Energy storage cluster internal power optimization distribution method based on master-slave coevolution algorithm

Abstract

The invention discloses an energy storage cluster internal power optimal distribution method based on a master-slave coevolution algorithm, and the method comprises the steps: determining the participation state and circulation depth of each lithium battery energy storage, determining each energy storage distribution coefficient under the circulation depth, and selecting the participation state of each lithium battery energy storage and each energy storage distribution coefficient as variables to be optimized; based on the to-be-optimized variables, constructing an energy storage power optimization model taking the minimum operation cost, the minimum abandoned wind and light quantity and the minimum loss load quantity as optimization targets; for a coupling relation of two optimization variables in power distribution, a master-slave co-evolution method suitable for energy storage internal power distribution is provided, the two optimization variables are optimized and solved, and a final Pareto optimal individual is determined. Compared with the prior art, the method has the advantages that the variable dimension and the solving complexity can be reduced, the multi-objective optimization problem in the distribution process can be solved, synchronous optimization of the operation cost, the new energy consumption and the load supply is realized, and the rationality of the optimal solution is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of optimized operation of new power systems, and specifically relates to a method for optimizing internal power distribution of an energy storage cluster based on a master-slave collaborative evolutionary algorithm. Background Art

[0002] Large-scale energy storage systems are important flexible resources that support the flexible operation of new power systems. At present, the day-ahead optimization scheduling of large-scale energy storage systems mainly focuses on the planned power of the energy storage cluster as a whole, and pays less attention to the charging and discharging characteristics of each sub-energy storage within the cluster. Considering the economic requirements of power distribution within the energy storage cluster, it is necessary to model the cost per kilowatt-hour of each sub-energy storage more accurately on the premise that the scheduling results are feasible. However, the current problem is that if the impact of cycle depth and discharge power on the life of energy storage is considered in depth, the cost per kilowatt-hour of the entire life cycle will be nonlinearly related to variables such as energy storage power and cycle depth; at the same time, when optimizing the overall power allocation, it should be ensured that the allocation results meet the overall requirements of day-ahead scheduling for large-scale energy storage systems, and avoid each sub-energy storage from entering an extreme operating state as much as possible, which will also add more complex constraints. However, the linear programming algorithm in the existing technology cannot cope with nonlinear constraints, it is difficult to use the coordination relationship between sub-energy storage, it cannot preset reasonable allocation conditions for model dimensionality reduction, lacks a rapid decision-making mechanism, is prone to some extreme solutions and dimensionality disaster problems, and cannot reasonably correct the operating boundaries and constraints based on the current optimization results. At the same time, in modern power systems, dispatching objectives are becoming increasingly diverse, and traditional linear programming algorithms are also difficult to achieve parallel optimization of multiple objectives. Therefore, linearization algorithms are no longer applicable to internal power optimization allocation problems.

[0003] For multivariate nonlinear power optimization allocation problems, intelligent algorithms represented by evolutionary algorithms and particle swarm algorithms have better flexibility. For various nonlinear relationships and logical judgment problems, search modes and decision-making mechanisms can be preset, and multiple target values ​​can be calculated at the same time and used as screening basis, thereby achieving multi-objective parallel optimization. The optimization results can be stored. When there are slight changes in the system, there is no need to re-solve the problem, and the optimization can be iterated directly on the original optimal solution. Summary of the invention

[0004] Purpose of the invention: In response to the problems pointed out in the background technology, the present invention discloses a method for optimizing power allocation within an energy storage cluster based on a master-slave collaborative evolution algorithm. The traditional collaborative evolution algorithm is improved by combining the multivariable coupling relationship, and a new intelligent algorithm - the master-slave collaborative evolution algorithm is proposed. The algorithm is used for power allocation within an energy storage cluster, providing technical support for optimizing power allocation of large-scale energy storage clusters in new power systems.

[0005] Technical solution: The present invention discloses a method for optimizing power allocation within an energy storage cluster based on a master-slave collaborative evolutionary algorithm, comprising the following steps:

[0006] Step 1: Combined with the optimal operation mode of energy storage, determine the participation state and cycle depth of each lithium battery energy storage, the participation state represents a virtual variable indicating whether each lithium battery participates in a certain cycle, and the cycle depth is the ratio of the cycle capacity to the rated capacity; determine the distribution coefficient of each energy storage under the cycle depth, that is, the ratio of its own cycle capacity to the cycle capacity of all energy storage, and select the participation state J of each lithium battery energy storage and the distribution coefficient Coe of each energy storage as the variables to be optimized;

[0007] Step 2: Based on the variables to be optimized determined in step 1, a power optimization model is constructed, wherein the power optimization model aims to optimize the energy storage power by minimizing the operating cost, minimizing the amount of wind and solar power abandoned, and minimizing the amount of load loss;

[0008] Step 3: For the coupling relationship between the participation state J of each lithium battery energy storage in the power allocation in step 1 and the allocation coefficient Coe of each energy storage, the variable representing whether each energy storage participates in the cycle is taken as the dominant population, and the power allocation coefficient is taken as the subordinate population. The power allocation coefficient is generated on the basis of the variable of whether each energy storage participates in the cycle. With the optimization objective in step 2, the two variables to be optimized are optimized and solved to determine the final Pareto optimal individual.

[0009] Furthermore, in step 1:

[0010] The participation status J of each lithium battery energy storage represents a virtual variable indicating whether each lithium battery participates in a certain cycle;

[0011] The distribution coefficient Coe of each energy storage represents the distribution coefficient of each lithium battery in a certain cycle;

[0012] The two variables J and Coe are not independent. Only when the i-th energy storage participates in the k-th cycle, J i,k =1, Coe i,k The value of Coe is meaningful. i,k Should be in a specific J i,k The variables are divided into two groups. The virtual variable J of the participation status is the dominant population, and the distribution coefficient Coe is its subordinate population. The value of Coe is determined by J, and there is a master-slave relationship between the two groups.

[0013] Furthermore, the specific process of determining the participation status of each lithium battery energy storage is as follows:

[0014] Step 1.1: First, assume that only lithium batteries participate in the charge and discharge cycle, and determine the average charging efficiency of lithium batteries under the current circumstances based on the total output of all energy storage in the region. Charge energy to be processed Average discharge efficiency and the discharge energy to be processed And the average cycle depth of lithium battery energy storage Calculate the cost per kilowatt-hour of each lithium battery based on the average cycle depth;

[0015] Step 1.2: Among the compressed air energy storage and hydrogen energy storage that are not currently involved in the cycle, determine whether there are individuals with a kilowatt-hour cost lower than the kilowatt-hour cost of a lithium battery, add them to this charge and discharge cycle in order of their kilowatt-hour costs from small to large, and make them run at full power. The remaining power is allocated to all lithium batteries, and the kilowatt-hour cost of lithium batteries is recalculated until the costs of all hydrogen energy storage and compressed air energy storage are higher than the kilowatt-hour cost of any lithium battery;

[0016] Step 1.3: Further determine the participation status of lithium batteries based on the participation of hydrogen energy storage and compressed air energy storage in the charge and discharge cycle.

