] (19);

[0134] Where: is the charging and discharging incentive cost of the EV cluster.

[0135] Therefore, a deterministic model of EV and generator units jointly participating in frequency regulation is constructed, and the frequency regulation cost is used as the Minimum is the optimization goal, and the objective function can be expressed as:

[0136] (20);

[0137] Where: F is the cost of joint frequency regulation; T is the entire research period; is the number of frequency regulation units; The time interval.

[0138] Taking into account the characteristics of unit operation, EV charging characteristics and power balance, the constraints of the EV cluster and unit joint frequency regulation model are established, as follows:

[0139] (1) Frequency regulation mileage constraints:

[0140] (twenty one);

[0141] Where: and They are the upward and downward frequency modulation output mileage of the EV cluster in period t, respectively; and They are respectively the upward and downward frequency modulation output mileage of the unit in period t; For the period The number of time intervals of the internal frequency modulation signal; For the FM market The mth frequency modulation indication signal issued at time, .

[0142] (2) Single EV power constraints:

[0143] (twenty two);

[0144] Where: , They are Time period The charging and discharging power of an EV; , They are Time period The upper and lower limits of charging and discharging power of EVs; It is a 0-1 variable, and its value is 1. Time period The EV is in charging state, and its value is 0. Time period The EV is in a discharging state, which is used to restrict the EV from being charged and discharged at the same time.

[0145] (3) Single EV energy operation constraints:

[0146] (twenty three);

[0147] Where: , They are Moment The upper and lower limits of the SOC of the EV; for Moment The SOC of an EV;

[0148] (4) EV cluster power constraints:

[0149] (twenty four);

[0150] Where: For EV clusters Frequency modulation capacity of the time period; , They are the upper and lower limits of the charging and discharging power of the EV cluster respectively.

[0151] (5) Unit output constraints:

[0152] (25);

[0153] Where: and are the maximum output and minimum output of unit g respectively;

[0154] (6) Unit climbing constraints:

[0155] (26);

[0156] (7) Unit downhill climbing constraints:

[0157] (27);

[0158] Where: and They are the upward and downward climbing of unit g in period t respectively; and They are respectively the maximum upward and downward climbing limits of the unit's output when it is operating in g mode. and They are respectively the maximum upward and downward climbing limits of unit g when the start-stop state changes.

[0159] (8) Start and stop time constraints:

[0160] (28);

[0161] (29);

[0162] Where: is the actual operating time of unit g; and are the minimum start time and minimum shutdown time of unit g respectively; The actual operating time of the unit.

[0163] (9) Logical constraints on the start and stop status of the unit:

[0164] (30);

[0165] (10) Unit frequency regulation power constraints:

[0166] (31);

[0167] (32);

[0168] (11) Constraints on the combined active power balance of EV clusters and units:

[0169] (33);

[0170] Where: is the frequency modulation power required by the system in period t.

[0171] S32: Establish a distributed robust uncertainty model for the joint frequency regulation of EVs and units. When the frequency regulation demand capacity distribution of the EV cluster is unknown, help the EV cluster to formulate the frequency regulation capacity at each time, so that the EV cluster frequency regulation benefit is maximized, which can minimize the joint frequency regulation cost of the EV cluster and the unit. Therefore, the above-mentioned joint frequency regulation minimum cost problem can be transformed into helping the EV cluster report the frequency regulation capacity at each time to maximize its frequency regulation benefit. sub-problem of .

[0172] First, in order to fully consider the uncertainty of the frequency modulation signal, the objective function is updated to a stochastic optimization model. According to the obtained frequency modulation signal scenario set, the objective function of the stochastic optimization model taking into account the uncertainty of the frequency modulation signal is established as shown in formula (34), where the frequency modulation benefit expression of the EV cluster is shown in formula (35).

