A scheduling method of a gravity energy storage system based on multi-objective optimization

Abstract

The present application belongs to the technical field of charging and discharging scheduling of energy storage system, in order to solve the problem of poor scheduling effect existed in traditional technology, a kind of scheduling method of gravity energy storage system based on multi-objective optimization is disclosed, the present application constructs the multi-optimization target energy storage optimization function of energy storage system, and then, using multiple charging and discharging strategies, an initial population is generated;Then, the improved non-dominated sorting genetic algorithm is used to perform iterative genetic operation on the initial population, so as to obtain the pareto optimal solution set of the energy storage optimization function;Finally, according to the pareto optimal solution set, the optimal charging and discharging strategy can be determined;Wherein, the improved non-dominated sorting algorithm can improve the effectiveness of the charging and discharging strategy obtained, can maximize the economic benefit of the system and minimize the charging and discharging times to the greatest extent, and at the same time ensure the load balance, therefore, the method improves the charging and discharging scheduling effect of gravity energy storage system.

Description

Technical Field

[0001] This invention relates to the field of energy storage system charging and discharging scheduling technology, and specifically to a scheduling method for a gravity energy storage system based on multi-objective optimization. Background Technology

[0002] Currently, with the large-scale integration of renewable energy into the power grid, the power system has placed higher demands on the scheduling and optimization of energy storage devices. Among them, gravity energy storage systems, as a new type of energy storage, have gradually gained attention and widespread application in the field of large-scale energy storage due to their advantages such as long cycle life, environmental friendliness, and low operation and maintenance costs.

[0003] In terms of scheduling optimization, existing technologies mainly employ the following three types of methods: First, optimization methods based on linear or nonlinear programming, which can obtain analytical solutions in well-constrained models and have a strong theoretical foundation; second, methods based on heuristic intelligent algorithms, such as genetic algorithms (GA) and particle swarm optimization (PSO), which can adapt to complex nonlinear problems and have a certain global search capability; and third, methods based on deep reinforcement learning, which can autonomously learn scheduling strategies through continuous interaction and have a certain generalization capability.

[0004] While the aforementioned methods have improved the optimization capabilities of energy storage scheduling to some extent, they still have many shortcomings in practical applications. Linear programming methods are difficult to adapt to problems with nonlinear objective functions or uncertain constraints, have high computational complexity, and lack flexibility. Heuristic algorithms, although highly adaptable, have limited effectiveness in multi-objective optimization problems, are prone to getting trapped in local optima, and have insufficient ability to maintain population diversity, resulting in a limited solution set coverage. Deep reinforcement learning algorithms are limited by problems such as long training time, high data dependence, and poor model stability, making them difficult to apply to practical scheduling engineering.

[0005] Therefore, based on the aforementioned shortcomings, an optimization scheduling method based on NSGA-II has been proposed. Although this method can be applied to multi-objective optimization problems, its non-dominated ranking is based on a strict Pareto dominance relationship. That is, if individual one is not inferior to individual two on all objectives and is superior on at least one objective, then individual one is considered to dominate individual two. However, in the gravity energy storage scheduling scenario, the physical units of multiple objective functions are different and their numerical scales differ greatly. Therefore, the aforementioned ranking method is prone to making the ranking results too sensitive or lacking in ranking discrimination, which will affect the effectiveness of the generated scheduling strategy.

[0006] Meanwhile, the traditional NSGA-II algorithm measures individual crowding based on a single target spatial distance (i.e., it only relies on two individuals before and after to measure crowding), which cannot accurately reflect the local density of the target space. This leads to poor population diversity, which further reduces the effectiveness of the scheduling strategy and ultimately results in poor scheduling performance of the gravity energy storage system. It is unable to maximize economic benefits, minimize the number of charge and discharge cycles, and ensure load balance. Therefore, based on the aforementioned shortcomings, how to provide a scheduling method for gravity energy storage systems with good scheduling performance has become an urgent problem to be solved. Summary of the Invention

[0007] The technical problem to be solved by this invention is the scheduling problem of gravity energy storage systems. The purpose is to provide a scheduling method for gravity energy storage systems based on multi-objective optimization, which solves the problem of poor scheduling effect of gravity energy storage systems in traditional technologies.

[0008] This invention is achieved through the following technical solution:

[0009] Firstly, a scheduling method for a gravity energy storage system based on multi-objective optimization is provided, including:

[0010] Obtain the energy storage parameters of the gravity energy storage system;

[0011] Based on the energy storage parameters, an energy storage optimization function is constructed with the optimization objectives of load balance, maximizing operational efficiency, and minimizing the number of charge and discharge cycles.

[0012] Multiple charge and discharge strategies for the gravity energy storage system are generated, and an initial population is generated using these strategies.

[0013] An improved non-dominated sorting genetic algorithm is used to perform iterative genetic operations on the initial population to obtain the Pareto optimal solution set of the energy storage optimization function after the iterative genetic operations. In each iteration of the genetic operation, the domination tolerance and the energy storage optimization function value of each individual are calculated based on the energy storage optimization function. The population at each iteration is then sorted non-dominated based on the domination tolerance and the energy storage optimization function value of each individual. The joint crowding degree of each individual is calculated based on the difference between the energy storage optimization function value and the difference between the structure state matrix of each individual and its corresponding nearest neighbor individuals. After the non-dominated sorting, individuals are selected based on the joint crowding degree to obtain the population at the next iteration. The structure state matrix of any nearest neighbor individual is used to characterize the stacking state matrix of the gravity blocks in the gravity energy storage system after the charging and discharging strategy corresponding to that nearest neighbor individual is executed.

[0014] Based on the Pareto optimal solution set, the optimal charging and discharging strategy of the gravity energy storage system is determined, and the operation of the gravity energy storage system is controlled based on the optimal charging and discharging strategy.

[0015] Based on the above-disclosed content, this invention constructs an energy storage optimization function with the optimization objectives of load balancing, maximizing operational efficiency, and minimizing the number of charge-discharge cycles. Then, multiple charge-discharge strategies for the gravity energy storage system are generated, and an initial population is created using these strategies. Next, an improved non-dominated sorting genetic algorithm is used to perform iterative genetic operations on the initial population to obtain the Pareto optimal solution set of the energy storage optimization function. Finally, the optimal charge-discharge strategy is determined based on the Pareto optimal solution set, thus allowing the operation of the gravity energy storage system to be controlled according to this optimal strategy.

[0016] In this invention, a dominance tolerance is introduced during non-dominated sorting, adding a relative dynamic tolerance to the non-dominated sorting of individuals. This allows the non-dominated sorting of individuals to be less reliant on Pareto dominance, thus making the algorithm more tolerant to approximate solutions within the physically feasible range and enhancing the separability of sorting individuals in dense solution sets. Simultaneously, the crowding degree of individuals is measured by the difference in objective function values ​​and the difference in the stacking state structure of gravity blocks in the gravity energy storage system. Based on this, the crowding degree calculation method not only ensures a wide distribution of individuals in terms of optimization objectives but also diversity in the execution paths of gravity blocks in the system. Therefore, using this crowding degree measurement method for individual selection increases population diversity. Thus, the optimal charging and discharging strategy obtained through the aforementioned improved non-dominated sorting genetic algorithm can maximize economic benefits, minimize the number of charging and discharging cycles, and simultaneously ensure load balance. This improves the charging and discharging scheduling effect of the gravity energy storage system, making it more in line with actual operational scheduling requirements.

[0017] In one possible design, the energy storage parameters include: the system efficiency, maximum system power, and maximum system energy storage capacity of the gravity energy storage system;

[0018] Based on the energy storage parameters, an energy storage optimization function is constructed with the objectives of load balancing, maximizing operational efficiency, and minimizing the number of charge-discharge cycles. This function includes:

[0019] Obtain the electricity price at different times, and generate an operational efficiency optimization function based on system efficiency and the electricity price at different times;

[0020] The load balance optimization function and the charge / discharge cycle optimization function of the gravity energy storage system are constructed.

[0021] Based on the system's maximum power and maximum energy storage capacity, charging and discharging constraints are constructed.

[0022] The energy storage optimization function is formed by using the operation efficiency optimization function, the load balance optimization function, the charge / discharge cycle optimization function, and the charge / discharge constraints.

[0023] In one possible design, an improved non-dominated sorting genetic algorithm is used to perform iterative genetic operations on the initial population, including:

[0024] Initialize the number of iterations n, and obtain the individual population at the nth iteration, where the initial value of n is 1, and when n is 1, the individual population at the nth iteration is the initial population;

[0025] Perform individual genetic operations on the individual population at the nth iteration to obtain the offspring population;

[0026] A candidate population is formed using the offspring population and the individual population at the nth iteration;

[0027] The domination tolerance and the energy storage optimization function value of each candidate individual in the candidate population are calculated.

[0028] By utilizing the domination tolerance and the energy storage optimization function value of each candidate individual, the candidate population is subjected to non-dominated sorting to obtain several non-dominated sorting sets.

[0029] Calculate the difference between the energy storage optimization function value and the structural state matrix of each candidate individual in each non-dominated sorting set and its corresponding nearest neighbor individual. Based on the difference between the energy storage optimization function value and the structural state matrix of each candidate individual and its corresponding nearest neighbor individual, calculate the joint congestion degree of each candidate individual.

[0030] Based on the joint crowding of each candidate individual, several non-dominated sorting sets are subjected to individual selection processing to obtain a new individual population of the same size as the initial population, and the new individual population is used as the individual population in the (n+1)th iteration.

