Multi-stage electric vehicle cluster regulation potential quantification method

Abstract

The invention relates to the technical field of power system stability, and discloses a multi-stage electric vehicle cluster regulation potential quantification method, which comprises the following steps of converting an electric vehicle cluster into generalized energy storage equipment, and performing feature screening on electric vehicle stop data; carrying out support vector machine regression modeling, and determining a capacity boundary and a power boundary of the generalized energy storage equipment; constructing a self-scheduling model of the electric vehicle cluster based on economic incentive and user irrational consumption behavior influence, and determining a self-scheduling load of the electric vehicle cluster; calculating the rescheduling power and rescheduling capacity of the generalized energy storage equipment; and integrating to obtain a schedulable potential quantification parameter set, and quantifying the scheduling potential of the electric vehicle cluster according to the parameter set. According to the method, the schedulable potential of the electric vehicle cluster can be accurately evaluated, and various parameters such as the construction site, the number and the charging rate of the charging stations can be planned according to needs, so that resources are saved, and the service quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system stability, and in particular to a multi-stage electric vehicle cluster regulation potential quantification method. Background Art

[0002] Data shows that by the end of June 2024, the number of new energy vehicles in my country has reached 24.72 million, accounting for 7.18% of the total number of vehicles in the country. The electric vehicle market has shown a rapid growth trend in recent years and has become the world's largest electric vehicle market. The charging demand generated by large-scale application of electric vehicles will further increase the demand for electricity during the peak load period of the power system, causing problems such as insufficient energy supply for equipment. On the other hand, the promotion and application of electric vehicles has promoted the integration of the two major systems of transportation and energy, and introduced flexible and high-quality demand-side resources for the regulation of the power system.

[0003] The main reason why electric vehicles can participate in the optimization and regulation of the power system as high-quality demand-side resources is that the on-board batteries of electric vehicles have charging and discharging functions. After being connected to the power system through charging piles during parking periods, they can be regarded as general energy storage (GES) units. The charging and discharging power of EVs (Electric Vehicles) can be regulated through demand response measures to achieve a reasonable distribution of EV charging loads, thereby playing a role in load peak shaving and valley filling on the grid side, absorbing renewable energy, and providing auxiliary services. Therefore, accurately evaluating the dispatchable potential of EV clusters is of great significance to optimizing the power system.

[0004] At present, the calculation methods of the dispatchable potential of electric vehicles can be mainly divided into two categories: data-driven and model-driven. The data-driven method mainly uses machine learning algorithms represented by neural networks and long short-term memory networks (LSTM) to predict the charging and discharging power curves of EVs to characterize the behavior of EVs participating in system regulation. However, the dispatchable potential of electric vehicles is not a set of definite charging and discharging power vectors, but an envelope space composed of all possible charging and discharging decisions. The existing data-driven methods fail to consider the implicit constraints between variables, and the resulting dispatch plans may be difficult to implement. The model-driven method is mainly based on the idea of ​​mathematical modeling to deduce the energy consumption behavior of electric vehicles under the influence of multiple factors, but due to the existence of heterogeneity problems, the model-driven method is prone to a priori errors.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In view of the problems in the related art, the present invention proposes a multi-stage electric vehicle cluster control potential quantification method to overcome the above-mentioned technical problems existing in the existing related technology.

[0007] To this end, the specific technical solution adopted by the present invention is as follows:

[0008] A multi-stage method for quantifying the potential of electric vehicle cluster regulation includes the following steps:

[0009] S1. Based on the Minkowski summation principle, the electric vehicle cluster is transformed into a generalized energy storage device, and the electric vehicle stop entropy is used to perform feature screening on the electric vehicle stop data;

[0010] S2. Using the improved Harris Eagle optimization algorithm, support vector machine regression modeling is performed on the screened electric vehicle parking data to determine the capacity boundary and power boundary of the generalized energy storage device;

[0011] S3. Based on economic incentives and the influence of users' irrational consumption behaviors, a self-dispatching model of electric vehicle clusters is constructed, and the self-dispatching load of electric vehicle clusters is determined according to the self-dispatching model;

[0012] S4. Calculate the redispatch power and redispatch capacity of the generalized energy storage device based on the capacity boundary and power boundary of the generalized energy storage device and the self-dispatch load of the electric vehicle cluster;

[0013] S5. Based on the redispatch power and redispatch capacity of the generalized energy storage device, a set of quantitative parameters of dispatchable potential is integrated, and the dispatch potential of the electric vehicle cluster is quantified based on the parameter set.

[0014] Furthermore, the method of converting the electric vehicle cluster into a generalized energy storage device based on the Minkowski summation principle and using the electric vehicle stop entropy to perform feature screening on the electric vehicle stop data includes the following steps:

[0015] S11. Aggregate the electric vehicle cluster at the aggregator level using the Minkowski summation process to obtain a generalized energy storage device and extract key parameters of the generalized energy storage device;

[0016] S12. Based on the conditional entropy and information entropy of electric vehicles, the uncertainty and information content of each feature on the electric vehicle parking data are evaluated, and the entropy value and information gain value of each feature are calculated;

[0017] S13, determining an information gain threshold according to the information gain value, and judging whether the information gain value is greater than the information gain threshold, if so, retaining the electric vehicle stop data, if not, deleting the electric vehicle stop data.

[0018] Furthermore, the calculation formula for the key parameters of the generalized energy storage device is:

[0019]

[0020] In the formula, and They represent the charging dispatching power and discharging dispatching power of the generalized energy storage equipment in period t, and They represent the charging dispatching power and discharging dispatching power of electric vehicle n in period t, respectively. EV represents the set of electric vehicles in the generalized energy storage device, and They represent the maximum charging dispatch power and the maximum discharging dispatch power of the generalized energy storage device in period t, and They represent the upper limit of charging power and the upper limit of discharging power of electric vehicle n, respectively. are Boolean variables representing the charging state and discharging state of electric vehicle n at time t, respectively. GES,t represents the power of the generalized energy storage device in period t, S n,t represents the battery power of electric vehicle n at time t, and They represent the minimum and maximum power of the generalized energy storage device in period t, and They represent the minimum and maximum battery capacity of electric vehicle n, respectively. A Boolean variable representing the parking state of electric vehicle n at time t, ΔS GES,t represents the change in power of the generalized energy storage device due to the change in the grid-connected state of the electric vehicle during period t, S n,arrival , S n,leave They represent the initial power of electric vehicle n and the power of electric vehicle n when it leaves the power grid respectively.

[0021] Furthermore, the improved Harris Eagle optimization algorithm is used to perform support vector machine regression modeling on the screened electric vehicle parking data to determine the capacity boundary and power boundary of the generalized energy storage device, including the following steps:

[0022] S21. constructing a support vector machine regression model based on the screened electric vehicle parking data, and optimizing the support vector machine regression model using an improved Harris Eagle optimization algorithm to obtain a support vector machine regression model based on Harris Eagle optimization;

[0023] S22. Use the support vector machine regression model based on Harris Eagle optimization to predict the electric vehicle cluster shutdown data on the trading day, and calculate the capacity boundary and power boundary of the generalized energy storage device at each moment based on the electric vehicle cluster shutdown data on the trading day.