[0017] Furthermore, the process of calculating the cost per kilowatt-hour of each lithium battery is as follows:

[0018] The total planned output of all large-scale energy storage in this area obtained by scheduling on the day before is P t SUM , 1≤t≤nT, n is the number of energy storage cycles per year, T is the energy storage cycle, and the planned wind and solar power abandonment is P t cur , 1≤t≤nT, then the total output of all regional energy storage is P t ideal :

[0019] P t ideal =P t SUM -P t cur , 1≤t≤nT

[0020] If only lithium batteries are involved in the cycle, the average charging efficiency Charge energy to be processed Average discharge efficiency and the discharge energy to be processed Make an estimate:

[0021]

[0022] Among them, Ce i , η ci , η di Respectively represent the rated capacity, charging efficiency, and discharging efficiency of the i-th energy storage, T C The subscript set of all charging times in this cycle, TD Represents the subscript set of all discharge moments;

[0023] Average cycle depth of lithium battery energy storage And the cost per kilowatt-hour is as follows:

[0024]

[0025] in, represents the available capacity of the i-th energy storage in the current cycle; p i represents the calculated cost per kWh of the i-th energy storage at the current cycle depth; M is a maximum value, indicating that when the average cycle depth is greater than 1, the lithium battery energy storage can no longer meet the capacity requirements, and it is not feasible to use only lithium batteries for charging and discharging, taking a maximum penalty cost; f i p Represents the functional relationship between the i-th energy storage cycle depth and the cost per kilowatt-hour.

[0026] Furthermore, the average cycle depth and the cost per kilowatt-hour are updated as follows:

[0027]

[0028] in, represents the updated average cycle depth of lithium battery energy storage, p i ' represents the updated cost per kWh of the i-th energy storage, and G represents the number of the added compressed air energy storage or hydrogen energy storage.

[0029] Furthermore, the power optimization model in step 2 is specifically as follows:

[0030]

[0031] Among them, f c represents the operating cost, f cur Indicates the amount of abandoned wind power, f loss Indicates the amount of load loss, dod i.k represents the cycle depth in the kth cycle of the i-th storage day, DOD N,i represents the rated cycle depth of the i-th energy storage, CIR represents the number of cycles on the day; g(·) represents the cost coefficient, which is a function of the cycle depth; LCOS represents the cost per kilowatt-hour considering the influence of the cycle depth, It indicates the rated cost of electricity when the energy storage is in rated operation state, ignoring energy loss; m is the number of energy storage units, N is the rated value, P t ideal ' is the ideal power of the energy storage cluster to meet the dispatching demand at time t, P t up represents the maximum adjustable power of all controllable units in the region at time t, Pt down Indicates the maximum adjustable power of all controllable units in the area at time t, P t act It represents the energy storage regulation power, and Δt is the energy storage regulation time.

[0032] Furthermore, the specific process of optimizing and solving the two optimization variables using the master-slave collaborative evolution method in step 3 is as follows:

[0033] Step 3.1: Set the initial number of individuals in the population to npop. First, initialize the dominant population J, randomly generate npop individuals, and then generate npop subordinate population individuals Coe based on each dominant population individual J.

[0034] Step 3.2: Initialize the dominant population individuals, randomly determine whether the i-th energy storage participates in the k-th cycle, and i,k Assign values ​​to determine whether the energy storage capacity participating in the cycle can meet the capacity and power requirements. If so, output the dominant population individual; if not, find the dominant population individual that currently decides not to participate in the k-th cycle and re-randomly assign values ​​to its participation status virtual variable;

[0035] Step 3.3: Initialization of subordinate population individuals. Under each dominant population individual, the Coe of its subordinate population individuals is initialized, including randomly generating the initialization cycle depth of each energy storage near the average cycle depth, determining the initial distribution coefficient of each energy storage under the initial cycle depth, that is, the ratio of its own cycle capacity to the cycle capacity of all energy storage, correcting the distribution coefficient, normalizing the distribution coefficient, and updating the energy storage power, energy, and cycle depth based on the distribution coefficient;

[0036] Step 3.4: Crossover operation, which allows subordinate populations under the same dominant population individual to cross internally. After the crossover, the power, energy, and cycle depth of each energy storage under the allocation method are updated. Then, the dominant population individual selects excellent subordinate population individuals to form a complete individual, and crosses between the complete individuals.

[0037] Step 3.5: Mutation operation: first mutate the subordinate populations under each dominant population individual, then mutate the dominant population. After the individuals in the dominant population mutate, the subordinate populations need to be adjusted accordingly. After the two mutations, the energy storage power, energy, and cycle depth are updated.

[0038] Step 3.6: Selection and creation. First, the distribution coefficient of each subordinate population individual is used to obtain the energy storage power, energy, and cycle depth of each energy storage. Then, the operating cost, abandoned wind and solar power, and load loss corresponding to each subordinate population are calculated respectively. Then, the non-dominated sorting algorithm is used to select npopc1 excellent individuals. Under each dominant population individual, npopc2 subordinate population individuals are randomly added. The generation method is the same as the initialization of the subordinate population individuals. After the subordinate population is screened, the dominant population is screened. First, each dominant population individual is ordered to select the Pareto optimal individual from its subordinate population to form a complete individual. Then, the complete individual is screened using the non-dominated sorting algorithm to screen out npopj1 excellent individuals. If all the complete individuals formed by a dominant population individual are not selected, the dominant population individual and its corresponding subordinate individuals are deleted. Then, npopj2 dominant population individuals are randomly added, and each dominant population individual is randomly equipped with npopc2 subordinate population individuals. The generation method is the same as the initialization of the dominant population individuals.

[0039] Step 3.7: After completing the screening and creation of new individuals, the existing population is used as the parent generation for iteration again until the termination condition is met, and the final two optimization variables and their corresponding power, energy, and cycle depth are output.

[0040] Furthermore, the subordinate population individual initialization specifically includes the following process:

[0041] Step 3.1.1: Initialize loop depth

[0042] Generate the initial cycle depth of each energy storage randomly around the average cycle depth:

[0043]

[0044] Among them, dod i,k ' represents the initial cycle depth, δ represents the maximum fluctuation amplitude, U() is a random uniform distribution function, represents the average cycle depth of all energy storage participating in the kth cycle;

[0045] Step 3.1.2: Initialize the distribution coefficient:

[0046]

[0047] Step 3.1.3: Correct the initial partition coefficient:

[0048]

[0049] Among them, Pe i is the effective power of the i-th energy storage, α m is the margin factor, and denote the capacity requirement and power requirement of the kth cycle respectively, Take the power with the largest absolute value in the cycle; Coe i,k " is the corrected distribution coefficient, Coe i,k ' is the initial distribution coefficient, J i,k is a dummy variable indicating whether the i-th energy storage participates in the k-th cycle;

[0050] Step 3.1.4: Normalization of the partition coefficient:

[0051]

[0052] Step 3.1.5: Update the energy storage power, energy, and cycle depth based on the allocation coefficient:

[0053]

[0054] Among them, T k represents the cycle period corresponding to the kth cycle of energy storage, P t ideal ' is the ideal power of the energy storage cluster to meet the dispatching demand at time t, It represents the power obtained by the i-th energy storage at time t according to the distribution coefficient, Indicates the discharge power allocated to it, Represents the charging power allocated to it, and calculates the energy stored in the energy storage at each moment:

[0055]

[0056] in, represents the energy stored at time t and t-1 of the i-th energy storage, η ci represents the charging efficiency of the i-th energy storage, η di represents the discharge efficiency of the i-th energy storage;

[0057] The power and stored energy are further corrected and the depth of each storage cycle is calculated:

[0058]

[0059] in, and They represent the charging power and discharging power of the i-th energy storage at time t after correction, represents the corrected stored energy at the ith energy storage time t, Indicates the corrected cycle depth.