[0031] Determine if the iteration stopping condition is met;

[0032] If not, then n is incremented by 1, and the individual population at the nth iteration is reacquired until the iteration stopping condition is met, so that the charging and discharging strategies corresponding to each individual in the new individual population at the time of the iteration stopping condition are taken as the Pareto optimal solution set.

[0033] In one possible design, individual genetic operations are performed on the population of individuals at the nth iteration to obtain the offspring population, including:

[0034] Calculate the energy storage optimization function value of each individual in the population at the nth iteration, and calculate the fitness of each individual based on the energy storage optimization function value.

[0035] From the fitness of each individual, the minimum fitness is selected, and the average fitness at the nth iteration is determined;

[0036] Based on the minimum fitness and the average fitness, the mutation probability of each individual at the nth iteration is calculated;

[0037] Obtain the maximum number of iterations, and calculate the population crossover probability at the nth iteration based on the maximum number of iterations;

[0038] Based on the crossover probability and the mutation probability of each individual in the nth iteration, crossover and mutation operations are performed sequentially on the individual population in the nth iteration to obtain the offspring population after the mutation operation.

[0039] In one possible design, the mutation probability of each individual at the nth iteration is calculated based on the minimum fitness and the average fitness, including:

[0040] For any individual in the population at the nth iteration, the mutation probability of that individual at the nth iteration is calculated using the following formula (1);

[0041] (1)

[0042] In the above formula (1), Let represent the mutation probability of any individual in the nth iteration. Indicates the maximum mutation probability. This represents the fitness of any individual. The minimum fitness and the average fitness are represented, respectively. Represents positive numbers;

[0043] Accordingly, based on the maximum number of iterations, the population crossover probability at the nth iteration is calculated, which includes:

[0044] The crossover probability at the nth iteration is calculated according to the following formula (2);

[0045] (2)

[0046] In the above formula (2), This represents the crossover probability of the population at the nth iteration. Indicates the maximum crossover probability. This indicates the maximum number of iterations.

[0047] In one possible design, based on the population crossover probability and the mutation probability of each individual at the nth iteration, individual crossover and mutation operations are sequentially performed on the individual population at the nth iteration to obtain the offspring population after the individual mutation operation, including:

[0048] For any individual in the population at the nth iteration, generate the first random number;

[0049] Determine whether the first random number is less than or equal to the population crossover probability;

[0050] If so, then randomly select three distinct individuals from the individual population at the nth iteration.

[0051] Based on the three distinct individuals selected, a crossover operation is performed on any one of the individuals to obtain the crossover individual corresponding to that individual. After all individuals have been queried, the crossover population is obtained.

[0052] For any crossover individual in the crossover population, generate a second random number;

[0053] Determine whether the second random number is less than or equal to the mutation probability of any crossover individual in the nth iteration;

[0054] If so, then based on the Levi step size, perform a mutation operation on any of the crossover individuals to obtain the mutated individuals corresponding to any of the crossover individuals, and after all the crossover individuals have been queried, use all the mutated individuals to form the offspring population.

[0055] In one possible design, the energy storage optimization function includes three optimization functions, wherein the dominance tolerance is calculated, including:

[0056] Based on the three optimization functions in the energy storage optimization function, calculate the function values ​​of the three optimization functions corresponding to each candidate individual in the candidate population;

[0057] From the function values ​​of any optimization function corresponding to each candidate individual, select the maximum and minimum function values;

[0058] Based on the maximum and minimum function values, the sub-domination tolerance of any optimization function is calculated, and the sub-domination tolerance of each optimization function is obtained after all optimization functions have been polled.

[0059] The domination tolerance is formed by using the sub-domination tolerance of each optimization function.

[0060] In one possible design, the energy storage optimization function value of any candidate individual includes the function values ​​of three optimization functions, and the dominance tolerance includes the sub-dominance tolerances corresponding to the three optimization functions.

[0061] Specifically, by utilizing dominance tolerance and the energy storage optimization function values ​​of each candidate individual, the candidate population is subjected to non-dominated ranking processing to obtain several non-dominated ranking sets, including:

[0062] Initialize the non-dominated sort set of the u-th level to an empty set, where the initial value of u is 1;

[0063] The b-th candidate individual is selected from the candidate population. Based on the function values ​​of the three optimization functions corresponding to the b-th candidate individual, and the sub-domination tolerance corresponding to each optimization function, it is determined whether the b-th candidate individual meets the domination determination condition, wherein the domination determination condition is: And exist Make , , This represents the b-th candidate individual corresponding to the first... The function value of an optimization function. Indicates the first element in the target set The first target individual corresponding to the first The function value of an optimization function, the objective set containing all candidate individuals in the candidate population except for the b-th candidate individual, and , Indicates the total number of target individuals;

[0064] If so, then add the b-th candidate individual to the non-dominated sorting set of the u-th level;

[0065] Increment b by 1 and reselect the b-th candidate individual from the candidate population until b equals B, thus obtaining the non-dominated sorting set of the u-th level, where the initial value of b is 1 and B is the population size of the candidate population.

[0066] Remove candidate individuals from the non-dominated sorting set at level u from the candidate population to obtain a new candidate population, and update the candidate population to the new candidate population;

[0067] Increment u by 1 and reinitialize the non-dominated sorting set of level u to an empty set until all candidate individuals in the candidate population have been divided, resulting in several non-dominated sorting sets.

[0068] In a possible design, the energy storage optimization function value of any candidate individual includes the function values ​​of three optimization functions;

[0069] This includes calculating the difference in energy storage optimization function values ​​between each candidate individual in each non-dominated sorting set and its nearest neighbor individuals, including:

[0070] For any candidate individual in any non-dominated sorted set, obtain the K nearest neighbors of that candidate individual;

[0071] Based on the function values ​​of the three optimization functions corresponding to any candidate individual and the function values ​​of the three optimization functions corresponding to each nearest neighbor individual, the difference between the energy storage optimization function values ​​of any candidate individual and the corresponding nearest neighbor individual is calculated according to the following formula (3);

[0072] (3)

[0073] In the above formula (3), This represents the degree of difference between the energy storage optimization function value of any candidate individual and its corresponding nearest neighbor individual. Represents the first individual corresponding to any of the candidate individuals. The function value of an optimization function. Represents the k-th nearest neighbor individual corresponding to the k-th nearest neighbor individual. The function value of an optimization function. Indicates the first The weight coefficients of each optimization function.

[0074] In a possible design, the energy storage optimization function value for any candidate individual includes the function values ​​of three optimization functions, wherein the difference between the structure-state matrix of each candidate individual in each non-dominated sorting set and its corresponding nearest neighbor individual is calculated, including:

[0075] For any candidate individual in any non-dominated sorting set, determine the stacking state matrix of the gravity blocks in the gravity energy storage system after the gravity energy storage system executes the charging and discharging strategy corresponding to the candidate individual, and use the stacking state matrix as the structural state matrix of the candidate individual.

[0076] Obtain the structural state matrix of each nearest neighbor individual corresponding to any candidate individual;

[0077] Based on the structural state matrix of any candidate individual and the structural state matrix of each of the nearest neighbors of any candidate individual, the difference between the structural state matrix of any candidate individual and the corresponding nearest neighbor individual is calculated according to the following formula (4);

[0078] (4)

[0079] In the above formula (4), This represents the degree of difference between the structural state matrix of any candidate individual and its corresponding nearest neighbor individual. This represents the first element in the structural state matrix corresponding to any candidate individual. Line 1 The elements of the column, and used to characterize the first element in a gravity energy storage system. Layer The number of gravity blocks in the column, This represents the structure state matrix corresponding to the k-th nearest neighbor individual. Line 1 Column elements, These represent the total number of rows and columns of the structural state matrix corresponding to any candidate individual, respectively. This represents the total number of neighboring individuals.

[0080] Secondly, a scheduling device for a gravity energy storage system based on multi-objective optimization is provided, comprising:

[0081] Acquisition unit, used to acquire energy storage parameters of gravity energy storage system;

[0082] The energy storage optimization function construction unit is used to construct an energy storage optimization function based on energy storage parameters, with the optimization objectives of load balance, maximizing operating efficiency, and minimizing the number of charge and discharge cycles.

[0083] The optimization unit is used to generate multiple charging and discharging strategies for the gravity energy storage system, and to generate an initial population using these strategies.

[0084] The optimization unit employs an improved non-dominated sorting genetic algorithm to perform iterative genetic operations on the initial population. After the iterative genetic operations, a Pareto optimal solution set for the energy storage optimization function is obtained. During each iterative genetic operation, the domination tolerance and the energy storage optimization function value of each individual are calculated based on the energy storage optimization function. The population at each iteration is then non-dominated sorted according to the domination tolerance and the energy storage optimization function value of each individual. The joint crowding degree of each individual is calculated based on the difference between the energy storage optimization function value and the difference between the structure state matrix of each individual and its nearest neighbor. After the non-dominated sorting, individuals are selected based on the joint crowding degree to obtain the population for the next iteration. The structure state matrix of any nearest neighbor individual is used to characterize the stacking state matrix of the gravity blocks in the gravity energy storage system after the execution of the charging and discharging strategy corresponding to that nearest neighbor individual.

[0085] The scheduling unit is used to determine the optimal charging and discharging strategy of the gravity energy storage system based on the Pareto optimal solution set, so as to control the operation of the gravity energy storage system based on the optimal charging and discharging strategy.

