Long-term and short-term cloud energy storage optimal configuration method considering lease market

Abstract

The invention discloses a long-term and short-term cloud energy storage optimal configuration method considering a lease market, and relates to the field of energy storage optimal configuration, and the method comprises the steps: carrying out the clustering of typical scenes of offshore wind power through employing a semi-supervised multi-kernel fuzzy clustering algorithm; according to the characteristics of hydrogen energy storage and battery energy storage and comprehensive performance and economic factors, a long-short-term cloud energy storage coordinated optimization operation strategy including hydrogen energy storage and battery energy storage operation is output; establishing a long and short term cloud energy storage double-layer optimization configuration model considering two-system lease prices and typical scenes; converting the long and short term cloud energy storage double-layer optimization configuration model into a single-layer optimization model based on an optimality condition; and solving the single-layer optimization model by adopting a fungus growth optimization algorithm improved based on a memory mechanism. According to the method, the long-term and short-term cloud energy storage double-layer optimization configuration model considering the lease market is constructed, and theoretical support and decision basis are provided for large-scale application and sustainable development of cloud energy storage through collaborative optimization processing.

Description

Technical Field

[0001] The present invention relates to a long-term and short-term cloud energy storage optimization configuration method considering the leasing market, and belongs to the field of energy storage optimization configuration. Background Art

[0002] In recent years, with the continuous development of new energy technologies and the deepening of power market reform, cloud energy storage, as a new energy storage business model, has gradually become an important means to support flexible power system scheduling and friendly access to new energy due to its advantages such as resource sharing, cost sharing, and flexible participation in market transactions. However, existing cloud energy storage configuration and operation strategies still face many challenges. First, offshore wind power output has significant temporal and spatial randomness. When processing large-scale annual operation scenario data, traditional clustering methods often suffer from low computational efficiency, sensitivity to noise, and insufficient clustering result stability. This limits the accuracy of typical scenario extraction and the reliability of subsequent optimization decisions.

[0003] Secondly, there are significant differences between battery energy storage and hydrogen energy storage in terms of response speed, charging and discharging efficiency, and cost structure in long- and short-term cloud energy storage. If the characteristics of the two are not fully explored and coordinated, it will be difficult to achieve efficient operation and economically optimal configuration in multiple long- and short-term time scales. In addition, most existing optimization models fail to fully consider the two-part rental electricity price mechanism and user-side assessment costs, resulting in limited benefits of cloud energy storage systems in actual applications and difficulty in balancing operational economy and rationality. At the optimization solution level, traditional heuristic algorithms are prone to falling into local optimality and lack effective memory and utilization of historical excellent solutions, further affecting the efficiency and effectiveness of solving large-scale optimization problems.

[0004] Therefore, it is particularly important to study a long-term and short-term cloud energy storage optimization configuration method considering the leasing market. Summary of the Invention

[0005] In response to the above problems, the present invention provides a long-term and short-term cloud energy storage optimization configuration method considering the leasing market.

[0006] Specifically, the present invention adopts the following technical solutions to solve the above technical problems:

[0007] A method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market, comprising:

[0008] Based on latent representation learning and information fusion, the objective function of a semi-supervised multi-core fuzzy clustering algorithm is constructed, and the semi-supervised multi-core fuzzy clustering algorithm is used to cluster typical scenarios of offshore wind power; according to the characteristics of hydrogen energy storage and battery energy storage, and comprehensive performance and economic factors, a long-term and short-term cloud energy storage coordinated optimization operation strategy including hydrogen energy storage and battery energy storage is output; a long-term and short-term cloud energy storage two-layer optimization configuration model considering two-part leasing prices and typical scenarios is established, and includes an upper model and a lower model; based on the optimality condition, the long-term and short-term cloud energy storage two-layer optimization configuration model is converted into a single-layer optimization model; the single-layer optimization model is solved using an improved fungal growth optimization algorithm based on a memory mechanism.

[0009] Compared with the prior art, the present invention has the following beneficial effects:

[0010] 1. The present invention solves the problem that existing fuzzy clustering algorithms have certain limitations when performing scene clustering. They have high computational complexity, especially when processing large-scale source-load data, the computing time and resource consumption increase significantly. Fuzzy clustering is sensitive to noise in the data, which may lead to inaccurate clustering results. In response to the limitations of the above-mentioned fuzzy clustering algorithms, the present invention proposes a semi-supervised multi-kernel fuzzy clustering algorithm based on latent representation learning and information fusion. Through a semi-supervised constraint term based on information fusion, the constraint distance is constructed using pairwise constraint information, and the membership is re-evaluated accordingly. At the same time, the square loss term is constructed using label information to impose constraints on the membership matrix. Introducing multi-kernel functions into the proposed algorithm can effectively solve the limitations of a single kernel function, improve the generalization ability and robustness of clustering, and automatically adjust the kernel function combination by optimizing the kernel weights, making the proposed clustering algorithm more stable. The clustering algorithm shows better stability and clustering effect in processing high-dimensional, complex distribution and noisy data scenarios (for annual operation scenarios of offshore wind power). The semi-supervised multi-kernel fuzzy clustering algorithm based on latent representation learning and information fusion solves the problems of long computation time, significantly increased resource consumption, and noise sensitivity when processing large-scale source-load data by traditional fuzzy clustering algorithms, and improves the accuracy and stability of clustering results. By introducing the latent representation learning mechanism, it effectively extracts the essential characteristics of the data and reduces data redundancy and noise interference. The introduction of the semi-supervised strategy uses a small amount of prior information to guide the clustering process. While improving the clustering performance, it significantly reduces the dependence on a large amount of labeled data, and has good scalability and practical application value.

[0011] 2. The present invention establishes a long-term and short-term cloud energy storage dual-layer optimization configuration model that considers the two-part leasing market. In view of the technical characteristics limitations of a single type of energy storage, a long-term and short-term cloud energy storage that combines battery energy storage and hydrogen energy storage is adopted. The upper model determines the configuration capacity of cloud energy storage with the goal of minimizing the annual investment, construction, and operating costs of cloud energy storage, comprehensively considering the investment and construction costs, peak-valley arbitrage income, and leasing income of cloud energy storage operators. The second stage considers the user's assessment cost. The leasing price is formulated according to the user's demand in each time period and the income of cloud energy storage and passed to the lower model. The lower model aims to minimize the user's annual assessment cost. Through the collaborative optimization of this model, on the one hand, cloud energy storage is encouraged to participate in auxiliary services such as grid peak and frequency regulation, enhance system flexibility, and smooth out wind power output fluctuations; on the other hand, it is promoted to participate in electricity spot market transactions, expand revenue sources, improve resource utilization and operational economy, thereby providing theoretical support and decision-making basis for the large-scale application and sustainable development of cloud energy storage. That is, a cloud energy storage system that considers the long-term and short-term multi-time scale characteristics is adopted, taking into account the advantages of battery energy storage and hydrogen energy storage. In view of the differences in their characteristics, a long-term and short-term energy storage collaborative optimization operation strategy is proposed, which comprehensively considers multiple factors such as response speed, charging and discharging efficiency, and operating costs. The proposed long-term and short-term energy storage operation strategy comprehensively considers the differentiated characteristics of hydrogen energy storage and battery energy storage, fully combining the advantages of both, effectively reducing system operating costs and improving reliability. The established two-layer optimization model takes into account the two-part electricity price, including capacity electricity price and electricity price, effectively reducing the operating cost and opportunity cost of cloud energy storage, while taking into account the user's assessment cost, further ensuring the rationality and economy of the cloud energy storage configuration results.

[0012] 3. Improve the fungal growth optimization algorithm based on the memory mechanism. Based on the Karush-Kuhn-Tucker optimality condition, the established two-layer model is converted into a single-layer optimization model, and the improved fungal growth optimization algorithm is used to solve the single-layer optimization model. The memory mechanism is introduced into the fungal growth optimization algorithm, and the two-layer model is solved based on the improved fungal growth optimization algorithm, which effectively solves the problems of traditional heuristic algorithms easily falling into local optimality and lacking effective integration of historical optimal solutions. By retaining and utilizing historical optimal solutions, some dimensions are reset and locally disturbed in a targeted manner when individuals are updated, and an information sharing mechanism is introduced between individuals to enhance the collaborative search ability of the group. While maintaining the global search capability, the algorithm improves the local development accuracy, effectively solves the problems of the original fungal growth optimization algorithm easily falling into local optimality, lacking the use of historical information, and slow convergence speed, and significantly improves the global optimization performance and stability of the algorithm. The algorithm can effectively solve the two-layer optimization model established by the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0014] Figure 1 is a flow chart of a method for optimizing long-term and short-term cloud energy storage configuration considering the leasing market according to an embodiment of the present invention;

[0015] Figure 2 This is one of the flow charts of the fungal growth optimization algorithm;

[0016] Figure 3 This is the second flowchart of the fungal growth optimization algorithm. DETAILED DESCRIPTION

[0017] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the present invention, a long-term and short-term cloud energy storage optimization configuration method considering the leasing market is proposed, including:

[0018] Based on latent representation learning and information fusion, the objective function of a semi-supervised multi-kernel fuzzy clustering algorithm was constructed and used to cluster typical offshore wind power scenarios. Based on the characteristics of hydrogen and battery energy storage, and taking into account performance and economic factors, a coordinated optimization strategy for long- and short-term cloud energy storage, including hydrogen and battery energy storage, was output. A two-layer optimization configuration model for long- and short-term cloud energy storage was established, including both upper and lower models, considering two-part leasing prices and typical scenarios. Based on optimality conditions, the two-layer optimization configuration model for long- and short-term cloud energy storage was converted into a single-layer optimization model. This single-layer optimization model was solved using an improved fungal growth optimization algorithm based on a memory mechanism.

