Addressing optimization method for establishing energy storage system by considering multiple factors

Abstract

An addressing optimization method for establishing an energy storage system by considering multiple factors belongs to the technical field of energy storage optimization configuration, and comprises the following steps: establishing a double-layer optimization configuration model of the energy storage system, the outer layer comprising investment cost and daily operation cost; the inner layer comprises system active power loss and system voltage deviation; establishing a double-layer optimization configuration model constraint condition according to the energy storage charge state; adopting an improved multi-target particle swarm algorithm to obtain the minimum installation capacity and the operation power of each energy storage; and taking the output power as the rated power, and optimizing a configuration scheme with the minimum investment cost as an energy storage system addressing scheme. According to the invention, a multi-strategy improved multi-objective particle swarm optimization (CMOPSO-MSI) is introduced to solve a multi-objective optimization model of hybrid energy storage. Through combination of various strategy improvements, the algorithm aims to overcome a series of problems that premature convergence is in a local non-dominated solution, inertia weight setting lacks guidance and the like, and better Pareto frontier distribution is realized.

Description

Technical Field

[0001] This invention belongs to the field of energy storage optimization configuration technology, and in particular relates to an addressing optimization method for establishing an energy storage system that takes into account multiple factors. Background Technology

[0002] Currently, energy storage systems suffer from short device lifespans and high costs, significantly hindering their large-scale application. Therefore, their configuration planning has become a hot research topic. Optimal configuration of distributed energy storage refers to establishing different mathematical models based on varying needs to determine the location and capacity of distributed energy storage, thus solving the problem of location and capacity determination. Like distributed power sources, the integration of distributed energy storage alters the initial structure of the distribution network, and different locations and capacities will affect the effectiveness of energy storage. Without optimal configuration, distribution network problems may remain unresolved, and even the safety and stability of the power grid may be compromised.

[0003] Regarding the selection of algorithms for optimal configuration, the site selection and capacity determination problem of distributed energy storage falls under the category of multi-objective optimization. As a current research hotspot, its application scenarios and various mathematical models all involve multiple objective functions. Traditional PSO algorithms for solving multi-objective optimization problems typically combine the Pareto dominance method. This method updates the individual optimal solution, the global optimal solution, and the Pareto optimal solution set by evaluating the dominance relationship between particles. However, the original multi-objective optimization particle swarm optimization algorithm faces problems such as premature convergence to local non-dominated solutions, lack of guidance in setting inertia weights, and poor Pareto front distribution due to inappropriate external file update strategies. Furthermore, it involves a large computational load and is difficult to obtain the global optimal solution. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an addressing optimization method for energy storage systems that takes into account multiple factors. The method establishes an addressing optimization model for energy storage systems with the goal of minimizing the average annual cost of energy storage device configuration within the planning period, and considers the efficiency, charge and discharge characteristics, SOC constraints, power balance constraints of the power distribution system, and branch power flow constraints of the energy storage system.

[0005] An addressing optimization method for an energy storage system that considers multiple factors includes the following steps, which are performed sequentially:

[0006] Step 1: Establish a two-layer optimal configuration model for the energy storage system. The outer layer includes investment costs and daily operating costs; the inner layer includes system active power losses and system voltage deviation.

[0007] Step 2: Establish constraints for a two-layer optimization configuration model based on the energy storage state of charge;

[0008] Step 3: Use an improved multi-objective particle swarm optimization algorithm to obtain the minimum installed capacity and operating power of each energy storage system; use the output power as the rated power, and optimize the configuration scheme with the lowest investment cost as the energy storage system addressing scheme.

[0009] The investment cost mentioned in step one is:

[0010]

[0011] In the formula: f1 is the investment cost; n is the number of nodes in the power system; C ip Cost per unit capacity; C i The power cost per unit capacity of an energy storage system; z i This indicates the installation location of the energy storage power station in the energy storage system. The z-node is installed at node i. i It is 1 if it is true, otherwise it is 0.

[0012] The daily operating cost is:

[0013]

[0014] In the formula: f2 is the daily operating cost; C om Operating and maintenance cost per unit of apparent power; P i and Q i These are the active power and reactive power of node i, respectively;

[0015] The active power loss of the system is:

[0016]

[0017] In the formula: f3 is the system active power loss; b is the total number of branches in the system; R k P is the impedance of branch k; k and Q k These represent the active and reactive loads at the end of branch k, respectively; U k This refers to the voltage at the end of the line.

