A multi-objective optimization method and device for a micro-grid energy storage network considering network characteristic parameters

Abstract

The application discloses a kind of network construction energy storage microgrid multi-objective optimization method and device considering network construction characteristic parameter, virtual inertia time constant, active-power-frequency droop coefficient, reactive-power-voltage droop coefficient and virtual damping coefficient are used as decision variable;With dynamic performance optimal, the strongest balance power fluctuation ability and the lowest operating economic cost are used as target to establish three-objective optimization model, and constraint condition is set;Artificial vulture multi-objective optimization algorithm is used to solve Pareto optimal solution set: initialize population and external file, select two leaders from external file, carry out position update to generate candidate solution after calculating satiety rate, carry out non-dominated screening after applying deep exploration disturbance, update external file and iterate to convergence output Pareto optimal solution set.The application realizes the global collaborative optimization of network construction characteristic parameter, dynamic security, power supply reliability and operating economy are considered, and is suitable for weak grid end microgrid scene.

Description

Technical Field

[0001] This invention relates to the field of microgrid operation technology, specifically a multi-objective optimization method and device for grid-connected energy storage microgrids that takes into account grid characteristic parameters. Background Technology

[0002] Microgrid systems using grid-based energy storage as the main power source have significant application value in remote mountainous areas, islands, and other end-of-grid scenarios. These systems typically use grid-based energy storage to provide voltage and frequency references, while distributed power sources such as photovoltaics and wind power, along with local loads, are connected to the main grid via long, weak interconnect lines or operate independently. Because microgrids have small short-circuit capacity and low inertia, the system's frequency and voltage are highly susceptible to load fluctuations and renewable energy source volatility. Therefore, the parameter configuration of the grid-based energy storage directly determines whether the system can operate safely, economically, and reliably.

[0003] Most existing methods only use the rated power and capacity of energy storage as optimization variables, while network characteristic parameters such as virtual inertia, droop coefficient, and damping coefficient rely on fixed values ​​based on experience, making it difficult to achieve optimal system performance. Furthermore, existing technologies often use economic efficiency or dynamic performance as the sole optimization objective, considering other important performance aspects only as constraints, failing to simultaneously address dynamic performance, power balance, and economic efficiency. During microgrid optimization operation, these three factors exhibit significant interrelationships, making it difficult to obtain a balanced and feasible configuration scheme through single-objective optimization.

[0004] To address the aforementioned issues, this invention uses the grid-connected characteristic parameters of the grid-connected energy storage microgrid as key optimization variables, establishes a collaborative optimization model with three objectives: dynamic performance, power balance, and economy, and employs the artificial vulture multi-objective optimization algorithm to solve the problem, thereby obtaining an optimized operation scheme for the grid-connected energy storage microgrid. Summary of the Invention

[0005] The technical problem to be solved by this invention is a multi-objective optimization method and device for grid-connected energy storage microgrids that takes into account grid characteristic parameters. It breaks through the limitations of empirical values ​​of grid characteristic parameters, realizes balanced and coordinated optimization of dynamic performance, power balance and economy, and efficiently obtains a uniformly distributed Pareto optimal solution set through an improved multi-objective optimization algorithm, providing a scientific, comprehensive and engineering-applicable optimization configuration scheme for grid-connected energy storage microgrids.

[0006] A multi-objective optimization method for grid-connected energy storage microgrids that takes into account grid-connection characteristic parameters includes the following steps:

[0007] Step S1: Obtain the system operating parameters of the grid-connected energy storage microgrid, and establish grid characteristic parameter decision variables based on the system operating parameters: virtual inertia time constant H, active power-frequency droop coefficient Dp, reactive power-voltage droop coefficient D q and virtual damping coefficient Dd Together, they serve as optimization decision variables, forming a four-dimensional network characteristic parameter decision vector X=[H, D] p D q D d ] T ;

[0008] Step S2: Based on the network characteristic parameter decision vector, establish a multi-objective optimization model for the network-connected energy storage microgrid: with the objectives of optimizing the dynamic performance of the network-connected energy storage microgrid system, maximizing the ability to balance power fluctuations, and minimizing the economic cost of system operation, establish dynamic performance objective function f1, dynamic power balance function f2, and economic objective function f3 respectively, and set constraints including power balance constraints, energy storage operation constraints, power exchange and voltage constraints, new energy output constraints, and boundary constraints of network characteristic parameters;

[0009] Step S3: The artificial vulture multi-objective optimization algorithm is used to solve the multi-objective optimization model of the grid-connected energy storage microgrid to obtain the Pareto optimal solution set: the four-dimensional grid-connected characteristic parameter decision vector X is used as the optimization variable, and the minimization of the dynamic performance objective function f1, dynamic power balance function f2 and economic objective function f3 is taken as the optimization direction. Based on the constraints, the feasibility of each individual is judged, and the three-objective Pareto optimal solution set is obtained.

[0010] Furthermore, the dynamic performance objective function f1 is composed of a weighted average of four sub-indices: the minimum frequency, the maximum rate of change of frequency, the integral of the absolute value of frequency deviation, and the steady-state voltage deviation. The expression is as follows:

[0011] ;

[0012] in: This is the lowest point of the system frequency; This represents the maximum rate of change of the system frequency. Integral the absolute value of the system frequency deviation; denoted as , where is the steady-state voltage deviation of the system; f0 is the rated frequency of the system; Δω(t) is the time series obtained by integrating the system frequency deviation every 1ms; T is the duration of the disturbance simulation; α1, α2, α3, and α4 are normalized weighting coefficients.

[0013] The dynamic power balance function f2 is composed of a weighted average of the frequency over-limit accumulation, voltage over-limit accumulation, and power shortage, and its expression is:

[0014] ;

[0015] Where: E f E represents the cumulative amount of frequency exceeding the limit. v E represents the cumulative amount of voltage exceeding the limit. shedFor lack of power supply; N d Number of typical daily scenes; D m T represents the number of representative days of a typical day; m This represents the set of simulation time periods for a typical day; Δf(t) and ΔV(t) represent the frequency and voltage deviations for each time period of a typical day; δf lim δV lim These represent the allowable normal fluctuation ranges for frequency and voltage, respectively; γ f γ v γ shed P represents the normalized weighting coefficients. shed (t) represents the load shedding power at time t, Δt=1h;

[0016] The economic objective function f3 is composed of a weighted average of annual operating costs, wind and solar curtailment rates, and energy storage equivalent lifetime loss costs, and its expression is:

[0017] ;

[0018] Where: C annual Annual operating cost; η cur C represents the wind and solar curtailment rate. degrad Cost of energy storage equivalent lifetime loss; C m The operating cost for a typical day; E cur,m E avail,m These represent the amount of wind and solar power curtailment on a typical day and the maximum available power of new energy sources; C bat The unit capacity investment cost of the battery; E lifetime E represents the total charge throughput over the battery's entire lifespan. through,m represents the total charge / discharge throughput of energy storage on a typical day; w1, w2, and w3 are normalized weighting coefficients.