[0019] In one embodiment, based on latent representation learning and information fusion, an objective function of a semi-supervised multi-core fuzzy clustering algorithm is constructed, and clustering of typical offshore wind power scenarios using the semi-supervised multi-core fuzzy clustering algorithm includes:

[0020] Construct a label matrix and a weight matrix, and adjust the objective function of the fuzzy clustering algorithm after supervision and constraint according to the label matrix, weight matrix, membership matrix and adjustment parameters; based on the objective function of the fuzzy clustering algorithm, construct a semi-supervised multi-core clustering loss function that introduces multi-core distance, and use the Lagrange multiplier method to solve the semi-supervised multi-core clustering loss function to obtain a semi-supervised multi-core clustering loss function in the form of a Lagrangian function; decompose the semi-supervised multi-core clustering loss function in the form of a Lagrangian function into a data-to-cluster center membership subproblem and a cluster center subproblem, and derive them to obtain the update formulas for the data-to-cluster center membership and the cluster center update formula respectively. ; Based on a deep autoencoder with reconstruction loss, low-dimensional representation of source load data is obtained through latent representation learning to learn the latent structure of source load data; the mean square error is used to measure the information loss between the original source load data and the reconstructed data, and the information loss function is determined; the kernel density estimation is used to estimate the probability distribution of the original source load data and the latent representation respectively, and the distribution consistency loss function is constructed in the form of divergence; the objective function of the semi-supervised multi-kernel clustering loss function, information loss function and distribution consistency loss function are combined to obtain the objective function of the semi-supervised multi-kernel fuzzy clustering algorithm; the semi-supervised multi-kernel fuzzy clustering algorithm is used to cluster typical scenarios of offshore wind power.

[0021] In one embodiment, a label matrix and a weight matrix are constructed, and the objective function of the fuzzy clustering algorithm after obtaining the supervision constraint is adjusted according to the label matrix, the weight matrix, the membership matrix, and the adjustment parameter, including:

[0022] Based on the labeled data and unlabeled data, a label matrix is ​​constructed; weights are assigned to each data according to the label matrix, and a weight matrix is ​​obtained; the adjustment parameter is used to represent the supervision constraint of the data and balance the error between the membership matrix and the label matrix; based on the label matrix, weight matrix, membership matrix and adjustment parameter, the objective function of the fuzzy clustering algorithm is adjusted to obtain the objective function of the fuzzy clustering algorithm after supervision constraint:

[0023]

[0024] Where J1(U,V) is the objective function of the fuzzy clustering algorithm after supervision constraint; n dt is the sample size, c e is the optimal number of clusters; w ij is the element in the i-th row and j-th column of the weight matrix, ||x i -υ j || is the i-th data x i To the jth cluster center υ j The Euclidean distance, u ij Represents the i-th data x i For the jth cluster center υ jThe membership degree, l ij is the element in the i-th row and j-th column of the label matrix, o is the balance parameter, is the adjustment parameter.

[0025] In one embodiment, based on the objective function of the fuzzy clustering algorithm, a semi-supervised multi-core clustering loss function is constructed by introducing the multi-core distance, and the Lagrange multiplier method is used to solve the semi-supervised multi-core clustering loss function. The semi-supervised multi-core clustering loss function in the form of a Lagrangian function includes:

[0026] The multi-core distance is introduced into the objective function of the fuzzy clustering algorithm to obtain the semi-supervised multi-core clustering loss function. The Lagrange multiplier method is used to solve the semi-supervised multi-core clustering loss function to obtain the semi-supervised multi-core clustering loss function in the form of Lagrange function:

[0027]

[0028] Where, J s (U, V) is the semi-supervised multi-kernel clustering loss function in the form of Lagrangian function, is the multi-core distance from the data point to the cluster center, λ i is the Lagrange multiplier; n dt is the sample size, c e is the optimal number of clusters; w ij is the element in the i-th row and j-th column of the weight matrix, u ij Represents the i-th data x i For the jth cluster center υ j The membership degree, l ij is the element in the i-th row and j-th column of the label matrix, o is the balance parameter, is the adjustment parameter.

[0029] In one embodiment, when the objective function of the semi-supervised multi-core fuzzy clustering algorithm is obtained by integrating the semi-supervised multi-core clustering loss function, the information loss function, and the distribution consistency loss function, the objective function of the semi-supervised multi-core fuzzy clustering algorithm is:

[0030] minJ=J lrl +χ1J a +χ2J s ;

[0031] Where J is the total loss function, J lrl is the information loss function, J a is the distribution consistency loss function, J s is the clustering loss function; χ1 and χ2 are different balance parameters.

[0032] In one embodiment, based on the characteristics of hydrogen energy storage and battery energy storage, and taking into account performance and economic factors, a long-term and short-term cloud energy storage coordinated optimization operation strategy including hydrogen energy storage and battery energy storage operation is output, including:

[0033] The economic balance power of energy storage is determined based on the unit energy cost of hydrogen energy storage and battery energy storage. When the economic balance power of energy storage is higher than the minimum operating threshold of hydrogen energy storage, the economic balance power of energy storage is used as the strategy switching threshold. When the economic balance power of energy storage is lower than the minimum operating threshold of hydrogen energy storage, the minimum power threshold of hydrogen energy storage is used as the strategy switching threshold. When the operating power of the cloud energy storage system is lower than the strategy switching threshold, the battery energy storage independently bears the charging and discharging needs of the cloud energy storage system. When the system operating power is higher than or equal to the strategy switching threshold, it switches to hydrogen energy storage operation.

[0034] In one embodiment, establishing a two-tier optimization configuration model for long-term and short-term cloud energy storage considering two-part rental prices and typical scenarios includes:

[0035] Construct an upper-level model of the long-term and short-term cloud energy storage dual-layer optimization configuration model, with the upper-level model aiming to minimize the equal-annual comprehensive cost of cloud energy storage; construct a lower-level model of the long-term and short-term cloud energy storage dual-layer optimization configuration model that considers the leasing market and typical scenarios, with the lower-level model aiming to minimize the user's annual assessment cost.

[0036] In one embodiment, an upper layer model of a long-term and short-term cloud energy storage dual-layer optimization configuration model is constructed. The upper layer model aims to minimize the equivalent annual value comprehensive cost of cloud energy storage and includes:

[0037] Calculate the equivalent annual investment and construction costs, operation and maintenance costs, energy consumption costs, replacement costs, annual rental income, peak-valley arbitrage income, and recovery value of battery energy storage; calculate the equivalent annual investment and construction costs, operation and maintenance costs, energy consumption costs, replacement costs, annual rental income, peak-valley arbitrage income, and recovery value of hydrogen energy storage; and construct the objective function of the upper model of the long-term and short-term cloud energy storage dual-layer optimization configuration model:

[0038]

[0039] Where, F up C is the objective function of the upper model of the long-term and short-term cloud energy storage dual-layer optimization configuration model; B and C H are the annual comprehensive costs of battery energy storage and hydrogen energy storage respectively; C TCC,B and C TCC,H are the equivalent annual investment and construction costs of battery energy storage and hydrogen energy storage respectively; C OC,B and C OC,H are the operation and maintenance costs of battery energy storage and hydrogen energy storage respectively; C ES,B and C ES,Hare the energy consumption costs of battery energy storage and hydrogen energy storage respectively; C RL,B and C RL,H are the replacement costs of battery energy storage and hydrogen energy storage respectively; I LEA,B and I LEA,H are the annual rental income of battery energy storage and hydrogen energy storage respectively; I IP,B and I IP,H are the peak-valley arbitrage benefits of battery energy storage and hydrogen energy storage respectively; RC,B and I RC,H The recovery value of battery energy storage and hydrogen energy storage; construct the constraint function of the upper model of the long-term and short-term cloud energy storage two-layer optimization configuration model, including power balance constraints, configuration power and capacity constraints of battery energy storage and hydrogen energy storage, SOC constraints and LOH constraints.

[0040] The power balance constraint is:

[0041]

[0042] Where U y (t) is the voltage of node y during period t; θ xy (t) is the voltage phase difference between node x and node y during period t; G xy and B xy is the equivalent conductance and susceptance of the line between node x and node y, N nodes is the number of nodes, P x (t) and Q x (t) are the active power and reactive power of node x at time t;

[0043] The configuration power and capacity constraints of battery energy storage and hydrogen energy storage are:

[0044]

[0045] Where, The battery energy storage configuration capacity E is BESS Upper and lower limits and configuration power P BESS Upper and lower limits; They are the upper and lower limits of the water electrolysis power, the upper and lower limits of the fuel cell power, and the upper and lower limits of the hydrogen storage tank capacity;

[0046] The SOC constraint is:

[0047] SOC min ≤SOC≤SOC max ;

[0048] Where, SOC max and SOC min They are the upper and lower limits of the SOC of the battery energy storage;

[0049] The LOH constraints are:

[0050] LOH min ≤LOH≤LOH max ;

[0051] In the formula, LOH max and LOH min They are the upper and lower limits of LOH of hydrogen energy storage respectively.

[0052] In one embodiment, a lower-level model of a two-tier optimization configuration model for long-term and short-term cloud energy storage is constructed, taking into account the leasing market and typical scenarios. The lower-level model aims to minimize the user's annual assessment cost and includes:

[0053] The objective function of the lower model of the long-term and short-term cloud energy storage dual-layer optimization configuration model is constructed as follows:

[0054]

[0055] Where, F down is the objective function of the lower model, C EXA is the annual assessment cost of users including wind farms, C FB The cost of assessing the deviation of power generation forecast for users, C LEA is the user's annual cloud storage rental cost, N s is the number of typical scenes, π n is the number of days when the nth scenario occurs, T is the total time period in any day, and t is the time period; c EXA (t) is the assessment cost of users with wind farms at time t, c FB (t) is the power generation forecast deviation assessment cost for the user at time t, I LEA,B The user's battery energy storage system rental cost, I LEA,H The user's hydrogen energy storage system rental cost; Construct the constraint function of the lower model of the long-term and short-term cloud energy storage dual-layer optimization configuration model, including wind power output constraints and wind curtailment constraints;

[0056] Among them, the wind power output constraints are:

[0057]

[0058] Where: P WT (t) is the wind power output during period t, The upper limit of wind power output;

[0059] The wind curtailment constraint is:

[0060]

[0061] Where R WT is the annual wind curtailment rate, is the maximum annual wind curtailment rate.