[0018] The system voltage deviation is:

[0019]

[0020] In the formula: f4 is the system voltage deviation; U N This is the rated voltage at the end of the line.

[0021] The constraints described in step two include power generation output constraints, node power constraints, energy storage charge and discharge constraints, and energy storage state of charge constraints.

[0022] Step 3 describes an improved multi-objective particle swarm optimization algorithm that includes the following steps:

[0023] 301: Run the data and set the parameters of the CMOPSO-MSI algorithm; set the population size, grid dimension, and maximum external document capacity; randomly initialize the velocity and position of the population particles; calculate the fitness value of each particle; store the non-dominated solution in the external document according to the Pareto dominance; determine the initial individual optimal value pbest and the global optimal value gbest of the particles; set the number of iterations T = 1.

[0024] 302: Inertia weight in dynamic optimization and update algorithm based on improved strategy of adaptive nonlinear decreasing inertia weight;

[0025] 303: Update particle velocity and position components, and calculate particle fitness values;

[0026] 304: Determine whether a particle needs Gaussian mutation. If the mutation probability is greater than the random number, perform Gaussian mutation on the particle according to the formula; otherwise, the particle position remains unchanged.

[0027] 305: Recalculate fitness values, update individual optimal solutions, maintain external documents based on adaptive grid density and roulette wheel algorithm, and select the global optimal solution from them;

[0028] 306: Determine if the algorithm meets the termination condition. If it does, use the optimal Pareto solution set in the external archive as the final search result; otherwise, proceed to step 302 for the next iteration.

[0029] Through the above design scheme, this invention can bring the following beneficial effects: It provides an addressing optimization method for energy storage systems that considers multiple factors, introducing a multi-strategy improved multi-objective particle swarm optimization algorithm (CMOPSO-MSI) to solve the multi-objective optimization model of hybrid energy storage. By combining multiple strategy improvements, this algorithm aims to overcome a series of problems such as premature convergence to local non-dominated solutions and lack of guidance in setting inertia weights, achieving a better Pareto front distribution. It obtains the optimal Pareto solution set that balances the cost of energy storage configuration with its overall stability. Attached Figure Description

[0030] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0031] Figure 1 This is a schematic diagram of a two-layer optimization configuration model for an energy storage system, which is an addressing optimization method for an energy storage system that takes into account multiple factors according to the present invention.

[0032] Figure 2 This is a schematic diagram of the solution process for a method for optimizing the addressing of an energy storage system that considers multiple factors, according to the present invention.

[0033] Figure 3This is a schematic diagram of an improved particle swarm optimization algorithm for addressing an energy storage system that considers multiple factors, according to the present invention.

[0034] Figure 4 This is a schematic diagram of the improved particle swarm optimization algorithm individual update strategy for an energy storage system addressing optimization method that considers multiple factors, as described in this invention.

[0035] Figure 5 This is a schematic diagram illustrating an improved particle swarm optimization algorithm for adaptive grid congestion estimation, which considers multiple factors in establishing an addressing optimization method for an energy storage system according to the present invention.

[0036] Figure 6 This invention provides an improved particle swarm optimization algorithm for addressing energy storage systems, taking into account multiple factors, and includes an external file update flowchart.

[0037] Figure 7 This is a schematic diagram illustrating the improved particle swarm optimization algorithm's external file size maintenance strategy for an addressing optimization method that considers multiple factors in establishing an energy storage system, as described in this invention.

[0038] Figure 8 This is a schematic diagram of the improved particle swarm optimization algorithm for an energy storage system addressing optimization method that considers multiple factors, as described in this invention. Detailed Implementation

[0039] An addressing optimization method for energy storage systems that considers multiple factors, such as Figures 1-8 As shown, the steps include the following steps, which are performed sequentially:

[0040] Step 1: Establish a two-layer optimal configuration model for the energy storage system. The outer layer includes investment costs and daily operating costs; the inner layer includes system active power losses and system voltage deviation.

[0041] Step 2: Establish constraints for a two-layer optimization configuration model based on the energy storage state of charge;

[0042] Step 3: Use an improved multi-objective particle swarm optimization algorithm to obtain the minimum installed capacity and operating power of each energy storage system; use the output power as the rated power, and optimize the configuration scheme with the lowest investment cost as the energy storage system addressing scheme.