[0060] Furthermore, the crossover operation is specifically as follows:

[0061] Step 3.4.1: Slave population crossover, when the slave population Coe performs internal crossover:

[0062]

[0063] in, is the intermediate variable of the distribution coefficient, represents the distribution coefficient of the kth stage of the i-th energy storage in the new individual formed after crossover, and They represent the distribution coefficients of the kth stage of the i-th energy storage in the two parent individuals, and CS represents the energy storage number selected for exchange:

[0064] CS=randsample(CR·npop,npop)

[0065] Among them, randsample(n,m) is a sampling function, which means randomly selecting n integers from 1 to m, and CR represents the crossover rate;

[0066] Step 3.4.2: Dominant population crossover:

[0067]

[0068] in, Indicates the dummy variable representing whether the i-th energy storage in the new individual formed by the crossover participates in the k-th cycle, and Represents the corresponding dummy variable of the selected parent individual.

[0069] Furthermore, the specific process of the mutation operation is as follows:

[0070] Step 3.5.1: Subordinate population variation, within the subordinate group led by the same dominant individual:

[0071]

[0072] in, is the intermediate variable of the distribution coefficient, represents the maximum value of the distribution coefficient of the i-th energy storage in the k-th stage, represents the distribution coefficient of the i-th energy storage in the new individual formed after mutation at the k-th stage, represents the distribution coefficient of the i-th energy storage in the selected parent individual at the k-th stage, δ M represents the variable step length, U() is a random uniform distribution function, and MS represents the set of energy storage numbers selected for mutation:

[0073] MS=randsample(MR·npop,npop)

[0074] Among them, randsample(n,m) is a sampling function, which means randomly selecting n integers from 1 to m, and MR represents the mutation rate;

[0075] Step 3.5.2: Dominant population mutation, the process is as follows:

[0076] (a) Randomly regenerate the participating state variables of the variable energy storage:

[0077]

[0078] Among them, rand(0,1) represents a random value between 0 and 1. If it is 1, it means that the i-th energy storage participates in the k-th cycle, and if it is 0, it means that it does not participate; A dummy variable indicating whether the i-th energy storage participates in the k-th cycle after the mutation;

[0079] (b) Determine whether the energy storage capacity involved in the cycle can meet the capacity demand and power demand. If the following conditions are met at the same time, the output of individual Otherwise, proceed to operation (c), with the following conditions:

[0080]

[0081] in, and They represent the capacity demand and power demand of the kth cycle, Pe i is the effective power of the i-th energy storage, Ce i represents the rated capacity of the i-th energy storage;

[0082] (c) Find the individuals in MS that are not participating in the kth cycle, re-randomly select the values ​​of their participation state variables, and then go to operation (b):

[0083]

[0084] in, represents the set of energy storage subscripts in the MS that currently decides not to participate in the kth cycle;

[0085] (d) Regenerate the distribution coefficient of the stored energy in the MS:

[0086]

[0087] in, represents the average cycle depth of energy storage in MS at the kth cycle, represents the initial cycle depth of the i-th energy storage in the MS at the k-th cycle, Indicates the use of Determine the maximum allocation coefficient of each energy storage, It represents the final distribution coefficient of the mutated new individual, and δ represents the maximum fluctuation range.

[0088] Beneficial effects:

[0089] 1. The present invention improves the traditional co-evolution algorithm by combining the multivariable coupling relationship, and then proposes a new intelligent algorithm - the master-slave co-evolution algorithm, and uses it for power distribution within the energy storage cluster, providing technical support for the power optimization distribution of large-scale energy storage clusters in new power systems.

[0090] 2. The present invention combines the optimal operation mode of energy storage to determine the optimization direction and decision-making method, reduce the variable dimension, and avoid extreme solutions or infeasible solutions. When determining the variable dimension, by setting the virtual variables of whether the compressed air energy storage and hydrogen energy storage variables to be optimized participate in this round of charge and discharge cycle, further determine whether the lithium battery energy storage participates in this round of charge and discharge cycle. When determining the participation status, it is determined whether the compressed air energy storage and hydrogen energy storage have a lower cost per kilowatt-hour. If so, the one with the lowest cost is added to the charge and discharge cycle and made to run at full power. The remaining power is allocated to all lithium batteries, and the lithium battery cost per kilowatt-hour is recalculated. In this way, the costs of all hydrogen energy storage and compressed air energy storage are higher than the cost per kilowatt-hour of any lithium battery. In this way, it can be determined whether hydrogen energy storage and compressed air energy storage participate in the operation. In the later optimization process, only two types of variables to be optimized for lithium batteries need to be considered, which reduces the optimization process.

[0091] 3. The cycle depth of the present invention is the ratio of the current charge and discharge capacity to the rated capacity, and the allocation coefficient is the ratio of the overall power of the energy storage cluster to each energy storage device in the cluster. The cycle depth is used to calculate the cost per kilowatt-hour, because the cost per kilowatt-hour of a lithium battery changes with the cycle depth, and the allocation coefficient must not only consider the cost but also match the overall power of the energy storage. The cycle depth will affect the cost per kilowatt-hour and the power allocation coefficient, and there is a coupling relationship between them. The coupling relationship between the two facilitates the use of a master-slave co-evolutionary algorithm for optimization in the present invention. By analyzing the relationship between various types of energy storage charge and discharge power, cycle depth, cycle life, and cost per kilowatt-hour, the present invention proposes an operating cost accounting method under different cycle depths, and pre-processes the power conditions of compressed air and hydrogen energy storage based on the cost per kilowatt-hour, effectively reducing the variable dimension and computational complexity to be optimized.

[0092] 4. Aiming at the coupling relationship between the two major optimization variables in power distribution, the present invention takes the variables representing whether each energy storage participates in the cycle as the dominant population, and the power distribution coefficient as the subordinate population. The latter is generated on the basis of the former. The proposed master-slave collaborative evolution algorithm suitable for energy storage power distribution can cope with the multi-objective optimization problems in the distribution process, efficiently solve the multi-objective optimization problems, and realize the synchronous optimization of operating costs, new energy consumption and load supply. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a flowchart of the master-slave collaborative evolution algorithm framework of the present invention;

[0094] Figure 2 This is a relationship diagram between the energy storage cycle depth and cycle life of a typical lithium battery of the present invention;

[0095] Figure 3 The energy storage energy and power of the embodiment of the present invention;

[0096] Figure 4 The optimal solution corresponding to the running cost graph generated in each round of iteration by the master-slave collaborative evolution algorithm in the embodiment of the present invention;

[0097] Figure 5 This is a diagram of the energy storage power after optimization using the master-slave collaborative evolution algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0098] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0099] The present invention discloses a method for optimizing power allocation within an energy storage cluster based on a master-slave collaborative evolutionary algorithm, comprising the following steps:

[0100] Step 1: Combined with the optimal operation mode of energy storage, determine the participation state and cycle depth of each lithium battery energy storage, the participation state represents a virtual variable indicating whether each lithium battery participates in a certain cycle, and the cycle depth is the ratio of the cycle capacity to the rated capacity; determine the distribution coefficient of each energy storage under the cycle depth, that is, the ratio of its own cycle capacity to the cycle capacity of all energy storage, and select the participation state J of each lithium battery energy storage and the distribution coefficient Coe of each energy storage as the variables to be optimized.

[0101] The variables to be optimized in traditional energy storage models usually include the storage energy, charge and discharge status, charging power, discharging power and state conversion indicator variables of each energy storage at each time. There are many variables. If the same modeling method is used in the intelligent algorithm, it is difficult to give full play to the flexibility of the intelligent algorithm. Therefore, it is necessary to first combine the optimal operation mode of energy storage, determine the optimization direction and decision-making method, reduce the variable dimension, and avoid extreme solutions or infeasible solutions.

[0102] The present invention sets the lithium battery energy storage to include two types of variables to be optimized: (1) a virtual variable indicating whether to participate in the current charge and discharge cycle; and (2) the cycle depth of the energy storage in the current cycle. At the same time, the compressed air energy storage and hydrogen energy storage variables to be optimized are set as virtual variables indicating whether to participate in the current charge and discharge cycle.

