A Cooperative Planning Method for Energy Storage Capacity of Energy Bases Based on Copula-Multi-Objective Optimization

By characterizing the correlation between wind and solar power output and the dynamic evolution of SOC through the Copula function, and combining a multi-objective optimization model and a hybrid evolutionary algorithm, the problem of balancing economy, absorption capacity and grid connection volatility in energy storage configuration is solved, and efficient energy storage capacity planning is achieved.

CN123073786APending Publication Date: 2026-11-10华电(海西)新能源有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611455902.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-18
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing energy storage capacity planning methods do not accurately depict the dependent structure of wind and solar power output in scenario modeling, and do not simultaneously consider the dynamic evolution constraints of SOC and the power-capacity coupling constraints. The solution algorithm is prone to getting trapped in local optima and lacks an adaptive balancing mechanism, resulting in poor economic efficiency of energy storage configuration schemes, insufficient new energy absorption capacity, and large fluctuations in grid-connected power.

Method used

A three-objective collaborative planning model is constructed by adopting a copula function-based model of wind and solar power output joint scenario, combined with the dynamic evolution of SOC of energy storage system and power-capacity coupling constraints. The model is solved by the hybrid evolutionary algorithm of NSGA-II and MOPSO, and the iterative process is adjusted by an adaptive hybrid ratio strategy. Finally, the optimal model is selected by fuzzy decision-making.

Benefits of technology

It improved the accuracy of wind and solar power output correlation characterization, optimized the economy of energy storage configuration and the renewable energy absorption rate, reduced grid connection volatility, achieved a dynamic balance between global exploration and local development, and improved the operating performance of the energy storage system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123073786A_ABST
    Figure CN123073786A_ABST
Patent Text Reader

Abstract

This invention discloses a collaborative planning method for energy storage capacity in energy bases based on Copula-multi-objective optimization. It employs kernel density estimation to fit the marginal distributions of wind and solar power output, connects them via a Copula function to obtain a joint probability distribution model, generates typical daily power output scenarios using the conditional distribution method, and compresses the scenario scale using the Kantorovich distance backward reduction algorithm to obtain a set of typical scenarios with probability weights. Using the rated power and capacity of the energy storage system as decision variables, and taking the lowest annual value of the total life-cycle cost, the highest renewable energy absorption rate, and the lowest grid connection volatility as optimization objectives, a multi-objective collaborative planning model is constructed, considering the dynamic evolution constraints of energy storage SOC, power-capacity coupling constraints, and feasible region constraints. The Pareto front is solved using a multi-objective hybrid evolutionary algorithm, and the optimal energy storage configuration is determined through fuzzy decision-making. This invention is applicable to energy storage planning for large-scale renewable energy bases with capacities of tens of millions of kilowatts or more.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of power system planning technology, specifically involving the optimized configuration of energy storage capacity in large-scale new energy bases, and particularly involving a collaborative planning method for energy storage based on Copula function scenario modeling and multi-objective optimization algorithms. Background Technology

[0002] my country's planned installed capacity of new energy bases in desert and Gobi areas has exceeded 200 million kilowatts. The output of wind power and photovoltaic power is highly random and intermittent, and the energy storage configuration scheme has a significant impact on the economic benefits and operational performance of the bases.

[0003] Copula functions are commonly used tools for characterizing multivariate correlations. Gumbel-Copula excels at capturing upper-tail dependencies, Clayton-Copula focuses on lower-tail dependencies, and Frank-Copula exhibits tail symmetry (both upper and lower tail dependency coefficients are equal to 0), making it more suitable for modeling correlations in non-extreme intervals. However, existing planning models often employ single-objective or bi-objective optimization, rarely incorporating the life-cycle cost of energy storage, renewable energy absorption rate, and grid connection volatility into the objective system simultaneously. This results in a difficulty in achieving a balance between economic efficiency, absorption capacity, and grid connection quality in the optimization results. Most models also fail to simultaneously incorporate SOC dynamic evolution constraints and power-capacity coupling constraints, making it difficult to reflect the true operating characteristics of energy storage systems. Regarding solution algorithms, single NSGA-II or MOPSO algorithms are prone to premature convergence in high-dimensional decision spaces, and existing hybrid strategies lack adaptive adjustment mechanisms, making it difficult to maintain a balance between global exploration and local development.