[0173] (34);

[0174] (35);

[0175] Where: S is the total number of FM signal scenes after clustering; s is the scene; is the probability of the scenario occurring; and To coordinate the frequency regulation capacity matrix of the units and the frequency regulation capacity matrix of the EV cluster reported by the control center; The cost of adjusting the frequency of the unit; The benefits of frequency regulation for EV clusters; and The charging incentive price matrix and the discharging incentive price matrix for each time period; Provide a charging price matrix for EV clusters at different time periods; It is the EV cluster frequency modulation benefit matrix for each time period.

[0176] Then, in view of the uncertainty of EV frequency regulation capacity, the present invention uses CVaR to determine the risk of frequency regulation revenue reduction caused by the EV frequency regulation capacity reported by the aggregator, so as to update the frequency regulation revenue expression of the EV cluster. When the frequency regulation capacity distribution of the EV cluster at different times is unknown, the distributed robust optimization based on CVaR is adopted, taking into account the risk of frequency regulation revenue reduction in the worst case. At this time, the frequency regulation revenue expression of the EV cluster is shown in formula (36).

[0177] (36);

[0178] Where: is a column vector confidence level matrix, where the elements of each row are , represents the risk aversion degree, The bigger the size, the smaller the risk. To consider the worst-case EV cluster frequency regulation benefit

[0179] Finally, in order to fully consider the uncertainty of EV charging behavior of CVaR, the objective function of the updated stochastic optimization model is a distributed robust model based on CVaR constraints. Due to the uncertainty of the EV cluster frequency regulation demand capacity, the EV cluster frequency regulation benefits are uncertain, which in turn leads to the uncertainty of the joint frequency regulation costs of the EV cluster and the unit. Therefore, the objective function of the distributed robust model of the EV cluster and the unit joint frequency regulation based on CVaR constraints is updated as shown in formula (37):

[0180] (37);

[0181] S33: Model reconstruction and solution:

[0182] (1) According to the sub-problem of the objective function of formula (37), the optimization expression for maximizing the benefit of EV cluster participating in frequency regulation is established as shown in formula (38).

[0183] (38);

[0184] Where: is the threshold matrix introduced when calculating the conditional risk value.

[0185] (2) According to the maximum-minimum problem and the saddle point theorem, equation (38) is equivalent to a two-level optimization equation, as shown in equation (39).

[0186] (39);

[0187] The inner layer optimization problem in formula (38) is extracted as shown in formula (39).

[0188] (39);

[0189] Combined with formula (8), formula (39) can be written as a cone optimization problem, as shown in formula (40).

[0190] (40);

[0191] Where: Frequency modulation capacity distribution for EV clusters The distribution function of is the integrand variable.

[0192] (3) According to Lagrange’s theorem, Equation (40) can be expressed as an unconstrained problem, as shown in Equation (41).

[0193] (41);

[0194] Where: is the introduced Lagrange multiplier matrix; For vector and vector The number product between .

[0195] (4) Construct the dual problem of equation (41). The specific expression is shown in (42).

[0196] (42);

[0197] because The frequency regulation capacity distribution of EV clusters distribution function, if equation (42) has a solution, then for any matrix variable , the conditions satisfied are shown in formula (43).

[0198] (43);

[0199] (5) Combined with equation (35), we establish the dual problem of equation (40), whose specific expression is shown in (44).

[0200] (44);

[0201] (6) According to known mathematical conclusions: there exists any value on a straight line that is greater than any value on the parabola tangent to the straight line. By converting the constraints in equation (44) and combining them with the outer optimization problem in equation (39), the equivalent problem of equation (39) can be obtained as shown in equation (45).

[0202] (45);

[0203] Where: is the dual coefficient matrix.

[0204] (7) Due to the distribution of frequency modulation capacity that aggregators can control in different time periods If is unknown, then the constraints in equation (45) are infinite, which is not conducive to solving the original problem. Formula (45) is written as a second-order cone programming problem, and the constraints are written as finite, as shown in equation (46).

[0205] (46);

[0206] Solving equation (46) yields the frequency regulation capacity of the EV cluster set by the aggregator: Substitute the result into equation (46) and combine it with the unit operation constraints to solve the frequency regulation capacity of the unit at each time. .