[0024] Furthermore, the method of optimizing the support vector machine regression model using the improved Harris Hawk optimization algorithm to obtain the support vector machine regression model based on Harris Hawk optimization includes the following steps:

[0025] Determine the population size, initial velocity of particles, initial position, learning factor, weight factor, number of iterations, penalty parameter and the range of kernel function parameters;

[0026] Initialize the population size, maximum number of iterations, convergence threshold of the Harris Eagle algorithm, and use the mean square error or root mean square error as the objective function;

[0027] The population of Harris Eagle algorithm is initialized by using Tent chaotic mapping, the fitness value of each individual is calculated, and the position and population structure of the individual are updated based on the nonlinear energy factor value and encirclement strategy combined with the fitness value;

[0028] Determine whether the maximum number of iterations has been reached or the fitness has converged. If not, continue iterating. If so, determine the optimal parameter combination of the support vector machine regression model based on the convergence result, and construct a support vector machine regression model based on Harris Eagle optimization based on the optimal parameter combination.

[0029] Furthermore, the use of Tent chaotic mapping to initialize the population of the Harris Hawk algorithm, calculating the fitness value of each individual, and updating the position and population structure of the individual based on the nonlinear energy factor value and the encirclement strategy in combination with the fitness value includes the following steps:

[0030] The Tent chaotic map is used to generate a chaotic sequence, the maximum and minimum values ​​of the positions of the individuals in the search space, and the initial positions of the individuals in the search space to obtain the initial population.

[0031] Based on the objective function, calculate the fitness value of each individual in the initial population, and determine the current best individual based on the calculation results;

[0032] Determine whether the individual fitness value is greater than the preset fitness threshold. If so, select the soft encirclement strategy; if not, select the hard encirclement strategy;

[0033] Calculate the escape energy factor value according to the energy factor, and determine whether the escape energy factor value is less than 1. If so, select the local search update method, if not, select the global search update method;

[0034] Based on the selected update method and encirclement strategy, the position and population structure of each individual are updated in combination with the current best individual.

[0035] Furthermore, the expression of chaotic order is:

[0036]

[0037] The expression of the initial position of the population individuals in the search space is:

[0038]

[0039] The position update formula of the soft encirclement strategy is:

[0040] X new =X best +r1·(X best -X i )

[0041] The position update formula of the hard encirclement strategy is:

[0042] X new =X best +r2·(X best -X i )

[0043] The calculation formula of escape energy factor is:

[0044] E1=E×E0

[0045]

[0046] In the formula, represents the random number in the population after the I+1th iteration, Z k represents a random number in the population, u represents a random number between [0,1], represents the initial position of the population individuals in the search space after the Ith iteration, They represent the maximum and minimum values ​​of the individual positions of the population in the search space after the Ith iteration, respectively, and X new represents the new position calculated based on the current optimal solution or update strategy, X best represents the current global optimal solution, X i represents the current position of individual i, r1 represents the random number used to control the amplitude of movement, r2 represents the random number used to control the amplitude and direction of movement, El represents the escape energy factor, E represents the energy factor, E0 represents the initial energy, t represents the current number of iterations, T represents the maximum number of iterations, and α represents the rate of decay or growth, which determines how the energy changes with time t.

[0047] Furthermore, the self-dispatching model of the electric vehicle cluster is constructed based on economic incentives and the influence of the user's irrational consumption behavior, and the self-dispatching load of the electric vehicle cluster is determined according to the self-dispatching model, including the following steps:

[0048] S31. Based on the irrational behavior of electric vehicle users, the user's expectation of the responsiveness of the incentive price of the demand response program is modeled as a linear function of the incentive price.

[0049] S32, determining the deviation range of the electric vehicle user response, using a random variable that obeys a normal distribution to describe the probability distribution of the responsiveness deviation, and obtaining the maximum deviation value of the electric vehicle user response according to a truncation process;

[0050] S33. Based on the expected value of the responsiveness of electric vehicle users to the incentive price of the demand response plan and the maximum deviation of the electric vehicle user response, a demand response function between the electric vehicle responsiveness and the uncertain incentive price is modeled to obtain a self-dispatching model of the electric vehicle cluster, and the self-dispatching load of the electric vehicle cluster is determined based on the self-dispatching model.

[0051] Furthermore, the expected value of the responsiveness of electric vehicle users to the incentive price of the demand response program is expressed as:

[0052]

[0053] The expression of the deviation range of electric vehicle user response is:

[0054]

[0055] The objective function expression of the self-dispatching model of electric vehicle clusters is:

[0056]

[0057] In the formula, λ EV,t represents the expected value of EV users’ responsiveness to the incentive price of the demand response program, represents the maximum response degree of electric vehicle users to the incentive price of the demand response program, π inc,t represents the incentive price, represents the incentive price corresponding to the maximum responsiveness of electric vehicle users, d EV,t represents the maximum deviation of the electric vehicle user response, k1 and k2 represent the two actual proportional coefficients of the maximum deviation of the electric vehicle user response, Indicates the inflection point of the incentive price, represents a function that calculates the escape rate (or escape intensity) based on these input parameters to determine the key dynamics in the escape process, Represents a logical indicator, ξ n represents a random variable with truncated standard normal distribution, Indicates the logical negation of a logical indicator.

[0058] Furthermore, the calculation formula for the redispatch power of the generalized energy storage device is:

[0059]

[0060] The calculation formula for the redispatch capacity of generalized energy storage equipment is:

[0061]

[0062] The expression of the schedulable potential quantitative parameter set is:

[0063]

[0064] In the formula, are the upward regulation power and downward regulation power of the generalized energy storage device at time t, are the self-scheduled charging power and discharging power of the generalized energy storage device at time t, are the upper regulation capacity and lower regulation capacity of the generalized energy storage device at time t, Represents the self-dispatching capacity of the generalized energy storage device at time t.

[0065] The beneficial effects of the present invention are:

[0066] 1) The present invention proposes to use the support vector machine regression model optimized by the improved Harris Eagle algorithm to predict the model parameters of the data filtered by the driving entropy of electric vehicles. Compared with Bi-LSTM and LSTM, this method produces less mean square error and root mean square error, thus surpassing other models. The present invention also proposes a self-dispatching model that considers the impact of economic incentives and irrational consumption behaviors of EV users on the participation of electric vehicle users in demand response. Compared with the self-dispatching results of the traditional self-dispatching model that assumes that users are completely rational and can achieve reasonable load distribution, the prediction results obtained by this model are more in line with the actual energy consumption behavior of EV users. The GES redispatching power and redispatching capacity calculated based on the self-dispatching load obtained by this model can provide a more reliable reference basis for aggregators during electricity market transactions.