[0004] To address the aforementioned issues, a collaborative planning method for energy storage capacity is needed that can accurately characterize the correlation between wind and solar power output, fully consider the operational characteristics of energy storage, and effectively solve multi-objective optimization problems. Summary of the Invention

[0005] The technical problem solved by this invention is that existing energy storage capacity planning methods do not accurately depict the wind and solar power output dependent structure in scenario modeling. The planning model does not simultaneously take into account the dynamic evolution constraints of SOC and the power-capacity coupling constraints. The solution algorithm is prone to getting trapped in local optima and lacks an adaptive balancing mechanism, resulting in poor economic efficiency of energy storage configuration schemes, insufficient new energy absorption capacity, and large fluctuations in grid-connected power.

[0006] To address the aforementioned technical problems, this invention proposes a collaborative planning method for energy storage capacity in large-scale new energy bases based on Copula multi-objective optimization, comprising the following steps: S1. Copula-based joint scene modeling of wind and solar power output Collect historical wind power output data from major new energy bases and historical photovoltaic power output data ,in For the sample size, and The first Wind power and photovoltaic output at each sampling point (unit: MW).

[0007] After data cleaning, normalize using the following formula: ; In the formula, and These are the per-unit values ​​(dimensionless) of wind power and solar power output, respectively. and These are the rated installed capacities of wind farms and photovoltaic power plants, respectively (unit: MW).

[0008] After removing outliers using the KS test, the kernel density is estimated and the marginal distribution is fitted using the following formula: ; In the formula, For point The kernel density estimate at that location; For sample size; This is a bandwidth parameter that controls the smoothness of the estimation. Let be a Gaussian kernel function, satisfying ; For the first Individual sample observations. Bandwidth parameter. Adaptive selection using Silverman rules: ,in The standard deviation of the sample is 1. The interquartile range is the sample interval.

[0009] According to Sklar's theorem, the edge distribution functions of wind power and photovoltaic power can be connected into a joint distribution function using the Copula function: ; The probability integral transform is: , , and These are the cumulative distribution functions of wind power and solar power output, respectively; For Copula functions, These are the relevant parameters.

[0010] Copula functions can be selected from various types, including Frank-Copula, Gaussian-Copula, and Clayton-Copula. Taking Frank-Copula as an example, its function form is as follows: ; In the formula, and These are the uniform variables of wind power and photovoltaic power output after probability integral transformation, respectively. These are parameters related to the Frank-Copula. The Frank-Copula exhibits tail symmetry (upper tail dependence coefficient). Lower tail dependency coefficient This is suitable for characterizing moderate correlation in non-extreme intervals.

[0011] Estimate using maximum likelihood estimation method Then, using Kendall's rank correlation coefficient Verify the correlation between wind and solar power output: ; One of the first-order Debye functions: .when hour, , indicating a negative correlation.

[0012] Copula random samples are generated using the conditional distribution method. about The conditional distribution is: ; Sampling steps: (1) Generate independent uniform random numbers , ,in (2) Solve for the second independent uniform random number used for conditional sampling; The solution is obtained numerically using Newton's iteration method, with the convergence condition being... tolerance Maximum number of iterations (3) Inverse Transformation , A typical daytime scene in the original output space is obtained.

[0013] Finally, the number of scenes is compressed to a preset size using the Kantorovich distance backward reduction algorithm. The Kantorovich distance is defined as:

[0014] in, For the original scene set, This is the reduced set of scenes; For the first The probability weights of each original scene; , and The first In the first scenario Per-unit values ​​of wind and solar power output for each time period This represents the total number of time periods on a typical day. The backward reduction algorithm uses a greedy strategy to iterate until the target number of scenes is reached, resulting in a set of typical scenes with probability weights. ,in For the first A typical scenario's output sequence For its probability weights, This represents the total number of scenes after reduction.

[0015] S2. Construction of a Multi-Objective Collaborative Planning Model for Energy Storage Capacity Based on the rated power of the energy storage system (Unit: kW) and rated capacity (Unit: kWh) is used as the decision variable to establish a three-objective collaborative programming model.

[0016] Objective 1: Annual Value of Total Life Cycle Cost Minimum

[0017] in, The annual value of the total life cycle cost (unit: yuan / year); The annual value of the initial investment. , Cost per unit power of energy storage (RMB / kW). Cost per unit capacity of energy storage (RMB / kWh) Rated energy storage capacity (kW). Rated energy storage capacity (kWh). This is the capital recovery factor. The discount rate is... Project operation cycle (years); To replace the cost of annual value, , Battery cycle life (years); Annual maintenance costs , Annual maintenance fee rate; For retirement and other annual values, , Cost per unit capacity for decommissioning and disposal (RMB / kWh).