[0207] Due to the combination of CVaR, the confidence level in the original problem is a nonlinear term. In this chapter, we first solve the confidence level The minimum EV cluster and unit joint frequency regulation cost under the condition that the constraints remain unchanged, and then the gradient descent method is used to dynamically find the optimal confidence level , until the error of the frequency regulation capacity of the constructed EV cluster is less than The constructed error function J is shown in formula (47).

[0208] (47);

[0209] Where: Bring the iterative confidence level to the previous step.

[0210] According to the error function, the gradient update strategy is shown in formula (48).

[0211] (48).

[0212] The beneficial effects of this embodiment are as follows:

[0213] (1) The stochastic optimization method and the CVaR-based distributed robust optimization method in the present invention can handle the uncertainty of the frequency modulation signal and EV charging behavior in the proposed model. The evaluation indicators PICP and CWC of KDE interval prediction reach 0.334 and 213.04, respectively.

[0214] (2) The distributed robust optimization scheduling model for the joint frequency regulation of EV clusters and units that fully considers the dual uncertainties in the present invention can minimize the economic benefits obtained from participating in the frequency regulation market and help aggregators determine the reportable EV cluster frequency regulation capacity under the premise of having the highest expected profit, standard deviation of profit and expected conditional loss.

[0215] (3) The calculation results of the present invention show that when only traditional units are considered to participate in market frequency regulation, the frequent start and stop of the generator sets brings the worst economic effect; when only EV clusters are considered to participate in market frequency regulation, a large number of EVs are required to meet the frequency regulation demand; when EV clusters and units are considered to participate in market frequency regulation together, the rapid regulation of EV clusters compensates for the delay of traditional unit regulation, and the cost of frequency regulation is reduced by 22.45% compared with unit frequency regulation.

[0216] The contents involved in the above embodiment are described below in conjunction with a preferred embodiment.

[0217] like Figure 1-Figure 12 As shown, the present invention analyzes the effectiveness of the distributed robust optimization model of EV cluster and unit joint frequency regulation considering double uncertainty under the incentive mechanism based on the operating data and frequency regulation indication signal of the PJM market in the United States in 2023. Assume that there are 1,000 EVs with charging demand in one day, among which the proportion of private cars, taxis and other vehicles is shown in Table 1.

[0218] Table 1 Proportion of various vehicle models

[0219]

[0220] EV aggregators regulate EVs that respond to CIM and DIM. The rated charging power of EVs is 30kW, and the charging and discharging efficiencies are both 90%. Assume that the charging price given by the aggregator is 0.2$ / (MW·h) less than the day-ahead price. According to the pre-obtained statistical data on lithium battery costs, the EV battery cost is set to 200$ / (MW·h). Taking into account the PJM official website and frequency regulation performance score statistics, the EV battery performance score and the unit frequency regulation performance score are set to 0.9. Based on the pre-obtained survey data, the proportion of EVs that respond to incentives in the cluster under different cost savings is obtained, as shown in the attached figure. Figure 2 shown. Figure 2 In the figure, the bar with a horizontal axis of 0 represents the response ratio of the EV cluster when the cost saving is 0; the bar with a horizontal axis of 5 represents the response ratio of the EV cluster when the cost saving is not more than $5; the bar with a horizontal axis of 6 represents the response ratio of the EV cluster when the cost saving is not more than $6; and so on, the bar with a horizontal axis of >20 represents the response ratio of the EV cluster when the cost saving is more than $20. The aggregator estimates the cost saved by EV response incentives based on the given charging price and the peak-to-valley difference of electricity price, and calculates the proportion of EVs responding to CIM and DIM by conversion.

[0221] In order to fully demonstrate the coordinated allocation of unit combination capacity, the selected unit combination optimization model data are shown in Table 2.