[0019] Furthermore, step S3 employs the Artificial Vulture multi-objective optimization algorithm to solve the multi-objective optimization model of the grid-connected energy storage microgrid, specifically including the following sub-steps:

[0020] Step S31: Initialize the population and algorithm parameters, including population size N, external file A, and maximum number of iterations T. max ;

[0021] Step S32: Determine if the maximum number of iterations has been reached. If so, proceed to step S38 and output the external file A as the final Pareto optimal solution set; otherwise, proceed to step S33.

[0022] Step S33: Select two leaders X from external file A using a crowding-inverse roulette wheel. B1 and X B2 ;

[0023] Step S34: Calculate the satiety rate for adaptive adjustment of the exploration and development ratio;

[0024] Step S35: Perform position update for each individual in the current population to generate a new generation of candidate solutions, and calculate the objective value and feasibility label of each candidate solution based on the objective function and constraints of step S2;

[0025] Step S36: Apply random deep exploration perturbation to the candidate solutions to generate derived solutions and uniformly perform non-dominated screening to enhance local convergence ability;

[0026] Step S37: Merge the parent and child populations, perform fast non-dominated sorting and crowding screening to generate the next generation population, while updating the external archive and maintaining a uniform distribution of the solution set, and return to step S32.

[0027] Step S38: Output the set of three-objective Pareto optimal solutions from external file A for selecting a compromise configuration.

[0028] Furthermore, the population generation in step S31 includes the following steps:

[0029] S311. Population Initialization: Using the network characteristic parameters as the position vectors of individuals in the population, each component is uniformly and randomly generated under the constraints of the network characteristic parameters in step S2. The expression is as follows:

[0030] The initial population size is N=50;

[0031] ;

[0032] Where: r i，j (j=1,2,3,4) is a random number between [0,1];

[0033] S312. Individual fitness assessment: Dynamic performance simulation and annual operation simulation estimation are performed for each individual in the population to obtain three objective functions f1, f2, and f3 for each individual in the population.

[0034] For each individual in the population, a feasibility assessment is performed to check whether the data in each time period exceeds the limits:

[0035] ;

[0036] Where: Feasible(X) i ) is a feasibility function for an individual in the population. A value of 1 indicates that all constraints are met, while a value of 0 indicates that the constraints are exceeded, and the individual is discarded.

[0037] S313, Multi-objective dominance sorting and external file initialization: Perform fast non-dominated sorting on the initially generated population, and stratify individuals according to the constraint feasibility of step S2 and the dominance relationship of objective functions f1, f2, and f3 to obtain a series of frontiers including the first frontier and the second frontier. The first frontier consists of all feasible and non-dominated individuals in the current population, representing the non-dominated solution set that is optimal in terms of dynamic performance, power balance, and economic cost. Store the first frontier in the external file A to complete the external file initialization.

[0038] Furthermore, the leader selection method in step S33 is as follows:

[0039] For each objective function f k , Sort all solutions in file A in ascending order according to the objective value. For each solution in file A, calculate its crowding distance d in the three-objective space. j :

[0040] ;

[0041] in: , These are the values ​​of the previous and next neighbors of the j-th individual after sorting, respectively, on the k-th target. , Let these be the maximum and minimum values ​​of the k-th target in the file;

[0042] Using a roulette wheel game with inverse crowding ratio, select two different individuals X from file A. B1 and X B2 :

[0043] Individual X j The probability of being selected is:

[0044] ;

[0045] Where: ε is a small positive number to avoid the denominator being zero; |A| is the number of individuals in the current external archive.

[0046] Furthermore, the formula for calculating the satiety rate in step S34 is as follows:

[0047] ;

[0048] Where: t is the current iteration number; T max is the maximum number of iterations; r1 is a random number uniformly distributed between 0 and 1; z is a parameter that changes linearly with iteration, decreasing linearly from the initial value of 0 to the final value of -2; w is a constant, taken as 2.5, used to control the fluctuation range.

[0049] Furthermore, the individual location update method in step S35 is as follows:

[0050] S351. Determine the reference individual: Randomly select the current reference solution with probabilities L1=0.8 and L2=0.2.

[0051] , where r2 is a uniformly random number between 0 and 1;

[0052] S352, Exploration Phase: A random number r3 is generated each time an individual is updated. Its value follows a uniform distribution within the interval [0,1] and is used for strategy selection. If the random number r3 < 0.6, the exploration phase begins.

[0053] Generate random numbers r4∈[0,1] for strategy selection, then the candidate solutions X generated in this stage new for:

[0054] , where: X ref The current reference individual's position vector; F is the satiety rate; X i r5 and r6 are mutually independent uniform random numbers between 0 and 1; l b u b These are the lower and upper bound vectors for the network characteristic parameters, respectively;

[0055] S353, Development Phase: If the random number r3 ≥ 0.6, then proceed to the development phase;

[0056] If the satiety rate |F| ≥ 0.5, then proceed to the first sub-stage. Generate random numbers r7 ∈ [0, 1] for sub-strategy selection;

[0057] When the random number r7 < 0.5, a food competition strategy is executed, and the candidate solutions X generated in this stage are... new Update as follows:

[0058] Where: r8 is a uniformly random number between 0 and 1;

[0059] When the random number 0.5 ≤ r7 ≤ 1, a rotational flight strategy is executed, and the candidate solutions X generated in this stage are... new Update as follows:

[0060] Where: r9 is a uniformly random number between 0 and 1; S is the spiral radius control parameter, which is set to 0.2;

[0061] If the satiety rate |F| < 0.5, then proceed to the second sub-stage and generate a random number r. 10 ∈[0,1], used for sub-policy selection;

[0062] When the random number r 10When the value is less than 0.5, a food source accumulation strategy is implemented, and the candidate solutions X generated in this stage are... new Update as follows:

[0063] ;

[0064] When the random number 0.5 ≤ r 10 When the value is ≤1, a fierce competition strategy is implemented, and the candidate solutions X generated in this stage are... new Update as follows:

[0065] , where levy(d) is a random vector following a Lévy distribution with shape parameter β=1.5.

[0066] Furthermore, the method for in-depth exploration in step S36 is as follows: for each new individual X new Generate random number r 11 ∈[0,1], if r 11 If the value is less than 0.5, then the following deep refinement steps will be performed on the individual:

[0067] S361. Generate mean perturbation points: Randomly select three distinct individuals X from the current population. r1 X r2 X r3 Calculate their mean x avg =(X r1 +X r2 +X r3 ) / 3, and construct the first candidate solution X new1 :

[0068] , where: r 12 A uniformly random number between 0 and 1;

[0069] S362, Adaptive Disturbance Generation: Dynamically adjusts the search radius and sets the attenuation coefficient w. local =exp(-rand·t / T max ), where rand is an independent random number between 0 and 1, used to control the decay rate; then generate a random number r. 13 ∈[0,1], the second candidate solution X is generated according to the following bifurcation rule. new2 :

[0070] , where: r 14 r 15 r 16 These are mutually independent uniform random numbers between 0 and 1;

[0071] S363. Perform secondary refinement: Generate random number r 17∈[0,1], if r 17 <w local Then for X new2 Perform the development phase operation again to obtain the third candidate solution X. new3 ;

[0072] S364. Non-dominated screening: Calculate X new2 and X new3 Compare the three objective function values ​​with the original X. new The dominance relationship between them is used to add all non-dominant solutions of the three to the temporary child set Q. t .