[0018] The above formula is a simplified theoretical model of life cycle cost. In actual engineering applications, each cost is calculated accurately using an engineering financial model, which takes into account factors such as construction period interest, equipment input tax deduction, and residual value recovery. The deviation between the calculated value of the simplified formula and the accurate model is within the acceptable range for engineering.

[0019] Objective 2: Renewable energy consumption rate maximum

[0020] in, and The first In the scenario, the first Actual wind and solar power output for each time period (unit: kW); and These are the power of wind curtailment and solar curtailment (unit: kW); The time interval is expressed in hours (h). The numerator represents the actual amount of renewable energy power consumed, and the denominator represents the total renewable energy power that can be generated. A higher consumption rate indicates more efficient utilization of renewable energy.

[0021] Objective 3: Grid-connected volatility Minimum

[0022] in, No. In the scenario, the first Grid-connected power during each time period (unit: kW); This refers to the total installed capacity of the new energy base (unit: kW). This target measures the sum of the absolute values ​​of the changes in grid-connected power over adjacent time periods; the smaller the value, the smoother the grid-connected power.

[0023] Constraints: (1) Power balance constraint: ; in, and These represent the charging and discharging power of energy storage (unit: kW), both non-negative variables. The equation indicates that the total output of new energy sources equals the sum of grid-connected power, energy storage charging power, and curtailed wind and solar power, minus the energy storage discharging power.

[0024] (2) Feasibility region constraints for decision variables: ; in, , These are the lower and upper limits of the rated power of energy storage (unit: kW); , These are the lower and upper limits of the rated capacity of energy storage, respectively (unit: kWh).

[0025] (3) SOC dynamic evolution constraints SOC recurrence equation: ; in, For the first In the scenario, the first State of charge for each time period (per unit, 0-1); and These represent the charging efficiency and discharging efficiency (dimensionless). This equation describes the change of the stored energy state over time.

[0026] SOC upper and lower bound constraints: ,in and These are the lower and upper limits of the state of charge (usually 0.10 and 0.90), respectively, to avoid overcharging and over-discharging.

[0027] SOC initial and final value equality constraint: This ensures that the energy storage system maintains energy balance during each typical daily operating cycle, enabling daily cyclic operation.

[0028] Charge and discharge power upper limit constraints: ; in, It is a binary variable. =1 indicates a charging state. =0 indicates a discharge state. This constraint limits the charging and discharging power to no more than the rated power of the energy storage.

[0029] Charge and discharge mutual exclusion constraint: This ensures that the energy storage device can be charged and discharged simultaneously.

[0030] (4) Energy storage duration constraints: ; in This indicates the rated continuous discharge duration of the energy storage system (in hours). This constraint ensures that the energy storage duration is between 2 and 6 hours, which is in line with the current mainstream engineering practices for electrochemical energy storage systems.

[0031] S3, multi-objective hybrid evolutionary algorithm for solution A multi-objective hybrid evolutionary algorithm is employed to solve the multi-objective collaborative programming model. Taking the NSGA-II and MOPSO hybrid algorithm as an example, NSGA-II is responsible for global exploration through fast non-dominated sorting and crowding distance calculation, performing simulated binary crossover and polynomial mutation to generate offspring. MOPSO is responsible for local fine-grained search through a particle velocity-position update mechanism. The two algorithms alternate iteratively and share non-dominated solutions in an external archive, which maintains the uniformity of the Pareto front distribution through a crowding distance strategy.

[0032] The mathematical expression for the adaptive mixing ratio adjustment strategy is: ; In the formula, For the first The percentage of NSGA-II iterations in a generation; This represents the current iteration number; As the switching point, This represents the maximum number of iterations. The steepness coefficient controls the transition rate; =0.7, =0.3 represents the upper and lower limits of the NSGA-II iteration ratio, respectively. (Preliminary) ≈ =0.7, NSGA-II dominates to expand the global search range; later ≈ =0.3, MOPSO dominates to accelerate local convergence.