[0062] In one embodiment, solving a single-layer optimization model using an improved fungus growth optimization algorithm based on a memory mechanism includes:

[0063] Based on the memory mechanism, the fungal growth optimization algorithm is improved, local perturbations and supplements are implemented in the selected dimensions, and group search is promoted through information sharing between individuals. The improved fungal growth optimization algorithm is used to solve the single-layer optimization model. The solution of the single-layer optimization model using the improved fungal growth optimization algorithm includes:

[0064] Input cloud storage system parameters and wind power output data, configure the parameters, maximum number of iterations, and population size of the improved fungal growth optimization algorithm; perform memory strategy updates on the non-forgotten dimensions, and perform forgetfulness supplementary updates and dream sharing updates on the forgotten dimensions; execute the mycelium tip growth stage, mycelium branching stage, and spore germination stage until convergence conditions are met.

[0065] In one embodiment, a fungus growth optimization algorithm is improved based on a memory mechanism, local perturbations and supplements are implemented in selected dimensions, and group search is promoted through information sharing between individuals, including:

[0066] Initialize the solution of the fungal growth optimization algorithm; during the initialization phase, the individual with the best fitness in the population is used as the initial global memory; by introducing global memory or local memory into individual position updates, each individual is ensured to continuously move closer to the optimal solution during the search process;

[0067] In the forgetting supplementation phase, the position of the individual on the selected forgetting dimension set is locally perturbed and updated; the perturbation amplitude is gradually attenuated with the number of iterations, and the perturbation amplitude is gradually reduced by the cosine attenuation factor. The formula for local perturbation update is:

[0068]

[0069] Where, For the i ter At the +1th iteration, the τth individual is in The location of the dimension, For the i ter The global optimal solution at the iteration, Υ is the set of forgotten dimensions, For the The perturbation vector of individual τ in dimension;

[0070] In the memory sharing stage, the information of other individuals in the population is used for supplementary updates to promote knowledge sharing between individuals. The update formula is:

[0071]

[0072] Where r is a random number in [0,1], p is the sharing probability threshold, and m is the index of an individual randomly selected from the population; For the i ter At the +1th iteration, the τth individual is in The location of the dimension, For the i ter +1 iteration, the mth individual is in The location of the dimension, For the i ter At the iteration, the mth individual Dimensional location;

[0073] During the hyphae tip growth stage, the global optimal solution is found through growth rate and direction analysis and chemotaxis behavior analysis, while controlling the trade-off between exploration and exploitation. During the hyphae branching and spore germination stages, the search space is explored in multiple directions to escape from the local optimal solution, and the memory sharing update mechanism is used to update the dimensions of individuals during iteration. During the memory update stage, the individual with the best fitness is selected as the global optimal solution, and the global memory is updated.

[0074] In one embodiment, during the hyphal tip growth stage, the global optimal solution is achieved by analyzing growth rate and direction and chemotaxis behavior, and controlling the trade-off between exploration and exploitation. The method includes:

[0075] The growth rate in the fungal growth optimization algorithm is matched with the fitness of the solution, and the fitness value is multiplied by a random number to make the growth rate fluctuate. The growth rate calculation formula after adding fluctuations is:

[0076]

[0077] Where F is the adjustment factor of the normalized fitness value after the introduction of random perturbations, ξ is the factor that regulates the exploration and exploitation operators during the optimization process, i ter is the current iteration number, is the maximum number of iterations, r1 is a random number in [0,1], E is a factor used to adjust the exploration and exploitation operators during the optimization process, and f τ is the fitness value of the τth solution, f σ is the fitness value of the σth solution, M is the number of hyphae; the growth direction is determined by randomly selecting two solutions from the current population and calculating the difference between the two solutions. The growth direction vector expression is:

[0078]

[0079] Where, is the growth direction vector, and are two different solutions randomly selected from the current population;

[0080] According to the growth rate and growth direction of the mycelium, a new growth vector is calculated and combined with the current growth vector of the mycelium to generate a new mycelium growth vector. The expression of the update process is:

[0081]

[0082] Where, For the The random value generated in the dimension, d is the number of dimensions, r2 is a random number in [0,1], is the new growth vector, For the i ter The growth vector at the iteration;

[0083] In the first state of chemotaxis analysis, the growth direction is guided to the most nutrient-rich location based on the preset probability mechanism. The mathematical model of the mycelium's growth direction toward the current optimal solution or the randomly selected solution is:

[0084]

[0085] Where, is a random number uniformly distributed within the interval [0,1], r3 is a random number within [0,1], ζ is a parameter between 0 and 1, and β is a randomly selected value. is the growth direction of the τth mycelium; at the same time, based on the fitness value, the development operator is associated with the fitness value of each solution. The higher the fitness value, the faster the convergence speed; the mycelium growth update definition expression according to the first state is:

[0086]

[0087] Where η τ is the growth step of mycelium under the current development behavior;

[0088] In the second state of the chemotaxis behavior analysis, the expression of the second state is:

[0089]

[0090] Where, is a random number in the interval [0,1], The growth direction of the τth hypha is towards the individual randomly selected from the population Location, For the i ter At the +1th iteration, the τth individual is in The location of the dimension, For the At the same time, by adding the exploration step, the mycelium continues to explore in the search space;

[0091] The expression for switching between the first and second states in the chemotaxis behavior assay is:

[0092]

[0093] Where r 10 、r 11 、r 12 are random numbers in different intervals [0,1], is the adjustment factor, r 13 、r 14 is a random number in the interval [0,1];

[0094] By normalizing the fitness value of each solution, the hyphae located in the worst position try to apply the development operator and move towards the nutrient-rich area; the hyphae located in the nutrient-rich area continue to explore; when the normalized fitness value is less than the defined exploration rate, the exploration operation is performed, otherwise the development operation is performed.

[0095] In one embodiment, during the hyphae branching and spore germination stages, the search space is explored in multiple directions to escape from the local optimal solution, and the memory sharing update mechanism is used to update the dimensions of the individual during iteration. The following steps are performed: in the first mode during the hyphae branching stage, two solutions are randomly selected from the current population, and the growth direction of the new hyphae is determined based on the difference between the two solutions; in the second mode during the hyphae branching stage, the growth direction of the new hyphae is determined based on the difference between the current optimal solution and the solution randomly selected from the current population; the growth direction of the new hyphae is comprehensively determined based on the first mode and the second mode during the hyphae branching stage; in the hyphae branching stage, the growth rate of hyphae generated by lateral branches is calculated as follows:

[0096]

[0097] Where, E L is the growth rate of hyphae generated by lateral branches, r 16 and r 17 is a random number in different intervals [0,1], M is the number of mycelium, f σ is the fitness value of the σth solution, f τ is the fitness value of the τth solution; the growth rate is introduced into the growth direction of the new hyphae determined comprehensively, and the expression of the adjusted growth direction of the new hyphae is obtained as follows:

[0098]

[0099] Where r2, r3, r 15 are random numbers in different intervals [0,1], For the The random values ​​generated in the dimensions are uniformly distributed in the interval [0,1], where d is the number of dimensions. is the growth direction of new hyphae in the first mode, is the growth direction of new hyphae in the first mode, For the i ter At the τth iteration, the τth individual The location of the dimension, For the i ter At the +1th iteration, the τth individual is in Dimensional location;

[0100] In the spore germination stage, new solutions are randomly generated in the search space, and the exploration and development are balanced based on the mean of the current optimal solution and the random solution. The expression of the spore germination process is:

[0101]

[0102] Where, X g is a randomly selected value between -1 and 1, r2 and r5 are random numbers in different intervals [0,1], i ter is the current iteration number, is the maximum number of iterations, For the i ter At the τth iteration, the τth individual The location of the dimension, For the i ter At the τth iteration, the τth individual The location of the dimension, For the i ter At the +1th iteration, the τth individual is in Dimensional location; For the The optimal individual of dimension; is randomly selected from the population The position of the dimension, E is the factor used to adjust the exploration and exploitation operators during the optimization process; a random threshold is preset, and updates of mycelial branching and spore germination are performed on some dimensions. For dimensions that meet the conditions, the original position is retained; otherwise, the dimension is updated based on the memory sharing update mechanism.

[0103] In order to facilitate understanding of the above technical solutions of the present invention, the above technical solutions of the present invention are further explained from the perspectives of architecture and principle as follows.

[0104] Step S1: Define label matrix and weight matrix

[0105] First, the FCM clustering algorithm (Fuzzy Clustering Algorithm) attempts to divide the data into c e optimal clusters and minimize the following objective function:

[0106]

[0107] Where: J0(U,V) is the objective function of the fuzzy clustering algorithm, U and V are the membership matrix and cluster center matrix respectively; n dt is the sample size; u ij Represents the i-th data x i For the jth cluster center υ j The membership degree, u ij The larger the value, the greater the x i The more likely it is to be classified into the jth cluster; s represents the fuzzy factor. A larger s will cause the data point to be more inclined to belong to a single cluster, while a smaller s will make the membership distribution more uniform, and the data point can have a higher membership and thus belong to multiple clusters; ||x i -υ j || is the i-th data x i To the jth cluster center υ j The Euclidean distance of .

[0108] Secondly, the definition of the label matrix L is proposed, as shown in formula (2):

[0109]

[0110] Where, X v is the labeled data; l ij is the element in the i-th row and j-th column of the label matrix.

[0111] Label matrix L = [L v L u ] consists of two parts, one corresponding to the labeled data and the other corresponding to the unlabeled data. Then, the definition of the weight matrix W is proposed, as shown in formula (3):

[0112]

[0113] Where, X u is the unlabeled data; β1 and β2 are different constants, where β1∈(0,1], β2≥1; w ij is the element in the i-th row and j-th column of the weight matrix.

[0114] Finally, take the parameter Indicates whether the data has supervision constraints, as shown in formula (4):

[0115]

[0116] Where, is a tuning parameter used to balance the error between U and L, indicating that the supervision constraint only applies to labeled data. U represents the membership matrix.

[0117] The present invention modifies formula (1) into the following objective function, as shown in formula (5):

[0118]

[0119] Where J1(U,V) is the objective function of the fuzzy clustering algorithm after supervision constraint; o is the balance parameter; u ij Represents the i-th data x i For the jth cluster center υ j The degree of membership.