[0086] Thirdly, another scheduling device for a gravity energy storage system based on multi-objective optimization is provided. Taking the device as an electronic device as an example, it includes a memory, a processor, and a transceiver that are connected in sequence. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the scheduling method for the gravity energy storage system based on multi-objective optimization as described in the first aspect or any possible design of the first aspect.

[0087] Fourthly, a storage medium is provided, on which instructions are stored, which, when executed on a computer, perform the scheduling method of the gravity energy storage system based on multi-objective optimization as described in the first aspect or any possible design of the first aspect.

[0088] Fifthly, a computer program product containing instructions is provided, which, when executed on a computer, causes the computer to perform the scheduling method for a gravity energy storage system based on multi-objective optimization as described in the first aspect or any possible design of the first aspect.

[0089] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0090] (1) The scheduling method for gravity energy storage system provided by the present invention introduces a domination tolerance during non-dominated sorting, which adds a relative dynamic tolerance to the non-dominated sorting of individuals, thereby increasing the separability of individuals in the dense solution set. At the same time, the difference in objective function value and the difference in the stacking state structure of gravity blocks in the gravity energy storage system are used to measure the congestion of individuals. Based on this, compared with the traditional single target spatial distance congestion measurement method, the present invention not only makes individuals widely distributed in the optimization target, but also has diversity in the execution path of gravity blocks in the system, thereby increasing the diversity of the population after individual selection. Thus, the optimal charging and discharging strategy obtained by the aforementioned improved non-dominated sorting genetic algorithm can maximize the economic benefits of the system, minimize the number of charging and discharging, and ensure load balance at the same time. Based on this, the charging and discharging scheduling effect of the gravity energy storage system is improved, making it more in line with the actual operation scheduling requirements.

[0091] (2) To further enhance the exploration-utilization balance in the search process, this invention designs a dynamic probability control strategy, wherein the mutation probability is dynamically adjusted according to the individual fitness. The worse the individual quality, the higher its mutation probability, which helps to improve the exploration ability of poor individuals. The crossover probability decreases with the number of generations of evolution. In the early stage, a high crossover activity is maintained to promote population diffusion, and in the later stage, the crossover frequency is reduced to achieve steady convergence. Thus, this mechanism can effectively regulate the search strategy of the algorithm at different evolutionary stages, thereby improving the overall stability of optimization.

[0092] (3) To address the problems of limited search scale and easy trapping in local optima in traditional SBX crossover and polynomial mutation, this invention adopts differential evolution crossover operator to drive the generation of new individuals with random individuals in the population, thereby improving the global perturbation capability. At the same time, during mutation, the Levy flight mechanism is introduced to generate a larger amplitude of variable asynchronous length through long-tail probability distribution, that is, to introduce irregular jump search path on the basis of conventional local search. In this way, the probability of escaping the local optimum trap can be increased, thereby expanding the search range. Based on this, the effectiveness of the generated charging and discharging strategy can be further improved. Attached Figure Description

[0093] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0094] Figure 1 A flowchart illustrating the steps of a scheduling method for a gravity energy storage system based on multi-objective optimization, as provided in an embodiment of the present invention.

[0095] Figure 2 A schematic diagram of the structure of a scheduling device for a gravity energy storage system based on multi-objective optimization, provided in an embodiment of the present invention;

[0096] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0097] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are for explanation only and are not intended to limit the invention. It should be understood that although terms such as "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of this invention.

[0098] Example:

[0099] See Figure 1As shown in this embodiment, the scheduling method for a gravity energy storage system based on multi-objective optimization constructs an energy storage optimization function with load balancing, maximizing operational efficiency, and minimizing the number of charge-discharge cycles as optimization objectives. Then, multiple charge-discharge strategies for the gravity energy storage system are generated, and an initial population is created using these strategies. Next, an improved non-dominated sorting genetic algorithm is used to iteratively genetically process the initial population to obtain the Pareto optimal solution set of the energy storage optimization function. Finally, the optimal charge-discharge strategy is determined based on the Pareto optimal solution set, allowing the system operation to be controlled accordingly. This method introduces dominance tolerance during non-dominated sorting, increasing the relative dynamic tolerance for individual non-dominated sorting and thus improving the sorting distinguishability of individuals. Simultaneously, a... By integrating the differences in energy storage optimization function values ​​between an individual and its neighbors, as well as the corresponding differences in system state structure, a congestion measurement method is used to calculate individual congestion and select individuals. Based on this, the diversity and structural differences of the solution set can be maintained, thereby improving the diversity of the population. Therefore, this method can improve the effectiveness of the derived charging and discharging strategy, maximizing the economic benefits of the system, minimizing the number of charging and discharging cycles, and ensuring load balance. Thus, this method improves the charging and discharging scheduling effect of gravity energy storage systems, making it very suitable for large-scale application and promotion. For example, this method can be run on the energy storage scheduling end side. It is understood that the aforementioned execution entity does not constitute a limitation on the embodiments of this application. Accordingly, the operation steps of this method can be, but are not limited to, the steps S1 to S5 below.

[0100] S1. Obtain the energy storage parameters of the gravity energy storage system; In this embodiment, the energy storage parameters may include, but are not limited to, the system efficiency, maximum power and maximum energy storage capacity of the gravity energy storage system. They may also include the number of layers and columns of gravity blocks in the gravity energy storage system. That is, each layer of the gravity energy storage system has several gravity blocks arranged in a matrix form, so as to store or release energy by lifting or placing gravity blocks.

[0101] After obtaining the system's energy storage parameters, the system optimization objectives and constraints can be constructed so that the optimal charging and discharging strategy of the system can be obtained by solving the optimization objectives. The process of constructing the optimization objectives and constraints can be, but is not limited to, the steps shown in step S2 below.

[0102] S2. Based on the energy storage parameters, construct an energy storage optimization function with the optimization objectives of load balance, maximizing operational efficiency, and minimizing the number of charge and discharge cycles. In specific implementation, for example, but not limited to, the following steps S21 to S24 can be used to construct the aforementioned energy storage optimization function.

[0103] S21. Obtain the electricity price at different times, and generate an operation benefit optimization function based on the system efficiency and the electricity price at different times; in specific implementation, for example, but not limited to, constructing the operation benefit optimization function according to the following formula (5).

[0104] (5)

[0105] In the above formula (5), This represents the function for optimizing operational efficiency. This represents the discharge power of the gravity energy storage system at time t. This represents the charging power of the gravity energy storage system in time period t. Indicates system efficiency. This represents the electricity price during time period t. In this embodiment, the unit of time period is hours, that is, the charging and discharging control of the system is performed in hours.

[0106] Thus, the aforementioned formula (5) indicates that one of the optimization objectives of the gravity energy storage system is to maximize the operational efficiency; where, after completing the construction of the operational efficiency optimization function, the other two optimization objective functions of the gravity energy storage system can be constructed, and the process is shown in step S22 below.

[0107] S22. Construct the load balance optimization function and the charge / discharge cycle optimization function of the gravity energy storage system; in specific applications, for example, but not limited to, the following formula (6) can be used to construct the load balance optimization function.

[0108] (6)

[0109] In the above formula (6), This represents the load balancing optimization function. This represents the optimized load of the gravity energy storage system at time t, where, ,and Let represent the original load of the gravity energy storage system at time t; thus, as can be seen from the aforementioned formula (6), the goal of this optimization function is to minimize the load peak-to-valley difference, that is, to smooth the load curve, reduce the load peak-to-valley difference, achieve load balance, and improve grid stability; where the original load is the sum of the power of all loads supplied by the gravity energy storage system.

[0110] Similarly, after constructing the load balancing optimization function, the charge and discharge cycle optimization function can be constructed, and its construction formula is shown in the following formula (7).

[0111] (7)

[0112] In the above formula (7), This represents the function for optimizing the number of charge / discharge cycles. This represents the charging and discharging decision variables of the gravity energy storage system at time t. This represents the charging and discharging decision variables of the gravity energy storage system at time t+1. Let be a symbolic function, where, when When the value is greater than 0, it indicates that the gravity energy storage system is in discharge operation at time t. When the value is less than 0, it indicates that the gravity energy storage system is in charging operation during time period t. When the value equals 0, it indicates that the gravity energy storage system is in standby mode at time t; where, when When the value is greater than 0, the sign function outputs 1. When the value equals 0, the sign function outputs 0, while when... When the value is less than 0, the sign function outputs -1.

[0113] Based on this, the goal of the aforementioned formula (7) is to minimize the number of charge and discharge cycles. That is, frequent charge and discharge operations will affect the lifespan of the energy storage system. Therefore, it is necessary to optimize the scheduling strategy to reduce unnecessary equipment operation cycles and thus extend the equipment lifespan.

[0114] After constructing the functions corresponding to the three optimization objectives using the above formulas (5)-(7), the constraint conditions corresponding to the three optimization objectives can be established so that the subsequent generated charging and discharging strategies conform to the physical constraints of the system; wherein, the construction process of the charging and discharging constraint conditions is as shown in step S23 below.

[0115] S23. Based on the system's maximum power and maximum energy storage capacity, construct the charge and discharge constraints. In this embodiment, the charge and discharge constraints can be constructed according to, but are not limited to, the following formula (8).