[0033] This strategy allows the hybrid algorithm to fully utilize NSGA-II's global exploration capabilities to quickly locate the approximate region of the Pareto front in the early stages of iteration, and leverage MOPSO's local exploration capabilities for a more refined search of the front in the later stages of iteration. Compared to the fixed-ratio hybrid strategy, the adaptive strategy significantly improves both IGD and Spread metrics.

[0034] The iteration terminates when any of the following conditions are met: (1) The number of iterations reaches the preset maximum value. (2) Non-dominated solutions in external archives in continuous No updates were made within the generation, among which (3) The Hypervolume index of the Pareto front in continuous The rate of change within a generation is less than the threshold. , .

[0035] Inequality constraints are incorporated into the objective function using an improved adaptive penalty function method, equality constraints are reduced using variable elimination, and SOC exceeding limits are addressed using a heuristic repair strategy: if Then force set And reduce discharge power; if Then force set And reduce the charging power. Output the Pareto solution set after iterative convergence. ,in The number of non-dominated solutions.

[0036] S4. Fuzzy Decision-Making Comprehensive Optimization For each Pareto solution, fuzzy satisfaction is calculated according to the objective dimension. Minimizing objectives (such as annual LCC values, grid connection volatility) uses a decreasing membership function, while maximizing objectives (such as renewable energy absorption rate) uses an increasing membership function. ; in, For the first The Pareto solution is at the th . Function values ​​on each target; and These are the maximum and minimum values ​​of the objective in the Pareto solution set, respectively. The value ranges from 0 to 1, with values ​​closer to 1 indicating better performance of the target.

[0037] The overall satisfaction rate is calculated as the arithmetic mean. ; in, The target number (in this invention) =3).

[0038] The optimal compromise solution is: ; in This represents the overall satisfaction level of the m-th Pareto solution.

[0039] Select the solution with the highest overall satisfaction Recommended configuration.

[0040] In practical engineering applications, a weighted comprehensive satisfaction level can be introduced based on the decision-maker's preferences. ,in, For the first The weights of each objective and . Beneficial effects

[0041] Compared with existing technologies, this invention uses the Copula function to characterize the wind and solar power output interdependent structure, improving the accuracy of scene modeling; it incorporates the dynamic evolution of SOC and power-capacity coupling constraints into the planning model, so that the configuration results can be directly used for intraday scheduling; through three-objective collaborative optimization, it improves the renewable energy consumption rate and reduces grid connection volatility while controlling costs; and it adopts an adaptive hybrid evolutionary algorithm to achieve a dynamic balance between global exploration and local development, resulting in a solution quality superior to a single algorithm. Attached Figure Description

[0042] Figure 1 This is an overall flowchart of the new energy large-scale base energy storage capacity collaborative planning method based on Copula multi-objective optimization described in this invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0044] Example 1 Taking a large-scale new energy base in a remote desert area in western China as an example, the wind power capacity, photovoltaic capacity, and total installed capacity are detailed in the project parameters. Wind power and photovoltaic output data for this base were collected at 15-minute intervals throughout 2023.

[0045] S1. Copula-based joint scene modeling of wind and solar power output Normalize the historical power output data to per-unit values ​​using the following formula: ; After removing outliers using the KS test, the sample size n = 35040 (365 days × 96 points / day).

[0046] Silverman bandwidth parameter during kernel density estimation =0.042 (wind power) =0.038 (photovoltaics).

[0047] After performing a probability integral transform, the Frank-Copula parameters are estimated using the maximum likelihood estimation method. Corresponding to Kendall rank correlation coefficient The correlation is weakly negative, which is consistent with the complementary pattern of "strong sunlight during the day and large wind power at night" in the "sand and barren" region.

[0048] Frank-Copula, Gaussian-Copula, and Clayton-Copula were used to fit the same set of historical data, and the goodness of fit of each Copula function to the sample data was evaluated by the Kolmogorov-Smirnov (KS) test. The smaller the KS test statistic, the better the fit. Table 1 shows a comparison of the tail characteristics of different Copula functions and their application scenarios. The tail characteristics of Frank-Copula make it more suitable for correlation modeling in non-extreme intervals.

[0049] Table 1. Comparison of goodness of fit for different Copula functions ; The Frank-Copula KS test statistic was the smallest (0.031), significantly better than the other two Copula functions (p<0.05), indicating that the Frank-Copula more accurately characterizes the dependent structure of wind and solar power output in the "Shagohuang" large-scale base during non-extreme periods. This result is consistent with the tail symmetry of the Frank-Copula and also fits the actual statistical characteristics of wind and solar power output in the "Shagohuang" region, where non-extreme periods dominate and extreme periods account for a low proportion.