[0067] 2) By evaluating the dispatching potential of electric vehicle clusters, the present invention enables grid operators to plan grid loads in advance to avoid grid failures caused by excessive loads, or conversely use electric vehicles to provide energy storage services for the grid to ensure stable grid operation. Accurately evaluating the dispatching potential of electric vehicle clusters can plan the construction location, number, charging rate and other parameters of charging stations as needed, thereby helping to save resources and improve service quality.

[0068] 3) On the one hand, the present invention can clarify that the dispatching potential of electric vehicles can help better arrange charging time periods so as to utilize renewable energy for charging; on the other hand, if the dispatching potential of electric vehicles is large, it can be used as a flexible load, which is conducive to stabilizing power fluctuations caused by renewable energy sources such as wind and solar energy. Electric vehicles can not only obtain energy from the power grid, but also transmit idle energy back to the power grid through intelligent charging and feedback power functions, forming a two-way energy exchange system. In summary, the significance of the present invention in establishing a multi-stage electric vehicle cluster control potential quantification method is that it can help all relevant parties to efficiently utilize resources, improve service quality, and ultimately promote the end-to-end sustainable development of electric vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0070] Figure 1 is a flow chart of a multi-stage electric vehicle cluster control potential quantification method according to an embodiment of the present invention;

[0071] Figure 2 4 is a schematic diagram of a multi-stage electric vehicle cluster control potential quantification method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0072] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0073] According to an embodiment of the present invention, a multi-stage electric vehicle cluster regulation potential quantification method is provided.

[0074] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1-Figure 2 As shown, the multi-stage electric vehicle cluster control potential quantification method according to an embodiment of the present invention includes the following steps:

[0075] S1. Based on the Minkowski summation principle, the electric vehicle cluster is transformed into a generalized energy storage device, and the electric vehicle stop entropy is used to perform feature screening on the electric vehicle stop data;

[0076] Specifically, in the first stage, the Minkowski summation principle is used to transform the electric vehicle (EV) cluster into a generalized energy storage (GES), and the maximum charging and discharging power, power boundary and power change of the generalized energy storage device in time period t are obtained as key parameters. In order to predict the key parameters of GES, the electric vehicle stop entropy is first used to perform feature selection on the electric vehicle stop data, screen the important features in the electric vehicle stop data, and remove the interference features.

[0077] Since the charging modes of electric vehicles can be divided into fast charging mode and slow charging mode, and in the fast charging mode, the electric vehicle is charged at the maximum charging power and the charging load is uncontrollable, only the electric vehicle in the slow charging mode is considered in this embodiment.

[0078] By clarifying the initial state of charge (SOC), parking start time and parking end time of a single electric vehicle, an individual charging and discharging model of the electric vehicle can be established:

[0079]

[0080] In the formula, and They represent the charging and discharging dispatching power of electric vehicle n in period t respectively; and They represent the upper limit of charging and discharging power of electric vehicle n; S n,t and S n,t-1 They represent the battery power of electric vehicle n in period t and the previous period respectively; η ch and η dis They represent the charge and discharge efficiency respectively; t represents the unit time interval; η ref Indicates the discharge compensation coefficient, which is determined by the discharge loss; and represents the battery power safety boundary of electric vehicle n; are the parking start time and parking end time of electric vehicle n, respectively. The electric vehicle is charged and discharged between the parking start time and the parking end time to meet the energy demand of the user; is a Boolean variable representing the parking state of electric vehicle n at time t. It takes the value 1 when it is in the parking state, otherwise it takes the value 0; They are Boolean variables representing the charging state and discharging state of the electric vehicle n at time t. When it is in the charging state, In the discharge state It means that electric car n cannot be charged and discharged at the same time at time t.

[0081] There are a large number of EVs in the power system. If the charging data of all EV users is directly reported to participate in system regulation, it can be seen from the individual electric vehicle modeling in S1 that a large amount of high-dimensional variable data will be generated during the dispatch process, which is not conducive to the optimization of the dispatch plan; and the capacity of a single EV is difficult to reach the entry threshold of the power market; in addition, exposing a large number of EV user data to the power system will cause user privacy and security issues. Therefore, aggregators represented by EV charging service providers should aggregate EV clusters before participating in power system dispatch.

[0082] Minkowski summation is an addition method applicable to Euclidean space, and its expression is shown in formula (3). Its physical essence is the expansion set of multiple spaces. With the help of Minkowski summation, electric vehicle groups can be aggregated into generalized energy storage devices.

[0083]

[0084] In the formula, A and B are variable spaces with the same domain; a and b are elements in variable spaces A and B respectively; It is the Minkowski sum of two variable spaces, covering a larger space.

[0085] The method of converting the electric vehicle cluster into a generalized energy storage device based on the Minkowski summation principle and using the electric vehicle stop entropy to perform feature screening on the electric vehicle stop data includes the following steps:

[0086] S11. The Minkowski summation process is used to aggregate the electric vehicle cluster at the aggregator level to obtain the generalized energy storage device and extract the key parameters of the generalized energy storage device, as shown in formula (4):

[0087]

[0088] In the formula, and They represent the charging dispatching power and discharging dispatching power of the generalized energy storage equipment in period t, and They represent the charging dispatching power and discharging dispatching power of electric vehicle n in period t, respectively. EV represents the set of electric vehicles in the generalized energy storage device, and They represent the maximum charging dispatch power and the maximum discharging dispatch power of the generalized energy storage device in period t, and They represent the upper limit of charging power and the upper limit of discharging power of electric vehicle n, respectively. are Boolean variables representing the charging state and discharging state of electric vehicle n at time t, respectively. GES,t represents the power of the generalized energy storage device in period t, Sn,t represents the battery power of electric vehicle n at time t, and They represent the minimum and maximum power of the generalized energy storage device in period t, and They represent the minimum and maximum battery capacity of electric vehicle n, respectively. A Boolean variable representing the parking state of electric vehicle n at time t, ΔS GES,t represents the change in power of the generalized energy storage device due to the change in the grid-connected state of the electric vehicle during period t, S n,arrival , S n,leave They represent the initial power of electric vehicle n and the power of electric vehicle n when it leaves the power grid respectively.

[0089] In this embodiment, the process of selecting features of electric vehicle stop data using electric vehicle stop entropy first uses the stop entropy as a metric to evaluate the uncertainty and information content of each feature on the stop data. By calculating the entropy value of each feature, important features that have a greater impact on user charging behavior can be identified, while removing interference features with low information content and limited contribution to model prediction.

[0090] Specifically, the level of parking entropy reflects the importance of the feature in the user's parking behavior. Features with higher entropy values ​​mean that the user's parking behavior is more diverse and can provide more information; while features with lower entropy values ​​may indicate that their role in the prediction model is insignificant. On this basis, by setting appropriate thresholds, important features are screened out to ensure that the retained data can more accurately reflect the user's charging needs and behavior patterns. This feature selection process not only improves the training efficiency of subsequent models, but also enhances the accuracy of prediction results, laying a solid data foundation for the subsequent Harris Eagle optimization algorithm and support vector machine regression modeling. By removing interfering features, the complexity of the model is reduced, thereby improving its generalization ability and stability.