[0103] In order to further simplify the variables to be optimized, the participation status of compressed air energy storage and hydrogen energy storage is analyzed in advance. The planned total output of all large-scale energy storage in this area obtained by scheduling on the day before is P t SUM (1≤t≤nT), n is the number of energy storage cycles per year, T is the energy storage cycle period, and the planned wind and solar power abandonment is P t cur (1≤t≤nT), then ideally the total output of all energy storage in this area is P t ideal It can be calculated according to formula (1):

[0104] P t ideal =P t SUM -P t cur , 1≤t≤nT (1)

[0105] If only lithium batteries are involved in the cycle, the average charging efficiency Charge energy to be processed Average discharge efficiency and the discharge energy to be processed It can be estimated using formulas (2) to (5) respectively:

[0106]

[0107] Among them, Ce i , η ci , η di Respectively represent the rated capacity, charging efficiency, and discharging efficiency of the i-th energy storage. C The subscript set of all charging times in this cycle, T D Represents the subscript set of all discharge times.

[0108] The average cycle depth of lithium battery energy storage at this time As shown in formula (6), the cost per kilowatt-hour of each lithium battery calculated according to this depth is shown in formula (7):

[0109]

[0110] in, represents the available capacity of the i-th energy storage in the current cycle; p i represents the calculated cost per kWh of the i-th energy storage at the current cycle depth; M is a maximum value, indicating that when the average cycle depth is greater than 1, the lithium battery energy storage can no longer meet the capacity requirements, and it is not feasible to use only lithium batteries for charging and discharging, taking a maximum penalty cost; f i p Represents the functional relationship between the i-th energy storage cycle depth and the cost per kilowatt-hour.

[0111] If there are individuals whose kilowatt-hour cost is less than the kilowatt-hour cost of a lithium battery in the compressed air energy storage and hydrogen energy storage that are not currently involved in the cycle, they will be added to the current cycle in order of kilowatt-hour cost from small to large, and the average cycle depth and kilowatt-hour cost of the lithium battery will be updated according to equations (8) and (9) until the kilowatt-hour cost of all individuals in the remaining energy storage is not less than the kilowatt-hour cost of any lithium battery:

[0112]

[0113] in, represents the updated average cycle depth of lithium battery energy storage, p i ' represents the updated cost per kWh of the i-th energy storage, and G represents the number of the added compressed air energy storage or hydrogen energy storage.

[0114] In the subsequent optimization, it is only necessary to determine the participation status and cycle depth of each lithium battery energy storage. In order to ensure that the sum of the power of each energy storage meets the overall power requirement, the power distribution coefficient among each energy storage can be determined first. The power of each lithium battery energy storage can be directly determined by multiplying the current power to be distributed by the respective distribution coefficient. The cycle capacity can be determined by multiplying the distribution coefficient by the capacity required for this cycle. The ratio of the cycle capacity to the rated capacity is the cycle depth.

[0115] Finally, there are two variables to be optimized in the algorithm: (1) the participation status J of each lithium battery energy storage represents a virtual variable indicating whether each lithium battery participates in a certain cycle; (2) the allocation coefficient Coe of each energy storage represents the allocation coefficient of each lithium battery in a certain cycle. The two variables are not independent. Only when the i-th energy storage participates in the k-th cycle, that is, J i,k =1, Coe i,k The value of Coe is meaningful. i,k Should be in a specific J i,kTherefore, the variable can be divided into two groups. The virtual variable J of the participation status is the dominant population, and the distribution coefficient variable Coe is its subordinate population. The value of Coe needs to be determined according to J. The two groups have a master-slave relationship, that is, they can evolve internally and cooperate with each other to evolve together.

[0116] Step 2: Based on the variables to be optimized determined in step 1, a power optimization model is constructed, wherein the power optimization model aims to optimize the energy storage power by minimizing the operating cost, minimizing the amount of wind and solar power abandoned, and minimizing the amount of load loss.

[0117] When redistributing lithium battery power, we should, under the premise of ensuring load supply, absorb as much as possible the new energy output that was abandoned in the previous scheduling, while reducing the system operating cost. Therefore, the optimization goals include: minimizing operating costs, minimizing the amount of abandoned wind and solar power, and minimizing the amount of lost load.

[0118] (1) Minimize operating costs

[0119] The cycle depth of lithium battery energy storage can affect the cycle life, which in turn affects the amount of power that can be processed during the entire life cycle, and ultimately affects the cost of electricity. The relationship between the cycle life and cycle depth of a typical lithium battery is as follows: Figure 2 shown.

[0120] Power functions, exponential functions or polynomials are often used to fit the relationship between cycle depth and battery life. Their mathematical models are shown in equations (10) to (12):

[0121]

[0122] Among them, α 1 , α 2 , β 1 , β 2 are model parameters, a k Represents the coefficient of the k-th term in the polynomial, K represents the highest degree, L N Indicates the cycle life of lithium batteries at rated cycle depth, DOD N represents the rated cycle depth, and dod represents the cycle depth during actual operation.

[0123] Assuming that the number of daily cycles remains unchanged, the service life T of lithium battery energy storage can be calculated according to formula (13):

[0124]

[0125] Assuming that the annual maintenance cost M(dod) and the annual number of cycles n do not change with the year, the calculation method of the cost per kilowatt-hour considering the influence of the cycle depth is as follows:

[0126]

[0127] Among them, C E , C P They represent the unit capacity cost and unit power cost of energy storage construction, d represents the discharge time, that is, the ratio of energy storage capacity to rated power, r represents the discount rate, η represents the total cycle efficiency, and P c represents the unit charging cost, P punish It represents the penalty cost corresponding to the loss of unit energy due to energy loss during the energy storage cycle.

[0128] The second term in the above equation represents the cost of energy loss caused by charging and processing the lithium battery. It has nothing to do with the cycle depth and is a constant value, denoted as P IN , and the first term is related to the loop depth, denoted as P VAR (dod), the cost per kilowatt-hour can be simplified as formula (15):

[0129] LCOS=P VAR (dod)+P IN (15)

[0130] The total energy storage cost in one cycle is as follows:

[0131]

[0132] in, represents the energy stored in the i-th lithium battery in one cycle, dod i Represents the cycle depth of energy storage of the i-th lithium battery.

[0133] The rated cost per kilowatt-hour of energy storage under rated operating conditions, ignoring energy loss, is shown in formula (17).

[0134]

[0135] The P of lithium battery VAR (dod) can be quickly calculated using the respective cycle depth and rated cost per kWh according to formula (18):

[0136]

[0137] Among them, g(dod) represents the cost coefficient, which is a function of the cycle depth, as shown in formula (19):

[0138]

[0139] Where T(dod) represents the function of the energy storage cycle period with respect to the cycle depth variable dod.

[0140] Therefore, when optimizing the total operating cost, the sub-objective function is as follows:

[0141]

[0142] In the formula, m is the number of energy storage, N is the rated value, dod i.k represents the cycle depth in the kth cycle of the i-th energy storage, DOD N,i It represents the rated cycle depth of the i-th energy storage, and CIR represents the number of cycles on that day.