[0116] (8)

[0117] In the above formula (8), This indicates the charge / discharge constraint condition. This represents the charging and discharging decision variables of the gravity energy storage system at time t. This indicates the maximum power of the system. This represents the system energy storage capacity of the gravity energy storage system at time t. This indicates the maximum energy storage capacity of the system. This represents the element in the i-th row and j-th column of the stacking state matrix of gravity blocks in the gravity energy storage system, used to characterize the number of gravity blocks in the i-th layer and j-th column of the gravity energy storage system. This indicates the maximum load-bearing capacity of each layer on the equipment in a gravity energy storage system where weight blocks are arranged. Represents any symbol.

[0118] Furthermore, In the formula, This represents the system energy storage capacity of the gravity energy storage system at time t-1. This represents the discharge power of the gravity energy storage system at time t. This represents the charging power of the gravity energy storage system in time period t. Indicates system efficiency, and The term represents the energy storage interval time. In this embodiment, the first term in formula (8) is the power constraint, the second term is the energy state constraint of the system, and the third term is the distribution constraint of the gravity block.

[0119] Thus, through the aforementioned step S23, the constraints of the optimization objectives can be constructed; then, the functions corresponding to the aforementioned three optimization objectives can be combined to form the energy storage optimization function of the system, as shown in step S24 below.

[0120] S24. The energy storage optimization function is formed by using the operation efficiency optimization function, the load balance optimization function, the charge / discharge cycle optimization function, and the charge / discharge constraint conditions.

[0121] Therefore, based on the aforementioned steps S21 to S24, after constructing the functions corresponding to the three optimization objectives of the gravity energy storage system (i.e., the energy storage optimization function includes the aforementioned three optimization functions), the optimal charging and discharging strategy that satisfies the charging and discharging constraints can be obtained by solving the energy storage optimization function. Thus, the charging and discharging of the gravity energy storage system can be controlled based on the optimal charging and discharging strategy.

[0122] In this embodiment, an improved non-dominated sorting genetic algorithm is used to solve the aforementioned multi-objective energy storage optimization function, as shown in steps S3 and S4 below.

[0123] S3. Generate multiple charging and discharging strategies for the gravity energy storage system, and use these strategies to generate an initial population. In practice, any given charging and discharging strategy is represented as the charging and discharging strategy vector of the gravity energy storage system over 24 hours, where each element ranges from [-1, 1]. For example, a charging and discharging strategy might be: [-0.8, -0.6, -0.4, 0, 0, 0.5, 1.0, 0.8, 0.6, 0, 0, 0, -0.3, -0.5, 0, 0, 0.4, 0.6, 0.9, 1.0, 0, 0, -0.7, -1.0]. In the charging and discharging strategy, negative numbers represent charging, for example, -0.8 means charging at 80% of the maximum charging power in the first hour, and positive values ​​represent discharging, for example, 0.5 means discharging at 50% of the maximum discharging power in the sixth hour; while 0 means that the system is in standby mode for that hour, neither charging nor discharging; thus, the aforementioned charging and discharging strategy is essentially encoded using real numbers and generated using a random uniform distribution at the beginning; based on this, each initial individual corresponds to a charging and discharging strategy, thereby forming an initial population.

[0124] After generating the initial population, the improved non-dominated sorting genetic algorithm can be used to iterate the population, continuously updating the charging and discharging strategy during the continuous genetic iteration process, so as to obtain the optimal charging and discharging strategy after the genetic iteration operation; wherein, the population iteration process is shown in step S4 below.

[0125] S4. An improved non-dominated sorting genetic algorithm is used to perform iterative genetic operations on the initial population to obtain the Pareto optimal solution set of the energy storage optimization function after the iterative genetic operations.

[0126] In this embodiment, the non-dominated sorting genetic algorithm is improved. The improvement in non-dominated sorting and crowding calculation is mainly as follows: During each iteration of the genetic operation, the domination tolerance and the energy storage optimization function value of each individual are calculated based on the energy storage optimization function. The population at each iteration is then non-dominated sorted according to the domination tolerance and the energy storage optimization function value of each individual. The joint crowding of each individual is calculated based on the difference between the energy storage optimization function value and the difference between the structure state matrix of each individual and its corresponding nearest neighbor individuals. After non-dominated sorting, individuals are selected based on the joint crowding to obtain the population at the next iteration. The structure state matrix of any nearest neighbor individual is used to characterize the stacking state matrix of the gravity blocks in the gravity energy storage system after the execution of the charging and discharging strategy corresponding to that nearest neighbor individual. In this embodiment, the aforementioned improvements are described in detail in the following iteration process.

[0127] Optionally, the detailed process of the population genetic iteration of the improved non-dominated sorting genetic algorithm disclosed below may be, but is not limited to, the steps S41 to S49 below.

[0128] S41. Initialize the number of iterations n, and obtain the individual population at the nth iteration, wherein the initial value of n is 1, and when n is 1, the individual population at the nth iteration is the initial population.

[0129] After obtaining the individual population at the nth iteration, individual genetic operations can be performed on the individual population at the nth iteration to obtain the offspring population; wherein, the individual genetic operations are as shown in step S42 below.

[0130] S42. Perform individual genetic operations on the individual population at the nth iteration to obtain the offspring population; in specific applications, this mainly involves performing individual mutation and crossover operations on the individual population at the nth iteration; in this embodiment, to enhance the exploration-utilization balance in the search process, an adaptive crossover and mutation probability adjustment mechanism based on fitness and generation is constructed to effectively control the search strategy of the algorithm at different evolutionary stages, thereby improving the overall optimization stability; at the same time, in view of the problems of limited search scale and easy getting trapped in local optima in traditional SBX crossover and polynomial mutation, this embodiment also provides an improved crossover and mutation method to increase the probability of escaping the local optimum trap.

[0131] The detailed individual variation and crossover operations are shown in steps S42a to S42e below.

[0132] S42a. Calculate the energy storage optimization function value of each individual in the population at the nth iteration, and calculate the fitness of each individual based on the energy storage optimization function value. In this embodiment, for any individual at the nth iteration, substitute the charging and discharging strategy vector corresponding to that individual into the aforementioned three optimization functions to obtain three optimization function values. Then, normalize each optimization function value and perform a weighted sum of the three normalized optimization function values ​​to obtain the fitness of that individual. For example, the weighting coefficient of the three optimization function values ​​is 1 / 3.

[0133] Thus, after obtaining the fitness of each individual in the population at the nth iteration, the mutation probability of each individual at the nth iteration can be calculated, as shown in steps S42b and S42c below.

[0134] S42b. From the fitness of each individual, select the minimum fitness and determine the average fitness at the nth iteration; in specific implementation, after obtaining the minimum fitness and average fitness at the nth iteration, the mutation probability of each individual can be calculated based on this, as shown in step S42c below.

[0135] S42c. Based on the minimum fitness and the average fitness, calculate the mutation probability of each individual in the nth iteration; in this embodiment, take any individual in the population in the nth iteration as an example for illustration, and the following formula (1) can be used to calculate the mutation probability of any individual in the nth iteration.

[0136] (1)

[0137] In the above formula (1), Let represent the mutation probability of any individual in the nth iteration. Indicates the maximum mutation probability. This represents the fitness of any individual. The minimum fitness and the average fitness are represented, respectively. Represents positive numbers (used to prevent division by zero errors); optional, maximum mutation probability and sum. All are preset values.

[0138] Thus, after calculating the mutation probability of each individual in the nth iteration based on the aforementioned formula (1), the population crossover probability in the nth iteration can be calculated, as shown in step S42d below.

[0139] S42d. Obtain the maximum number of iterations and calculate the population crossover probability at the nth iteration based on the maximum number of iterations. In specific applications, for example, the aforementioned population crossover probability can be calculated according to the following formula (2).

[0140] (2)

[0141] In the above formula (2), This represents the crossover probability of the population at the nth iteration. Indicates the maximum crossover probability. This represents the maximum number of iterations; in this embodiment, the maximum crossover probability and the maximum number of iterations are also preset values.

[0142] As can be seen from the aforementioned formulas (1) and (2), the mutation probability is dynamically adjusted according to the individual fitness. The worse the individual quality, the higher its mutation probability, which helps to improve the exploration ability of poor individuals. The crossover probability decreases with the number of generations. In the early stage, the crossover activity is maintained to promote population diffusion, and in the later stage, the crossover frequency is reduced to achieve steady convergence. Based on this, the mechanism can effectively regulate the search strategy of the algorithm at different evolutionary stages, thereby improving the overall stability of optimization.

[0143] Therefore, after calculating the population crossover probability using the aforementioned formula (2), the individual mutation and crossover of each individual can be carried out by combining the mutation probability of each individual; wherein, the individual mutation and crossover process is as shown in step S42e below.

[0144] S42e. Based on the population crossover probability and the mutation probability of each individual at the nth iteration, perform individual crossover and individual mutation operations on the individual population at the nth iteration in sequence, so as to obtain the offspring population after the individual mutation operation.

[0145] In practice, we will take any individual in the population at the nth iteration as an example to illustrate the crossover and mutation operations. The process is shown in steps S42e1 to S42e7 below.

[0146] S42e1. For any individual in the population at the nth iteration, generate a first random number; in this embodiment, the value range of the first random number is (0,1); thus, after obtaining the first random number, it can be compared with the population crossover probability to determine whether any individual can undergo individual crossover operation.

[0147] The judgment process is shown in step S42e2 below.

[0148] S42e2. Determine whether the first random number is less than or equal to the population crossover probability; in specific applications, if the first random number is greater than the population crossover probability, then keep any individual and do not perform individual crossover; conversely, if the first random number is less than or equal to the population crossover probability, then individual crossover is required, as shown in steps S42e3 and S42e4.