[0091] The entropy of electric vehicle parking includes the information entropy that measures the uncertainty of information and the conditional entropy that measures the uncertainty of the target variable under given characteristics:

[0092] The calculation formula of the conditional entropy is:

[0093]

[0094] The calculation formula of the information entropy is:

[0095]

[0096] Where H(Y|X) represents the conditional entropy of the random variable Y under the condition of known X; m represents the total number of conditions (the number of possible categories); i is the index of the value of Y, ranging from 1 to n; j is the index of the value of X, ranging from 1 to m; p(x j ) represents the event X=x j The probability of; n represents the total number of random variables (the number of possible categories); x j is the specific value of X, that is, one of the possible categories of X; p(y i |x j ) means that when X=x j Under the condition Y=y i The probability of; H(Y) is the information entropy of the random variable Y, indicating the uncertainty of Y itself; p(y i ) represents the category y i Probability of occurrence;

[0097] The information gain IG reflects the predictive ability of feature X to target variable Y by calculating the subtraction of information entropy. The calculation formula of the information gain threshold is:

[0098] IG(Y,X)=H(Y)-H(Y|X) (7)

[0099] Where IG(Y,X) represents the information gain of random variable Y under the condition of known X;

[0100] Based on the calculated information gain value, a threshold can be set to select features whose information gain exceeds the threshold as important features. You can also sort by the number of selected features and select the top k features.

[0101] S12. Based on the conditional entropy and information entropy of electric vehicles, the uncertainty and information content of each feature on the electric vehicle parking data are evaluated, and the entropy value and information gain value of each feature are calculated;

[0102] S13, determining an information gain threshold according to the information gain value, and judging whether the information gain value is greater than the information gain threshold, if so, retaining the electric vehicle stop data, if not, deleting the electric vehicle stop data.

[0103] S2. Using the improved Harris Eagle optimization algorithm, support vector machine regression modeling is performed on the screened electric vehicle parking data to determine the capacity boundary and power boundary of the generalized energy storage device;

[0104] Specifically, in the first stage, the Harris Eagle optimization algorithm was improved to improve the optimization efficiency of the model, the initial population and energy factor update methods of the Harris Eagle optimization algorithm were improved, and the Harris Eagle algorithm was introduced with the artificial bee colony algorithm (ABC) to enhance its regional aggregation ability and local search efficiency. This improvement not only improves the adaptability of the algorithm at different stages, but also enhances its balance between local optimal solutions and global optimal solutions and enhances local search efficiency. Through these improvements, the optimized Harris Eagle algorithm can process the screened parking data more efficiently, improve the prediction accuracy of the support vector machine regression model, and provide more accurate estimates for the key parameters of the GES model. Then, the improved Harris Eagle optimization algorithm is used to perform support vector machine regression modeling on the screened data, so as to accurately define the capacity and power limits of the GES.

[0105] In this embodiment, the Harris Hawk optimization algorithm is improved: As a new type of biological inspiration algorithm, the Harris Hawk algorithm has the characteristics of few algorithm parameters, high search accuracy, and fast convergence speed, so it is widely used in path planning, model parameter optimization, image processing and other fields. The algorithm mainly consists of three stages: global exploration stage, transformation stage and local development stage, and simulates the predation state of Harris Hawk population on prey through different strategies.

[0106] In the global exploration phase, the Harris Hawk algorithm simulates the state of the population when detecting prey according to the two strategies in formula (8).

[0107]

[0108] Where t is the number of population iterations; X(t) is the position of the Harris hawk population after the tth iteration; X rand (t) represents a position in the population randomly selected in the tth iteration; q represents the probability factor; X rabbit (t) is the position of the prey after the tth iteration; X m (t) is the average position of Harris hawk individuals after the tth iteration; r1, r2, r3, and r4 are all random numbers in [0,1]; Ub and Lb are the upper and lower limits of the algorithm search space, respectively.

[0109] The Harris Hawk algorithm changes from global exploration to local exploitation according to the escape energy E of the prey. When |E|>1, the Harris Hawk algorithm performs the global exploration phase, and when |E|<1, it performs the local exploitation phase.

[0110]

[0111] Where E0 is the initial energy of the prey, ranging from [-1,1]; T is the maximum number of iterations of the algorithm.

[0112] In the Harris Hawk algorithm, λ is used to represent the probability of successful prey escape. When λ≥0.5, it means that the prey has failed to escape successfully; λ<0.5 means that the probability of successful prey escape is high. The Harris Hawk algorithm uses the prey energy E and the escape probability λ to jointly adjust the predation strategy of the Harris Hawk population, thereby establishing the global exploration and local development stages of the Harris Hawk algorithm.

[0113] (1) Improve the population initialization method:

[0114] In the process of solving optimization problems, the Harris Hawk algorithm usually uses a randomized initial population. Therefore, it is difficult for the population to be evenly distributed in the search space, resulting in a decrease in the diversity of the population. Since chaotic mapping has the typical characteristics of randomness, ergodicity, and nonlinearity, when initializing the population, chaotic mapping can not only maintain the diversity of the population, but also prevent the algorithm from falling into a local optimum. Among them, the Tent chaotic mapping can generate uniformly distributed initial values ​​in the search space and improve the uniformity of the population distribution. Therefore, using the Tent chaotic mapping instead of randomly initialized populations can effectively improve the diversity of the Harris Hawk population and enhance the global search capability of the algorithm. Chaotic sequence Z I k The mathematical model is:

[0115]

[0116] In the formula, represents the random number in the population after the I+1th iteration, Z k Represents the random number in the population, which is used to describe the numerical model of chaotic sequence. u represents a random number between [0,1];

[0117] Chaotic sequence generated by Tent chaotic mapping and the maximum value of individual positions of the population in the search space and minimum value Generate the initial position X of the population individuals in the search space I k , defined as:

[0118]

[0119] In the formula, represents the initial position of the population individuals in the search space after the Ith iteration, Respectively represent the maximum and minimum values ​​of the individual positions of the population in the search space after the Ith iteration;

[0120] Using Tent chaotic mapping to initialize the population can improve the diversity of Harris's hawk population.

[0121] (2) The artificial bee colony algorithm (ABC) is introduced to enhance the regional aggregation capability and local search efficiency of Harris Hawk:

[0122] Artificial Bee Colony Algorithm (ABC) Bee Dance Mechanism:

[0123] Worker bee update, each worker bee is updated according to its own fitness and the current best food source location:

[0124]

[0125] Where φ is a random factor; X best is the current global best solution; X bee For worker bee positions; The location where the worker bee will appear next.

[0126] Scout bees, which periodically convert worker bees with poor fitness into scout bees, which will randomly explore new areas:

[0127] X scout =X min +(X max -X min )·r (13)

[0128] In the formula, r is a random number; X scout Indicates the location of the scout bee; X max and X min Represents the boundary of the search space.