[0143] (2) Minimize the amount of wind and solar power abandoned and minimize the amount of load loss

[0144] The power of compressed air and hydrogen storage in this area at time t on the same day is P t H&A , then the ideal power P of the energy storage cluster to meet the dispatching demand at time t is t ideal Calculate according to formula (21):

[0145] P t ideal '=P t SUM -P t cur -P t H&A , 1≤t≤nT (21)

[0146] In the process of optimizing allocation, the two types of controllable units, hydropower and thermal power, can adjust their output around the planned power without changing the planned start and stop status of the units, and use auxiliary energy storage for adjustment. When obtaining the adjustable range, the maximum power of the unit and the thermal power ramp limit should be considered. The maximum adjustable power of all controllable units in the area can be expressed by formula (22), and the adjustable power can be expressed by formula (23):

[0147]

[0148] Among them, P t up It represents the maximum adjustable power of all controllable units in the area at time t, P t down Indicates the maximum power that can be reduced for all controllable units in the area at time t, X i,t is a virtual variable representing the operating status of the i-th thermal power unit at time t, which is 1 when it is operating and 0 when it is shut down; i max The actual maximum power of the i-th thermal power plant, P i th,max represents the maximum power of the i-th thermal power unit, Indicates its minimum output ratio, r i It represents the climbing rate, represents the planned power of the ith thermal power unit at time t, nth represents the total number of thermal power units in the area, represents the maximum power that the i-th hydropower unit can provide at time t, represents the minimum power that the i-th hydropower unit can provide at time t, It represents the planned power of the ith hydropower unit at time t, and nh represents the total number of hydropower units in the area.

[0149] When the energy storage power satisfies formula (24), the deviation between the total energy storage output and the ideal output can be filled by the auxiliary adjustment of the controllable unit, so that at time t, neither wind nor solar power can be abandoned and the load power demand can be met:

[0150] P t ideal '-P t up ≤P t act ≤P t ideal '+P t down (twenty four)

[0151] P t act Indicates the energy storage regulation power. If P t act ≥P t ideal '+P t down , the energy storage discharge power is large or the charging power is small. Even if all controllable units are at the lowest power level that can be achieved at this moment, the output of new energy cannot be effectively utilized, resulting in abandoned wind and solar power. The sub-objective function of minimizing abandoned wind and solar power can be expressed by formula (25):

[0152]

[0153] When P t act ≤P t ideal '-P t up , the energy storage discharge power is too small or the charging power is too large. Even if all controllable units are at the highest power level that can be achieved at this moment, the overall output of the region will still be lower than the load demand, resulting in load loss. The sub-objective function of minimizing load loss can be expressed by formula (26):

[0154]

[0155] Among them, Δt is the energy storage adjustment time.

[0156] Step 3: For the coupling relationship between the two major control variables in the power allocation in step 1, a master-slave co-evolution method is proposed. The variables representing whether each energy storage participates in the cycle are taken as the dominant population, and the power allocation coefficient is taken as the subordinate population. The power allocation coefficient is generated on the basis of the variables representing whether each energy storage participates in the cycle. With the optimization objective in step 2, the two major optimization variables are optimized and solved to determine the final Pareto optimal individual.

[0157] Aiming at the coupling relationship between the two major control variables in power allocation, a master-slave co-evolution algorithm is proposed. The algorithm framework is as follows: Figure 1 As shown in the figure, the variables representing whether each energy storage participates in the cycle are taken as the dominant population, and the power allocation coefficient is taken as the subordinate population, and the latter is generated on the basis of the former. At the same time, the operation method is improved to reduce the variable dimension and solution complexity, and improve the rationality of the optimization solution.

[0158] 1. Initialization

[0159] Set the initial number of individuals in the population to npop, then first initialize the dominant population J, randomly generate npop individuals, and then based on each dominant population individual J, generate npop subordinate population individuals Coe as its subordinate population.

[0160] First, the initialization of the dominant population individuals:

[0161] (1) Randomly determine whether the i-th energy storage participates in the k-th cycle, and calculate J according to formula (27) i,k Assignment:

[0162]

[0163] Among them, J i,k A dummy variable indicating whether the i-th energy storage participates in the k-th cycle. rand(0,1) represents a random value between 0 and 1. If it is 1, it means that the i-th energy storage participates in the k-th cycle. If it is 0, it means that it does not participate.

[0164] (2) Determine whether the energy storage capacity involved in the cycle can meet the capacity and power requirements

[0165] If both equations (28) and (29) are satisfied, then the individual is output; otherwise, proceed to operation (3):

[0166]

[0167] in, and denote the capacity requirement and power requirement of the kth cycle respectively, The power with the largest absolute value in this cycle can be taken as shown in formula (30):

[0168]

[0169] Among them, P t ideal ' represents the ideal power of the energy storage cluster to meet the dispatching requirements at time t.

[0170] (3) Find the individual who currently decides not to participate in the kth cycle, and re-randomly select the value of its participation status dummy variable, as shown in formula (31), and then go to operation (2):

[0171]

[0172] in, Represents the set of subscripts of individuals that have not previously participated in the k-th cycle.

[0173] Second, the initialization of subordinate individuals. Under each dominant individual, the Coe of its subordinate population is initialized. The processing flow of each individual is as follows:

[0174] (1) Initialize loop depth

[0175] The initial cycle depth of each energy storage is randomly generated near the average cycle depth, as shown in formula (32):

[0176]

[0177] Among them, dod i,k ' represents the initial cycle depth, δ represents the maximum fluctuation amplitude, U() is a random uniform distribution function, represents the average cycle depth of all energy storage participating in the kth cycle, which can be calculated according to formula (33):

[0178]

[0179] (2) Initialize the allocation coefficient

[0180] The initial distribution coefficient of each energy storage is determined at the initial cycle depth, that is, the ratio of its own cycle capacity to the cycle capacity of all energy storage, as shown in formula (34):

[0181]

[0182] (3) Correction of initial distribution coefficient

[0183] In order to avoid unreasonable allocation coefficients, which may lead to the situation that after power allocation, the energy to be processed by some energy storage is greater than its own capacity, or the charging and discharging power is greater than its own rated power, the allocation coefficient needs to be corrected in one step, such as equations (35) and (36):

[0184]

[0185] Among them, Pei is the effective power of the i-th energy storage, α m is the margin coefficient, which prevents the capacity or power from exceeding the limit again after the allocation coefficient is normalized. i,k " is the corrected distribution coefficient, Coe i,k ' is the initial distribution coefficient, J i,k It is a dummy variable indicating whether the i-th energy storage participates in the k-th cycle.

[0186] (4) Normalization of distribution coefficient

[0187] In order to ensure that the sum of the normalized coefficients of all energy storage allocation coefficients is 1, so as to achieve the complete allocation of the total energy storage power, the allocation coefficients need to be normalized, as shown in formula (37):

[0188]

[0189] (5) Update energy storage power, energy, and cycle depth based on the allocation coefficient

[0190] After obtaining the allocation coefficient, the power allocated to each energy storage can be determined according to the allocation coefficient, as shown in equations (38) to (40):

[0191]

[0192] Among them, T k represents the cycle period corresponding to the kth cycle of energy storage, It represents the power obtained by the i-th energy storage at time t according to the distribution coefficient, Indicates the discharge power allocated to it, Represents the charging power allocated to it, and the energy stored in the energy storage at each moment can be calculated, as shown in formula (41):

[0193]

[0194] in, represents the energy stored at time t and t-1 of the i-th energy storage, η ci represents the charging efficiency of the i-th energy storage, η di Represents the discharge efficiency of the i-th energy storage.

[0195] In order to avoid the situation where the extreme difference of energy storage is greater than the capacity of the energy storage itself, resulting in some power being unable to be realized, it is necessary to further correct the power and storage energy, and calculate the depth of each energy storage cycle, as shown in equations (42) to (45). The corrected energy storage charging and discharging power and storage energy can meet the operating constraints of the energy storage itself, so there is no need to set a penalty function for violating the constraints.

[0196]

[0197]

[0198] in, and They represent the charging power and discharging power of the i-th energy storage at time t after correction, represents the corrected stored energy at the ith energy storage time t, Indicates the corrected cycle depth.