[0073] Furthermore, the method for updating external files in step S37 is as follows:

[0074] For the temporary child set Q t For each new individual, accurately evaluate its three objective function values;

[0075] Parental population P t Merge with all offspring individuals to form R t =P t ∪Q t , for R t Performing a fast nondominated sort yields several frontiers;

[0076] Starting from the first frontier, individuals are selected into P frontier by frontier. t+1 This continues until the total number of selected individuals reaches N. If adding a certain frontier would result in the total number of individuals exceeding N, then the crowding distance between all individuals at that frontier is calculated, and individuals with higher crowding are selected first to maintain the uniformity of the solution set distribution.

[0077] R t All non-dominated solutions are stored in external file A;

[0078] The updated external file A is re-sorted non-dominated, retaining only the new first frontier; if the file size |A|>N, the grid mechanism is enabled: the target space is divided into equidistant grids in each dimension, the number of individuals in each hypercube is counted, and the individuals in the most crowded grid are deleted in a loop until the file size meets the set value.

[0079] A multi-objective optimization device for grid-connected energy storage microgrids, taking into account grid characteristic parameters, is used to execute the method described above, comprising:

[0080] The decision variable establishment module is used to obtain the system operating parameters of the grid-connected energy storage microgrid, and establish grid characteristic parameter decision variables based on the system operating parameters: virtual inertia time constant H, active power-frequency droop coefficient Dp, reactive power-voltage droop coefficient D q and virtual damping coefficient D dTogether, they serve as optimization decision variables, forming a four-dimensional network characteristic parameter decision vector X=[H, D] p D q D d ] T ;

[0081] The optimization model building module is used to establish a multi-objective optimization model for the grid-connected energy storage microgrid based on the grid-connected characteristic parameter decision vector: with the objectives of optimizing the dynamic performance of the grid-connected energy storage microgrid system, maximizing the ability to balance power fluctuations, and minimizing the economic cost of system operation, dynamic performance objective function f1, dynamic power balance function f2, and economic objective function f3 are established respectively, and constraints including power balance constraints, energy storage operation constraints, power exchange and voltage constraints, new energy output constraints, and grid-connected characteristic parameter boundary constraints are set.

[0082] The optimization solution module is used to solve the multi-objective optimization model of the grid-connected energy storage microgrid using the artificial vulture multi-objective optimization algorithm to obtain the Pareto optimal solution set: taking the four-dimensional grid-connected characteristic parameter decision vector X as the optimization variable, minimizing the dynamic performance objective function f1, the dynamic power balance function f2, and the economic objective function f3 as the optimization direction, and judging the feasibility of individuals based on the constraints, the three-objective Pareto optimal solution set is obtained.

[0083] Compared with the prior art, the present invention has the following significant advantages and beneficial effects:

[0084] 1. The optimized variable system is more comprehensive, giving full play to the performance potential of grid-type energy storage.

[0085] Existing methods only use the rated power and capacity of energy storage as optimization variables, and the network characteristic parameters such as virtual inertia, droop coefficient, and damping coefficient rely on empirically fixed values. This invention makes a breakthrough by incorporating the virtual inertia time constant H, active-frequency droop coefficient Dp, reactive-voltage droop coefficient Dp, and virtual damping coefficient D... d By incorporating these variables into the optimization decision variables, a four-dimensional network characteristic parameter decision vector is formed, enabling the dynamic response characteristics of network-based energy storage to be globally and collaboratively optimized according to the actual operating scenario. This breaks through the performance bottleneck of traditional experience-based values ​​and fully leverages the frequency and voltage support potential of network-based energy storage in weak grid scenarios.

[0086] 2. Three objectives are optimized in a coordinated manner, taking into account dynamic security, power supply reliability and operational economy.

[0087] Existing technologies often use economic efficiency or dynamic performance as a single optimization objective, treating other performance aspects only as constraints, making it difficult to obtain a balanced and feasible configuration scheme. This invention constructs three mutually constraining optimization objectives: a dynamic performance objective function (quantifying the ability to maintain frequency and voltage safety and quickly recover stability under extreme disturbances), a dynamic power balance function (assessing the ability to balance source-load power fluctuations throughout the year), and an economic objective function (optimizing the economical operating level based on safety and balance). This achieves synergistic optimization of the three objectives, effectively overcoming the problem of neglecting one aspect for another caused by single-objective optimization, and providing a rich set of Pareto optimal configuration schemes for engineering decision-making.

[0088] 3. An improved multi-objective optimization algorithm for artificial vultures, which combines solution accuracy and convergence speed.

[0089] To address the high-dimensional nonlinearity of the three-objective optimization model, this invention employs an improved artificial vulture multi-objective optimization algorithm. By using a crowding-inverse roulette wheel leader selection mechanism, non-dominated solutions in sparse regions are preferentially selected as leaders, effectively maintaining the uniformity of the solution set distribution. An adaptive satiety rate adjustment strategy achieves a dynamic balance between exploration and exploitation ratios. Furthermore, by adding a random depth exploration perturbation step to the standard algorithm framework, local convergence capability is enhanced, effectively avoiding getting trapped in local optima. These improvements enable the algorithm to maintain population diversity while ensuring convergence accuracy, obtaining a uniformly distributed and fully covered Pareto optimal front.

[0090] 4. It has strong engineering practicality and is suitable for microgrid scenarios at the end of weak power grids.

[0091] This invention addresses the actual operating scenarios of microgrids in remote mountainous areas (energy storage grid-connected control, photovoltaic and wind power grid integration, and power supply via high-impedance lines). The optimization process fully considers the impact of load mutations and new energy fluctuations during actual operation. It comprehensively evaluates the performance of the configuration scheme by combining dynamic performance simulation with annual operation simulation estimation. The optimization results can directly guide the engineering parameter tuning of grid-connected energy storage microgrids, demonstrating significant engineering application value and promising prospects for promotion. Attached Figure Description

[0092] Figure 1 This is a typical microgrid topology diagram of a remote mountainous area.

[0093] Figure 2 This is a flowchart of a multi-objective optimization method for a grid-connected energy storage microgrid that takes into account grid characteristic parameters, according to an embodiment of the present invention.

[0094] Figure 3 This is a flowchart illustrating the process of using the artificial vulture algorithm to solve the multi-objective optimization model of the grid-connected energy storage microgrid in an embodiment of the present invention.