[0149] S42e3. If so, then randomly select three distinct individuals from the population at the nth iteration.

[0150] S42e4. Based on the three selected distinct individuals, perform a crossover operation on any one of the individuals to obtain the crossover individual corresponding to that individual, and obtain the crossover population after all individuals have been queried.

[0151] In practice, for example, but not limited to, the following formula (9) can be used to perform cross operations on any individual.

[0152] (9)

[0153] In the above formula (9), This represents the intersection individual corresponding to any of the individuals. This indicates that three distinct individuals have been selected. Denotes the differential scaling factor, where, , used to adjust the cross strength.

[0154] Thus, by using the aforementioned formula (9), the crossover operation of each individual in the individual population at the nth iteration can be completed, thereby obtaining the crossover population; then, the individual mutation operation can be performed by combining the mutation probability of each individual; here, we will take any crossover individual in the crossover population as an example to illustrate the process as shown in the following steps S42e5~S42e7.

[0155] S42e5. For any crossover individual in the crossover population, generate a second random number; in this embodiment, the value range of the second random number is also (0,1); thus, after obtaining the second random number, it can be compared with the mutation probability of the crossover individual in the nth iteration to determine whether an individual mutation operation can be performed, wherein the determination process is as shown in step S42e6 below.

[0156] S42e6. Determine whether the second random number is less than or equal to the mutation probability of any crossover individual in the nth iteration; in specific implementation, if the second random number is greater than the mutation probability of any crossover individual in the nth iteration, then the crossover individual is retained, that is, no mutation operation is performed; otherwise, a mutation operation is required; further, for example, the mutation probability of any crossover individual in the nth iteration is the mutation probability of the original individual corresponding to that crossover individual in the nth iteration; if individual C in the individual population in the nth iteration, after performing an individual crossover operation, results in crossover individual C1, then the mutation probability of crossover individual C1 in the nth iteration is the mutation probability of individual C in the nth iteration.

[0157] The individual variation operation is shown in step S42e7 below.

[0158] S42e7. If so, then based on the Levi step size, perform a mutation operation on any of the crossover individuals to obtain the mutated individuals corresponding to any of the crossover individuals, and after all the crossover individuals have been queried, use all the mutated individuals to form the offspring population.

[0159] In specific applications, for example, but not limited to, the following formula (10) can be used to perform mutation operations on any of the crossover individuals to obtain the corresponding mutated individuals.

[0160] (10)

[0161] In the above formula (10), This represents the variant individual corresponding to any of the aforementioned crossover individuals. This refers to any of the individuals mentioned. Indicates the step size coefficient. This represents the Levy step size; in this embodiment, ,Right now It follows a heavy-tailed distribution, and its probability density function is: ;in, Indicates the magnitude of the disturbance. The value is usually set to 1.5, while the step size coefficient is used to control the range of variation.

[0162] Therefore, through the aforementioned steps S42a to S42e and their sub-steps, this embodiment employs a differential evolution crossover operator to drive the generation of new individuals by the difference between random individuals, thereby improving the global perturbation capability. At the same time, the Levy flight mutation mechanism is introduced, which generates a larger amplitude of variable asynchronous length through a long-tail probability distribution. That is, an irregular jump search path is introduced on the basis of conventional local search. Based on this, the probability of escaping the local optimum trap can be increased, thereby expanding the search range and achieving the goal of improving the global search capability.

[0163] After completing the individual genetic operations of the individual population at the nth iteration, the parent population (i.e., the individual population at the nth iteration) can be combined to form a candidate population, as shown in step S43 below.

[0164] S43. Use the offspring population and the individual population at the nth iteration to form a candidate population.

[0165] After combining the parent and offspring populations, non-dominated sorting can be performed. This involves prioritizing candidate solutions using fast non-dominated sorting to ensure the rationality of the Pareto front and to maintain solution diversity by utilizing the joint crowding of individuals, thereby preventing solutions from concentrating in a specific region. In this embodiment, domination tolerance is introduced into the non-dominated sorting of individuals, thereby increasing the relative dynamic tolerance for the non-dominated sorting of individuals and improving the sorting distinguishability of individuals. Specifically, the calculation process of domination tolerance is shown in step S44 below.

[0166] S44. Calculate the domination tolerance and the energy storage optimization function value of each candidate individual in the candidate population; in specific applications, substitute the charging and discharging strategy vector corresponding to each candidate individual into the aforementioned three optimization functions to obtain the energy storage optimization function value of each candidate individual; then, the domination tolerance can be calculated by combining the energy storage optimization function value of each candidate individual; the specific calculation process is shown in steps S44a to S44d below.

[0167] S44a. Based on the three optimization functions in the energy storage optimization function, calculate the function values ​​of the three optimization functions corresponding to each candidate individual in the candidate population.

[0168] After obtaining the function values ​​of the three optimization functions corresponding to each candidate individual (i.e., the function value of the operation benefit optimization function, the function value of the load balancing optimization function, and the function value of the charge and discharge times optimization function), the sub-dominance tolerance corresponding to each optimization function can be calculated based on these values, as shown in steps S44b and S44c below.

[0169] S44b. Select the maximum and minimum function values ​​from the function values ​​of any optimization function corresponding to each candidate individual. In this embodiment, taking the operational efficiency optimization function as an example, each candidate individual corresponds to a function value of the operational efficiency optimization function. For example, if there are 10 individuals, then 10 function values ​​of the operational efficiency optimization function will be obtained. Therefore, the maximum and minimum function values ​​are selected from the 10 function values ​​of the operational efficiency optimization function.

[0170] Then, based on the selected maximum and minimum function values, the sub-dominance tolerance corresponding to the operational efficiency optimization function can be calculated. The calculation process is shown in step S44c below.

[0171] S44c. Calculate the sub-domination tolerance of any optimization function based on the maximum and minimum function values, and obtain the sub-domination tolerance of each optimization function after all optimization functions have been polled.

[0172] In practical applications, the formulas for calculating the sub-dominance tolerance of various optimization functions are as follows:

[0173] (11)

[0174] In formula (11), Indicates the first Sub-dominance tolerance of an optimization function Indicates the first The relative tolerance coefficient of an optimization function. Each of the candidate individuals represents the first... The maximum and minimum function values ​​among the function values ​​of the optimization function, where, ,and .

[0175] Thus, based on the aforementioned formula (11), after calculating the sub-domination tolerance of each optimization function, the domination tolerance can be obtained by combining them, as shown in step S44d below.

[0176] S44d. The domination tolerance is formed by utilizing the sub-domination tolerance of each optimization function.

[0177] Therefore, after calculating the domination tolerance through the aforementioned steps S44a to S44d, the non-dominated ranking of candidate populations can be performed based on the domination tolerance, as shown in step S45 below.

[0178] S45. Using the dominance tolerance and the energy storage optimization function value of each candidate individual, perform non-dominated sorting on the candidate population to obtain several non-dominated sorting sets; in specific applications, for example, but not limited to, the following steps S45a to S45f can be used to complete the non-dominated sorting of the candidate population.

[0179] S45a. Initialize the non-dominated sort set of the u-th order to an empty set, where the initial value of u is 1.

[0180] After the initialization process of the first-level non-dominated sort set is completed, the first-level non-dominated sort can be performed, as shown in steps S45b to S45d below.

[0181] S45b. Select the b-th candidate individual from the candidate population, and based on the function values ​​of the three optimization functions corresponding to the b-th candidate individual, and the sub-domination tolerance corresponding to each optimization function, determine whether the b-th candidate individual meets the domination determination condition, wherein the domination determination condition is: And exist Make , , This represents the b-th candidate individual corresponding to the first... The function value of an optimization function. Indicates the first element in the target set The first target individual corresponding to the first The function value of an optimization function, the objective set containing all candidate individuals in the candidate population except for the b-th candidate individual, and , This represents the total number of target individuals.

[0182] In this embodiment, the b-th candidate individual in the candidate population is compared with all other candidate individuals in the candidate population to determine whether the b-th candidate individual meets the aforementioned dominance determination condition compared with the other candidate individuals. If it does, the b-th candidate individual is added to the first-level non-dominated sorting set as a first-level non-dominated sorting individual. The process is shown in step S45c below.

[0183] S45c. If so, the b-th candidate individual is added to the non-dominated sorting set of the u-th level. In this embodiment, if the b-th candidate individual does not meet the aforementioned dominance judgment condition, b is directly incremented by 1, that is, the judgment of the next candidate individual is performed. In this way, the individuals are traversed according to the principle until all individuals in the candidate population have been traversed, and then the non-dominated sorting set of the first level can be obtained. The process is shown in step S45d below.

[0184] S45d. Increment b by 1 and reselect the b-th candidate individual from the candidate population until b equals B, thus obtaining the non-dominated sorting set of the u-th level, where the initial value of b is 1 and B is the population size of the candidate population.

[0185] After obtaining the first-level non-dominated sort set, the candidate individuals that have been divided can be deleted from the candidate population, that is, the candidate individuals in the non-dominated sort set of the u-th level can be deleted. Then, the individuals are re-classified based on the dominance determination condition in step S42b above, until all candidate individuals have been classified, and then several non-dominated sort sets can be obtained. The cyclic division process of individuals is shown in steps S45e and S45f below.