[0222] Table 2 Parameters of unit commitment optimization model

[0223]

[0224] In order to construct the probability scenario set of FM signal data curves, the present invention uses the historical FM signal data of 2023 and divides it by the maximum frequency deviation ±0.2Hz as normalization. The K-means clustering method is used to cluster the 365 FM signal data curves after normalization. The number of cluster centers is Take 10, add Figure 3 The clustered FM signal scenarios are given. The scenario probabilities corresponding to each FM signal scenario are: 0.27%, 0.82%, 20.27%, 58.08%, 2.74%, 0.27%, 8.22%, 8.77%, 0.27% and 0.27% respectively. It can be found that in the historical FM signal data of the PJM market in the United States in 2023, the dominant FM signal scenarios are mainly scenario 3, scenario 4, scenario 5, scenario 7 and scenario 8, and other scenarios also play an indispensable role.

[0225] 2 Cluster Frequency Modulation Capacity Analysis

[0226] The Monte Carlo method is used to simulate 1,000 EVs, with a research time interval of 1 hour, and a day is divided into 24 time periods. Since the distribution of EV charging demand in each time period is unknown, in order to fully consider the impact of EV charging behavior uncertainty on the joint frequency regulation cost of EV and unit, the mean and variance of the charging demand of the EV cluster in each time period are simulated by multiple Monte Carlo simulations to achieve approximate results. The present invention simulates 1,000 sets of charging behavior data, of which 800 sets are used as training sets, 100 sets are used as test sets, and 100 sets are used as validation sets. The LSTM regression method is used to train the charging demand of the EV cluster in each time period under the training set. As the number of iterations changes, the root mean square error and loss value of the training set during the training process are shown in the attached figure. Figure 4 As shown in Figure 2, it can be found that when the number of iterations is about 300, the root mean square error converges to 0.2; and when the number of iterations is about 200, the loss value tends to 0, indicating that the training set data has been trained.

[0227] Table 3 Comparative error of interval prediction

[0228]

[0229] According to the trained model, the present invention uses Gaussian interval prediction

[23] , KDE interval prediction and Boostrap interval prediction

[24] The EV cluster frequency modulation capacity is predicted, and the PICP and CWC evaluation indicators are used to compare the prediction results of the above interval prediction methods at a 60% confidence interval. The prediction comparison error is shown in Table 3. Compared with the results of KDE interval prediction, the PICP of the Gauss interval prediction result is reduced by 0.029, and the evaluation index CWC is higher than the KDE interval prediction; the PICP of the Boostrap interval prediction result is reduced by 0.015, and the evaluation index CWC is higher. The larger the PICP index and the smaller the CWC, the better the confidence interval coverage probability. Therefore, KDE interval prediction is used to predict the EV cluster frequency modulation capacity. The probability distribution of the training set is shown in the attached figure. Figure 5 As shown in the attached figure, the cumulative probability density function is Figure 6 The results of different confidence intervals are shown in the attached Figure 7 As shown, the upper limit matrix and lower limit matrix of EV cluster uncertainty frequency regulation demand capacity under different confidence intervals are obtained.

[0230] 3 Analysis of Joint Frequency Modulation Results

[0231] In order to fully demonstrate the effectiveness of the distributed robust optimization model of the joint frequency regulation of EV clusters and units considering double uncertainties under the incentive mechanism, the present invention uses Python language to call Cplex to solve the distributed robust optimization model of the joint frequency regulation of EV clusters and units.

[0232] At 0:00, the reported EV cluster frequency regulation capacity under different confidence levels and discharge incentive prices is shown in the attached figure. Figure 8 As shown in the figure, it can be seen that when the confidence level is small, the aggregator is more inclined to report the capacity with high risk and high income, and when the confidence level continues to increase, the EV cluster frequency regulation capacity reported by the aggregator at this moment continues to decrease. When it is greater than 0.5, the change trend of the EV cluster frequency regulation capacity reported by the aggregator at that moment becomes smaller. At the same time, when the discharge incentive price is set to 13$ / (MW·h), the EV cluster frequency regulation capacity reported is the largest. Fig. 9 The confidence level is given When is equal to 0.5 and the discharge incentive price is equal to 13$ / (MW·h), the mean and variance of the charging demand of the EV cluster in each time period.