[0095] Figure 4 This is a graph showing the fitness optimization results of an embodiment of the present invention.

[0096] Figure 5 This is a trajectory diagram for optimizing the network characteristic parameters in an embodiment of the present invention. Detailed Implementation

[0097] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0098] Regarding the appendix Figure 1 The invention presents an operational scenario for a microgrid in a remote mountainous area (where energy storage uses grid-based control, photovoltaic and wind power are connected to the grid via grid-connected converters, and electricity is supplied to a 400V local load via a 5 km high-impedance line). This embodiment proposes a multi-objective optimization configuration method for grid-based energy storage microgrids, taking into account grid-based characteristic parameters. This method incorporates the grid-based characteristic parameters of the grid-based energy storage microgrid into the optimization variables, establishes a multi-objective optimization model considering dynamic performance, power balance, and economy, and solves the model using a multi-objective artificial vulture algorithm to obtain an optimized operation scheme for the grid-based energy storage microgrid with multi-objective coordination.

[0099] like Figure 2 As shown, the method includes the following steps:

[0100] Step S1: Establish decision variables for network characteristic parameters:

[0101] The virtual inertia time constant H and the active-frequency droop coefficient D are used. p Reactive power-voltage droop coefficient D q and virtual damping coefficient D d Together, they serve as optimization decision variables, forming a four-dimensional network characteristic parameter decision vector X=[H, D] p D q D d ] T .

[0102] Step S2: Establish a multi-objective optimization model for grid-connected energy storage microgrids:

[0103] (1) Establish a multi-objective optimization model with the objectives of achieving optimal dynamic performance, strongest power fluctuation balancing capability, and lowest economic operating cost of the grid-connected energy storage microgrid system. The three objective functions are:

[0104] Dynamic performance objective function: Used to quantify the ability of a grid-connected energy storage microgrid system to maintain frequency and voltage safety and quickly recover stability under extreme disturbances. It consists of a weighted average of four sub-indicators: minimum frequency, maximum rate of frequency change, integral of absolute frequency deviation, and steady-state voltage deviation. The expression for the dynamic performance objective function f1 is:

[0105] ;

[0106] in: This is the lowest point of the system frequency; This represents the maximum rate of change of the system frequency. Integral the absolute value of the system frequency deviation; denoted as α, where α is the steady-state voltage deviation of the system; f0 is the rated frequency of the system; Δω(t) is the time series obtained by integrating the system frequency deviation every 1ms; T is the duration of the disturbance simulation; and α1, α2, α3, and α4 are normalized weighting coefficients.

[0107] Dynamic power balance function: Used to evaluate the system's ability to balance source-load power fluctuations throughout the year. It is weighted by frequency, cumulative voltage over-limit deviation, and power shortage. The expression for the dynamic power balance function f2 is:

[0108] ;

[0109] Where: E f E represents the cumulative amount of frequency exceeding the limit. v E represents the cumulative amount of voltage exceeding the limit. shed For lack of power supply; N d Number of typical daily scenes; D m T represents the number of representative days of a typical day; m This represents the set of simulation time periods for a typical day; Δf(t) and ΔV(t) represent the frequency and voltage deviations for each time period of a typical day; δf lim δV lim These represent the allowable normal fluctuation ranges for frequency and voltage, respectively; γ f γ v γ shed P represents the normalized weighting coefficients. shed (t) represents the load shedding power at time t, where Δt = 1h.

[0110] Economic objective function: Used to optimize the economic operation level of the system while maintaining safety and balance. It consists of three sub-objectives: annual operating cost, wind and solar curtailment rate, and equivalent lifetime loss cost of energy storage. The expression for the economic objective function f3 is:

[0111] ;

[0112] Where: C annualAnnual operating cost; η cur C represents the wind and solar curtailment rate. degrad Cost of energy storage equivalent lifetime loss; C m The operating cost for a typical day; E cur,m E avail,m These represent the amount of wind and solar power curtailment on a typical day and the maximum available power of new energy sources; C bat The unit capacity investment cost of the battery; E lifetime E represents the total charge throughput over the battery's entire lifespan. through,m represents the total charge / discharge throughput of energy storage on a typical day; w1, w2, and w3 are normalized weighting coefficients.

[0113] (2) Establish constraints, including:

[0114] Power balance constraints:

[0115] , where: P pv (t) represents the active power output of photovoltaic power generation; P wind (t) represents the active power output of wind power generation; P ESS，dis (t) represents the energy storage discharge power; P ESS，ch (t) represents the energy storage charging power; P L (t) represents the total active power of the system load.

[0116] Energy storage operation constraints:

[0117] Charge / discharge power limit: 0≤P ESS，dis (t)≤P bat , 0≤P ESS，ch (t)≤P bat , where: P bat This is the rated power of the energy storage.

[0118] State of charge limitation: SOC min ≤SOC(t)≤SOC max Where: SOC(t) is the state of charge of the energy storage battery; SOC min State of Charge (SOC) is the minimum value for the state of charge of an energy storage battery. max This represents the upper limit of the state of charge (SOC) of the energy storage battery.

[0119] Power exchange and voltage constraint:

[0120] , 0.93≤V PCC (t)≤1.07 (pu), where: The permissible active power limit for switching is determined based on line parameters and transformer capacity. This refers to the permissible reactive power limit determined based on line parameters and transformer capacity; V PCC (t) represents the per-unit value of the voltage at the point of common coupling.

[0121] Constraints on new energy output:

[0122] , ,in: This refers to the rated installed capacity of the photovoltaic power generation system. This refers to the rated installed capacity of the wind turbine generator set.

[0123] Boundary constraints on network characteristic parameters:

[0124] H min ≤H≤H max , ≤D p ≤ , ≤D q ≤ , ≤D d ≤ , where: H min H represents the lower limit of the virtual inertia time constant. max This represents the upper limit of the virtual inertia time constant. This is the lower limit of the active power-frequency droop factor. This represents the upper limit of the active power-frequency droop factor. This is the lower limit of the reactive power-voltage droop coefficient. This represents the upper limit of the reactive power-voltage droop coefficient. This represents the lower limit of the virtual damping coefficient. This represents the upper limit of the virtual damping coefficient.

[0125] Step S3: Solve the multi-objective optimization model of the grid-connected energy storage microgrid using the artificial vulture algorithm:

[0126] The four-dimensional decision vector X=[H, D] established in step S1 is... p D q D d ] T As optimization variables, minimizing the three objective functions f1, f2, and f3 established in step S2 is the optimization direction. Based on all constraints listed in step S2, the feasibility of each individual is determined, and the Pareto optimal solution set is solved using the Artificial Vulture multi-objective optimization algorithm. Figure 3 As shown, it includes the following steps:

[0127] Step S31: Initialize the population and algorithm parameters, including population size N, external file A, and maximum number of iterations T. max wait;

[0128] Step S32: Determine if the maximum number of iterations has been reached. If so, proceed to step S38 and output the external file A as the final Pareto optimal solution set; otherwise, proceed to step S33.