[0199] 2. Cross

[0200] The subordinate populations under the same dominant individual are allowed to cross internally, and then the individuals of the dominant population are allowed to select excellent subordinate individuals to form complete individuals, and cross between the complete individuals.

[0201] First, the subordinate population crossover. When the subordinate population Coe performs internal crossover, the specific method is as shown in equations (46) to (47):

[0202]

[0203] in, is the intermediate variable of the distribution coefficient, represents the distribution coefficient of the kth stage of the i-th energy storage in the new individual formed after crossover, and They represent the distribution coefficients of the kth stage of the i-th energy storage in the two parent individuals, and CS represents the energy storage number selected for exchange, which can be selected according to formula (48):

[0204] CS=randsample(CR·npop,npop) (48)

[0205] Among them, randsample(n,m) is a sampling function, which means randomly selecting n integers from 1 to m, and CR represents the crossover rate.

[0206] After completing the exchange of the allocation coefficients, the power, energy, and cycle depth of each energy storage under the allocation method must be updated according to equations (38) to (45).

[0207] Second, the dominant population crosses over. When the dominant population crosses over, it is necessary to first select a Pareto optimal individual from its subordinate population to form a complete solution. The method of determining the energy storage number for exchange is the same as that when the subordinate population crosses over. The specific method of the overall crossover is as shown in equations (49) to (51):

[0208]

[0209] in, Indicates the dummy variable representing whether the i-th energy storage in the new individual formed by the crossover participates in the k-th cycle, and Represents the corresponding dummy variable of the selected parent individual.

[0210] After the overall exchange is completed, the power, energy, and cycle depth of each energy storage must also be updated according to equations (38) to (45).

[0211] 3. Mutation

[0212] Mutation can increase the diversity of solutions and avoid optimization from entering the local optimum. The subordinate populations under each dominant population individual can be mutated first, and then the dominant population can be mutated. Since the dominant population will affect the subordinate population, after the individuals in the dominant population mutate, the subordinate population needs to be adjusted accordingly.

[0213] First, subordinate population mutation: within a subordinate group led by an individual from the same dominant population, new subordinate individuals can be generated according to equations (52) and (53).

[0214]

[0215]

[0216] in, is the intermediate variable of the distribution coefficient, represents the maximum value of the distribution coefficient of the i-th energy storage in the k-th stage, represents the distribution coefficient of the i-th energy storage in the new individual formed after mutation at the k-th stage, represents the distribution coefficient of the i-th energy storage in the selected parent individual at the k-th stage, δ M MS represents the number of the energy storage selected for mutation, and the selection method is as shown in formula (54).

[0217] MS=randsample(MR·npop,npop) (54)

[0218] Here, MR represents the mutation rate.

[0219] After the distribution coefficient is changed, the energy storage power, energy, and cycle depth must also be updated according to equations (38) to (45).

[0220] Second, the dominant population mutation. The mutation of the dominant population is relatively complex. Each dominant population individual needs to randomly select a Pareto optimal individual from its subordinate population to assist in its own mutation. The process is similar to the initialization of the dominant population individual, but only the corresponding variables of the mutated energy storage need to be operated. The process is as follows:

[0221] (1) The participating state variables of the variable energy storage are randomly regenerated according to formula (55):

[0222]

[0223] Among them, MS is the set of energy storage numbers to be mutated, and its selection method is the same as that of the subordinate population mutation. A dummy variable indicating whether the i-th energy storage participates in the k-th cycle after mutation.

[0224] (2) Determine whether the energy storage capacity involved in the cycle can meet the capacity and power requirements

[0225] If both equations (56) and (57) are satisfied, then the output of the individual Otherwise, proceed to operation (3).

[0226]

[0227]

[0228] (3) Find the individuals in the MS that are not participating in the kth cycle, re-randomly select the values ​​of their participation state variables according to equation (58), and then go to operation (2)

[0229]

[0230] in, Represents the energy storage index set in the MS that currently decides not to participate in the k-th cycle.

[0231] (4) The distribution coefficient of the stored energy in the MS is regenerated as shown in equations (59) to (62).

[0232]

[0233] in, represents the average cycle depth of energy storage in MS at the kth cycle, represents the initial cycle depth of the i-th energy storage in the MS at the k-th cycle, Indicates that Substituting the maximum distribution coefficient of each energy storage into equation (35), Represents the final distribution coefficient of the new individual mutated.

[0234] After the change, the energy, power and cycle depth of the energy storage are still updated according to equations (38) to (45).

[0235] 4. Choice and Creation

[0236] Selection and creation: First, screen the subordinate groups under each dominant population individual, and then screen the dominant group.

[0237] (1) First, the distribution coefficients of the subordinate population individuals are used to obtain the state variables of each energy storage, such as energy storage power, energy, and cycle depth. Then, the operating cost, wind and solar curtailment, and load loss corresponding to each subordinate population are calculated, and the non-dominated sorting algorithm is used to select npopc1 excellent individuals. In order to ensure the creative and innovative ability of the group, npopc2 subordinate individuals can be randomly added under each dominant population individual, and their generation method is the same as the initialization of the subordinate individuals.

[0238] (2) After completing the screening of subordinate groups, the dominant population needs to be screened. First, each dominant population individual selects the Pareto optimal individual from its subordinate population to form a complete individual, and then the complete individual is screened using the non-dominated sorting algorithm to screen out npopj1 excellent individuals. If all the complete individuals formed by a dominant population individual are not selected, the dominant population individual and its corresponding subordinate individuals are deleted. Similarly, in order to ensure the overall innovation ability, npopj2 dominant population individuals need to be randomly added, and each dominant population individual is randomly equipped with npopc2 subordinate individuals, and the generation method is the same as the initialization of the dominant population individuals.

[0239] (3) After completing the screening and creation of new individuals, the existing population is used as the parent generation for iteration again until the termination condition is met.

[0240] Combined with the typical example system, the energy status and planned output of all energy storage in a certain area are as follows: Figure 1 As shown. There are two charge and discharge cycles on the day. The first cycle is from 0:00 to 10:00 on the day. The energy storage returns to the initial energy state after a charge and discharge at about 10:00 in the morning, and the second cycle is from 11:00 to 24:00. The energy required to be processed in the two cycles is less than the total capacity of all energy storage, and the maximum power is less than the total power of energy storage. Therefore, the two cycles do not require the participation of all energy storage, and the units can be optimized. Taking the energy state of energy storage obtained by the day-ahead dispatch as a reference, it is calculated that the total energy required to be processed in the two cycles is approximately 5645MWh and 6640MWh respectively; if only lithium battery energy storage is used in the cycle, the average cycle depth should be: 37.52% and 44.13%.

[0241] At this cycle depth, the comprehensive cost per kilowatt-hour of lithium battery energy storage, which has the highest cost per kilowatt-hour, is 0.35 and 0.40 yuan / kWh respectively, which is lower than the comprehensive cost per kilowatt-hour of compressed air energy storage and hydrogen energy storage. Therefore, compressed air and hydrogen energy storage are not required for both cycles in this scenario.

[0242] The master-slave collaborative evolution algorithm is used to optimize the power distribution. Since the wind and solar power consumption pressure in this scenario is relatively small, the variable range of energy storage power is relatively large and flexible. In the subsequent optimization, it can still ensure that the wind and solar power are fully consumed without load loss, that is, f cur and f loss are all 0. Therefore, it is only necessary to compare and analyze the operating costs before and after optimization. During initialization, the first dominant population individual can be set to all lithium batteries participating in two cycles, and the first subordinate population individual can be set to all energy storage to obtain the distribution coefficient in proportion to the capacity as the optimization starting point, and the optimization benchmark is given. The remaining individuals are randomly initialized. Taking the optimization results after 400 iterations as an example, the optimal solution of each iteration is recorded. The operating cost corresponding to the optimal solution generated in each round of iteration is as follows: Figure 4 ,It can be seen that with the continuous iteration of the optimization algorithm, ,the operation has been effectively reduced while ensuring the wind and solar power ,consumption and load supply.