[0129] In the update process of Harris Hawk, the aggregation information of bees is introduced:

[0130]

[0131] In the formula, is the new position of individual i; X i is the current position of individual i; J is the jump intensity; X new is the new position calculated based on the current optimal solution or update strategy, which may be calculated by other individuals (such as worker bees); X bee is the current position of the worker bee; β is the adjustment factor, which is used to control the influence of the worker bee information on the individual position update.

[0132] The Harris Hawk's encirclement strategy is selected according to the fitness of the bees. If the fitness of the worker bees is high, the soft encirclement strategy is used (to improve aggregation). The position update formula of the soft encirclement strategy is:

[0133] X new =X best +r1·(X best -X i ) (15)

[0134] Where, X new represents the new position calculated based on the current optimal solution or update strategy, X best represents the current global optimal solution, X i represents the current position of individual i, r1 represents the random number used to control the movement amplitude;

[0135] If the fitness of the worker bee is low, use the hard encirclement strategy (to improve exploration). The position update formula of the hard encirclement strategy is:

[0136] X new =X best +r2·(X best -X i ) (16)

[0137] In the formula, r2 represents a random number used to control the amplitude and direction of movement;

[0138] By introducing the aggregation ability of bees, the Harris Eagle algorithm can perform local search more effectively and avoid falling into local optimality. Such improvements not only enhance the overall performance of the algorithm, but also increase the flexibility of the algorithm.

[0139] (3) Improve the energy factor update method:

[0140] In the basic Harris Eagle optimization algorithm, the escape energy factor El is used to control the algorithm's transition from global search to local search. If |El|≥1, the algorithm enters the global search stage, and if |El|<1, the algorithm enters the local search stage. The energy factor E is updated by linearly decreasing from 2 to 0, which will cause the algorithm to fall into local search in the later stages of iteration. Therefore, a nonlinear energy factor E update method is proposed, which allows the algorithm to perform local search in the later stages of iteration while retaining the possibility of global search. The energy factor E update formula and the escape energy factor El update formula are as follows:

[0141]

[0142] El=E×E0 (18)

[0143] Where El represents the escape energy factor, E represents the energy factor, E0 represents the initial energy, t represents the current number of iterations, T represents the maximum number of iterations, and α represents the rate of decay or growth, which determines how the energy changes with time t.

[0144] The improved E decreases slowly in the early stage of the algorithm, which is conducive to global search in the early stage of the algorithm and avoids the "premature phenomenon" of the algorithm. The improved E decreases faster in the later stage of the algorithm, improving the local search performance and search efficiency in the later stage of the algorithm; the new energy factor update formula proposed can well balance the global search and local search of the algorithm, so that the Harris Eagle algorithm can perform global search and local search of the algorithm well in the entire iterative process.

[0145] The improved nonlinear energy E update formula enables the Harris Eagle algorithm to mainly perform global search in the early stage, global search and local search in the middle stage, and retain the possibility of global search in the late stage while ensuring local search.

[0146] In this embodiment, vector machine regression (SVR) uses an error control function ε (insensitive loss function). When the deviation between the model output value f(x) and the true value y is greater than ε, the model loss is calculated. Under the control of the insensitive loss function ε, by introducing the penalty parameter C and the relaxation factor ξ i , To solve the hyperplane, The relaxation factor represents the deviation of the actual value below the lower bound of the ε insensitive band, ξ i The relaxation factor represents the deviation of the actual value above the lower limit of the ε insensitive band. The model is based on statistical learning theory, with a simple structure, high classification accuracy and strong generalization ability. The SVR prediction model formula is:

[0147]

[0148] In the formula, α i and are two Lagrange multipliers, K(x,x i ) is the kernel function, b is the bias, which is a constant term in the support vector machine. It adjusts the position of the regression hyperplane. b(4) can be understood as the final value of b. n represents the number of support vectors, x represents the input feature vector of the test data, and x i represents the eigenvector of the i-th support vector.

[0149] The radial basis function is used as the kernel function of SVR. The penalty parameter C and the kernel function parameter γ are the key to determine the quality of the prediction model. By using the improved Harris Eagle algorithm to optimize the parameters of SVR and constructing an improved Harris Eagle algorithm-support vector machine regression model, we can expect to obtain better prediction performance, and at the same time use the behavior strategy of Harris Eagle to achieve more competitive results in complex optimization problems.

[0150] The method of using the improved Harris Eagle optimization algorithm to perform support vector machine regression modeling on the screened electric vehicle parking data to determine the capacity boundary and power boundary of the generalized energy storage device includes the following steps:

[0151] S21. constructing a support vector machine regression model based on the screened electric vehicle parking data, and optimizing the support vector machine regression model using an improved Harris Eagle optimization algorithm to obtain a support vector machine regression model based on Harris Eagle optimization;

[0152] Specifically, the method of optimizing the support vector machine regression model using the improved Harris Hawk optimization algorithm to obtain the support vector machine regression model based on Harris Hawk optimization includes the following steps:

[0153] 1) Initialize the solution space: determine the population size N and the initial velocity v of the particles i , initial position x i , the value range of learning factors c1 and c2, weight factor ω, number of iterations M, penalty parameter C and kernel function parameter γ;

[0154] 2) Initialize the Harris Hawk algorithm: Initialize the population size, maximum number of iterations, convergence threshold and other parameters of the Harris Hawk algorithm. These parameters will affect the algorithm's ability to search for optimal parameters;

[0155] 3) Define the objective function: In SVR prediction, the mean square error or root mean square error is usually used as the objective function;

[0156] The root mean square error (RMSE) is defined as the difference between the predicted and actual values ​​and it shows the square root of the residual variance:

[0157]

[0158] The Mean Squared Error (MSE) is defined as the average of the squared differences between the predicted and actual observed values ​​for ‘N’ samples:

[0159]

[0160] In the formula, R t Represents the actual value; F t represents the predicted value; N is the number of samples;

[0161] 4) Initialize the population: Use Tent chaotic mapping to initialize the population of the Harris Hawk algorithm. These individuals represent different combinations in the SVR parameter space.

[0162] 5) Fitness evaluation: Calculate the fitness value corresponding to each individual (i.e., parameter combination) and evaluate it based on the performance of SVR on the training set;

[0163] 6) Iterative optimization: Through the iterative optimization process, the improved Harris Eagle algorithm searches and adjusts individuals (parameter combinations). In each iteration, the position of the individuals and the structure of the entire population are updated to better adapt to the problem space, specifically:

[0164] Use Tent chaos map to generate initial individual positions to increase diversity. Calculate the fitness of all individuals and find the current best individual. In each iteration, update the position of each individual according to the current best individual.

[0165] Choose to use soft or hard surround strategy for position update according to the value of nonlinear energy factor. Add information of worker bees to enhance local search. Replace individuals with poor performance and retain individuals with high fitness according to the fitness of the new position. Let worker bees perform random exploration when the fitness is low to enhance population diversity.

[0166] The value of the energy factor is dynamically adjusted according to the iterative process to balance global search and local search. When the maximum number of iterations is reached or the fitness converges, the algorithm stops and the best solution is output.