[0050] N=10000 typical daily scenes are generated using the conditional distribution method, and then compressed using the Kantorovich distance backward reduction algorithm. Table 2 shows the probability weights for each of the following typical scenarios.

[0051] Table 2 Typical Scenarios and Probability Weights ; S2. Construction of a Multi-Objective Collaborative Planning Model for Energy Storage Capacity Energy storage economic parameters: ,

[0052] , , Year, Charge and discharge efficiency State of charge limitation Feasible region of decision variables: MW,

[0053] The capital recovery factor CRF(0.06,25) is calculated to be 0.07823, and the replacement cost sinking fund factor SFF(0.06,10) is calculated to be 0.07587.

[0054] S3, multi-objective hybrid evolutionary algorithm for solution Algorithm parameters: Population size Maximum number of iterations Crossover probability Probability of mutation Cross-distribution index Variation distribution index Inertial weight Learning factor External archive capacity =200. Adaptive blending ratio parameter: .

[0055] Adaptive mixing ratio The evolution process is shown in Table 3.

[0056] Table 3 Adaptive Mixing Ratio Evolutionary examples ; When using Newton's iterative method to solve for the inverse function of a conditional distribution, the initial value is taken as... Convergence tolerance Maximum number of iterations Under typical operating conditions, convergence can be achieved in 3 to 5 iterations.

[0057] After 500 iterations, the solution converged. The convergence criterion was that the external archive had not been updated for 50 consecutive generations. A total of M=50 non-dominated solutions were obtained from the Pareto solution set. Five representative solutions are shown in Table 4.

[0058] Table 4 Representative Pareto Frontier Schemes ; All schemes have an energy storage duration of 4.0 hours, indicating that under the current parameters, 4 hours is the balance between economic efficiency and technical feasibility.

[0059] With option B ( =600MW, Taking a 2400 MWh LCC as an example, the costs of each LCC are shown in Table 5.

[0060] Table 5. Composition of Life Cycle Cost (LCC) for Option B ; Note: The costs in Table 5 are calculated precisely using the engineering financial model, taking into account factors such as construction period interest, equipment input tax deduction, and residual value recovery. There are deviations from the theoretical values ​​of the simplified formula, but these deviations are within the acceptable range for the project.

[0061] S4. Fuzzy Decision-Making Comprehensive Optimization The fuzzy satisfaction assessment of each option is shown in Table 6.

[0062] Table 6. Fuzzy Satisfaction Assessment of Each Scheme ; Option B achieved the highest overall satisfaction (0.724) and is recommended as the optimal configuration. Compared to Option A, Option B significantly improves the renewable energy absorption rate and reduces grid connection volatility while maintaining a manageable increase in cost. The Pareto frontier SpacingMetric is 0.035.

[0063] Example 2 To verify the solution performance of the hybrid algorithm, a comparative experiment was conducted with NSGA-II and MOPSO. The ZDT series of standard multi-objective benchmark functions were used, and each algorithm was run independently 30 times and the statistical average was taken.

[0064] Table 7 Algorithm Performance Comparison Results ; The hybrid algorithm outperforms the two individual algorithms in all three metrics of IGD, Spread, and Spacing Metric, and its average running time falls between the two, thus verifying the computational efficiency of the hybrid strategy.

[0065] To verify the effectiveness of the adaptive hybrid scaling strategy in improving solution quality, a comparative experiment was conducted between the adaptive strategy and the fixed scaling strategy (NSGA-II:MOPSO=5:5), with each strategy running independently 30 times and the statistical average taken. The results are shown in Table 8.

[0066] Table 8 Comparison between Adaptive Strategy and Fixed Proportion Strategy ; The adaptive alternating iteration strategy dynamically adjusts the iteration ratio of NSGA-II and MOPSO, enabling the algorithm to focus on global exploration in the early stages of iteration and local development in the later stages. Compared with the fixed ratio strategy, it improves both IGD and Spread metrics.