[0129] Step S33: Select two leaders X from external file A using a crowding-inverse roulette wheel. B1 and X B2 ;

[0130] Step S34: Calculate the satiety rate for adaptive adjustment of the exploration and development ratio;

[0131] Step S35: Perform position update for each individual in the current population to generate a new generation of candidate solutions, and calculate the objective value and feasibility label of each candidate solution based on the objective function and constraints of step S2;

[0132] Step S36: Apply random deep exploration perturbation to the candidate solutions to generate derived solutions and uniformly perform non-dominated screening to enhance local convergence ability;

[0133] Step S37: Merge the parent and child populations, perform fast non-dominated sorting and crowding screening to generate the next generation population, while updating the external archive and maintaining a uniform distribution of the solution set, and return to step S32.

[0134] Step S38: Output the set of three-objective Pareto optimal solutions from external file A for selecting a compromise configuration.

[0135] Specifically, population generation in step S31 includes the following steps:

[0136] S311, Population Initialization

[0137] Using the network construction characteristic parameters as the position vectors of individuals in the population, each component is uniformly and randomly generated under the constraints of the network construction characteristic parameters in step S2. The expression for this is:

[0138] The initial population size is N=50.

[0139] ;

[0140] Where: r i，j (j=1,2,3,4) is a random number between [0,1].

[0141] S312, Individual Fitness Assessment

[0142] Dynamic performance simulation and annual operation simulation estimation were performed for each individual in the population to obtain three objective functions f1, f2, and f3 for each individual in the population.

[0143] Specifically, dynamic performance simulation refers to substituting the grid-connected characteristic parameters of individuals in the population into the grid-connected energy storage microgrid system model, setting a 20% load surge to simulate extreme disturbances, recording and calculating each sub-index in the objective function f1, and obtaining the value of the objective function f1 for that individual population.

[0144] Specifically, annual operation simulation estimation refers to selecting several representative operation days using a typical daily scenario clustering method, simulating daily operation with a step size of 1 hour, recording data such as frequency deviation, voltage deviation, load shedding power, energy storage charging and discharging power, state of charge, tie line switching power, and common coupling point voltage for each typical day and time period, and calculating and obtaining the values ​​of the individual objective functions f2 and f3 for the population.

[0145] Feasibility assessments are performed for each individual in the population, and data at different time periods are checked to see if they exceed limits.

[0146] ;

[0147] Where: Feasible(X) i ) is a feasibility function for an individual in the population. A value of 1 indicates that all constraints are met, while a value of 0 indicates that the constraints are exceeded, and the individual is discarded.

[0148] S313, Multi-target Domination Sorting and External File Initialization

[0149] A fast non-dominated sort is performed on the initially generated population. Individuals are stratified according to the feasibility of the constraints in step S2 and the dominance relationships of the objective functions f1, f2, and f3, resulting in a series of frontiers, such as the first frontier and the second frontier. The first frontier consists of all feasible and non-dominated individuals in the current population, representing the non-dominated solution set that is optimal in terms of dynamic performance, power balance, and economic cost. The first frontier is stored in external file A, completing the external file initialization.

[0150] Specifically, the leader selection method in step S33 is as follows:

[0151] For each objective function f k , Sort all solutions in file A in ascending order according to the objective value. For each solution in file A, calculate its crowding distance d in the three-objective space. j :

[0152]

[0153] in: , These are the values ​​of the previous and next neighbors of the j-th individual after sorting, respectively, on the k-th target. , Let be the maximum and minimum values ​​of the k-th target in the file.

[0154] Using a roulette wheel game with inverse crowding ratio, select two different individuals X from file A. B1 and X B2 :

[0155] Individual X j The probability of being selected is:

[0156]

[0157] Where: ε is a small positive number to avoid the denominator being zero; |A| is the number of individuals in the current external archive.

[0158] Specifically, the formula for calculating the satiety rate in step S34 is as follows:

[0159]

[0160] Where: t is the current iteration number; T max is the maximum number of iterations; r1 is a random number uniformly distributed between 0 and 1; z is a parameter that changes linearly with iteration, decreasing linearly from the initial value of 0 to the final value of -2; w is a constant, taken as 2.5, used to control the fluctuation range.

[0161] Specifically, the individual location update method in step S35 is as follows:

[0162] S351. Identify Reference Individuals

[0163] The current reference solution is randomly selected with probabilities L1=0.8 and L2=0.2:

[0164] , where r2 is a uniformly random number between 0 and 1.

[0165] S352, Exploration Phase

[0166] Each time an individual is updated, a random number r3 is generated, whose value follows a uniform distribution within the interval [0,1], and is used for strategy selection. If the random number r3 < 0.6, the exploration phase begins.

[0167] Generate random numbers r4∈[0,1] for strategy selection, then the candidate solutions X generated in this stage new for:

[0168] , where: X ref The current reference individual's position vector; F is the satiety rate; X i r5 and r6 are mutually independent uniform random numbers between 0 and 1; l b u b These are the lower and upper bound vectors for the network characteristic parameters, respectively.

[0169] S353, Development Phase

[0170] If the random number r3 ≥ 0.6, then proceed to the development stage.

[0171] If the satiety rate |F| ≥ 0.5, then proceed to the first sub-stage. Generate random numbers r7 ∈ [0, 1] for sub-strategy selection.

[0172] When the random number r7 < 0.5, a food competition strategy is executed, and the candidate solutions X generated in this stage are... new Update as follows:

[0173] , where r8 is a uniform random number between 0 and 1.

[0174] When the random number 0.5 ≤ r7 ≤ 1, a rotational flight strategy is executed, and the candidate solutions X generated in this stage are... new Update as follows:

[0175] Where: r9 is a uniform random number between 0 and 1; S is the spiral radius control parameter, which is set to 0.2.

[0176] If the satiety rate |F| < 0.5, then proceed to the second sub-stage. Generate random number r. 10 ∈[0,1], used for sub-policy selection.

[0177] When the random number r 10 When the value is less than 0.5, a food source accumulation strategy is implemented, and the candidate solutions X generated in this stage are... new Update as follows:

[0178]

[0179] When the random number 0.5 ≤ r 10 When the value is ≤1, a fierce competition strategy is implemented, and the candidate solutions X generated in this stage are... new Update as follows:

[0180] , where levy(d) is a random vector following a Lévy distribution with shape parameter β=1.5.

[0181] Specifically, the method for in-depth exploration in step S36 is as follows:

[0182] For each new individual X new Generate random number r 11 ∈[0,1]. If r 11 If the value is less than 0.5, then the following deep refinement steps will be performed on the individual.

[0183] S361, Generate mean perturbation points

[0184] Three distinct individuals X are randomly selected from the current population. r1 X r2 X r3 Calculate their mean x avg =(X r1 +X r2 +X r3 ) / 3. And construct the first candidate solution X. new1 :

[0185] , where: r 12 A uniformly random number between 0 and 1.