[0120] Step S2: Introducing multi-core distance into the clustering loss function

[0121] The kernel functions used in the present invention include: Gaussian kernel, linear kernel, polynomial kernel, etc., as shown in formulas (6)-(8):

[0122] 1) Gaussian kernel:

[0123]

[0124] Where x i and x j are the i-th and j-th data points respectively; σ is the scale parameter.

[0125] 2) Linear kernel:

[0126]

[0127] 3) Polynomial kernel:

[0128]

[0129] Where c a is the adjustment parameter used to translate the result of the inner product calculation; e xp is the degree of the polynomial. After the multi-core function is introduced, the multi-core distance from the data point to the cluster center is shown in formula (9):

[0130]

[0131] Where, is the multi-kernel distance from the data point to the cluster center; K is the total number of selected kernel functions; υ j is the jth cluster center; κ kis the kth kernel function; φ represents the mapping from the original data space to the high-dimensional space.

[0132] At this time, the semi-supervised multi-core clustering loss function with multi-core distance is introduced as shown in formula (10):

[0133]

[0134] Where, J s (U, V) The objective function of the fuzzy clustering algorithm with supervisory constraints is introduced based on the consideration of supervisory constraints.

[0135] Step S3: Get the updated formula of membership degree and cluster center based on Lagrange multiplier method

[0136] The Lagrange multiplier method is used to solve Equation (10), which is rewritten as a Lagrange function, as shown in Equation (11):

[0137]

[0138] Where λ i is the Lagrange multiplier. At this time, Equation (11) is decomposed into solving u ij 、υ j sub-problems;

[0139] 1) Apply equation (11) to u ij Take the derivative and set it to 0, and you can get u ij The update formula is shown in formula (12):

[0140]

[0141] Where, is the constraint distance; φ k (x i ) is the data point x i Mapping in kernel space; φ k (υ j ) is the cluster center υ j Mapping in kernel space.

[0142] 2) Apply equation (11) to υ j Taking the derivative and setting it to 0, we can get υ j The update formula is shown in formula (13):

[0143]

[0144] Where, is the kernel weight of the kth kernel function, and the calculation formula is shown in formula (14):

[0145]

[0146] Where A k is the sum of the weighted square errors of all data points relative to their respective category centers under the kth kernel function; For the kth φ The sum of the weighted squared errors of all data points relative to their respective category centers under the kernel function.

[0147] Step S4: Define information loss function

[0148] In order to obtain the key features of the data and reduce the interference of redundant information, a low-dimensional representation of the data is obtained through potential representation learning to learn the potential structure of the data. This process is implemented by a deep autoencoder with reconstruction loss. The mean square error is used to measure the difference between the original data X and the reconstructed data. The information loss between them is taken as the objective function, as shown in formula (15):

[0149]

[0150] Where, J lrl is the information loss function; f e is the encoder function; f d Decoder function.

[0151] Step S5: Define distribution consistency loss function

[0152] The introduction of distribution consistency loss ensures that the distribution characteristics of the original data and its potential representation are consistent. The kernel density estimation is used to estimate the probability distribution of the original data and the potential representation respectively, and the distribution consistency loss is constructed in the form of divergence, which is then introduced into the objective function, as shown in Equation (16):

[0153]

[0154] Where, J a is the distribution consistency loss function; Φ(x i ) and Φ(z i ) is the probability density function based on the Gaussian kernel; x i and z i are the data points in the original dimension and low-dimensional space respectively.

[0155] Therefore, the objective function of the clustering algorithm proposed in this invention is shown in formula (17):

[0156] minJ=J lrl +χ1J a +χ2J s ; (17)

[0157] Where J is the total loss function; J sis the clustering loss function; χ1 and χ2 are different balance parameters.

[0158] Step S6: Propose long-term and short-term cloud energy storage operation strategies

[0159] 1) Limit hydrogen energy storage to low power operation to ensure safety

[0160] Hydrogen energy storage not only significantly reduces operating efficiency and increases unit energy costs under low-power conditions, but also poses potential safety risks such as gas leakage and system instability. To prevent hydrogen energy storage from operating in unfavorable conditions for extended periods, the minimum power threshold for hydrogen energy storage activation is set at 20% of its rated power. This means that hydrogen energy storage will only operate when the system power demand exceeds this value, thus avoiding inefficient and dangerous operating conditions at the source.

[0161] 2) Introducing “Energy Storage Economic Balance Power” as the Strategy Switching Threshold

[0162] In order to balance the economy and operation efficiency of the system, the energy storage economic balance power P is defined as shift , meaning that when hydrogen energy storage and battery energy storage operate at this power point, their unit energy costs are equal. This power point not only reflects the economic switching limit between the two types of energy storage devices, but also serves as the sole switching threshold for operational strategies, guiding the actual selection of charging and discharging equipment.

[0163] 3) Using energy storage economic balance power as the switching point of operation strategy

[0164] ① If the economic balance power of energy storage is higher than the minimum operating threshold of hydrogen energy storage, the operating efficiency of hydrogen energy storage is high and the safety is guaranteed, and it can be directly used as the strategy switching threshold.

[0165] ② If the power is lower than the minimum operating threshold of hydrogen energy storage, in order to ensure system safety, the strategy uses the minimum power threshold of hydrogen energy storage as the strategy switching threshold.

[0166] 4) Strategic operation mode based on energy storage economic balance power

[0167] ① Mode 1: When the system operating power is lower than the balance power, the battery energy storage independently bears the system's charging and discharging needs, and the hydrogen energy storage does not operate.

[0168] ② Mode 2: When the system operating power is not less than the balance power, it switches to hydrogen energy storage operation to meet the system's charging and discharging needs, and the battery energy storage exits operation.

[0169] Step S7: Establish a two-tier optimization configuration model for cloud energy storage considering the two-part rental price

[0170] (1) Upper model

[0171] 1) Objective function:

[0172]

[0173] Where C B and C H is the annual comprehensive cost of energy storage; C TCC,B and C TCC,H = The annual investment and construction cost of battery energy storage and hydrogen energy storage; C OC,B and C OC,H The operation and maintenance costs of battery energy storage and hydrogen energy storage; C ES,B and C ES,H is the energy consumption cost of battery energy storage and hydrogen energy storage; C RL,B and C RL,H The replacement cost of battery energy storage and hydrogen energy storage; I LEA,B and I LEA,H The annual rental income of battery energy storage and hydrogen energy storage; IP,B and I IP,H Peak-valley arbitrage benefits for battery energy storage and hydrogen energy storage; I RC,B and I RC,H The recovery value of battery energy storage and hydrogen energy storage.

[0174] The equivalent annual investment and construction cost of battery energy storage is shown in formula (19):

[0175]

[0176] Where N B is the number of battery energy storage configurations; c in is the inverter cost; c bat For battery cost; For the The configured power of the battery energy storage unit; For the The configuration capacity of the battery energy storage; CRF,B The annual capital recovery rate of battery energy storage is calculated as follows:

[0177]

[0178] Where r is the discount rate; y B The service life of the battery energy storage.

[0179] The equivalent annual investment and construction cost of hydrogen energy storage is shown in formula (21):

[0180]

[0181] Where N H is the configuration quantity of hydrogen energy storage; c PME is the unit power cost of the proton exchange membrane electrolyzer; c FCis the unit power cost of the proton exchange membrane fuel cell; c ht is the unit capacity cost of the hydrogen storage tank; For the Configuration power of each proton exchange membrane electrolyzer; For the Configuration power of proton exchange membrane fuel cells; For the Configuration capacity of hydrogen storage tank; CRF,H The annual capital recovery rate of hydrogen energy storage is calculated as follows:

[0182]

[0183] Where y H The useful life of hydrogen energy storage.

[0184] The operation and maintenance cost of battery energy storage is shown in formula (23):

[0185]

[0186] Where, τ OC,in and τ OC,bat are the maintenance factors of the inverter and battery respectively.

[0187] The operation and maintenance cost of hydrogen energy storage is shown in formula (24):

[0188]

[0189] Where, τ OC,PME , τ OC,FC and τ OC,ht They are the maintenance factors of proton exchange membrane electrolyzer, proton exchange membrane fuel cell and hydrogen storage tank respectively.

[0190] The energy consumption cost of battery energy storage is shown in formula (25):

[0191]

[0192] Where, ρ pur (t) and ρ sell (t) are the electricity purchase and electricity sales prices of the electric-hydrogen hybrid energy storage system during period t; and Respectively The charge / discharge power of the battery energy storage.

[0193] The energy consumption cost of hydrogen energy storage is shown in formula (26):

[0194]

[0195] Where, The first The working efficiency of the proton exchange membrane electrolyzer; and Respectively The charging / discharging power of hydrogen energy storage.

[0196] The replacement cost of battery energy storage is shown in formula (27):

[0197]

[0198] Where n bat is the total number of battery replacements, n bat =y B / y bat ;y bat Is the battery life; l bat For battery l bat Replacement times; k B The annual decline in battery storage installation costs.

[0199] The replacement cost of hydrogen energy storage is shown in formula (28):

[0200]

[0201] Where n PEM is the total number of replacements of the proton exchange membrane electrolyzer, n PEM =y H / y PEM ;y PEM is the life of the proton exchange membrane electrolyzer; l PEM For the proton exchange membrane electrolyzer PEM Replacement times; n FC is the total number of FC replacements, n FC =y H / y FC ;y FC is the life of the proton exchange membrane fuel cell; l FC Proton exchange membrane fuel cell FC Replacement times; k H The annual reduction rate of hydrogen energy storage installation cost.

[0202] The peak-valley arbitrage benefits of battery energy storage are shown in formula (29):

[0203]

[0204] The peak-valley arbitrage benefits of hydrogen energy storage are shown in formula (30):

[0205]

[0206] The annual rental income of battery energy storage is shown in formula (31):

[0207]

[0208] Where, ρ cap (t) and ρ power (t) are the capacity rental price and electricity rental price of cloud energy storage in period t respectively; and Respectively The rental capacity and total charging and discharging power of the battery energy storage system.