[0167] 7) Convergence check: Check whether the algorithm has converged to the optimal solution or reached a sufficient number of iterations. When the convergence conditions are met, terminate the algorithm;

[0168] 8) Optimal parameter extraction: Based on the final convergence result, the optimal SVR parameter combination is extracted, which corresponds to the SVR model with the best performance on the given data set.

[0169] 9) Model training and validation: Use the obtained optimal parameter combination to retrain the SVR model on the complete training set and validate it on an independent validation set to ensure the generalization performance of the final model.

[0170] S22. Use the support vector machine regression model based on Harris Eagle optimization to predict the electric vehicle cluster shutdown data on the trading day, and calculate the capacity boundary and power boundary of the generalized energy storage device at each moment based on the electric vehicle cluster shutdown data on the trading day.

[0171] Specifically, the vector machine regression (SVR) model verified by training effect can be used to predict the parameters of the GES model. The specific steps are: use historical data to predict the electric vehicle cluster parking data on the trading day, and then calculate the capacity boundary and power boundary of the GES at each moment according to formula (4), where the prediction result of the initial SOC will be used for self-dispatching load calculation.

[0172] S3. Based on economic incentives and the influence of users' irrational consumption behaviors, a self-dispatching model of electric vehicle clusters is constructed, and the self-dispatching load of electric vehicle clusters is determined according to the self-dispatching model;

[0173] The EVA platform integrates charging stations in a certain area and provides electric vehicle users with information about charging stations, including location, number of available chargers, and charging prices. At the same time, EVA also needs to design appropriate demand response programs, manage chargers at charging stations to achieve collaborative charging, and aggregate electric vehicles to provide auxiliary services. As for charging stations, they provide charging infrastructure and make profits from charging service fees, electricity fees (especially charging stations owned by power companies), and government subsidies.

[0174] The goal of EV users during the charging process is to maximize the cost-effectiveness of electricity while meeting their daily electricity needs. However, they may not be able to fully grasp the fluctuations in electricity prices and cannot formulate the most economical charging plan. In addition, due to the influence of irrational psychological factors such as the desire to complete charging as soon as possible in order to use the vehicle earlier and anxiety about the amount of electricity, it is shown that EV users are not rational decision makers with complete information during the charging process, but rather limited rational people affected by multiple factors. Therefore, this embodiment will consider these non-completely rational factors to construct a self-dispatching model for electric vehicle groups.

[0175] Specifically, in the second stage, a self-dispatching model of electric vehicle clusters is established considering the impact of economic incentives and irrational consumption behaviors of EV users, and finally the self-dispatching load is obtained.

[0176] The self-dispatching model of the electric vehicle cluster is constructed based on economic incentives and the influence of irrational consumption behavior of users, and the self-dispatching load of the electric vehicle cluster is determined according to the self-dispatching model, which includes the following steps:

[0177] Considering the irrational behavior of electric vehicle users, the expected value of electric vehicle users’ responsiveness to the incentive price of the demand response program is EV,t It can be modeled as an incentive price π inc,t A linear function of

[0178]

[0179] In the formula, λ EV,t represents the expected value of EV users’ responsiveness to the incentive price of the demand response program, represents the maximum response of electric vehicle users to the incentive price of the demand response plan, that is, the upper limit of electric vehicle users' response to price incentives, π inc,t represents the incentive price, represents the incentive price corresponding to the maximum responsiveness of electric vehicle users. In practice, It depends on the willingness of electric vehicle users to participate in demand response. If the willingness of electric vehicle users to respond is high, the parameters of the charging station will be lower.

[0180] Then, λ EV,t The maximum deviation of electric vehicle user responses is d due to the triangle membership function modeling based on the consumer psychology model. EV,t Defined as:

[0181]

[0182] Where, d EV,t represents the maximum deviation of the electric vehicle user response, k1 and k2 represent the two actual proportional coefficients of the maximum deviation of the electric vehicle user response, Indicates the inflection point of the incentive price.

[0183] Finally, after defining the deviation range of the electric vehicle responsiveness, we further use the normal distribution random variable To describe the probability distribution of responsiveness deviation, where N represents normal distribution and σ t represents the standard deviation at time t. In addition, to avoid negative λ EV,t ,ξ t Should be limited to (-3σ t ,3σ t ) range, use truncation (this is a common treatment to ensure that 95% of the values ​​will fall within the range of μ±3σ based on the assumption of normal distribution, μ represents the mean of the normal distribution, and truncation can avoid unrealistic negative values), then we can have d EV,t =3σ t .

[0184] In addition, considering that the actual responsiveness of the driver is less than 100%, the specified To ensure a positive λ EV,t and in It means that when the incentive price π inc,t Arrive at a specific point The standard deviation of It means that when the incentive price π inc,t The standard deviation when it reaches its maximum possible value. Combining equations (22) and (23), the demand response function between electric vehicle responsiveness and the incentive price considering uncertainty can be modeled in a general form.

[0185] The objective function expression of the self-dispatching model of electric vehicle clusters is:

[0186]

[0187] In the formula, represents a function that calculates the escape rate (or escape intensity) based on these input parameters to determine the key dynamics in the escape process, Represents a logical indicator, ξ n represents a random variable with truncated standard normal distribution, Indicates the logical negation of a logical indicator.

[0188] Finally, an EV cluster self-dispatching model considering non-completely rational behavior factors is formed, and its objective function is shown in equation (24), and the constraints are shown in equations (1) and (2). This model simulates the process of EV users formulating self-dispatching charging and discharging plans under the influence of non-completely rational behavior factors, and finally obtains the self-dispatching load of the EV cluster.

[0189] S4. Calculate the redispatch power and redispatch capacity of the generalized energy storage device based on the capacity boundary and power boundary of the generalized energy storage device and the self-dispatch load of the electric vehicle cluster;

[0190] Specifically, the calculation formula for the redispatch power of the generalized energy storage device is:

[0191]

[0192] The calculation formula for the redispatch capacity of generalized energy storage equipment is:

[0193]

[0194] In the formula, are the upward regulation power and downward regulation power of the generalized energy storage device at time t, are the self-scheduled charging power and discharging power of the generalized energy storage device at time t, are the upper regulation capacity and lower regulation capacity of the generalized energy storage device at time t, represents the self-dispatching capacity of the generalized energy storage device at time t;

[0195] S5. Based on the redispatch power and redispatch capacity of the generalized energy storage device, a set of quantitative parameters of dispatchable potential is integrated, and the dispatch potential of the electric vehicle cluster is quantified based on the parameter set.

[0196] Specifically, based on the redispatch power and redispatch capacity of generalized energy storage equipment, combined with the parameter set of GES in S1, the dispatchable potential quantitative parameter set is further integrated By reporting this parameter set, the aggregator can enable the dispatch center to understand the operating status of the GES and the available upward and downward adjustment resources and conduct transactions, ultimately realizing the quantification and exploration of the dispatchable potential of the electric vehicle cluster.