[0186] S362, Adaptive Perturbation Generation

[0187] Dynamically adjust the search radius and set the attenuation coefficient w. local =exp(-rand·t / T max (rand is an independent random number between 0 and 1, used to control the decay rate). Then generate another random number r. 13 ∈[0,1], the second candidate solution X is generated according to the following bifurcation rule. new2 :

[0188] , where: r 14 r 15 r 16 These are mutually independent uniform random numbers between 0 and 1.

[0189] S363, Perform secondary refining

[0190] Generate random number r 17 ∈[0,1]. If r 17 <w local Then for X new2 Perform the development phase operation again to obtain the third candidate solution X. new3 .

[0191] S364, Non-dominated screening

[0192] Calculate X new2 and X new3 Compare the three objective function values ​​with the original X. new The dominance relationship between them. Add all non-dominant solutions of the three to the temporary child set Q. t .

[0193] Specifically, the method for updating external files in step S37 is as follows:

[0194] For the temporary child set Q tFor each new individual, accurately evaluate its three objective function values;

[0195] Parental population P t Merge with all offspring individuals to form R t =P t ∪Q t For R t Performing a fast nondominated sort yields several frontiers.

[0196] Starting from the first frontier, individuals are selected into P frontier by frontier. t+1 This continues until the total number of selected individuals reaches N. If adding a certain frontier would result in the total number of individuals exceeding N, then the crowding distance between all individuals at that frontier is calculated, and individuals with higher crowding are selected first to maintain the uniformity of the solution set distribution.

[0197] R t All non-dominated solutions are stored in external file A;

[0198] The updated external file A is re-sorted non-dominated, retaining only the new first frontier; if the file size |A|>N, the grid mechanism is enabled: the target space is divided into equidistant grids in each dimension, the number of individuals in each hypercube is counted, and the individuals in the most crowded grid are deleted in a loop until the file size meets the set value.

[0199] To verify the effectiveness of the method proposed in this invention, a typical microgrid system at the edge of a mountainous area was used as the simulation object. A simulation model of a grid-type energy storage microgrid was constructed, and the grid characteristic parameters were optimized based on the artificial vulture algorithm.

[0200] Reference Figure 1 Based on the topology shown, a simulation model is established including a grid-type energy storage converter, an equivalent power source of photovoltaic / wind power, and a 400V local load. For this model, the four-dimensional decision vector defined in step S1 is used as the optimization variable, and the upper and lower limits of each variable are set as follows: .

[0201] The artificial vulture algorithm proposed in step S3 is used for parameter optimization. In this embodiment, the population size N=8 and the maximum number of iterations T are set. max =10. In each iteration, the program automatically loads the network parameters to be evaluated into the Simulink model and sets an extreme disturbance of 20% load increase to simulate the most unfavorable working condition. The fitness value is calculated using the formula described in step S2, and individuals are marked as feasible according to the constraints set in step S2.

[0202] Through multiple iterations of optimization, the Artificial Vulture algorithm can quickly converge to a set of optimal network characteristics parameters. The convergence process and parameter optimization trajectory of the algorithm are as follows: Figure 4 and Figure 5 As shown. Figure 4 This indicates that the optimal fitness value tends to stabilize when iterating to the 3rd generation, demonstrating the algorithm's strong global search capability and fast convergence speed. Figure 5 The study shows the changing trends of the four decision variables with the number of iterations, eventually stabilizing at a relatively optimal equilibrium state, indicating that the coupling relationship between the parameters is effectively coordinated.

[0203] This invention also provides a multi-objective optimization device for a grid-connected energy storage microgrid that takes into account grid-connected characteristic parameters, for executing the method described above, including:

[0204] The decision variable establishment module is used to obtain the system operating parameters of the grid-connected energy storage microgrid, and establish grid characteristic parameter decision variables based on the system operating parameters: virtual inertia time constant H, active power-frequency droop coefficient Dp, reactive power-voltage droop coefficient D q and virtual damping coefficient D d Together, they serve as optimization decision variables, forming a four-dimensional network characteristic parameter decision vector X=[H, D] p D q D d ] T ;

[0205] The optimization model building module is used to establish a multi-objective optimization model for the grid-connected energy storage microgrid based on the grid-connected characteristic parameter decision vector: with the objectives of optimizing the dynamic performance of the grid-connected energy storage microgrid system, maximizing the ability to balance power fluctuations, and minimizing the economic cost of system operation, dynamic performance objective function f1, dynamic power balance function f2, and economic objective function f3 are established respectively, and constraints including power balance constraints, energy storage operation constraints, power exchange and voltage constraints, new energy output constraints, and grid-connected characteristic parameter boundary constraints are set.

[0206] The optimization solution module is used to solve the multi-objective optimization model of the grid-connected energy storage microgrid using the artificial vulture multi-objective optimization algorithm to obtain the Pareto optimal solution set: taking the four-dimensional grid-connected characteristic parameter decision vector X as the optimization variable, minimizing the dynamic performance objective function f1, the dynamic power balance function f2, and the economic objective function f3 as the optimization direction, and judging the feasibility of individuals based on the constraints, the three-objective Pareto optimal solution set is obtained.

[0207] This invention has the following features and effects:

[0208] 1. Fully leverage the frequency and voltage support potential of grid-based energy storage.

[0209] By using network characteristic parameters as optimization variables for global optimization, the system's dynamic performance indicators, such as the lowest frequency point, frequency change rate, and steady-state voltage deviation, under extreme disturbances such as load mutations are significantly better than empirical fixed parameter schemes, effectively improving the disturbance resistance capability of microgrids at the end of weak power grids.

[0210] 2. Achieve a balance between dynamic security, power supply reliability, and operational economy.

[0211] The three-objective synergistic optimization overcomes the problem of neglecting one aspect for another in single-objective optimization. Under the premise of ensuring the dynamic safety of the system, it can take into account the annual source-load power balance capability (reducing frequency / voltage overruns and load shedding) and the operational economy (reducing annual operating costs, wind and solar curtailment rates and energy storage lifespan losses), and finally output multiple balanced and feasible Pareto optimal configuration schemes for engineering decision-making.

[0212] 3. High solution accuracy and fast convergence speed

[0213] The improved artificial vulture multi-objective optimization algorithm achieves a good balance between convergence speed and population diversity through three mechanisms: leader sparsity selection, adaptive exploration-exploitation regulation, and local in-depth refinement. It can efficiently obtain a uniformly distributed Pareto optimal front for three objectives, and the algorithm is robust and suitable for high-dimensional nonlinear optimization problems.

[0214] 4. The project has significant practical value.

[0215] This invention targets the actual operation scenarios of microgrids at the end of weak power grids, such as remote mountainous areas and islands. The optimization process covers dynamic performance simulation and year-round operation simulation. The results can directly guide the parameter tuning of grid-connected energy storage converters, providing a scientific decision-making basis for the planning, design and operation control of grid-connected energy storage microgrids, and has broad prospects for engineering promotion.