[0043] like Figure 1 The two-layer optimal configuration model of the energy storage system shown below, along with its constraints, are as follows:

[0044] The investment cost in the outer optimization model is:

[0045]

[0046] In the formula: f1 is the investment cost; n is the number of nodes in the power system; C ip Cost per unit capacity; Ci The power cost per unit capacity of an energy storage system; z i This indicates the installation location of the energy storage power station in the energy storage system. The z-node is installed at node i. i The value is 1 if it is not 1, otherwise it is 0. Daily operating costs are:

[0047]

[0048] In the formula: f2 is the daily operating cost; C om Operating and maintenance cost per unit of apparent power; P i and Q i These represent the active power and reactive power of node i, respectively.

[0049]

[0050] In the formula: f3 is the system active power loss; b is the total number of branches in the system; R k P is the impedance of branch k; k and Q k These represent the active and reactive loads at the end of branch k, respectively; U k This represents the voltage at the end of the line.

[0051] The voltage deviation is:

[0052]

[0053] In the formula: f4 is the system voltage deviation; U N This is the rated voltage at the end of the line.

[0054] Unlike State of Energy (SOE), a more refined consideration of the energy storage state is needed when balancing economic and technical considerations. Battery energy storage is limited by its chemical characteristics; internal resistance exists, preventing 100% energy release. Therefore, the State of Charge (SOC) is used. The ratio of remaining battery energy to the total system energy is defined as the percentage of remaining charge (SOC). The charge / discharge characteristics of an energy storage system battery are determined by factors such as battery capacity, battery power exchange, charging and discharging efficiency, battery self-discharge rate, and battery SOC. The SOC of energy storage during charging:

[0055]

[0056] During discharge:

[0057]

[0058] In the formula: SOC(t) and SOC(t-1) represent the state of charge of the energy storage system at times t and t-1, respectively; Q represents the capacity of the energy storage battery; ΔT is the time interval; η d η is the self-discharge rate of the energy storage battery. cFor battery charging efficiency; η f This refers to the battery's discharge efficiency.

[0059] The constraints are as follows:

[0060] 1. Power generation output constraints

[0061] 0≤P W ≤P Wmax

[0062] In the formula: P Wmax This represents the maximum constraint value for wind power output.

[0063] 2. Node power constraints

[0064]

[0065] In the formula: P Wi Q Wi The active and reactive power supplied to the i-node power supply. Bi The active power is compensated for by the energy storage of node i. fi Q fi For the active and reactive loads of the node. U i U j , i and j are the voltage amplitudes at nodes i and j, respectively; Y is the system branch admittance matrix.

[0066] 3. Energy storage charging and discharging constraints

[0067] -P Bmax ≤P B (t)≤P Bmax

[0068] In the formula: P Bmax The maximum charge and discharge power limited by the battery.

[0069] 4. Energy storage state of charge constraints

[0070] SOC min ≤SOC(t)≤SOC max

[0071] Where: SOC min and SOC max These represent the upper and lower limits of the battery's state of charge.

[0072] like Figures 2-8 As shown,

[0073] The steps of the algorithm to solve the model are as follows:

[0074] (1) Run the data and set the parameters of the CMOPSO-MSI algorithm. Set the population size, grid dimension, and maximum external document capacity. Randomly initialize the velocity and position of the population particles. Calculate the fitness value of each particle. Store the non-dominated solution in the external document according to the Pareto dominance. Determine the initial individual optimal value pbest and the global optimal value gbest of the particles. Let the number of iterations T = 1.

[0075] (2) The inertial weight in the dynamic optimization and update algorithm is improved by an adaptive nonlinear decreasing strategy based on the inertial weight.

[0076] (3) Update the particle velocity and position components and calculate the particle fitness value.

[0077] (4) Determine whether the particle needs Gaussian mutation. If the mutation probability is greater than the random number, perform Gaussian mutation on the particle according to the formula. Otherwise, the particle position remains unchanged.

[0078] (5) Recalculate the fitness value, update the individual optimal solution, maintain the external document according to the adaptive grid density and roulette wheel algorithm, and select the global optimal solution from it.

[0079] (6) Determine whether the algorithm meets the termination condition. If it does, take the optimal Pareto solution set in the external file as the final search result; otherwise, go to step (2) for the next iteration.

[0080] The specific improvements are as follows: This invention adopts an individual optimal update strategy:

[0081] An improved individual optimality update strategy is adopted to determine the initial individual optimal value pbest and the global optimal value gbest of particles. In multi-objective particle swarm optimization, the individual optimal solution reflects the best solution found by each particle in its search history. By continuously updating the individual optimal solutions, particles can be guided to move towards the optimal solution space, and the diversity of the population can be maintained during the search process. The Pareto dominance rule is used to update the individual optimality. In multi-objective minimization optimization problems, for individuals x1 and x2 ∈ X, if all objective values ​​of individual x1 are no worse than those of x2, then individual x1 is said to Pareto dominate individual x2.