[0233] In order to fully demonstrate the effectiveness of the distributed robust optimization method based on CVaR for reporting the frequency regulation capacity of EV clusters, the minimum frequency regulation cost deterministic model is compared, and the expected benefits, standard deviation of benefits, and expected conditional losses of the two models are given as shown in Table 4. Compared with the minimum frequency regulation cost deterministic model, the distributed robust optimization frequency regulation benefit model based on CVaR is more suitable for coping with different frequency regulation signal scenarios, and has higher expected benefits, standard deviation of benefits, and expected conditional losses.

[0234] Table 4 Comparison of EV cluster frequency regulation benefits under two models

[0235]

[0236] When the confidence level is 0.5 and the discharge incentive price is equal to 13$ / (MW·h), Fig.10 The cluster frequency regulation capacity reported by EV aggregators in various time periods is shown. It can be seen that the EV cluster frequency regulation capacity is low during 4:00 to 10:00, because most EVs connected to the grid have completed charging or are in use during this period, so there are fewer EVs coming to charge during this period. In contrast, the EV frequency regulation capacity reported in other time periods is higher, and the frequency regulation capacity is the largest during 21:00 to 4:00, because the charging demand during this period is the most concentrated, so choose to connect to the grid during this period.

[0237] The following three examples are set to analyze the impact of EV cluster participation in joint frequency regulation on frequency regulation costs:

[0238] Case 1, only considers traditional units participating in power market frequency regulation;

[0239] Case 2, only considers EV clusters participating in power market frequency regulation;

[0240] Example 3 considers the joint participation of EV clusters and units in power market frequency regulation.

[0241] The calculation examples of the above three cases all aim to minimize the frequency regulation cost of the coordination control center. In order to ensure that the EV cluster in Example 2 can independently participate in the frequency regulation of the power market, Example 2 is no longer limited to the set 1,000 EV cluster size, that is, it is assumed that there are enough EVs to respond to the market frequency demand during this period.

[0242] Table 5 Frequency modulation results under three examples

[0243]

[0244] Due to the attached Fig.10 The reported frequency regulation capacity of the EV cluster during the period from 10:00 to 12:00 is quite different, so Table 3 gives the total frequency regulation cost of Example 1, Example 2 and Example 3 during the period from 10:00 to 12:00, as well as the various cost or benefits of the traditional units and the EV cluster. As can be seen from Table 5: The total frequency regulation cost of Example 3 is $21,648.19 lower than that of Example 1 and $3,482.05 lower than that of Example 2. The comparison of various cost or benefits between Example 1 and Example 3 shows that the startup cost and shutdown cost of the unit in Example 3 are lower than those in Example 1, and the main differences are in coal consumption cost, no-load cost and frequency regulation cost. Since Example 3 involves the participation of the EV cluster, the EV cluster frequency regulation cost, charging cost and incentive cost are significantly higher than those in Example 1. However, since EVs assume part of the frequency regulation capacity of the original unit, the coal consumption cost is reduced by $3303.91 and the no-load cost is reduced by $46.00, making the total frequency regulation cost of Example 3 lower than that of Example 1. The comparison of various costs or benefits between Example 2 and Example 3 shows that the frequency regulation capacity provided by a single EV is much smaller than the frequency regulation capacity of the unit. In Example 2, in order to meet the frequency regulation signal from 10:00 to 12:00, the number of EVs invested is approximately 5 times that of Example 3, and the total frequency regulation cost is higher than that of Example 3. Although the total frequency regulation cost of Example 2 is $18166.14 lower than that of Example 1, it is unrealistic to rely solely on a large number of EVs to replace the frequency regulation of the unit.

[0245] Attached Fig.11 and Fig.12 The comparison of the joint frequency regulation capacity of the EV cluster and the unit from 10:00 to 12:00 is given in Example 1 and Example 3. Since the probability of scenario 4 is the highest, Fig.11 and Fig.12 The FM capacity in the above example is the FM signal in scenario 4. Fig.11 It can be seen that at each time, unit 1 is the main provider of frequency regulation capacity, and units 2 and 3 mainly play a regulating role. There are even cases where the frequency regulation demand is not met at certain times (such as: time 10:08, time 11:16, etc.). Fig.12 It can be seen that at each moment, the frequency regulation capacity of the system is jointly provided by the EV cluster and different generator sets. Before 10:30, the frequency regulation capacity provided by the EV cluster was relatively small; then it gradually increased, and even at 11:06, facing the sudden increase in the frequency regulation demand of the system, the EV cluster became the main frequency regulation capacity provider, avoiding the same Fig.11 Units 2 and 3 in the system frequently operate and start and stop.