[0216] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-objective optimization method for grid-connected energy storage microgrids considering grid-connected characteristic parameters, characterized in that, Includes the following steps: Step S1: Obtain the system operating parameters of the grid-connected energy storage microgrid, and establish grid characteristic parameter decision variables based on the system operating parameters: virtual inertia time constant H, active power-frequency droop coefficient Dp, reactive power-voltage droop coefficient D q and virtual damping coefficient D d Together, they serve as optimization decision variables, forming a four-dimensional network characteristic parameter decision vector X=[H, D] p D q D d ] T ; Step S2: Based on the network characteristic parameter decision vector, establish a multi-objective optimization model for the network-connected energy storage microgrid: with the objectives of optimizing the dynamic performance of the network-connected energy storage microgrid system, maximizing the ability to balance power fluctuations, and minimizing the economic cost of system operation, establish dynamic performance objective function f1, dynamic power balance function f2, and economic objective function f3 respectively, and set constraints including power balance constraints, energy storage operation constraints, power exchange and voltage constraints, new energy output constraints, and boundary constraints of network characteristic parameters; Step S3: The artificial vulture multi-objective optimization algorithm is used to solve the multi-objective optimization model of the grid-connected energy storage microgrid to obtain the Pareto optimal solution set: the four-dimensional grid-connected characteristic parameter decision vector X is used as the optimization variable, and the minimization of the dynamic performance objective function f1, dynamic power balance function f2 and economic objective function f3 is taken as the optimization direction. Based on the constraints, the feasibility of each individual is judged, and the three-objective Pareto optimal solution set is obtained.

2. The multi-objective optimization method for grid-connected energy storage microgrids considering grid-connected characteristic parameters according to claim 1, characterized in that, The dynamic performance objective function f1 is composed of a weighted average of four sub-indices: the minimum frequency, the maximum rate of frequency change, the integral of the absolute value of frequency deviation, and the steady-state voltage deviation. Its expression is: ； in: This is the lowest point of the system frequency; This represents the maximum rate of change of the system frequency. Integral the absolute value of the system frequency deviation; denoted as , where is the steady-state voltage deviation of the system; f0 is the rated frequency of the system; Δω(t) is the time series obtained by integrating the system frequency deviation every 1ms; T is the duration of the disturbance simulation; α1, α2, α3, and α4 are normalized weighting coefficients. The dynamic power balance function f2 is composed of a weighted average of the frequency over-limit accumulation, voltage over-limit accumulation, and power shortage, and its expression is: ； Where: E f E represents the cumulative amount of frequency exceeding the limit. v E represents the cumulative amount of voltage exceeding the limit. shed For lack of power supply; N d Number of typical daily scenes; D m T represents the number of representative days of a typical day; m This represents the set of simulation time periods for a typical day; Δf(t) and ΔV(t) represent the frequency and voltage deviations for each time period of a typical day; δf lim δV lim These represent the allowable normal fluctuation ranges for frequency and voltage, respectively; γ f γ v γ shed P represents the normalized weighting coefficients. shed (t) represents the load shedding power at time t, Δt=1h; The economic objective function f3 is composed of a weighted average of annual operating costs, wind and solar curtailment rates, and energy storage equivalent lifetime loss costs, and its expression is: ； Where: C annual Annual operating cost; η cur C represents the wind and solar curtailment rate. degrad Cost of energy storage equivalent lifetime loss; C m The operating cost for a typical day; E cur,m E avail,m These represent the amount of wind and solar power curtailment on a typical day and the maximum available power of new energy sources; C bat The unit capacity investment cost of the battery; E lifetime E represents the total charge throughput over the battery's entire lifespan. through,m represents the total charge / discharge throughput of energy storage on a typical day; w1, w2, and w3 are normalized weighting coefficients.

3. The multi-objective optimization method for grid-connected energy storage microgrids considering grid-connected characteristic parameters according to claim 2, characterized in that, Step S3 employs the Artificial Vulture multi-objective optimization algorithm to solve the multi-objective optimization model of the grid-connected energy storage microgrid, specifically including the following sub-steps: Step S31: Initialize the population and algorithm parameters, including population size N, external file A, and maximum number of iterations T. max ; Step S32: Determine if the maximum number of iterations has been reached. If so, proceed to step S38 and output the external file A as the final Pareto optimal solution set; otherwise, proceed to step S33. Step S33: Select two leaders X from external file A using a crowding-inverse roulette wheel. B1 and X B2 ; Step S34: Calculate the satiety rate for adaptive adjustment of the exploration and development ratio; Step S35: Perform position update for each individual in the current population to generate a new generation of candidate solutions, and calculate the objective value and feasibility label of each candidate solution based on the objective function and constraints of step S2; Step S36: Apply a random deep exploration perturbation to the candidate solutions to generate derived solutions and uniformly perform non-dominated screening to enhance local convergence ability; Step S37: Merge the parent and child populations, perform fast non-dominated sorting and crowding screening to generate the next generation population, while updating the external archive and maintaining a uniform distribution of the solution set, and return to step S32. Step S38: Output the set of three-objective Pareto optimal solutions from external file A for selecting a compromise configuration.

4. The multi-objective optimization method for grid-connected energy storage microgrids considering grid characteristic parameters according to claim 3, characterized in that, Population generation in step S31 includes the following steps: S311. Population Initialization: Using the network characteristic parameters as the position vectors of individuals in the population, each component is uniformly and randomly generated under the constraints of the network characteristic parameters in step S2. The expression is as follows: The initial population size is N=50; ； Where: r i，j (j=1,2,3,4) is a random number between [0,1]; S312. Individual fitness assessment: Dynamic performance simulation and annual operation simulation estimation are performed for each individual in the population to obtain three objective functions f1, f2, and f3 for each individual in the population. For each individual in the population, a feasibility assessment is performed to check whether the data in each time period exceeds the limits: ； Where: Feasible(X) i ) is the feasibility function for an individual in the population. A value of 1 indicates that all constraints are met, while a value of 0 indicates that the constraints are exceeded, and the individual is discarded. S313, Multi-objective dominance sorting and external file initialization: Perform fast non-dominated sorting on the initially generated population, and stratify individuals according to the constraint feasibility of step S2 and the dominance relationship of objective functions f1, f2, and f3 to obtain a series of frontiers including the first frontier and the second frontier. The first frontier consists of all feasible and non-dominated individuals in the current population, representing the non-dominated solution set that is optimal in terms of dynamic performance, power balance, and economic cost. Store the first frontier in the external file A to complete the external file initialization.

5. The multi-objective optimization method for grid-connected energy storage microgrids considering grid-connected characteristic parameters according to claim 3, characterized in that, The leader selection method in step S33 is as follows: For each objective function f k , Sort all solutions in file A in ascending order according to the objective value. For each solution in file A, calculate its crowding distance d in the three-objective space. j : ； in: , These are the values ​​of the previous and next neighbors of the j-th individual after sorting, respectively, on the k-th target. , Let these be the maximum and minimum values ​​of the k-th target in the file; Using a roulette wheel game with inverse crowding ratio, select two different individuals X from file A. B1 and X B2 : Individual X j The probability of being selected is: ； Where: ε is a small positive number to avoid the denominator being zero; |A| is the number of individuals in the current external archive.