[0243] like Figure 5 The energy storage power corresponding to the optimal solution after the 400th iteration is shown, and compared with the original planned output, the optimal upper and lower limits. The optimal upper limit refers to the ideal output plus the power that can be adjusted downward by thermal power at each moment; similarly, the optimal lower limit refers to the ideal output minus the power that can be adjusted upward by thermal power. Between the optimal upper and lower limits, the difference between energy storage and ideal power can be achieved through thermal power regulation to avoid load loss and wind and solar power abandonment.

[0244] The embodiment of the present invention adopts three methods to carry out simulation experiments, and the three methods are specifically as follows:

[0245] Method 1: Random operation, i.e. randomly selecting energy storage to participate in the cycle and randomly specifying the allocation coefficient;

[0246] Method 2: Directly allocate in proportion to capacity;

[0247] Method 3: Use the master-same co-evolution algorithm proposed in this application to solve the allocation.

[0248] The results of the three methods show that, under the premise that all three methods can achieve no loss of load and full absorption of wind and solar power, when method one is adopted, the total operating cost is about 4,800,000 to 5,500,000; when method two is adopted, the total operating cost is 4,653,482 yuan, which is about 3% to 15% lower than method one; and the operating cost of method three is 2,881,279 yuan, which is about 40% to 48% lower than method one and 37.8% lower than method two. The comparison results show that in the scenario of full absorption of new energy, under the premise of ensuring the absorption of new energy and load supply, this method uses the cost difference of different energy storage and the same energy storage under different operating conditions to further optimize power distribution and more effectively reduce operating costs.

[0249] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing power allocation within an energy storage cluster based on a master-slave collaborative evolutionary algorithm, characterized in that: The steps include: Step 1: Combined with the optimal operation mode of energy storage, determine the participation state and cycle depth of each lithium battery energy storage, the participation state represents a virtual variable indicating whether each lithium battery participates in a certain cycle, and the cycle depth is the ratio of the cycle capacity to the rated capacity; determine the distribution coefficient of each energy storage under the cycle depth, the ratio of its own cycle capacity to the cycle capacity of all energy storage, and select the participation state J of each lithium battery energy storage and the distribution coefficient Coe of each energy storage as the variables to be optimized; Step 2: Based on the variables to be optimized determined in step 1, a power optimization model is constructed, wherein the power optimization model aims to optimize the energy storage power by minimizing the operating cost, minimizing the amount of wind and solar power abandoned, and minimizing the amount of load loss; Step 3: For the coupling relationship between the participation state J of each lithium battery energy storage in the power allocation in step 1 and the allocation coefficient Coe of each energy storage, the variable representing whether each energy storage participates in the cycle is taken as the dominant population, and the power allocation coefficient is taken as the subordinate population. The power allocation coefficient is generated on the basis of the variable of whether each energy storage participates in the cycle. With the optimization objective in step 2, the two variables to be optimized are optimized and solved to determine the final Pareto optimal individual.

2. According to claim 1, a method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm is characterized in that: In step 1: The participation status J of each lithium battery energy storage represents a virtual variable indicating whether each lithium battery participates in a certain cycle; The distribution coefficient Coe of each energy storage represents the distribution coefficient of each lithium battery in a certain cycle; The two variables J and Coe are not independent. Only when the i-th energy storage participates in the k-th cycle, J i,k =1, Coe i,k The value of Coe is meaningful. i,k Should be in a specific J i,k The variables are divided into two groups. The virtual variable J of the participation status is the dominant population, and the distribution coefficient Coe is its subordinate population. The value of Coe is determined by J, and there is a master-slave relationship between the two groups.

3. A method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 1 or 2, characterized in that: In step 1, the specific process of determining the participation status of each lithium battery energy storage is as follows: Step 1.1: First, assume that only lithium batteries participate in the charge and discharge cycle, and determine the average charging efficiency of lithium batteries under the current circumstances based on the total output of all energy storage in the area. Charge energy to be processed Average discharge efficiency and the discharge energy to be processed And the average cycle depth of lithium battery energy storage Calculate the cost per kilowatt-hour of each lithium battery based on the average cycle depth; Step 1.2: Among the compressed air energy storage and hydrogen energy storage that are not currently involved in the cycle, determine whether there are individuals with a kilowatt-hour cost lower than the kilowatt-hour cost of a lithium battery, add them to this charge and discharge cycle in order of their kilowatt-hour costs from small to large, and make them run at full power. The remaining power is allocated to all lithium batteries, and the kilowatt-hour cost of lithium batteries is recalculated until the costs of all hydrogen energy storage and compressed air energy storage are higher than the kilowatt-hour cost of any lithium battery; Step 1.3: Further determine the participation status of lithium batteries based on the participation of hydrogen energy storage and compressed air energy storage in the charge and discharge cycle.

4. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 3 is characterized in that: The process of calculating the cost per kilowatt-hour of each lithium battery is as follows: The total planned output of all large-scale energy storage in this area obtained by scheduling on the day before is n is the number of energy storage cycles per year, T is the energy storage cycle period, and the planned wind and solar power abandonment is The total output of all regional energy storage If only lithium batteries are involved in the cycle, the average charging efficiency Charge energy to be processed Average discharge efficiency and the discharge energy to be processed Make an estimate: Among them, Ce i , η ci , η di Respectively represent the rated capacity, charging efficiency, and discharging efficiency of the i-th energy storage, T C The subscript set of all charging times in this cycle, T D Represents the subscript set of all discharge moments; Average cycle depth of lithium battery energy storage And the cost per kilowatt-hour is as follows: in, represents the available capacity of the i-th energy storage in the current cycle; p i represents the calculated cost per kWh of the i-th energy storage at the current cycle depth; M is a maximum value, indicating that when the average cycle depth is greater than 1, the lithium battery energy storage can no longer meet the capacity requirements, and it is not feasible to use only lithium batteries for charging and discharging, and a maximum penalty cost is taken; Represents the functional relationship between the i-th energy storage cycle depth and the cost per kilowatt-hour.

5. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 4 is characterized in that: The average cycle depth and cost per kilowatt-hour are updated as follows: in, represents the updated average cycle depth of lithium battery energy storage, p i ' represents the updated cost per kWh of the i-th energy storage, and G represents the number of the added compressed air energy storage or hydrogen energy storage.

6. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 1, characterized in that: The power optimization model in step 2 is specifically as follows: Among them, f c represents the operating cost, f cur Indicates the amount of abandoned wind power, f loss Indicates the amount of load loss, dod i.k represents the cycle depth in the kth cycle of the i-th storage day, DOD N,i represents the rated cycle depth of the i-th energy storage, CIR represents the number of cycles on the day; g(·) represents the cost coefficient, which is a function of the cycle depth; LCOS represents the cost per kilowatt-hour considering the influence of the cycle depth, It indicates the rated cost of electricity when the energy storage is in rated operation state, ignoring energy loss; m is the number of energy storage units, and N is the rated value. is the ideal power of the energy storage cluster to meet the dispatching demand at time t, It represents the maximum adjustable power of all controllable units in the area at time t. Indicates the maximum power that can be reduced for all controllable units in the area at time t. It represents the energy storage regulation power, and Δt is the energy storage regulation time.

7. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 2, characterized in that: The specific process of optimizing and solving the two optimization variables using the master-slave co-evolution method in step 3 is as follows: Step 3.1: Set the initial number of individuals in the population to npop. First, initialize the dominant population J, randomly generate npop individuals, and then generate npop subordinate population individuals Coe based on each dominant population individual J. Step 3.2: Initialize the dominant population individuals, randomly determine whether the i-th energy storage participates in the k-th cycle, and i,k Assign values ​​to determine whether the energy storage capacity participating in the cycle can meet the capacity and power requirements. If so, output the dominant population individual; if not, find the dominant population individual that currently decides not to participate in the k-th cycle and re-randomly assign values ​​to its participation status virtual variable; Step 3.3: Initialization of subordinate population individuals. Under each dominant population individual, the Coe of its subordinate population individuals is initialized, including randomly generating the initialization cycle depth of each energy storage near the average cycle depth, determining the initial distribution coefficient of each energy storage under the initial cycle depth, that is, the ratio of its own cycle capacity to the cycle capacity of all energy storage, correcting the distribution coefficient, normalizing the distribution coefficient, and updating the energy storage power, energy, and cycle depth based on the distribution coefficient; Step 3.4: Crossover operation, which allows subordinate populations under the same dominant population individual to cross internally. After the crossover, the power, energy, and cycle depth of each energy storage under the allocation method are updated. Then, the dominant population individual selects excellent subordinate population individuals to form a complete individual, and crosses between the complete individuals. Step 3.5: Mutation operation: first mutate the subordinate populations under each dominant population individual, then mutate the dominant population. After the individuals in the dominant population mutate, the subordinate populations need to be adjusted accordingly. After the two mutations, the energy storage power, energy, and cycle depth are updated. Step 3.6: Selection and creation. First, the distribution coefficient of each subordinate population individual is used to obtain the energy storage power, energy, and cycle depth of each energy storage. Then, the operating cost, abandoned wind and solar power, and load loss corresponding to each subordinate population are calculated respectively. Then, the non-dominated sorting algorithm is used to select npopc1 excellent individuals. Under each dominant population individual, npopc2 subordinate population individuals are randomly added. The generation method is the same as the initialization of the subordinate population individuals. After the subordinate population is screened, the dominant population is screened. First, each dominant population individual is ordered to select the Pareto optimal individual from its subordinate population to form a complete individual. Then, the complete individual is screened using the non-dominated sorting algorithm to screen out npopj1 excellent individuals. If all the complete individuals formed by a dominant population individual are not selected, the dominant population individual and its corresponding subordinate individuals are deleted. Then, npopj2 dominant population individuals are randomly added, and each dominant population individual is randomly equipped with npopc2 subordinate population individuals. The generation method is the same as the initialization of the dominant population individuals. Step 3.7: After completing the screening and creation of new individuals, the existing population is used as the parent generation for iteration again until the termination condition is met, and the final two optimization variables and their corresponding power, energy, and cycle depth are output.

8. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 7, characterized in that: The initialization of the subordinate population individuals specifically includes the following process: Step 3.1.1: Initialize loop depth Generate the initial cycle depth of each energy storage randomly around the average cycle depth: Among them, dod i,k ' represents the initial cycle depth, δ represents the maximum fluctuation amplitude, U() is a random uniform distribution function, represents the average cycle depth of all energy storage participating in the kth cycle; Step 3.1.2: Initialize the distribution coefficient: Step 3.1.3: Correct the initial partition coefficient: Among them, Pe i is the effective power of the i-th energy storage, α m is the margin factor, and denote the capacity requirement and power requirement of the kth cycle respectively, Take the power with the largest absolute value in the cycle; Coe i,k " is the corrected distribution coefficient, Coe i,k ' is the initial distribution coefficient, J i,k is a dummy variable indicating whether the i-th energy storage participates in the k-th cycle; Step 3.1.4: Normalization of the partition coefficient: Step 3.1.5: Update the energy storage power, energy, and cycle depth based on the allocation coefficient: Among them, T k represents the cycle period corresponding to the kth cycle of energy storage, is the ideal power of the energy storage cluster to meet the dispatching demand at time t, It represents the power obtained by the i-th energy storage at time t according to the distribution coefficient, Indicates the discharge power allocated to it, Represents the charging power allocated to it, and calculates the energy stored in the energy storage at each moment: in, represents the energy stored at time t and t-1 of the i-th energy storage, η ci represents the charging efficiency of the i-th energy storage, η di represents the discharge efficiency of the i-th energy storage; The power and stored energy are further corrected and the depth of each storage cycle is calculated: in, and They represent the charging power and discharging power of the i-th energy storage at time t after correction, represents the corrected stored energy at the ith energy storage time t, Indicates the corrected cycle depth.

9. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 7, characterized in that: The crossover operation is specifically as follows: Step 3.4.1: Slave population crossover, when the slave population Coe performs internal crossover: in, is the intermediate variable of the distribution coefficient, represents the distribution coefficient of the kth stage of the i-th energy storage in the new individual formed after crossover, and They represent the distribution coefficients of the kth stage of the i-th energy storage in the two parent individuals, and CS represents the energy storage number selected for exchange: CS=randsample(CR·npop,npop) Among them, randsample(n,m) is a sampling function, which means randomly selecting n integers from 1 to m, and CR represents the crossover rate; Step 3.4.2: Dominant population crossover: in, Indicates the dummy variable representing whether the i-th energy storage in the new individual formed by the crossover participates in the k-th cycle, and Represents the corresponding dummy variable of the selected parent individual.

10. The method for optimizing internal power allocation of an energy storage cluster based on a master-slave collaborative evolutionary algorithm according to claim 7, characterized in that: The specific process of the mutation operation is as follows: Step 3.5.1: Subordinate population variation, within the subordinate group led by the same dominant individual: in, is the intermediate variable of the distribution coefficient, represents the maximum value of the distribution coefficient of the i-th energy storage in the k-th stage, represents the distribution coefficient of the i-th energy storage in the new individual formed after mutation at the k-th stage, represents the distribution coefficient of the i-th energy storage in the selected parent individual at the k-th stage, δ M represents the variable step length, U() is a random uniform distribution function, and MS represents the set of energy storage numbers selected for mutation: MS=randsample(MR·npop,npop) Among them, randsample(n,m) is a sampling function, which means randomly selecting n integers from 1 to m, and MR represents the mutation rate; Step 3.5.2: Dominant population mutation, the process is as follows: (a) Randomly regenerate the participating state variables of the variable energy storage: Among them, rand(0,1) represents a random value between 0 and 1. If it is 1, it means that the i-th energy storage participates in the k-th cycle, and if it is 0, it means that it does not participate; A dummy variable indicating whether the i-th energy storage participates in the k-th cycle after the mutation; (b) Determine whether the energy storage capacity involved in the cycle can meet the capacity demand and power demand. If the following conditions are met at the same time, the output of individual Otherwise, proceed to operation (c), with the following conditions: in, and They represent the capacity demand and power demand of the kth cycle, Pe i is the effective power of the i-th energy storage, Ce i represents the rated capacity of the i-th energy storage; (c) Find the individuals in MS that are not participating in the kth cycle, re-randomly select the values ​​of their participation state variables, and then go to operation (b): in, represents the set of energy storage subscripts in the MS that currently decides not to participate in the kth cycle; (d) Regenerate the distribution coefficient of the stored energy in the MS: in, represents the average cycle depth of energy storage in MS at the kth cycle, represents the initial cycle depth of the i-th energy storage in the MS at the k-th cycle, Indicates the use of Determine the maximum allocation coefficient of each energy storage, It represents the final distribution coefficient of the mutated new individual, and δ represents the maximum fluctuation range.

Images