[0186] S45e. Remove candidate individuals from the non-dominated sorting set of level u from the candidate population to obtain a new candidate population, and update the candidate population to the new candidate population. In this embodiment, when u is 1, the candidate individuals from the non-dominated sorting set of the first level are deleted from the candidate population to obtain a new candidate population. At this time, in the next partition, the candidate population is updated to the new candidate population. Then, the aforementioned steps S45b to S45d are repeated to obtain the non-dominated sorting set of the second level. Next, the candidate population in the second partition (which is the population after the candidate individuals in the non-dominated sorting set of the first level have been deleted) is deleted, that is, the candidate individuals in the non-dominated sorting set of the second level are deleted, and a new candidate population is obtained again. Thus, the partitioning is carried out continuously in this way until all candidate individuals have been partitioned, and then several non-dominated sorting sets can be obtained.

[0187] The cyclic partitioning process is shown in step S45f below.

[0188] S45f. Increment u by 1 and reinitialize the non-dominated sorting set of the u-th level to an empty set until all candidate individuals in the candidate population have been divided, resulting in several non-dominated sorting sets.

[0189] Therefore, through the aforementioned steps S45a to S45f, this embodiment introduces dominance tolerance during non-dominated sorting, that is, adds a relative dynamic tolerance to each individual. In this way, it allows approximately equivalent solutions to be regarded as non-dominated within the engineering tolerance range, thereby improving the sorting robustness and Pareto front resolution. Based on this, this approach reflects the understanding and integration of "acceptable differences" in energy storage scheduling, making the algorithm more tolerant to approximate solutions within the physically feasible range, improving the algorithm's sorting ability in dense solution set regions, and thus alleviating problems such as solution set aggregation and search stagnation.

[0190] After completing the non-dominated ordination of the candidate population, the joint crowding degree of individuals can be calculated so that individual selection can be carried out based on the joint crowding degree; the calculation process of joint crowding degree is shown in step S46 below.

[0191] S46. Calculate the difference between the energy storage optimization function value and the structural state matrix of each candidate individual in each non-dominated sorting set and its corresponding nearest neighbor individual. Based on the difference between the energy storage optimization function value and the structural state matrix of each candidate individual and its corresponding nearest neighbor individual, calculate the joint congestion degree of each candidate individual.

[0192] In order to further maintain the diversity and structural differences of the solution set among individuals with the same non-dominance level, this embodiment adopts a K-nearest neighbor density estimation method that integrates the differences in objective function values ​​and the differences in system state structure, instead of the traditional crowding degree measurement method based on a single objective spatial distance. Specifically, this embodiment constructs a "perceptual crowding degree distance" for each candidate individual by introducing a weighted combination of objective spatial distance and system state spatial distance, so that the retained solutions are not only widely distributed at the optimization objective level, but also more diverse in physical execution paths (such as gravity block distribution).

[0193] Optionally, taking any candidate individual in any non-dominated sorting set as an example to illustrate the calculation process of the difference in the aforementioned energy storage optimization function value, it can be, but is not limited to, as shown in steps S46a and S46b below.

[0194] S46a. For any candidate individual in any non-dominated sorting set, obtain the K nearest neighbors of that candidate individual; in specific applications, for example, but not limited to, calculating the Euclidean distance between the candidate individual and each of the other candidate individuals in the non-dominated sorting set, and then sorting the other candidate individuals in the non-dominated sorting set in ascending order of Euclidean distance to obtain a sorting sequence; finally, taking the first K candidate individuals in the sorting sequence as the nearest neighbors of the candidate individual.

[0195] After obtaining the nearest neighbors of any candidate individual, the difference between the energy storage optimization function value of any candidate individual and its corresponding nearest neighbor individual can be calculated, as shown in step S46b below.

[0196] S46b. Based on the function values ​​of the three optimization functions corresponding to any candidate individual and the function values ​​of the three optimization functions corresponding to each nearest neighbor individual, and according to the following formula (3), calculate the difference between the energy storage optimization function values ​​of any candidate individual and its corresponding nearest neighbor individual.

[0197] (3)

[0198] In the above formula (3), It represents the difference between the energy storage optimization function value of any candidate individual and its corresponding nearest neighbor individual (in essence, it also represents the target spatial distance of any candidate individual). Represents the first individual corresponding to any of the candidate individuals. The function value of an optimization function. Represents the k-th nearest neighbor individual corresponding to the k-th nearest neighbor individual. The function value of an optimization function. Indicates the first The weight coefficients of each optimization function; in this embodiment, for example, the weight coefficient of the first optimization function (operational efficiency optimization function) is 0.6, the weight coefficient of the second optimization function (load balancing optimization function) is 0.2, and the weight coefficient of the third optimization function (charge and discharge cycles optimization function) is 0.2; of course, the weight coefficients of different optimization functions can be specifically set according to actual use, and are not limited to the above examples.

[0199] Thus, based on the aforementioned formula (3), the difference in energy storage optimization function values ​​between each candidate individual in each non-dominated sorting set and its nearest neighbor can be calculated.

[0200] Then, the difference between the structural state matrix of each candidate individual in each non-dominated sorting set and its corresponding nearest neighbor individual can be calculated, as shown in steps S46c to S46e below.

[0201] S46c. For any candidate individual in any non-dominated sorting set, determine the stacking state matrix of gravity blocks in the gravity energy storage system after the gravity energy storage system executes the charging and discharging strategy corresponding to the candidate individual, and use the stacking state matrix as the structural state matrix of the candidate individual; in this embodiment, it is equivalent to obtaining the number of gravity blocks in each column of each layer in the gravity energy storage system after executing the charging and discharging strategy corresponding to the candidate individual, thereby forming the stacking state matrix of gravity blocks, and using the stacking state matrix as the structural state matrix of the candidate individual; of course, the aforementioned system uses a simulation method to execute the aforementioned charging and discharging strategy, thereby obtaining the number of gravity blocks under each hour's charging and discharging operation (such as charging, discharging or standby), and finally obtaining the stacking state matrix.

[0202] After obtaining the structural state matrix of any candidate individual, the structural state matrix of the nearest neighbor individuals of any candidate individual can be obtained, as shown in step S46d below.

[0203] S46d. Obtain the structural state matrix of each nearest neighbor individual corresponding to any candidate individual; In this embodiment, the process of determining the nearest neighbor individual corresponding to any candidate individual can be referred to the aforementioned step S46a, and the structural state matrix of the nearest neighbor individual is obtained in the same way as the structural state matrix of any candidate individual, which will not be repeated here.

[0204] After obtaining the structural state matrix of each nearest neighbor, the difference degree of the structural state matrix can be calculated, as shown in step S46e below.

[0205] S46e. Based on the structural state matrix of any candidate individual and the structural state matrix of each of the nearest neighbors of any candidate individual, the difference between the structural state matrix of any candidate individual and the corresponding nearest neighbor individual is calculated according to the following formula (4).

[0206] (4)

[0207] In the above formula (4), This represents the degree of difference between the structural state matrix of any candidate individual and its corresponding nearest neighbor individual. This represents the first element in the structural state matrix corresponding to any candidate individual. Line 1 The elements of the column, and used to characterize the first element in a gravity energy storage system. Layer The number of gravity blocks in the column, This represents the structure state matrix corresponding to the k-th nearest neighbor individual. Line 1 Column elements, These represent the total number of rows and columns of the structural state matrix corresponding to any candidate individual, respectively. This represents the total number of neighboring individuals.

[0208] Thus, through the aforementioned steps S46a to S46e, the difference in energy storage optimization function values ​​and structural state matrix between each candidate individual in each non-dominated sorting set and its corresponding nearest neighbor individual can be calculated. Then, the joint crowding degree of each candidate individual in each non-dominated sorting set can be calculated, as follows:

[0209] Taking any candidate individual in any non-dominated sorting set as an example, the joint crowding degree of any candidate individual can be calculated using, but is not limited to, the following formula (12).

[0210] (12)

[0211] In the above formula (12), This represents the joint crowding degree of any candidate individual. Represents the crowding coefficient, where, This reflects whether the goal is dominant or the structure is dominant.

[0212] Therefore, through the aforementioned step S46 and its sub-steps, this embodiment introduces the K-nearest neighbor density estimation method to find the K nearest neighbor individuals for each candidate individual. The density of the local solution space is measured by the weighted sum of the differences between the objective function distance and the structural state matrix, thereby guiding the individual retention strategy. In this way, the mechanism can significantly improve the distribution balance and boundary coverage of the population in the objective space, thereby enhancing the representativeness of the optimal solution.

[0213] After calculating the joint crowding degree of each candidate individual in the non-dominated sorting set, individual selection can be performed, as shown in step S47 below.

[0214] S47. Based on the joint crowding degree of each candidate individual, perform individual selection processing on several non-dominated sorting sets to obtain a new individual population of the same size as the initial population, and use the new individual population as the individual population in the (n+1)th iteration; in specific applications, as explained above, several non-dominated sorting sets correspond to different levels, therefore, individual selection can be performed according to the order of the non-dominated sorting sets from high to low levels, and the specific process is as follows:

[0215] Step 1: Obtain the non-dominated sort set of level u, where the initial value of u is 1.

[0216] Step 2: Determine whether the number of candidate individuals in the non-dominated sorting set at level u is less than or equal to a preset number, where the preset number is the number of individuals in the initial population.

[0217] Step 3: If yes, select all candidate individuals from the non-dominated sorting set of level u as new individuals; otherwise, sort all candidate individuals in the non-dominated sorting set of level u according to the order of joint crowding from high to low, and select the top Y candidate individuals as new individuals, where Y is the preset number.

[0218] Step 4: Determine whether the number of newly selected individuals has reached the preset number.