[0246] Embodiment 2: This embodiment provides a robust frequency modulation device for electric vehicle cluster and generator set, including:

[0247] The frequency modulation unit is used to obtain real-time frequency modulation indication signal data, input it into a pre-built and optimized distributed robust optimization model, reconstruct the distributed robust optimization model into a second-order cone programming problem, and predict the result of the joint robust frequency modulation of the electric vehicle cluster and the unit;

[0248] The construction and optimization method of the distributed robust optimization model includes:

[0249] Generate a scene set of frequency modulation signals based on the acquired historical frequency modulation indication signal data;

[0250] Considering the charging and discharging incentive mechanism, the kernel density estimation interval is used to predict the results of the active regulation of electric vehicle clusters, and a distributed robust optimization model of conditional risk value is established;

[0251] The generated scene set of the frequency modulation signal is input into the distributionally robust optimization model for optimization to obtain an optimized distributionally robust optimization model.

[0252] The specific functional implementation of each of the above modules can be found in the relevant contents of the method in Example 1 and will not be elaborated here.

[0253] Embodiment 3, this embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in Embodiment 1 are implemented.

[0254] Embodiment 4: This embodiment provides a computer device, including:

[0255] Memory, for storing computer programs / instructions;

[0256] A processor, configured to execute the computer program / instructions to implement the steps of any one of the methods described in Example 1.

[0257] Embodiment 5: This embodiment provides a computer program product, including a computer program / instruction, which implements the steps of any method described in Embodiment 1 when executed by a processor.

[0258] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0259] It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, systems or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0260] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0261] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0262] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0263] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure rather than to limit its protection scope. Although the present disclosure has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that after reading the present disclosure, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the disclosed claims to be approved.

Claims

1. A method for joint robust frequency modulation of electric vehicle clusters and generator sets, characterized in that: include: Obtain real-time frequency modulation indication signal data, input it into the pre-built and optimized distributed robust optimization model, restructure the distributed robust optimization model into a second-order cone programming problem, and predict the result of the joint robust frequency modulation of the electric vehicle cluster and the unit; The construction and optimization method of the distributed robust optimization model includes: Generate a scene set of frequency modulation signals based on the acquired historical frequency modulation indication signal data; Considering the charging and discharging incentive mechanism, the kernel density estimation interval is used to predict the results of the active regulation of electric vehicle clusters, and a distributed robust optimization model of conditional risk value is established; The generated scene set of the frequency modulation signal is input into the distributionally robust optimization model for optimization to obtain an optimized distributionally robust optimization model.

2. The method for robust frequency modulation of electric vehicle cluster and generator set according to claim 1, characterized in that: The step of generating a scene set of a frequency modulation signal based on the acquired historical frequency modulation indication signal data includes: S11: Collect and normalize the historical frequency modulation indicator signal data of the PJM market. Randomly select from the historical FM indicator signal data as the initial cluster centers; S12: Calculate the Euclidean distance between all historical frequency modulation indicator signal data and each cluster center; and classify the data into the class represented by the cluster center closest to the data; S13: Calculate the mean of each category of historical frequency modulation indicator signal data and use it as the new cluster center of the category data, and calculate the average Euclidean distance of all historical frequency modulation indicator signal data to the cluster center; S14: Determine whether the cluster center and the average value of the Euclidean distance have changed. If so, continue with step S12; if not, end clustering and generate a scene set of the FM signal.