[0209] The annual rental income of hydrogen energy storage is shown in formula (32):

[0210]

[0211] Where, and Respectively The rental capacity and total charging and discharging power of Taiwan Hydrogen Energy Storage.

[0212] The recovery value of battery energy storage is shown in formula (33):

[0213]

[0214] Where, γ B is the recovery factor of the battery.

[0215] The recovery value of hydrogen energy storage is shown in formula (34):

[0216]

[0217] Where, γ PEM and γ FC are the recovery factors of proton exchange membrane electrolyzer and proton exchange membrane fuel cell, respectively.

[0218] 2) Constraints

[0219] Power balance constraints:

[0220]

[0221] Where U y (t) is the voltage of node y during period t; θ xy (t) is the voltage phase difference between node x and node y during period t; G xy and B xy are the equivalent conductance and susceptance of the line between node x and node y.

[0222] Configuration power and capacity constraints for battery energy storage and hydrogen energy storage:

[0223]

[0224] Where, These are the upper and lower limits of battery energy storage configuration capacity and the upper and lower limits of configuration power respectively.

[0225]

[0226] Where, They are the upper and lower limits of PEME configuration power (PEM water electrolysis system), the upper and lower limits of PEMFC configuration power (PEM fuel cell system), and the upper and lower limits of hydrogen storage tank configuration capacity.

[0227] SOC (State of Charge, a parameter that measures the current remaining battery power) constraints:

[0228] SOC min ≤SOC≤SOC max ; (38)

[0229] Where, SOC max and SOC min They are the upper and lower limits of the SOC of battery energy storage.

[0230] LOH constraints:

[0231] LOH min ≤LOH≤LOH max ; (39)

[0232] In the formula, LOH max and LOH min They are the upper and lower limits of LOH (LOH refers to Level of Hydrogen, i.e. the level or state of hydrogen) of hydrogen energy storage.

[0233] (2) Lower-level model

[0234] 1) Objective function:

[0235]

[0236] Where C EXA is the annual assessment cost of users including wind farms; C FB Assess the cost of power generation forecast deviation for users; C LEA The annual rental cost of cloud energy storage for users; N s is the number of typical scenarios; π n is the number of days when the nth scenario occurs; T is 24 hours.

[0237] 2) Constraints

[0238] Wind power output constraints:

[0239]

[0240] Where, P WT (t) is the wind power output during period t, It is the upper limit of wind power output.

[0241] Wind curtailment constraints:

[0242]

[0243] Where R WT is the annual wind curtailment rate, is the maximum annual wind curtailment rate.

[0244] Step S8: Improving fungal growth optimization algorithm based on memory mechanism

[0245] (1) Initialization phase

[0246] In the present invention, spores germinate at random locations with suitable growth conditions, thereby generating different hyphae. These hyphae then grow within the search space and explore to find areas with higher nutritional value. To this end, the proposed fungal growth optimization algorithm is initialized by randomly distributing M hyphae within the search space, where each hyphae consists of d dimensions and represents a solution to the optimization problem. Specifically, each hyphae (solution) is initialized by random initialization within a predetermined lower and upper bound, as shown in Equation (43):

[0247]

[0248] Where, is the lower bound vector of each dimension in the optimization problem; ⊙ is the Hadamard product operation of two vectors; X U is the upper bound vector of each dimension in the optimization problem; rand is a random number in [0,1]. After initializing the lower and upper bounds of all solutions, an M×d matrix is ​​generated, which represents the initial population, as shown in Equation (44):

[0249]

[0250] Where, X pop is the initial population matrix. All solutions in the population are evaluated one by one by fitness (objective function) and compared with each other to identify the eutrophic area, that is, the current optimal solution If the objective function needs to be minimized, then this solution has the lowest fitness.

[0251] (2) Memory initialization

[0252] In the initialization phase of the algorithm, it is first necessary to calculate the objective function value of each individual in the population, that is, to evaluate the fitness of each individual. The objective function value usually represents the quality or degree of superiority of the individual relative to the optimal solution in the current solution space. After calculating the objective function value of each individual, the individual with the best fitness (that is, the individual with the smallest or largest objective function value, depending on the problem type) is set as the initial global memory. Global memory is a key reference information that contains the location of the optimal solution in the current iteration and the fitness value of the solution. This optimal solution will serve as guidance information for subsequent updates, helping to guide the entire population towards the optimal solution during the search process. By saving the global optimal solution, the algorithm can continuously optimize the quality of the solution by comparing the difference between the current solution and the optimal solution in subsequent iterations, thereby effectively avoiding falling into local optimality and accelerating the convergence process, as shown in Equation (45):

[0253]

[0254] Where, is the initial optimal solution; θ 0 is the global optimal solution in the initial stage, that is, the memory vector.

[0255] (3) Reset phase

[0256] At the beginning of each iteration, the algorithm processes each individual in the population to ensure they converge toward a higher-quality solution space in the new iteration. Specifically, for each individual, the algorithm directly utilizes the current global or local optimal solution in the non-forgotten dimension. These optimal solutions serve as memory information, providing the direction of the most promising solutions in the current search space, guiding the individual toward these high-quality solutions during the iteration process. This mechanism ensures that individuals do not stray from the optimal solution, accelerating convergence and continuously improving the quality of the solution with each update. In this way, individuals can more effectively explore the feasible solution space, avoid ineffective exploration, and conduct refined local searches around high-quality solutions. During this process, only the non-forgotten dimension directly utilizes the current optimal memory information, while the forgotten dimension uses local perturbations to conduct appropriate exploration, maintaining population diversity and preventing premature convergence. This balanced exploration and exploitation approach allows the algorithm to effectively switch between global and local optimization, ensuring continuous tracking of the optimal solution. The update formula introduces global or local memory into the individual position update to ensure that each individual continuously moves closer to the optimal solution during the search process, as shown in formula (46):

[0257]

[0258] Where, For the i terAt the +1th iteration, the τth individual is in Dimensional location; For the i ter The global optimal solution (memory vector) at the iteration; Υ is the set of forgotten dimensions.

[0259] (4) Forgetting-Replenishing Stage

[0260] Within the selected set of "forgotten dimensions", the algorithm no longer uses the information of the current optimal solution, but instead performs local perturbation updates on the individual's position in these dimensions to simulate the organism's forgetting and self-repair process of some information. Specifically, by introducing a certain amount of random perturbation, the individual is allowed to deviate from the existing trajectory in these dimensions, thereby jumping out of the possible local optimum and enhancing the diversity and uncertainty of the search. At the same time, in order to avoid excessive perturbation causing the search direction to deviate from the target area, the perturbation amplitude is designed to gradually decay with the number of iterations. In other words, in the early stages of the algorithm, larger perturbations help to explore the solution space extensively; in the later stages, the perturbations gradually weaken to support more refined local development. This mechanism not only achieves a smooth transition from global search to local optimization, but also enhances the diversity of solutions while improving the stability of the algorithm. As shown in formula (47):

[0261]

[0262] Where, For the The perturbation vector of individual τ in dimension is as shown in formula (48):

[0263]

[0264] Where φ(τ) is the attenuation factor, which takes the form of a cosine function and gradually decreases with the increase of the number of iterations; is the maximum number of iterations; The upper limit of the number of iterations in the exploration phase is usually satisfied l τ To explore the lower limit of the space, is a random number in [0,1], The upper limit of the exploration space.

[0265] The forgetting mechanism introduces random perturbations in some dimensions, and gradually reduces the perturbation amplitude through the cosine attenuation factor, avoiding excessive perturbations that lead to unstable search. It can also maintain strong global exploration capabilities in the early stages and achieve refined development in the later stages.

[0266] (5) Memory sharing stage

[0267] After the individual has gone through the memory initialization and forgetting-supplementation phases, the algorithm will use the information of other individuals in the population to supplement and update, promoting knowledge sharing among individuals. This process allows each individual to rely not only on their own historical experience, but also to learn from the local advantages of other individuals, thereby improving the collaborative ability of the entire group during the search process. Through information sharing and updating, individuals can make full use of the high-quality solutions of other members in the population, avoiding the dilemma of a single individual falling into the local optimal solution. In this way, when performing a global search, the algorithm not only maintains the independence of the individual, but also effectively combines the local excellent solutions within the group to enhance the diversity and effectiveness of the overall search. This mechanism helps to improve the search efficiency of the group, reduce the probability of falling into the local optimal solution, and accelerate the convergence of the search to the global optimal solution. The update formula is shown in Equation (50):

[0268]

[0269] Where r is a random number in [0,1]; p is the sharing probability threshold; and m is the individual index randomly selected from the population.

[0270] (6) Mycelial tip growth stage

[0271] Hyphal tip growth is one of the most typical and critical behaviors in fungal growth and forms the basis for mimicking natural search behavior in fungal growth optimization algorithms. Under appropriate conditions, each hypha in a population expands in a straight line, exploring the search space and seeking nutrient-rich areas. However, certain environmental and chemical factors may cause hyphae to shift their growth direction. Furthermore, hyphal growth rates vary depending on multiple factors, such as fungal species, environmental conditions (humidity, temperature, and nutrient availability), and hyphal age. Typically, younger hyphae grow faster. Furthermore, hyphal growth may accelerate in nutrient-rich areas, such as near decomposing organic matter, potentially leading to exponential growth. Conversely, in nutrient-poor or unfavorable environments, hyphal growth may slow significantly. In the proposed fungal growth optimization algorithm, variations in hyphal growth rate are used to maximize the exploration operator, thereby avoiding local optima and ultimately achieving a global optimal solution.