[0219] Step 5: If not, increment u by 1 and reacquire the non-dominated sorting set of level u until the number of selected new individuals reaches the preset number, so as to use all selected new individuals to form a new individual population.

[0220] Therefore, by adopting a population renewal mechanism that combines elite retention with joint crowding assessment, we can ensure that the best individuals are retained and stably passed on to the next generation. After obtaining a new population of individuals, we can determine whether the iteration stopping condition is met, as shown in step S48 below.

[0221] S48. Determine whether the iteration stopping condition is met; In this embodiment, the iteration stopping condition may include, but is not limited to, n reaching the maximum number of iterations, or the fitness corresponding to the global optimal solution reaching the fitness threshold; Wherein, the global optimal solution is updated in each iteration. For example, in the first iteration, the global optimal solution is the individual with the highest fitness. In the second iteration, it is determined whether the maximum fitness in the second iteration is greater than the fitness corresponding to the global optimal solution. If so, the global optimal solution is updated to the individual corresponding to the maximum fitness in the second iteration. Otherwise, no update is performed. In this way, the global optimal solution is continuously updated until the fitness corresponding to the global optimal solution reaches the fitness threshold, at which point the genetic iteration can end; Wherein, the cyclic iteration process is as shown in step S49 below.

[0222] S49. If not, increment n by 1 and reacquire the individual population at the nth iteration until the iteration stopping condition is met, so that the charging and discharging strategies corresponding to each individual in the new individual population at the time of the iteration stopping condition are taken as the Pareto optimal solution set.

[0223] Therefore, through the aforementioned steps S41 to S49, the improved non-dominated sorting genetic algorithm can be used to solve for a set of optimal charging and discharging strategies (i.e., Pareto optimal solution set) that satisfy the charging and discharging constraints. Then, an optimal charging and discharging strategy can be determined from the Pareto optimal solution set to control the operation of the gravity energy storage system. The process is shown in step S5 below.

[0224] S5. Based on the Pareto optimal solution set, the optimal charging and discharging strategy of the gravity energy storage system is determined, and the operation of the gravity energy storage system is controlled based on the optimal charging and discharging strategy. In this embodiment, the charging and discharging strategy with the highest fitness among a set of optimal charging and discharging strategies that meet the charging and discharging constraints is selected as the optimal charging and discharging strategy. That is, the vectors corresponding to each charging and discharging strategy in the set of optimal charging and discharging strategies that meet the charging and discharging constraints are substituted into the aforementioned energy storage optimization function to obtain three optimization function values. Finally, a weighted sum is performed to obtain the corresponding fitness. In this way, the charging and discharging strategy with the highest fitness in the Pareto optimal solution set can be selected as the optimal charging and discharging strategy, thereby obtaining the charging and discharging decision of the gravity energy storage system every hour, and thus completing the charging and discharging scheduling of the gravity energy storage system.

[0225] In addition, in this embodiment, a Pareto front 3D scatter plot can be drawn to analyze the trade-off between economic benefits, load balance, and equipment operation frequency; the optimal scheduling scheme adapted to different operating conditions can be selected; and after executing the optimal charging and discharging strategy, a load curve comparison analysis can be performed to draw the load curves before and after optimization, observe the improvement of the load peak-valley difference, and calculate the proportion of reduction in the load peak-valley difference after optimization to verify whether the scheduling strategy is effective.

[0226] Therefore, through the multi-objective optimization-based scheduling method for gravity energy storage systems described in detail in steps S1 to S5 above, this invention constructs an energy storage optimization function with load balancing, maximizing operational efficiency, and minimizing the number of charge-discharge cycles as optimization objectives. Then, multiple charge-discharge strategies for the gravity energy storage system are generated, and an initial population is created using these strategies. Next, an improved non-dominated sorting genetic algorithm is used to iteratively genetically process the initial population to obtain the Pareto optimal solution set of the energy storage optimization function. Finally, the optimal charge-discharge strategy is determined based on the Pareto optimal solution set, allowing the operation of the gravity energy storage system to be controlled according to this optimal strategy. This invention innovates comprehensively in terms of sorting accuracy, solution set distribution, search jump capability, and evolutionary control mechanism, forming an optimized scheduling method suitable for the operating characteristics, target structure, and engineering requirements of gravity energy storage systems. It improves the effectiveness of the derived charge-discharge strategies, maximizing the system's economic benefits, minimizing the number of charge-discharge cycles, and simultaneously ensuring load balancing. Therefore, this invention improves the charge-discharge scheduling effect of gravity energy storage systems, making it highly suitable for large-scale application and promotion.

[0227] like Figure 2 As shown, the second aspect of this embodiment provides a hardware system for implementing the scheduling method of a gravity energy storage system based on multi-objective optimization as described in the first aspect of the embodiment, comprising:

[0228] The acquisition unit is used to acquire the energy storage parameters of the gravity energy storage system.

[0229] The energy storage optimization function construction unit is used to construct an energy storage optimization function based on energy storage parameters, with the optimization objectives of load balance, maximizing operational efficiency, and minimizing the number of charge and discharge cycles.

[0230] The optimization unit is used to generate multiple charging and discharging strategies for the gravity energy storage system, and to generate an initial population using these strategies.

[0231] The optimization unit employs an improved non-dominated sorting genetic algorithm to perform iterative genetic operations on the initial population. After the iterative genetic operations, a Pareto optimal solution set for the energy storage optimization function is obtained. During each iterative genetic operation, the domination tolerance and the energy storage optimization function value of each individual are calculated based on the energy storage optimization function. The population at each iteration is then non-dominatedly sorted according to the domination tolerance and the energy storage optimization function values ​​of each individual. Furthermore, the joint crowding degree of each individual is calculated based on the difference in energy storage optimization function values ​​and the difference in the structure state matrix between each individual and its nearest neighbors. After the non-dominated sorting, individuals are selected based on the joint crowding degree to obtain the population for the next iteration. The structure state matrix of any nearest neighbor individual is used to characterize the stacking state matrix of the gravity blocks in the gravity energy storage system after executing the charging and discharging strategy corresponding to that nearest neighbor individual.

[0232] The scheduling unit is used to determine the optimal charging and discharging strategy of the gravity energy storage system based on the Pareto optimal solution set, so as to control the operation of the gravity energy storage system based on the optimal charging and discharging strategy.

[0233] The working process, working details and technical effects of the device provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0234] like Figure 3 As shown, the third aspect of this embodiment provides another scheduling device for a gravity energy storage system based on multi-objective optimization. Taking the device as an electronic device as an example, it includes: a memory, a processor, and a transceiver that are connected in sequence. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the scheduling method for a gravity energy storage system based on multi-objective optimization as described in the first aspect of the embodiment.

[0235] For specific examples, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or first-in-last-out (FILO) memory, etc.; specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor, also known as the CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state.

[0236] In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. For example, the processor may not be limited to microprocessors of the STM32F105 series, reduced instruction set computer (RISC) microprocessors, x86 architecture processors, or processors with integrated neural network processing units (NPUs). The transceiver may be, but is not limited to, a Wi-Fi transceiver, a Bluetooth transceiver, a General Packet Radio Service (GPRS) transceiver, a ZigBee (a low-power LAN protocol based on the IEEE 802.15.4 standard) transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver. Furthermore, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0237] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0238] The fourth aspect of this embodiment provides a storage medium that stores instructions containing the scheduling method of the gravity energy storage system based on multi-objective optimization as described in the first aspect of the embodiment. That is, the storage medium stores instructions, and when the instructions are run on a computer, the scheduling method of the gravity energy storage system based on multi-objective optimization as described in the first aspect of the embodiment is executed.

[0239] The storage medium refers to a carrier for storing data, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0240] The working process, working details and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0241] The fifth aspect of this embodiment provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the scheduling method for a gravity energy storage system based on multi-objective optimization as described in the first aspect of this embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0242] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A scheduling method of a multi-objective optimization-based gravity energy storage system, characterized by, The method comprises the following steps: obtaining energy storage parameters of a gravity energy storage system; constructing an energy storage optimization function with load balance, maximum operation benefit and minimum charging and discharging times as optimization objectives according to the energy storage parameters; generating a plurality of charging and discharging strategies of the gravity energy storage system, and generating an initial population by using the plurality of charging and discharging strategies; performing iterative genetic operations on the initial population by using an improved non-dominated sorting genetic algorithm, so as to obtain a Pareto optimal solution set of the energy storage optimization function after the iterative genetic operations, wherein, when performing each iterative genetic operation, the domination tolerance and the energy storage optimization function value of each individual are calculated based on the energy storage optimization function, and the population at each iteration is non-dominantly sorted according to the domination tolerance and the energy storage optimization function value of each individual, and the joint crowding degree of each individual is calculated according to the difference between the energy storage optimization function value of each individual and the energy storage optimization function value of the neighbor individual corresponding to each individual and the difference of the structure state matrix, so that the population at the next iteration is obtained based on the joint crowding degree after the non-dominant sorting, and the structure state matrix of any neighbor individual is used to represent the stacking state matrix of the gravity blocks in the gravity energy storage system after the gravity energy storage system executes the charging and discharging strategy corresponding to the any neighbor individual; determining an optimal charging and discharging strategy of the gravity energy storage system based on the Pareto optimal solution set, so as to control the operation of the gravity energy storage system based on the optimal charging and discharging strategy.