Claims

1. A collaborative planning method for energy storage capacity in large-scale new energy bases based on Copula-multi-objective optimization, characterized in that, Includes the following steps: S1. Collect historical wind power output data and historical photovoltaic power output data of new energy bases. After preprocessing, use kernel density estimation to fit their respective marginal distribution functions, and then connect them with the Copula function to obtain a joint probability distribution model. Use the conditional distribution method to generate typical daily power output scenarios, and use the Kantorovich distance backward reduction algorithm to compress the number of scenarios to a preset scale to obtain a set of typical scenarios with probability weights. S2. Using the rated power and rated capacity of the energy storage system as decision variables, and taking the minimum annual value of the total life cycle cost of the energy storage system, the maximum new energy absorption rate, and the minimum grid connection volatility as optimization objectives, and considering power balance constraints, dynamic evolution constraints of energy storage state of charge, mutual exclusion constraints of charging and discharging, energy storage duration constraints, and feasible domain constraints of decision variables, a multi-objective collaborative planning model for energy storage capacity is established. S3. The multi-objective hybrid evolutionary algorithm is used to solve the multi-objective collaborative programming model. In the multi-objective hybrid evolutionary algorithm, the first evolutionary algorithm is responsible for fast non-dominated sorting and global exploration, and the second evolutionary algorithm is responsible for individual updating and local development. The two algorithms alternately run iteratively and share an external archive for storing non-dominated solutions. The uniformity of the Pareto front distribution is maintained through the crowding distance strategy. After iterative convergence, the Pareto solution set is output. S4. Calculate the fuzzy satisfaction level for each Pareto solution according to the objective dimension, select the solution with the highest overall satisfaction level as the optimal compromise solution, and output the recommended configuration values ​​for the rated power and rated capacity of energy storage.

2. The method according to claim 1, characterized in that, The Copula function mentioned in step S1 is the Frank-Copula function, and its mathematical expression is: ; in, The wind power output is a uniform variable after probability integral transformation. Let be the uniform variable of photovoltaic power output after probability integral transformation. For Copula-related parameters and .

3. The method according to claim 1, characterized in that, In step S2, the annualized value of the energy storage system's total life-cycle cost consists of four parts: the annualized value of the initial investment, the annualized value of the replacement cost, the annualized value of the operation and maintenance cost, and the annualized value of the decommissioning disposal. ; in, The annual value is the total life cycle cost. The annual value of the initial investment. ; To replace the cost of annual value, ; Annual maintenance costs ; For retirement and other annual values, .

4. The method according to claim 1, characterized in that, The SOC dynamic evolution constraints mentioned in step S2 include the SOC recursive equation, SOC upper and lower limit constraints, SOC initial and final value equality constraints, and charge / discharge power constraints: SOC recurrence equation: ; SOC upper and lower bound constraints: ; SOC initial and final value equality constraint: ; Charge and discharge power upper limit constraints: ; Charge-discharge mutual exclusion constraint: ; Energy storage duration constraints: 。 5. The method according to claim 1, characterized in that, The multi-objective hybrid evolutionary algorithm mentioned in step S3 is a hybrid algorithm of NSGA-II and MOPSO. NSGA-II serves as the first evolutionary algorithm, undertaking fast non-dominated sorting and global exploration, while MOPSO serves as the second evolutionary algorithm, undertaking particle velocity-position update and local exploration. The alternating iterative operation adopts an adaptive alternating iterative strategy, with the mixing ratio adaptively adjusted according to the iteration process. In the early stage, the first evolutionary algorithm has a higher proportion of iterations to expand the global search range, while in the later stage, the second evolutionary algorithm has a higher proportion of iterations to accelerate local convergence. The solution process adopts the following constraint handling strategies: inequality constraints are incorporated into the objective function using an improved adaptive penalty function method, equality constraints are reduced using variable elimination, and SOC exceeding the limit is addressed using a heuristic repair strategy.

6. The method according to claim 1, characterized in that, The method for calculating fuzzy satisfaction in step S4 is as follows: Normalized satisfaction is calculated for each objective of each Pareto solution, and the minimization objective is adopted... Maximize the target The overall satisfaction rate is calculated as the arithmetic mean. or weighted form The solution with the highest satisfaction level is selected as the recommended configuration.

7. The method according to claim 1, characterized in that, The expression for the kernel density estimator in step S1 is: ; in, For sample size, For bandwidth parameters, This is the Gaussian kernel function.

8. The method according to claim 5, characterized in that, The mathematical expression for the adaptive mixing ratio is: ; in, For the first The percentage of NSGA-II iterations in a generation; As the switching point, The maximum number of iterations, Steepness coefficient; and These are the upper and lower limits of the NSGA-II iteration ratio, respectively.