[0272] 1) Growth rate and direction

[0273] In the present invention, the growth rate in the fungal growth optimization algorithm is simulated by an exponential function, which is related to the fitness value of the current individual. Solutions with high fitness are considered to be in nutrient-rich areas, while solutions with low fitness represent nutrient-poor areas. Through this mechanism, the fungal growth optimization algorithm can match the growth rate with the quality of the solution, realizing dynamic exploration based on nutrient distribution. The growth rate based on the exponential function in the fungal growth optimization algorithm is shown in Equation (51):

[0274]

[0275] Where, f τ is the fitness value of the τth solution, which is divided by the sum of all individual fitness values ​​to achieve normalization, limiting its range to between 0 and 1. This can minimize the numerical amplification effect brought by the exponential function during the optimization process; f σ is the fitness value of the σth solution. Before normalization, the magnitude of the exponential value can be very large, easily causing the solution to jump outside the feasible search boundary of the optimization problem; after normalization, the scale of the exponential value is effectively controlled, allowing the solution to search more accurately near the boundary. In this formula, a higher fitness value indicates that the individual is in a more nutrient-rich area and its growth rate is faster, which helps to improve the role of the exploration operator; conversely, a lower fitness value indicates a poorer nutrition and slower mycelial growth rate, thereby further strengthening the development effect on the local area and helping to improve the performance of the development operator. This mechanism effectively balances the relationship between global exploration and local development by mapping nutrient levels and growth behavior.

[0276] For each fitness value, the growth rate is constant, which means that the exploration behavior of the fungal growth optimization algorithm lacks diversity and it is difficult to accurately explore the entire search space. In other words, when the growth rate of all individuals is determined solely by the fitness value, the expansion path of the hyphae may become regular, thereby reducing the exploration efficiency. However, multiplying the normalized fitness value by a random number between 0 and 1 can cause the growth rate to fluctuate, thereby breaking the original regularity. This fluctuation mechanism introduces uncertainty, so that individuals no longer follow a fixed growth pattern during the exploration phase, thereby improving the diversity of understanding and the flexibility of search. The growth rate after adding fluctuations is shown in Equation (52):

[0277]

[0278] Where F is the adjustment factor of the normalized fitness value after the introduction of random perturbations; ξ is the factor that regulates the “exploration” and “exploitation” operators during the optimization process; i ter is the current iteration number; is the maximum number of iterations. r1 is a random number in the range [0, 1]. E is a factor used to adjust the exploration and exploitation operators during the optimization process. Initially, this factor increases the mycelial growth rate to enable rapid exploration of the entire search space. As the optimization process progresses, this factor decreases, slowing the growth rate and enabling in-depth exploration of areas near the current solution, thereby enhancing the exploitation capability of the proposed algorithm.

[0279] The growth direction of hyphae is affected by sudden environmental and chemical signals. Therefore, in the fungal growth optimization algorithm, the growth direction of hyphae is represented in a random manner to simulate the possible changes in the growth direction of hyphae. In the fungal growth optimization algorithm, the growth direction is determined by randomly selecting two solutions from the current population and calculating the difference between them. If the difference is less than zero, the current hyphae will change its growth direction; otherwise, it will continue to grow in the current direction. The growth direction vector is shown in Equation (53):

[0280]

[0281] Where, and are two different solutions randomly selected from the current population.

[0282] The update rate of the current solution is related to the diversity of the population. A higher population diversity results in a greater difference between two randomly selected solutions, leading to a higher growth rate, maintaining population diversity and enabling more accurate exploration of the search space. Conversely, if the population diversity is low, the difference between two solutions is smaller, and the current mycelial growth rate is correspondingly reduced, which helps enhance the effectiveness of the development operator during the optimization process. This allows the mycelium to adjust its growth direction in the event of unexpected environmental or chemical signals.

[0283] First, the update rate of the current individual is affected by population diversity. When the distribution of solutions within a population is dispersed and the differences are large, it indicates high population diversity. In this case, the difference between two randomly selected solutions is also large. A larger difference means a wider search step, which helps individuals expand beyond their local area, increasing the breadth of the search and thus more fully exploring the entire solution space. When the population converges and the differences between solutions are small, the population diversity decreases, the differences between individuals are smaller, and the update step becomes smaller, which facilitates the algorithm to refine local areas and improve convergence accuracy. Second, because in real environments, the growth direction of fungal hyphae can be affected by sudden chemical signals or changes in the external environment, the fungal growth optimization algorithm can effectively simulate this uncertainty in growth paths by introducing random directional differences, enabling the hyphae to dynamically adapt to the environment.

[0284] Therefore, the present invention multiplies the difference between the two solutions by the growth rate to form the actual growth vector of the mycelium, which further determines its forward direction and step size in the search space. This operation not only combines the changing trend of the overall structure of the population, but also integrates the regulation of the growth rate by individual fitness, so that the mycelium can adaptively adjust its expansion speed while maintaining search diversity. Finally, this growth vector is used to update the individual position, so that it moves towards the possible high-quality solution area, which not only promotes extensive exploration of the solution space, but also lays the foundation for subsequent local development. This mechanism realizes the characteristic of mycelium behavior to respond quickly to resource changes in the natural environment, which is reflected in the optimization algorithm as the dynamic adjustment and adaptive control ability of the search strategy.

[0285] Finally, according to the calculated growth rate and growth direction of the τth mycelium, its new growth vector is shown in formula (54):

[0286]

[0287] The growth of hyphae may change only in some dimensions, rather than expanding uniformly in all dimensions. Therefore, in the fungal growth optimization algorithm, the current growth vector of the hyphae is combined with the newly calculated growth vector to generate a new hyphae growth vector. The update process follows the formula (55):

[0288]

[0289] Where, For the A random value generated in the dimension, obeying the uniform distribution in the interval [0,1]; r2 is a random number in [0,1]; the symbol [·] is an Iverson bracket, which separates the logical statements (such as ) is converted to a Boolean value: 1 if the condition is true, 0 otherwise.

[0290] 2) Chemotactic behavior

[0291] In the fungal growth optimization algorithm, the development operator is based on the simulation of chemotaxis behavior, which enables the hyphae to grow towards nutrient-rich or other specific areas according to chemical signals, so as to obtain a better solution as quickly as possible. The development operator design in the present invention is based on the simulation of two states. The first state assumes that the growth direction of the hyphae is towards the nutrient-rich area to enhance the in-depth development of the surrounding positions and improve the convergence speed. The current optimal solution not only represents the nutrient-rich area, but also some hyphae can communicate information with other hyphae, indicating that there are nutrients in their current position, thereby guiding other hyphae to grow towards this area. Equation (56) defines the growth direction of the hyphae towards a randomly selected solution in the population, which is used to simulate the ability of the hyphae to generate chemical signals and indicate nutrient accessibility:

[0292]

[0293] In the formula, r3 is a random number in [0,1]. This formula indicates that the growth direction of the τth hypha is towards the individual randomly selected from the population. The location of the fungus. In addition, in the first state, the hyphae may also migrate directly to the current optimal solution position, which is regarded as the most nutrient-rich area, thereby accelerating the convergence of the algorithm. However, always moving towards the optimal solution may cause the algorithm to converge prematurely, thereby reducing the overall performance of the fungal growth optimization algorithm when facing complex optimization problems. In order to solve this problem, the fungal growth optimization algorithm adopts a mechanism based on preset probability to guide the growth direction to the most nutrient-rich location. Finally, the mathematical model of the growth direction of the τth hyphae towards the current optimal solution or the randomly selected solution (corresponding to the first state) is shown in formula (57):

[0294]

[0295] Where, is a vector containing random numbers uniformly distributed in the interval [0,1]; r5 is a random number in the interval [0,1]; ζ is a parameter between 0 and 1, which is used to determine whether the hyphae grow towards the current optimal solution; β is a randomly selected value, which takes the value of 1 or -1. When -1 is selected, the growth direction will move away from the current optimal solution. It is assumed that the opposite direction may contain some more nutritious hyphae, which helps to avoid premature convergence.

[0296] In nutrient-rich areas, the growth rate of hyphae may increase significantly. Therefore, the further the hyphae move toward the current optimal solution or a randomly selected solution, the higher the growth rate. To enhance the exploitation capability in the first half of the optimization process, the nutrient allocation in this behavior is first defined randomly. Then, based on the fitness value, the exploitation operator is associated with the fitness value of each solution. The higher the fitness value, the faster the convergence speed, and vice versa. This behavior is shown in Equation (58):

[0297]

[0298] Where η τ is the growth step of mycelium under the current development behavior; r6 is a random number in [0,1].

[0299] The scale of the fitness value may be large, causing the updated solution to exceed the search boundary, so η τ Normalized to the interval [0,1], as shown in formula (59):

[0300]

[0301] Where r7 is a random number in the interval [0,1]. A random number from a uniform distribution in the interval [0,2] is added to the normalized nutrient distribution to maximize its influence in the optimization process. Finally, the growth update of the τth hyphae according to the first state is defined as follows:

[0302]

[0303] The second state is based on hyphae growing away from nutrient-rich areas, simulating their behavior in response to potentially harmful or inhibitory chemicals that could hinder normal hyphae growth. The mathematical model for this state is defined as follows:

[0304]

[0305] Where, is a random number in the interval [0,1]. In formula (62), the logical expression This is used to determine whether nutrients affect the growth rate of the τth mycelium in all dimensions. To enable mycelium to adapt to sudden changes in the environment, an additional exploration step is added to continue exploring the search space. The two formulas after adding the exploration step are as follows:

[0306]

[0307] Where, is a vector containing uniformly distributed random numbers in the interval [0,1]; r9 is a random number in the interval [0,1].

[0308]

[0309]

[0310] Where r 10 and r 11 is a random number in the interval [0,1].

[0311]

[0312] Where r 12 is a random number in the interval [0,1]; E p is a preset probability used to determine whether to apply an additional exploration step in the first state. Its value is between [0, 1] to reduce the problem of premature convergence or slow convergence caused by random processes. Since the mycelial growth direction may change suddenly, the switching between the first state and the second state is performed in a random manner, as shown in Equation (68):

[0313]

[0314] The trade-off between exploration and exploitation in hyphal tip growth behavior is based on the following assumption: hyphae in the worst position will try to apply the exploitation operator and move toward nutrient-rich areas, while hyphae already in nutrient-rich areas will continue to explore in order to find areas with better nutrition. This assumption is achieved by normalizing the fitness value of each solution to between 0 and 1, as shown in Equation (69):

[0315]

[0316] In the formula, ε is a number close to 0 to avoid division by 0; p τ is the normalized fitness value, if p τ Less than the exploration rate E defined in formula (70) r , then perform the exploration operation, otherwise, perform the development operation.