[0082] The update strategy using the individual optimal solution Pbest is as follows: For each individual, calculate its different objective values ​​and compare them with the individual optimal solution. If all objective values ​​of the current individual x are no worse than the individual optimal objective value, i.e., the current individual Pareto dominates the individual optimal solution, then update the individual optimal solution to the current individual position. If the current particle's position is constrained by its historical optimal solution, and this optimal solution occupies a dominant position in the multi-objective space, then keep the historical optimal solution unchanged, i.e., do not update it. If the current individual's position and the individual optimal solution are in a non-dominant relationship, i.e., they do not dominate each other, then a new individual optimal solution is randomly selected. The individual optimal update strategy process is as follows: Figure 3 and Figure 4 As shown.

[0083] This invention employs an adaptive nonlinear decreasing strategy for inertia weights:

[0084] The specific method for inertia weight in the dynamic optimization and update algorithm based on the adaptive nonlinear decreasing improvement strategy of inertia weight is as follows:

[0085] In multi-objective particle swarm optimization (PSO) algorithms, the inertia weight ω plays a crucial role in balancing global and local search, controlling escape capabilities, and ensuring stable convergence, significantly impacting the algorithm's performance and effectiveness. In the initial iteration phase, a larger weight value is necessary to extend the global search time and enhance global search capability. Conversely, in the later stages of iteration, a smaller weight value is required to extend the local search time and strengthen the search capability.

[0086] Based on this, a nonlinear improvement strategy for adaptive inertia weight is proposed to optimize the inertia weight in the algorithm:

[0087]

[0088] In the formula, alpha is the adaptive control parameter. This strategy can further optimize the search effect by adjusting the adaptive control parameter alpha.

[0089] This invention employs a global optimal solution selection strategy:

[0090] The selection of the global optimum has a crucial impact on the search efficiency and convergence of the algorithm. To select a leader particle with the global optimum, guide the algorithm towards that optimum, increase the likelihood of finding a Pareto front solution, and ensure a more even distribution of the final non-dominated solutions on the Pareto front, this paper proposes a global optimum selection strategy based on adaptive grid crowding and the Roulette Wheel Selection algorithm. This strategy ensures that the global optimum is not identical after each iteration.

[0091] The principle of the adaptive grid strategy is to partition the search space. For a multi-dimensional objective space, this requires dividing the external dataset into multiple hypercube spaces. For a two-objective optimization problem, the hyperplane space is a two-dimensional grid, dividing the search space into different parts, each corresponding to a grid cell. The grid is divided according to the size of the multi-dimensional objective, and individuals within each grid cell have the same objective function value. In the grid partitioning method, the global optimum (gbest) can be found through gridded search.

[0092] Step 1: Create an n-dimensional adaptive grid. Determine the partition points for each dimension based on the range of the population's objective function and the given number of grid cells. Each individual can be assigned to a corresponding grid cell, ensuring the grid covers the entire population, resulting in D1*D2*...*D n The grid.

[0093] Step 2: Calculate the upper and lower boundaries of the grid in the target space of the i-th target at time t. As shown in the following formula.

[0094]

[0095] in, These are the maximum and minimum values ​​of the i-th objective function at the t-th iteration, respectively. As the iterations continue, alpha represents the expansion rate, which controls the degree of outward expansion of the mesh boundary, and the mesh boundary is adaptively adjusted accordingly.

[0096] Step 3: Calculate the module length of the network. Wherein, the module length d of the i-th objective function mesh at the k-th iteration is... i as follows:

[0097]

[0098] In the formula, nGrid represents the number of grids.

[0099] Step 4: Traverse all grids and calculate the position in grid space.

[0100]

[0101] In the formula, mode(x / y) is a rounding function that can calculate the number of Pareto optimal solutions contained in each grid.

[0102] Step 5: Calculate the crowding level of particles within the grid. To better quantify the crowding level of particles within the grid, the following formula can be used to calculate the particle crowding. The number of non-dominant solutions within the grid is used as the criterion for the particle distribution crowding level.

[0103]

[0104] In the formula, V is the number of non-occupied solutions in the i-th grid of particle at the t-th iteration; G The ideal number of particles in each cell depends on the population size and the cell size in the solution space; This represents the crowding degree of particle i at the t-th iteration, with a value range of [0,1]; for example... Figure 5 The diagram shows a two-dimensional adaptive grid crowding estimation, where particle p is located at (3,4) in the grid space. If there are no particles in the grid, the crowding is 0.