[0197] After large-scale electric vehicles are connected to the power grid, if they are only treated as regular loads and superimposed on the traditional load curve, the operating costs of the units will undoubtedly increase significantly. Electric vehicle to grid interaction (V2G) refers to the planned guidance and management of electric vehicle charging and discharging, so that they can participate in power balance during idle periods, such as peak load regulation, frequency regulation, and wind and solar power generation, thereby creating benefits for the power grid, users and society. In fact, most electric vehicles are parked for more than 90% of the day, which provides the possibility of time for the power grid to dispatch electric vehicles. By participating in grid dispatch, electric vehicles can not only reduce system operating costs, but also assist the power grid in improving its wind power acceptance capacity.

[0198] At present, research on EV participation in the economic dispatch of the auxiliary service market often focuses on the interests of distribution network operators, EV aggregators EVA and EV users. Electric vehicle clusters participating in vehicle-grid interaction need to respond to system dispatch requirements and provide power support. Electric vehicle clusters need to report the dispatchable capacity and price for each period in the dispatch cycle. Therefore, on the basis of exploring the dispatchability of electric vehicles, effectively evaluating the dispatchable capacity of electric vehicle clusters has become the key to serving the power grid and EV users. According to the report released by the National Big Data Alliance of New Energy Vehicles, there are many types of electric vehicles in my country, including private cars, taxis, official cars, logistics vehicles and public buses. According to the current mainstream uses and travel characteristics of electric vehicles, we can divide them into commuting private cars, taxis and other functional vehicles. There are differences in the driving characteristics and charging characteristics of these three types of vehicles. As individuals, their adjustment capabilities and dispatchable capacities for participating in vehicle-grid collaboration are different.

[0199] The dispatchable capacity assessment of electric vehicle clusters refers to the assessment of the power supply capacity of electric vehicles in a cluster to determine its potential in supporting the energy demand of the power grid. This research is crucial for the popularization of electric vehicles and the stable operation of the power system. The dispatchable capacity assessment of electric vehicle clusters aims to determine the power supply capacity of electric vehicles in the cluster and its ability to adapt to different power system needs. This can help power system planners understand how to effectively use the energy storage potential of electric vehicles to improve the reliability and flexibility of the power system.

[0200] In summary, the study of the dispatchable capacity evaluation of electric vehicle clusters is of great significance in providing feasibility and reliability for power system planning and intelligent energy management. However, the existing research on the calculation model of the dispatchable potential of electric vehicles cannot take into account both accuracy and generalization, so this embodiment establishes an evaluation method that can accurately evaluate the dispatchable potential of electric vehicle clusters.

[0201] In summary, with the help of the above technical solution of the present invention, the present invention proposes to use the support vector machine regression model optimized by the improved Harris Eagle algorithm to predict the model parameters of the data filtered by the driving entropy of electric vehicles. Compared with Bi-LSTM and LSTM, this method produces less mean square error and root mean square error, thus surpassing other models. The present invention also proposes a self-dispatching model that considers the impact of economic incentives and irrational consumption behaviors of EV users on the participation of electric vehicle users in demand response. Compared with the self-dispatching results of the traditional self-dispatching model that assumes that users are completely rational and can achieve reasonable load distribution, the prediction results obtained by this model are more in line with the actual energy consumption behavior of EV users. The GES redispatching power and redispatching capacity calculated based on the self-dispatching load obtained by this model can provide a more reliable reference basis for aggregators in the power market transaction process.

[0202] In addition, by evaluating the dispatching potential of electric vehicle clusters, the present invention enables grid operators to plan grid loads in advance to avoid grid failures caused by excessive loads, or conversely use electric vehicles to provide energy storage services for the grid to ensure stable grid operation. Accurately evaluating the dispatching potential of electric vehicle clusters can plan the construction location, number, charging rate and other parameters of charging stations as needed, thereby helping to save resources and improve service quality.

[0203] In addition, on the one hand, the present invention can clarify that the dispatching potential of electric vehicles can help better arrange charging time periods so as to utilize renewable energy for charging; on the other hand, if the dispatching potential of electric vehicles is large, it can be used as a flexible load, which is conducive to stabilizing the power fluctuations caused by renewable energy sources such as wind energy and solar energy. Electric vehicles can not only obtain energy from the power grid, but also transmit idle energy back to the power grid through intelligent charging and feedback power functions, forming a two-way energy exchange system. In summary, the significance of the present invention in establishing a multi-stage electric vehicle cluster control potential quantification method is that it can help all relevant parties to efficiently utilize resources, improve service quality, and ultimately promote the end-to-end sustainable development of electric vehicles.

[0204] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0205] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-stage electric vehicle cluster control potential quantification method, characterized in that: The following steps are involved: S1. Based on the Minkowski summation principle, the electric vehicle cluster is transformed into a generalized energy storage device, and the electric vehicle stop entropy is used to perform feature screening on the electric vehicle stop data; S2. Using the improved Harris Eagle optimization algorithm, support vector machine regression modeling is performed on the screened electric vehicle parking data to determine the capacity boundary and power boundary of the generalized energy storage device; S3. Based on economic incentives and the influence of users' irrational consumption behaviors, a self-dispatching model of electric vehicle clusters is constructed, and the self-dispatching load of electric vehicle clusters is determined according to the self-dispatching model; S4. Calculate the redispatch power and redispatch capacity of the generalized energy storage device based on the capacity boundary and power boundary of the generalized energy storage device and the self-dispatch load of the electric vehicle cluster; S5. Based on the redispatch power and redispatch capacity of the generalized energy storage device, a set of quantitative parameters of dispatchable potential is integrated, and the dispatch potential of the electric vehicle cluster is quantified based on the parameter set.

2. A multi-stage electric vehicle cluster control potential quantification method according to claim 1, characterized in that: The method of converting the electric vehicle cluster into a generalized energy storage device based on the Minkowski summation principle and using the electric vehicle stop entropy to perform feature screening on the electric vehicle stop data includes the following steps: S11. Aggregate the electric vehicle cluster at the aggregator level using the Minkowski summation process to obtain a generalized energy storage device and extract key parameters of the generalized energy storage device; S12. Based on the conditional entropy and information entropy of electric vehicles, the uncertainty and information content of each feature on the electric vehicle parking data are evaluated, and the entropy value and information gain value of each feature are calculated; S13, determining an information gain threshold according to the information gain value, and judging whether the information gain value is greater than the information gain threshold, if so, retaining the electric vehicle stop data, if not, deleting the electric vehicle stop data.

3. A multi-stage electric vehicle cluster control potential quantification method according to claim 2, characterized in that: The calculation formula for the key parameters of generalized energy storage equipment is: In the formula, and They represent the charging dispatching power and discharging dispatching power of the generalized energy storage equipment in period t, and They represent the charging dispatching power and discharging dispatching power of electric vehicle n in period t, respectively. EV represents the set of electric vehicles in the generalized energy storage device, and They represent the maximum charging dispatch power and the maximum discharging dispatch power of the generalized energy storage device in period t, and They represent the upper limit of charging power and the upper limit of discharging power of electric vehicle n, respectively. are Boolean variables representing the charging state and discharging state of electric vehicle n at time t, respectively. GES,t represents the power of the generalized energy storage device in period t, S n,t represents the battery power of electric vehicle n at time t, and They represent the minimum and maximum power of the generalized energy storage device in period t, and They represent the minimum and maximum battery capacity of electric vehicle n, A Boolean variable representing the parking state of electric vehicle n at time t, ΔS GES,t represents the change in power of the generalized energy storage device due to the change in the grid-connected state of the electric vehicle during period t, S n,arrival , S n,leave They represent the initial power of electric vehicle n and the power of electric vehicle n when it leaves the power grid respectively.