3. The method for joint robust frequency modulation of electric vehicle clusters and generator sets according to claim 1 is characterized in that: The charging and discharging incentive mechanism is considered, and the kernel density estimation interval is used to predict the results of the electric vehicle cluster participating in the active regulation capacity, and a distributed robust optimization model of conditional risk value is established, including: Construct charging incentive mechanism and discharging incentive mechanism, and obtain charging incentive price and discharging incentive price; According to the charging incentive price and the discharging incentive price, the proportion of electric vehicles that respond to the charging incentive mechanism and the discharging incentive mechanism are obtained respectively; According to the obtained statistical data of basic state parameters of electric vehicles, the basic state parameter results of a single electric vehicle are sampled and obtained; According to the basic state parameter results of a single electric vehicle, the power constraint and energy operation constraint of the single electric vehicle are obtained by Monte Carlo simulation; Based on the power constraints and energy operation constraints of individual electric vehicles, multiple Monte Carlo simulations were performed according to the proportion of electric vehicles to obtain multiple preliminary results of the capacity of electric vehicle clusters participating in active regulation. The final electric vehicle cluster active regulation capacity results are obtained by using kernel density estimation interval prediction based on multiple preliminary electric vehicle cluster active regulation capacity results. The final electric vehicle cluster participates in active capacity regulation, and a distributed robust optimization model is established through the conditional risk value distributed robust optimization method.

4. The method for joint robust frequency modulation of electric vehicle clusters and generator sets according to claim 3 is characterized in that: The multiple groups of preliminary electric vehicle cluster participation active regulation capacity results are used to obtain the final electric vehicle cluster participation active regulation capacity results by kernel density estimation interval prediction. The calculation formula is as follows: ； ； Where: is the bandwidth parameter, which is used to control the width of the kernel function; is the number of data points in the test set; For a given bandwidth The estimated probability density function under ; For is the kernel function of the parameter; It is a sample composed of multiple preliminary groups of electric vehicle clusters participating in active capacity regulation results. is the cth observation in the sample; is the standardized kernel function; and They are the upper and lower limits of the prediction interval of the test set; The final result is the active regulation capacity of the electric vehicle cluster; is the confidence level; is the inverse function of the estimated probability density function.

5. The method for joint robust frequency modulation of electric vehicle clusters and generator sets according to claim 1 is characterized in that: The objective function of the distributed robust optimization model is to minimize the joint frequency regulation cost of the electric vehicle cluster and the unit under the regulation of the electric vehicle aggregator, wherein the joint frequency regulation cost of the electric vehicle cluster and the unit includes the startup cost, shutdown cost, no-load cost, coal consumption cost, frequency regulation cost and frequency regulation benefit of the unit; the frequency regulation cost, frequency regulation benefit, charging incentive cost and discharge incentive cost of the electric vehicle.

6. The method for joint robust frequency modulation of electric vehicle clusters and generator sets according to claim 5 is characterized in that: The objective function of the distributed robust optimization model is calculated as follows: ； In the formula, is the objective function, S is the total number of FM signal scenes after clustering; s is the scene; is the probability of the scenario occurring; and The frequency regulation capacity matrix of the unit and the frequency regulation capacity matrix of the EV cluster obtained in advance, To consider the worst-case EV cluster frequency regulation benefit.

7. An electric vehicle cluster and unit joint robust frequency modulation device, characterized in that: include: The frequency modulation unit is used to obtain real-time frequency modulation indication signal data, input it into a pre-built and optimized distributed robust optimization model, reconstruct the distributed robust optimization model into a second-order cone programming problem, and predict the result of the joint robust frequency modulation of the electric vehicle cluster and the unit; The construction and optimization method of the distributed robust optimization model includes: Generate a scene set of frequency modulation signals based on the acquired historical frequency modulation indication signal data; Considering the charging and discharging incentive mechanism, the kernel density estimation interval is used to predict the results of the active regulation of electric vehicle clusters, and a distributed robust optimization model of conditional risk value is established; The generated scene set of the frequency modulation signal is input into the distributionally robust optimization model for optimization to obtain an optimized distributionally robust optimization model.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

9. A computer device, characterized in that: include: Memory, for storing computer programs / instructions; A processor, configured to execute the computer program / instructions to implement the steps of the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images