[0105] Step 6: Calculate the probability of a grid being selected based on particle crowding. The probability of the nth grid being selected in the t-th iteration is:

[0106]

[0107] In the formula, Let be the particle crowding density in the nth grid at the t-th iteration, and nGrid be the number of grids containing particles. The selection probability is calculated by an exponential function (exp(-beta*N)), where beta is an algorithm parameter used to control the magnitude of the selection probability. This exponential function implements the strategy that the smaller the crowding density of the grid containing the particle, the higher its corresponding selection probability. The selection probability is obtained by normalizing the exponential function so that the sum of the selection probabilities of all grids is 1.

[0108] Step 7: Perform a roulette wheel selection. Given a random number r between [0,1], select from the particle swarm that satisfies the selection probability. The particles are used as the global optimal solution, and methods such as binary search are used to efficiently select particles; Step 7 is repeated until a limited number of optimal Pareto solutions are selected and stored in an external file.

[0109] This invention employs an external archive maintenance strategy.

[0110] An external archive is used to update and maintain non-dominated solutions. This external archive stores the Pareto optimal solution set, and its size cannot be infinitely large; therefore, maintaining the external archive is crucial. By limiting the archive size or dynamically adjusting it, the algorithm effectively manages storage space without wasting resources, preventing excessive expansion of the external archive and improving the algorithm's efficiency. A strategy for updating and maintaining the external archive based on grid technology and the roulette wheel selection algorithm is proposed, where the external archive update process is as follows: Figure 6 As shown.

[0111] The external archive maintenance process is as follows:

[0112] The first round of screening is conducted based on dominance relationships, eliminating inferior solutions and adding the remaining solutions to the external archive. For each newly generated non-inferior solution...

[0113] (1) If the size of the external files does not reach the specified size, then the non-inferior solutions will be directly added to the external files;

[0114] (2) Otherwise, if the new solution dominates some solutions in the external archive, these dominated solutions are removed, and the new solution is added to the external archive. The deletion mechanism in the multi-objective particle swarm optimization algorithm is implemented based on the proposed grid technique and roulette wheel strategy. During maintenance, to promote global search, the algorithm tends to select solutions with lower density values ​​(selection probability P). n Larger solutions are used as the next search target to ensure that the algorithm covers as much of the solution space as possible during the search process, rather than focusing on areas that are already relatively dense.

[0115] This strategy involves selecting a particle from the existing external file and deleting it to maintain the particle file size. The specific process of the strategy is as follows:

[0116] a) Extract the grid index of all particles from the external archive and store it in the GI vector.

[0117] b) Calculate the indices of all the cells occupied by the existing particles and remove duplicates to obtain different cells.

[0118] c) Calculate the number of particles (Ni) in each cell and store it in the N vector.

[0119] d) Based on a roulette wheel strategy, a selection probability Pn is used to select the cell containing the particle to be deleted. For non-dominated solutions in the external archive, the higher the particle density, the greater the selection probability. The external archive's size maintenance strategy is as follows: Figure 7 As shown, a roulette wheel strategy is used to randomly select a particle from the chosen grid, delete it from the external archive, and then re-grid. The probability of the nth grid being selected is:

[0120]

[0121] The purpose of this strategy is to maintain the size of the external file of particles to avoid excessive particles that would reduce the efficiency of the algorithm, while updating the external file to ensure the diversity of solution distribution and the global search capability of the algorithm.

[0122] This invention employs a Gaussian mutation strategy for populations.

[0123] Gaussian mutation originates from the field of evolutionary algorithms and is a mutation operation based on the normal distribution. In evolutionary algorithms, new individuals are generated by randomly perturbing the genes of individuals to improve population diversity and the algorithm's ability to avoid getting trapped in local optima. With the continuous updates and iterations of intelligent algorithms, Gaussian mutation has been applied to other optimization algorithms, such as particle swarm optimization and differential evolution.

[0124] To avoid the multi-objective particle swarm optimization algorithm converging to a local Pareto front, a Gaussian mutation mechanism is introduced into the traditional multi-objective particle swarm optimization algorithm. This mechanism adds a small-scale perturbation to the position of the internal particles, generating new solutions to increase the diversity and randomness of the population. This helps to avoid getting trapped in local optima, enhances the ability to search locally, and thus allows for a better search of the space.