4. A multi-stage electric vehicle cluster control potential quantification method according to claim 1, characterized in that: The improved Harris Eagle optimization algorithm is used to perform support vector machine regression modeling on the screened electric vehicle parking data to determine the capacity boundary and power boundary of the generalized energy storage device, including the following steps: S21. constructing a support vector machine regression model based on the screened electric vehicle parking data, and optimizing the support vector machine regression model using an improved Harris Eagle optimization algorithm to obtain a support vector machine regression model based on Harris Eagle optimization; S22. Use the support vector machine regression model based on Harris Eagle optimization to predict the electric vehicle cluster shutdown data on the trading day, and calculate the capacity boundary and power boundary of the generalized energy storage device at each time based on the electric vehicle cluster shutdown data on the trading day.

5. A multi-stage electric vehicle cluster control potential quantification method according to claim 4, characterized in that: The method of optimizing the support vector machine regression model by using the improved Harris Hawk optimization algorithm to obtain the support vector machine regression model based on Harris Hawk optimization includes the following steps: Determine the population size, initial velocity of particles, initial position, learning factor, weight factor, number of iterations, penalty parameter and the range of kernel function parameters; Initialize the population size, maximum number of iterations, convergence threshold of the Harris Eagle algorithm, and use the mean square error or root mean square error as the objective function; The population of Harris Eagle algorithm is initialized by using Tent chaotic mapping, the fitness value of each individual is calculated, and the position and population structure of the individual are updated based on the nonlinear energy factor value and encirclement strategy combined with the fitness value; Determine whether the maximum number of iterations has been reached or the fitness has converged. If not, continue iterating. If so, determine the optimal parameter combination of the support vector machine regression model based on the convergence result, and construct a support vector machine regression model based on Harris Eagle optimization based on the optimal parameter combination.

6. A multi-stage electric vehicle cluster control potential quantification method according to claim 5, characterized in that: The method of using Tent chaotic mapping to initialize the population of the Harris Hawk algorithm, calculating the fitness value of each individual, and updating the position and population structure of the individual based on the nonlinear energy factor value and the encirclement strategy in combination with the fitness value includes the following steps: The Tent chaotic map is used to generate a chaotic sequence, the maximum and minimum values ​​of the positions of the individuals in the search space, and the initial positions of the individuals in the search space to obtain the initial population. Based on the objective function, calculate the fitness value of each individual in the initial population, and determine the current best individual based on the calculation results; Determine whether the individual fitness value is greater than the preset fitness threshold. If so, select the soft encirclement strategy; if not, select the hard encirclement strategy; Calculate the escape energy factor value according to the energy factor, and determine whether the escape energy factor value is less than 1. If so, select the local search update method, if not, select the global search update method; Based on the selected update method and encirclement strategy, the position and population structure of each individual are updated in combination with the current best individual.

7. A multi-stage electric vehicle cluster control potential quantification method according to claim 6, characterized in that: The expression of chaotic order is: The expression of the initial position of the population individuals in the search space is: The position update formula of the soft encirclement strategy is: X new =X best +r1·(X best -X i ) The position update formula of the hard encirclement strategy is: X new =X best +r2·(X best -X i ) The calculation formula of escape energy factor is: E1=E×E0 In the formula, represents the random number in the population after the I+1th iteration, Z k represents a random number in the population, u represents a random number between [0,1], represents the initial position of the population individuals in the search space after the Ith iteration, They represent the maximum and minimum values ​​of the individual positions of the population in the search space after the Ith iteration, respectively, and X new represents the new position calculated based on the current optimal solution or update strategy, X best represents the current global optimal solution, X i represents the current position of individual i, r1 represents the random number used to control the movement amplitude, r2 represents the random number used to control the amplitude and direction of movement, El represents the escape energy factor, E represents the energy factor, E0 represents the initial energy, t represents the current number of iterations, T represents the maximum number of iterations, and α represents the rate of decay or growth.

8. A multi-stage electric vehicle cluster control potential quantification method according to claim 1, characterized in that: The self-dispatching model of the electric vehicle cluster is constructed based on economic incentives and the influence of the user's irrational consumption behavior, and the self-dispatching load of the electric vehicle cluster is determined according to the self-dispatching model, which includes the following steps: S31. Based on the irrational behavior of electric vehicle users, the user's expectation of the responsiveness of the incentive price of the demand response program is modeled as a linear function of the incentive price. S32, determining the deviation range of the electric vehicle user response, using a random variable that obeys a normal distribution to describe the probability distribution of the responsiveness deviation, and obtaining the maximum deviation value of the electric vehicle user response according to a truncation process; S33. Based on the expected value of the responsiveness of electric vehicle users to the incentive price of the demand response plan and the maximum deviation of the electric vehicle user response, a demand response function between the electric vehicle responsiveness and the uncertain incentive price is modeled to obtain a self-dispatching model of the electric vehicle cluster, and the self-dispatching load of the electric vehicle cluster is determined based on the self-dispatching model.

9. A multi-stage electric vehicle cluster control potential quantification method according to claim 8, characterized in that: The expected value of electric vehicle users’ responsiveness to the incentive price of the demand response program is expressed as: The expression of the deviation range of electric vehicle user response is: The objective function expression of the self-dispatching model of electric vehicle clusters is: In the formula, λ EV,t represents the expected value of EV users’ responsiveness to the incentive price of the demand response program, represents the maximum response degree of electric vehicle users to the incentive price of the demand response program, π inc,t represents the incentive price, represents the incentive price corresponding to the maximum responsiveness of electric vehicle users, d EV,t represents the maximum deviation of the electric vehicle user response, k1 and k2 represent the two actual proportional coefficients of the maximum deviation of the electric vehicle user response, Indicates the inflection point of the incentive price, represents a function that calculates the escape rate or escape intensity, Represents a logical indicator, ξ n represents a random variable with truncated standard normal distribution, Indicates the logical negation of a logical indicator.

10. A multi-stage electric vehicle cluster control potential quantification method according to claim 3, characterized in that: The calculation formula for the redispatch power of generalized energy storage equipment is: The calculation formula for the redispatch capacity of generalized energy storage equipment is: The expression of the schedulable potential quantitative parameter set is: In the formula, are the upward regulation power and downward regulation power of the generalized energy storage device at time t, are the self-scheduled charging power and discharging power of the generalized energy storage device at time t, are the upper regulation capacity and lower regulation capacity of the generalized energy storage device at time t, Represents the self-dispatching capacity of the generalized energy storage device at time t.

Images