2. The method of claim 1, wherein, The energy storage parameters include system efficiency, system maximum power and system maximum energy storage capacity of the gravity energy storage system. The construction of the energy storage optimization function with load balance, maximum operation benefit and minimum charging and discharging times as optimization objectives according to the energy storage parameters comprises the following steps: obtaining electricity prices at different time periods, and generating an operation benefit optimization function according to the system efficiency and the electricity prices at different time periods; constructing a load balance optimization function and a charging and discharging times optimization function of the gravity energy storage system; constructing charging and discharging constraint conditions based on the system maximum power and the system maximum energy storage capacity; composing the energy storage optimization function by using the operation benefit optimization function, the load balance optimization function, the charging and discharging times optimization function and the charging and discharging constraint conditions.

3. The method of claim 1, wherein, The iterative genetic operations on the initial population by using the improved non-dominated sorting genetic algorithm comprise the following steps: initializing the iteration number n and obtaining the individual population at the nth iteration, wherein the initial value of n is 1, and when n is 1, the individual population at the nth iteration is the initial population; performing individual genetic operations on the individual population at the nth iteration to obtain a child population; composing a candidate population by using the child population and the individual population at the nth iteration; calculating the domination tolerance and the energy storage optimization function value of each candidate individual in the candidate population; performing non-dominant sorting processing on the candidate population by using the domination tolerance and the energy storage optimization function value of each candidate individual, so as to obtain a plurality of non-dominant sorting sets; The difference degree of the energy storage optimization function value and the difference degree of the structure state matrix of each candidate individual in each non-dominated sorting set are calculated, and the joint crowding degree of each candidate individual is calculated according to the difference degree of the energy storage optimization function value and the difference degree of the structure state matrix of each candidate individual and the corresponding neighbor individual; According to the joint crowding degree of each candidate individual, individual selection processing is performed on the plurality of non-dominated sorting sets to obtain a new individual population of the same size as the initial population, and the new individual population is taken as the individual population at the n+1th iteration; It is judged whether the iteration stop condition is met or not; If not, n is increased by 1, and the individual population at the n th iteration is reacquired until the iteration stop condition is met, so that the charging and discharging strategy corresponding to each individual in the new individual population when the iteration stop condition is met is taken as the Pareto optimal solution set.

4. The method of claim 3, wherein, The individual genetic operation is performed on the individual population at the n th iteration to obtain a child population, including: The energy storage optimization function value of each individual in the individual population at the n th iteration is calculated, and the fitness of each individual is calculated according to the energy storage optimization function value of each individual; The minimum fitness is selected from the fitness of each individual, and the average fitness at the n th iteration is determined; According to the minimum fitness and the average fitness, the mutation probability of each individual at the n th iteration is calculated; The maximum iteration number is acquired, and the population crossover probability at the n th iteration is calculated according to the maximum iteration number; Based on the population crossover probability and the mutation probability of each individual at the n th iteration, the individual crossover operation and the individual mutation operation are sequentially performed on the individual population at the n th iteration, so that the child population is obtained after the individual mutation operation.

5. The method of claim 4, wherein, According to the minimum fitness and the average fitness, the mutation probability of each individual at the n th iteration is calculated, including: For any individual in the individual population at the n th iteration, the mutation probability of the individual at the n th iteration is calculated by using the following formula (1); （1）； In the above equation (1), denotes the mutation probability of the respective individual at the nth iteration, denotes the maximum mutation probability, denotes the fitness of the respective individual, denotes the minimum fitness and the average fitness, respectively, denotes a positive number; Accordingly, the population crossover probability at the n th iteration is calculated according to the maximum iteration number, including: The crossover probability at the n th iteration is calculated according to the following formula (2); （2）； In the above equation (2), denotes the population crossover probability at the n-th iteration, denotes the maximum crossover probability, denotes the maximum number of iterations.

6. The method of claim 4, wherein, Based on the population crossover probability and the mutation probability of each individual at the n th iteration, the individual crossover operation and the individual mutation operation are sequentially performed on the individual population at the n th iteration, so that the child population is obtained after the individual mutation operation, including: For any individual in the individual population at the n th iteration, a first random number is generated; It is judged whether the first random number is less than or equal to the population crossover probability; If yes, three mutually different individuals are randomly selected from the individual population at the n th iteration; According to the selected three mutually different individuals, the individual is crossed to obtain a crossover individual corresponding to the individual, and a crossover population is obtained after all individuals are polled; For any crossover individual in the crossover population, a second random number is generated; It is judged whether the second random number is less than or equal to the mutation probability of the crossover individual at the n th iteration; If yes, the individual is mutated to obtain a mutated individual corresponding to the individual, and the mutated individual is added to the crossover population. If yes, the any crossed individual is subjected to mutation operation based on a Levy step length, to obtain a mutated individual corresponding to the any crossed individual, and after polling all the crossed individuals, all the mutated individuals are used to form the offspring population.

7. The method of claim 3, wherein, The energy storage optimization function comprises three optimization functions, wherein the dominance tolerance is calculated, including: According to the three optimization functions in the energy storage optimization function, the function values of the three optimization functions corresponding to each candidate individual in the candidate population are calculated; From the function values of any optimization function corresponding to each candidate individual, the maximum function value and the minimum function value are screened out; According to the maximum function value and the minimum function value, the sub-dominance tolerance of the any optimization function is calculated, and after polling all the optimization functions, the sub-dominance tolerance of each optimization function is obtained; The sub-dominance tolerance of each optimization function is used to form the dominance tolerance.

8. The method of claim 3, wherein, The energy storage optimization function value of any candidate individual comprises the function values of the three optimization functions, and the dominance tolerance comprises the sub-dominance tolerances corresponding to the three optimization functions; Wherein, the candidate population is subjected to non-dominant sorting processing by using the dominance tolerance and the energy storage optimization function value of each candidate individual, to obtain a plurality of non-dominant sorting sets, including: Initialize the non-dominant sorting set of the u-th level as an empty set, wherein the initial value of u is 1; The b-th candidate individual is selected from the candidate population. Based on the function values ​​of the three optimization functions corresponding to the b-th candidate individual, and the sub-domination tolerance corresponding to each optimization function, it is determined whether the b-th candidate individual meets the domination determination condition, wherein the domination determination condition is: And exist Make , , This represents the b-th candidate individual corresponding to the first... The function value of an optimization function. Indicates the first element in the target set The first target individual corresponding to the first The function value of an optimization function, the objective set containing all candidate individuals in the candidate population except for the b-th candidate individual, and , Indicates the total number of target individuals; If yes, the b-th candidate individual is added to the non-dominant sorting set of the u-th level; b is incremented by 1, and the b-th candidate individual is selected again from the candidate population, until b is equal to B, to obtain the non-dominant sorting set of the u-th level, wherein the initial value of b is 1, and B is the population size of the candidate population; The candidate individuals in the non-dominant sorting set of the u-th level are removed from the candidate population to obtain a new candidate population, and the candidate population is updated as the new candidate population; u is incremented by 1, and the non-dominant sorting set of the u-th level is reinitialized as an empty set, until all candidate individuals in the candidate population are divided, to obtain a plurality of non-dominant sorting sets.

9. The method of claim 3, wherein, The energy storage optimization function value of any candidate individual comprises the function values of the three optimization functions; Wherein, the difference degree of the energy storage optimization function values of each candidate individual in each non-dominant sorting set and the corresponding neighbor individual is calculated, including: For any candidate individual in any non-dominant sorting set, K neighbor individuals corresponding to the any candidate individual are obtained; According to the function values of the three optimization functions corresponding to the any candidate individual and the function values of the three optimization functions corresponding to each neighbor individual, and according to the following formula (3), the difference degree of the energy storage optimization function values of the any candidate individual and the corresponding neighbor individual is calculated; （3）； In the above formula (3), This represents the degree of difference between the energy storage optimization function value of any candidate individual and its corresponding nearest neighbor individual. Represents the first individual corresponding to any of the candidate individuals. The function value of an optimization function. Represents the k-th nearest neighbor individual corresponding to the k-th nearest neighbor individual. The function value of an optimization function. Indicates the first The weight coefficients of each optimization function.

10. The method of claim 3, wherein, The energy storage optimization function value of any candidate individual comprises the function values of the three optimization functions, wherein the difference of the structure state matrix of each candidate individual in each non-dominant sorting set and the corresponding neighbor individual is calculated, including: For any candidate individual in any non-dominated sorting set, a pile-up state matrix of gravity blocks in the gravity energy storage system is determined after the gravity energy storage system executes a charging and discharging strategy corresponding to the any candidate individual, so as to take the pile-up state matrix as a structure state matrix of the any candidate individual; A structure state matrix of each neighbor individual corresponding to the any candidate individual is obtained; According to the structure state matrix of the any candidate individual and the structure state matrices of the neighbor individuals corresponding to the any candidate individual, a difference degree between the structure state matrix of the any candidate individual and the structure state matrices of the neighbor individuals corresponding to the any candidate individual is calculated according to the following formula (4). （4）； In the above formula (4), This represents the degree of difference between the structural state matrix of any candidate individual and its corresponding nearest neighbor individual. This represents the first element in the structural state matrix corresponding to any candidate individual. Line 1 The elements of the column, and used to characterize the first element in a gravity energy storage system. Layer The number of gravity blocks in the column, This represents the structure state matrix corresponding to the k-th nearest neighbor individual. Line 1 Column elements, These represent the total number of rows and columns of the structural state matrix corresponding to any candidate individual, respectively. This represents the total number of neighboring individuals.

Images