[0125] The specific steps of the Gaussian mutation strategy used in this paper are as follows:

[0126] Step 1: For each individual, calculate its historical best solution and global best solution.

[0127] Step 2: For each individual, randomly generate a Gaussian distribution mutation vector.

[0128] Step 3: Dynamically adjust the standard deviation of the Gaussian distribution and the variable λ based on the individual historical best and global best solutions.

[0129] Step 4: Calculate the individual's new position X according to the following formula. i,j

[0130] X' i,j =X i,j +λGauss(0,1)

[0131] In the formula, X′ i,j X represents the value of the j-th dimension of the i-th particle after Gaussian mutation; i,j λ represents the value of the j-th dimension of the i-th particle before Gaussian mutation; λ is used to adjust the variable length and is a constant.

[0132] Step 5: Set boundary restrictions for the new location.

[0133] Step 6: Calculate the fitness function value for the new location.

[0134] Step 7: Determine if the new position can be added to the non-dominated solution set. Repeat steps 2-7 until the stopping criterion is met.

[0135] As can be seen from the implementation steps of Gaussian mutation, introducing randomness and adding random perturbations to the target space helps the algorithm escape local optima, better explore the search space, and thus increase the algorithm's global search capability. This randomness helps prevent the algorithm from getting trapped in local optima and maintains diversity during the search process, thereby increasing the probability of finding the global optimum.

Claims

1. An addressing optimization method for an energy storage system that considers multiple factors, characterized in that: Includes the following steps, And the following steps are performed in sequence: Step 1: Establish a two-layer optimal configuration model. The outer layer includes investment costs and daily operating costs; the inner layer includes system active power loss and system voltage deviation. Step 2: Establish constraints for a two-layer optimization configuration model based on the energy storage state of charge; Step 3: Use an improved multi-objective particle swarm optimization algorithm to obtain the minimum installed capacity and operating power of each energy storage system; use the output power as the rated power, and optimize the configuration scheme with the lowest investment cost as the energy storage system addressing scheme.

2. The addressing optimization method for an energy storage system considering multiple factors as described in claim 1, characterized in that: The investment cost mentioned in step one is: In the formula: f1 is the investment cost; n is the number of nodes in the power system; C ip Cost per unit capacity; C i The power cost per unit capacity of an energy storage system; z i This indicates the installation location of the energy storage power station in the energy storage system. The z-node is installed at node i. i It is 1 if it is true, otherwise it is 0; The daily operating cost is: In the formula: f2 is the daily operating cost; C om Operating and maintenance cost per unit of apparent power; P i and Q i These are the active power and reactive power of node i, respectively; The active power loss of the system is: In the formula: f3 is the system active power loss; b is the total number of branches in the system; R k P is the impedance of branch k; k and Q k These represent the active and reactive loads at the end of branch k, respectively; U k This refers to the voltage at the end of the line. The system voltage deviation is: In the formula: f4 is the system voltage deviation; U N This is the rated voltage at the end of the line.

3. The addressing optimization method for an energy storage system considering multiple factors as described in claim 1, characterized in that: The constraints described in step two include power generation output constraints, node power constraints, energy storage charge and discharge constraints, and energy storage state of charge constraints.

4. The addressing optimization method for an energy storage system considering multiple factors as described in claim 1, characterized in that: Step 3 describes an improved multi-objective particle swarm optimization algorithm that includes the following steps: 301: Run the data and set the parameters of the CMOPSO-MSI algorithm; set the population size, grid dimension, and maximum external document capacity; randomly initialize the velocity and position of the population particles; calculate the fitness value of each particle; store the non-dominated solution in the external document according to the Pareto dominance; determine the initial individual optimal value pbest and the global optimal value gbest of the particles; set the number of iterations T = 1. 302: Inertia weight in dynamic optimization and update algorithm based on improved strategy of adaptive nonlinear decreasing inertia weight; 303: Update particle velocity and position components, and calculate particle fitness values; 304: Determine whether a particle needs Gaussian mutation. If the mutation probability is greater than the random number, perform Gaussian mutation on the particle according to the formula; otherwise, the particle position remains unchanged. 305: Recalculate fitness values, update individual optimal solutions, maintain external documents based on adaptive grid density and roulette wheel algorithm, and select the global optimal solution from them; 306: Determine if the algorithm meets the termination condition. If it does, use the optimal Pareto solution set in the external archive as the final search result; otherwise, proceed to step 302 for the next iteration.

Images