6. The multi-objective optimization method for grid-connected energy storage microgrids considering grid characteristic parameters according to claim 3, characterized in that, The formula for calculating satiety rate in step S34 is: ； Where: t is the current iteration number; T max is the maximum number of iterations; r1 is a random number uniformly distributed between 0 and 1; z is a parameter that changes linearly with iteration, decreasing linearly from the initial value of 0 to the final value of -2; w is a constant, taken as 2.5, used to control the fluctuation range.

7. The multi-objective optimization method for grid-connected energy storage microgrids considering grid characteristic parameters according to claim 3, characterized in that, The method for updating the individual location in step S35 is as follows: S351. Determine the reference individual: Randomly select the current reference solution with probabilities L1=0.8 and L2=0.

2. , where r2 is a uniformly random number between 0 and 1; S352, Exploration Phase: A random number r3 is generated each time an individual is updated. Its value follows a uniform distribution within the interval [0,1] and is used for strategy selection. If the random number r3 < 0.6, the exploration phase begins. Generate random numbers r4∈[0,1] for strategy selection, then the candidate solutions X generated in this stage new for: , where: X ref The current reference individual's position vector; F is the satiety rate; X i r5 and r6 are mutually independent uniform random numbers between 0 and 1; l b u b These are the lower and upper bound vectors for the network characteristic parameters, respectively; S353, Development Phase: If the random number r3 ≥ 0.6, then proceed to the development phase; If the satiety rate |F| ≥ 0.5, then proceed to the first sub-stage. Generate random numbers r7 ∈ [0, 1] for sub-strategy selection; When the random number r7 < 0.5, a food competition strategy is executed, and the candidate solutions X generated in this stage are... new Update as follows: Where: r8 is a uniformly random number between 0 and 1; When the random number 0.5 ≤ r7 ≤ 1, a rotational flight strategy is executed, and the candidate solutions X generated in this stage are... new Update as follows: Where: r9 is a uniformly random number between 0 and 1; S is the spiral radius control parameter, which is set to 0.2; If the satiety rate |F| < 0.5, then proceed to the second sub-stage and generate a random number r. 10 ∈[0,1], used for sub-policy selection; When the random number r 10 When the value is less than 0.5, a food source accumulation strategy is implemented, and the candidate solutions X generated in this stage are... new Update as follows: ； When the random number 0.5 ≤ r 10 When the value is ≤1, a fierce competition strategy is implemented, and the candidate solutions X generated in this stage are... new Update as follows: , where levy(d) is a random vector following a Lévy distribution with shape parameter β=1.

5.

8. The multi-objective optimization method for grid-connected energy storage microgrids considering grid characteristic parameters according to claim 3, characterized in that, The method for in-depth exploration in step S36 is as follows: for each new individual X new Generate random number r 11 ∈[0,1], if r 11 If the value is less than 0.5, then the following deep refinement steps will be performed on the individual: S361. Generate mean perturbation points: Randomly select three distinct individuals X from the current population. r1 X r2 X r3 Calculate their mean x avg =(X r1 +X r2 +X r3 ) / 3, and construct the first candidate solution X new1 : , where: r 12 A uniformly random number between 0 and 1; S362, Adaptive Disturbance Generation: Dynamically adjusts the search radius and sets the attenuation coefficient w. local =exp(-rand·t / T max ), where rand is an independent random number between 0 and 1, used to control the decay rate; then generate a random number r. 13 ∈[0,1], the second candidate solution X is generated according to the following bifurcation rule. new2 : , where: r 14 r 15 r 16 These are mutually independent uniform random numbers between 0 and 1; S363. Perform secondary refinement: Generate random number r 17 ∈[0,1], if r 17 <w local Then for X new2 Perform the development phase operation again to obtain the third candidate solution X. new3 ; S364. Non-dominated screening: Calculate X new2 and X new3 Compare the three objective function values ​​with the original X. new The dominance relationship between them is used to add all non-dominant solutions of the three to the temporary child set Q. t .

9. The multi-objective optimization method for grid-connected energy storage microgrids considering grid-connected characteristic parameters according to claim 3, characterized in that, The method for updating external files in step S37 is as follows: For the temporary child set Q t For each new individual, accurately evaluate its three objective function values; Parental population P t Merge with all offspring individuals to form R t =P t ∪Q t , for R t Performing a fast nondominated sort yields several frontiers; Starting from the first frontier, individuals are selected into P frontier by frontier. t+1 This continues until the total number of selected individuals reaches N. If adding a certain frontier would result in the total number of individuals exceeding N, then the crowding distance between all individuals at that frontier is calculated, and individuals with higher crowding are selected first to maintain the uniformity of the solution set distribution. R t All non-dominated solutions are stored in external file A; The updated external file A is re-sorted non-dominated, retaining only the new first frontier; if the file size |A|>N, the grid mechanism is enabled: the target space is divided into equidistant grids in each dimension, the number of individuals in each hypercube is counted, and the individuals in the most crowded grid are deleted in a loop until the file size meets the set value.

10. A multi-objective optimization device for grid-connected energy storage microgrids considering grid-connected characteristic parameters, used to execute the method according to any one of claims 1-9, characterized in that, include: The decision variable establishment module is used to obtain the system operating parameters of the grid-connected energy storage microgrid, and establish grid characteristic parameter decision variables based on the system operating parameters: virtual inertia time constant H, active power-frequency droop coefficient Dp, reactive power-voltage droop coefficient D q and virtual damping coefficient D d Together, they serve as optimization decision variables, forming a four-dimensional network characteristic parameter decision vector X=[H, D] p D q D d ] T ; The optimization model building module is used to establish a multi-objective optimization model for the grid-connected energy storage microgrid based on the grid-connected characteristic parameter decision vector: with the objectives of optimizing the dynamic performance of the grid-connected energy storage microgrid system, maximizing the ability to balance power fluctuations, and minimizing the economic cost of system operation, dynamic performance objective function f1, dynamic power balance function f2, and economic objective function f3 are established respectively, and constraints including power balance constraints, energy storage operation constraints, power exchange and voltage constraints, new energy output constraints, and grid-connected characteristic parameter boundary constraints are set. The optimization solution module is used to solve the multi-objective optimization model of the grid-connected energy storage microgrid using the artificial vulture multi-objective optimization algorithm to obtain the Pareto optimal solution set: taking the four-dimensional grid-connected characteristic parameter decision vector X as the optimization variable, minimizing the dynamic performance objective function f1, the dynamic power balance function f2, and the economic objective function f3 as the optimization direction, and judging the feasibility of individuals based on the constraints, the three-objective Pareto optimal solution set is obtained.

Images