[0317]

[0318] Where μ is a preset probability that controls the trade-off between the exploration operator and the exploitation operator. In the initial stage of the optimization process, all hyphae will perform exploration operations to fully search the solution space and find the most nutrient-rich locations. As the optimization process progresses, when the p of some hyphae increases, the number of τ >E r When , they will enter the development stage and develop towards the nutrient-rich area, and their position update is shown in formula (71):

[0319]

[0320] (7) Mycelial branching and spore germination stage

[0321] Mycelial branching behavior allows fungi to expand in multiple directions, thereby improving their resource utilization efficiency. Branching can be divided into apical branches and lateral branches. Lateral branches are particularly important, allowing fungi to explore in multiple directions, forming a powerful exploration operator, enabling the fungal growth optimization algorithm to escape the local optimal solution. The present invention designs this behavior based on two different modes to explore the search space as much as possible. The first mode determines the growth direction based on two solutions randomly selected from the current population, as shown in formula (72):

[0322]

[0323] In the formula, the growth direction of new hyphae is based on and If the former is greater than the latter, the difference is positive and is added to the current mycelium, causing the new mycelium to grow in the positive direction. Otherwise, the difference is negative and is subtracted from the current mycelium, causing the new mycelium to grow in the negative direction. The second mode determines the growth direction based on the current optimal solution and a solution randomly selected from the current population, as shown in Equation (73):

[0324]

[0325] Where, is a randomly selected solution from the current population. This formula helps the newly generated hyphae grow in different directions, thus exploring the search space as accurately as possible. Finally, the growth direction of the new hyphae is shown in Equation (74):

[0326]

[0327] Where r 15 is a random number in the interval [0,1], when r 15 When r > 0.5, the first mode has a greater impact on the growth of new hyphae. 15 When <0.5, the second mode has a greater impact on the growth of new hyphae.

[0328] In terms of growth rate, the growth rate of lateral hyphae is comparable to that of the original hyphae, especially under optimal conditions. However, the growth rate of lateral branches can be affected by environmental conditions such as nutrient availability, stress levels, and fungal species. To simulate this behavior, the growth rate of hyphae generated by lateral branches was calculated as follows:

[0329]

[0330] Where r 16 and r 17 is a random number in the interval [0,1], used to determine whether the growth rate of the new mycelium is similar to that of the original mycelium. 16 <r 17 , the growth rate of the new mycelium is close to that of the original mycelium; otherwise, the growth rate will be significantly reduced.

[0331] After introducing the growth rate, equation (74) is redefined as follows:

[0332]

[0333] Spore germination is the process of expanding the fungal population through spore reproduction. Spores germinate in a suitable environment, generate new hyphae, and continue to grow. In the fungal growth optimization algorithm, spore germination is simulated as randomly generating new solutions in the search space. These solutions are based on the mean of the current optimal solution and the random solution to balance exploration and exploitation, helping to avoid premature convergence and promote global search. The specific process is shown in Equation (77):

[0334]

[0335] Where, X g A randomly selected value between -1 and 1.

[0336] A random threshold p1 is preset, and branching or spore germination updates are performed on some dimensions. For dimensions that meet the conditions, the original position is retained. Otherwise, the update is based on the memory sharing mechanism, as shown in formula (78):

[0337]

[0338] Where r 18 is a random number in [0,1]; p1 is the random threshold for whether to perform branching or spore germination operations.

[0339] (8) Memory Update

[0340] At the end of each iteration, the algorithm evaluates the fitness values ​​of all individuals in the population to determine the best performing individual in the current iteration. By comparing the fitness values ​​of each individual, the individual with the best fitness is selected as the global optimal solution, and the global memory is updated. This process ensures the dynamic update of the optimal solution and can reflect the optimal state of the current search process in real time, thereby providing more accurate reference information for subsequent searches. In the process of updating the global memory, if the current global optimal solution is better than the historical solution, the historical optimal solution is replaced, and the new optimal solution is used as a reference to guide the individual to search in a better direction in subsequent iterations. Through this mechanism, the algorithm can continuously track the optimal solution and gradually approach the global optimal solution, thereby improving the overall search performance and convergence speed. As shown in formula (79):

[0341]

[0342] Where, For the i ter +1 times the global optimal solution.

[0343] (9) Model solution

[0344] Based on the Karush-Kuhn-Tucker optimality criteria, the established two-layer model was converted to a single-layer optimization model and solved using an improved fungal growth optimization algorithm. System parameters and wind power output data were input, and the parameters, maximum number of iterations, and population size of the improved fungal growth optimization algorithm were set. A memory strategy update was performed on the non-forgotten dimension; a forget-supplement update was performed on the forgotten dimension; a dream sharing update was performed on the forgotten dimension; hyphae tip growth, hyphae branching, and spore germination were performed. These steps were repeated until convergence conditions were met.

[0345] like Figure 2-Figure 3 As shown in the figure, it is a flow chart of the fungal growth optimization algorithm, where: Figure 2 The line segments A1, A2 and A3 in Figure 3 The line segments A1, A2 and A3 in are connected.

[0346] The above descriptions are only partial embodiments of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and modifications without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market, characterized in that: include: Based on latent representation learning and information fusion, the objective function of a semi-supervised multi-kernel fuzzy clustering algorithm is constructed, and the semi-supervised multi-kernel fuzzy clustering algorithm is used to cluster typical scenarios of offshore wind power. Based on the characteristics of hydrogen energy storage and battery energy storage, and taking into account performance and economic factors, a coordinated optimization operation strategy for long-term and short-term cloud energy storage, including hydrogen energy storage and battery energy storage, is output; Establish a two-tier optimization configuration model for long-term and short-term cloud energy storage that considers two-part leasing prices and typical scenarios, including both upper and lower tier models. Based on the optimality condition, the long-term and short-term cloud energy storage double-layer optimization configuration model is converted into a single-layer optimization model; the single-layer optimization model is solved using an improved fungus growth optimization algorithm based on the memory mechanism.

2. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 1, characterized in that: The objective function of the semi-supervised multi-core fuzzy clustering algorithm is constructed based on potential representation learning and information fusion, and the semi-supervised multi-core fuzzy clustering algorithm is used to cluster typical scenarios of offshore wind power, including: Construct the label matrix and weight matrix, and adjust the objective function of the fuzzy clustering algorithm after supervision constraints based on the label matrix, weight matrix, membership matrix and adjustment parameters; Based on the objective function of the fuzzy clustering algorithm, a semi-supervised multi-core clustering loss function is constructed by introducing multi-core distance, and the Lagrange multiplier method is used to solve the semi-supervised multi-core clustering loss function, and the semi-supervised multi-core clustering loss function in the form of Lagrangian function is obtained; The semi-supervised multi-kernel clustering loss function in the form of Lagrangian function is decomposed into the data-to-cluster center membership subproblem and the cluster center subproblem, and the derivatives are taken to obtain the update formulas of the data-to-cluster center membership and the cluster center update formula respectively; Based on a deep autoencoder with reconstruction loss, a low-dimensional representation of the source-load data is obtained through latent representation learning to learn the latent structure of the source-load data. The mean square error is used to measure the information loss between the original source load data and the reconstructed data, and the information loss function is determined; Kernel density estimation is used to estimate the probability distribution of the original source load data and the potential representation respectively, and a distribution consistency loss function is constructed in the form of divergence; The objective function of the semi-supervised multi-kernel fuzzy clustering algorithm is obtained by integrating the semi-supervised multi-kernel clustering loss function, information loss function and distribution consistency loss function; The semi-supervised multi-kernel fuzzy clustering algorithm is used to cluster typical scenarios of offshore wind power.

3. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 2, characterized in that: The label matrix and weight matrix are constructed, and the objective function of the fuzzy clustering algorithm after the supervision constraint is adjusted according to the label matrix, weight matrix, membership matrix and adjustment parameter includes: Based on the labeled data and unlabeled data, a label matrix is ​​constructed; a weight is assigned to each data according to the label matrix, and a weight matrix is ​​obtained; The error parameter is used to represent the supervision constraint of the data and balance the error between the membership matrix and the label matrix; According to the label matrix, weight matrix, membership matrix and adjustment parameters, the objective function of the fuzzy clustering algorithm is adjusted to obtain the objective function of the fuzzy clustering algorithm after supervision constraint: Where J1(U,V) is the objective function of the fuzzy clustering algorithm after supervision constraint; n dt is the sample size, c e is the optimal number of clusters; w ij is the element in the i-th row and j-th column of the weight matrix, ||x i -υ j || is the i-th data x i To the jth cluster center υ j The Euclidean distance, u ij Represents the i-th data x i For the jth cluster center υ j The membership degree, l ij is the element in the i-th row and j-th column of the label matrix, o is the balance parameter, is the adjustment parameter.

4. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 2, characterized in that: The objective function of the fuzzy clustering algorithm is based on which a semi-supervised multi-core clustering loss function is constructed by introducing the multi-core distance. The Lagrange multiplier method is used to solve the semi-supervised multi-core clustering loss function. The semi-supervised multi-core clustering loss function in the form of a Lagrange function is obtained. By introducing multi-core distance into the objective function of fuzzy clustering algorithm, we obtain the semi-supervised multi-core clustering loss function. The Lagrange multiplier method is used to solve the semi-supervised multi-core clustering loss function, and the semi-supervised multi-core clustering loss function in the form of Lagrange function is obtained: Where, J s (U, V) is the semi-supervised multi-kernel clustering loss function in the form of Lagrangian function, is the multi-core distance from the data point to the cluster center, λ i is the Lagrange multiplier; n dt is the sample size, c e is the optimal number of clusters; w ij is the element in the i-th row and j-th column of the weight matrix, u ij Represents the i-th data x i For the jth cluster center υ j The membership degree, l ij is the element in the i-th row and j-th column of the label matrix, o is the balance parameter, is the adjustment parameter.

5. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 2, characterized in that: When the objective function of the semi-supervised multi-core fuzzy clustering algorithm is obtained by integrating the semi-supervised multi-core clustering loss function, the information loss function and the distribution consistency loss function, the objective function of the semi-supervised multi-core fuzzy clustering algorithm is: min J=J lrl +χ1J a +χ2J s ; Where J is the total loss function, J lrl is the information loss function, J a is the distribution consistency loss function, J s is the clustering loss function; χ1 and χ2 are different balance parameters.

6. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 1, characterized in that: Based on the characteristics of hydrogen energy storage and battery energy storage, and taking into account performance and economic factors, the long-term and short-term cloud energy storage coordinated optimization operation strategies that include hydrogen energy storage and battery energy storage operations include: Determine the economic equilibrium power of energy storage based on the unit energy cost of hydrogen energy storage and battery energy storage; When the economic balance power of energy storage is higher than the minimum operating threshold of hydrogen energy storage, the economic balance power of energy storage is used as the strategy switching threshold; when the economic balance power of energy storage is lower than the minimum operating threshold of hydrogen energy storage, the minimum power threshold of hydrogen energy storage is used as the strategy switching threshold; When the operating power of the cloud energy storage system is lower than the strategy switching threshold, the battery energy storage will independently bear the charging and discharging needs of the cloud energy storage system; When the system operating power is higher than or equal to the strategy switching threshold, it switches to hydrogen energy storage operation.

7. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 1, characterized in that: The establishment of a two-tier optimization configuration model for long-term and short-term cloud energy storage considering two-part rental prices and typical scenarios includes: Constructing the upper layer model of the long-term and short-term cloud energy storage dual-layer optimization configuration model, the upper layer model aims to minimize the equal annual value comprehensive cost of cloud energy storage; Construct the lower model of the long-term and short-term cloud energy storage two-tier optimization configuration model that takes into account the leasing market and typical scenarios. The lower model aims to minimize the user's annual assessment cost.

8. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 7, characterized in that: The upper layer model of the long-term and short-term cloud energy storage dual-layer optimization configuration model aims to minimize the equivalent annual value comprehensive cost of cloud energy storage and includes: Calculate the equivalent annual investment and construction costs, operation and maintenance costs, energy consumption costs, replacement costs, annual rental income, peak-valley arbitrage income, and recovery value of battery energy storage; Calculate the equivalent annual investment and construction costs, operation and maintenance costs, energy consumption costs, replacement costs, annual rental income, peak-valley arbitrage income, and recovery value of hydrogen energy storage; The objective function of the upper model of the long-term and short-term cloud energy storage dual-layer optimization configuration model is: Where, F up The objective function of the upper model of the long-term and short-term cloud energy storage dual-layer optimization configuration model; C B and C H are the annual comprehensive costs of battery energy storage and hydrogen energy storage respectively; C TCC,B and C TCC,H are the equivalent annual investment and construction costs of battery energy storage and hydrogen energy storage respectively; C OC,B and C OC,H are the operation and maintenance costs of battery energy storage and hydrogen energy storage respectively; C ES,B and C ES,H are the energy consumption costs of battery energy storage and hydrogen energy storage respectively; C RL,B and C RL,H are the replacement costs of battery energy storage and hydrogen energy storage respectively; I LEA,B and I LEA,H are the annual rental income of battery energy storage and hydrogen energy storage respectively; I IP,B and I IP,H are the peak-valley arbitrage benefits of battery energy storage and hydrogen energy storage respectively; RC,B and I RC,H The recovery value of battery energy storage and hydrogen energy storage; Construct the constraint functions of the upper model of the long-term and short-term cloud energy storage dual-layer optimization configuration model, including power balance constraints, configuration power and capacity constraints of battery energy storage and hydrogen energy storage, SOC constraints, and LOH constraints; The power balance constraint is: Where U y (t) is the voltage of node y during period t; θ xy (t) is the voltage phase difference between node x and node y during period t; G xy and B xy is the equivalent conductance and susceptance of the line between node x and node y, N nodes is the number of nodes, P x (t) and Q x (t) are the active power and reactive power of node x at time t; The configuration power and capacity constraints of battery energy storage and hydrogen energy storage are: Where, The battery energy storage configuration capacity E is BESS Upper and lower limits and configuration power P BESS Upper and lower limits; They are the upper and lower limits of the water electrolysis power, the upper and lower limits of the fuel cell power, and the upper and lower limits of the hydrogen storage tank capacity; The SOC constraint is: SOC min ≤SOC≤SOC max ; Where, SOC max and SOC min They are the upper and lower limits of the SOC of the battery energy storage; The LOH constraints are: LOH min ≤NO≤NO max 100. In the formula, LOH max and LOH min They are the upper and lower limits of LOH of hydrogen energy storage respectively.

9. The method for optimizing the configuration of long-term and short-term cloud energy storage considering the leasing market according to claim 7, characterized in that: The lower layer model of the long-term and short-term cloud energy storage dual-layer optimization configuration model, which considers the leasing market and typical scenarios, aims to minimize the user's annual assessment cost and includes: The objective function of the lower model of the long-term and short-term cloud energy storage dual-layer optimization configuration model is constructed as follows: Where, F down is the objective function of the lower model, C EXA is the annual assessment cost of users including wind farms, C FB The annual power generation forecast deviation assessment cost for the user, C LEA is the user's annual cloud storage rental cost, N s is the number of typical scenes, π n is the number of days when the nth scenario occurs, T is the total time period in any day, and t is the time period; c EXA (t) is the assessment cost of users with wind farms at time t, c FB (t) is the power generation forecast deviation assessment cost for the user at time t, I LEA,B The user's battery energy storage system rental cost, I LEA,H The cost of leasing hydrogen energy storage systems for users; Construct the constraint functions of the lower model of the long-term and short-term cloud energy storage dual-layer optimization configuration model, including wind power output constraints and wind curtailment constraints; Among them, the wind power output constraints are: Where: P WT (t) is the wind power output during period t, The upper limit of wind power output; The wind curtailment constraint is: Where R WT is the annual wind curtailment rate, is the maximum annual wind curtailment rate.

10. The method for optimizing long-term and short-term cloud energy storage configuration considering the leasing market according to claim 1, characterized in that: The method of solving the single-layer optimization model by using the improved fungus growth optimization algorithm based on the memory mechanism includes: An improved fungal growth optimization algorithm is proposed based on a memory mechanism, which implements local perturbations and supplements in selected dimensions and promotes group search through information sharing between individuals. The improved fungal growth optimization algorithm is used to solve the single-layer optimization model; Among them, the improved fungal growth optimization algorithm is used to solve the single-layer optimization model, including: Input cloud storage system parameters and wind power output data, and configure the parameters, maximum number of iterations, and population size of the improved fungus growth optimization algorithm; Perform memory strategy update on the non-forgotten dimension, and perform forgetfulness supplement update and dream sharing update on the forgotten dimension; The hyphal tip growth stage, hyphal branching stage, and spore germination stage are executed until the convergence condition is met.

11. The method for optimizing long-term and short-term cloud energy storage configuration considering the leasing market according to claim 10, characterized in that: The improved fungal growth optimization algorithm based on the memory mechanism, the implementation of local perturbation and supplementation in the selected dimension, and the promotion of group search through information sharing between individuals include: Initialize the solution of the fungal growth optimization algorithm; in the initialization stage, the individual with the best fitness in the population is used as the initial global memory; By introducing global memory or local memory into individual position updates, it ensures that each individual continuously moves closer to the optimal solution during the search process; In the forgetting supplementation phase, the position of the individual on the selected forgetting dimension set is locally perturbed and updated; the perturbation amplitude is gradually attenuated with the number of iterations, and the perturbation amplitude is gradually reduced by the cosine attenuation factor. The formula for local perturbation update is: Where, For the i ter At the +1th iteration, the τth individual is in The location of the dimension, For the i ter The global optimal solution at the iteration, Υ is the set of forgotten dimensions, For the The perturbation vector of individual τ in dimension; In the memory sharing stage, the information of other individuals in the population is used for supplementary updates to promote knowledge sharing between individuals. The update formula is: Where r is a random number in [0,1], p is the sharing probability threshold, and m is the index of an individual randomly selected from the population; For the i ter At the +1th iteration, the τth individual is in The location of the dimension, For the i ter +1 iteration, the mth individual is in The location of the dimension, For the i ter At the iteration, the mth individual Dimensional location; During the hyphal tip growth stage, the global optimal solution is found by analyzing growth rate and direction, chemotaxis behavior, and controlling the trade-off between exploration and exploitation. During the hyphal branching and spore germination stages, the search space is explored in multiple directions to escape from the local optimal solution, and the dimensions of the individual are updated during iteration in conjunction with the memory sharing update mechanism. In the memory updating stage, the individual with the best fitness is selected as the global optimal solution, and the global memory is updated.

12. The method for optimizing long-term and short-term cloud energy storage configuration considering the leasing market according to claim 11, characterized in that: In the hypha tip growth stage, the global optimal solution is achieved by analyzing the growth rate and direction and the chemotaxis behavior, and controlling the trade-off between exploration and exploitation. The method includes: The growth rate in the fungal growth optimization algorithm is matched with the fitness of the solution, and the fitness value is multiplied by a random number to make the growth rate fluctuate. The growth rate calculation formula after adding fluctuations is: Where F is the adjustment factor of the normalized fitness value after the introduction of random perturbations, ξ is the factor that regulates the exploration and exploitation operators during the optimization process, i ter is the current iteration number, is the maximum number of iterations, r1 is a random number in [0,1], E is a factor used to adjust the exploration and exploitation operators during the optimization process, and f τ is the fitness value of the τth solution, f σ is the fitness value of the σth solution, M is the number of hyphae; The growth direction is determined by randomly selecting two solutions from the current population and calculating the difference between the two solutions. The growth direction vector expression is: Where, is the growth direction vector, and are two different solutions randomly selected from the current population; According to the growth rate and growth direction of the mycelium, a new growth vector is calculated and combined with the current growth vector of the mycelium to generate a new mycelium growth vector. The expression of the update process is: Where, For the The random value generated in the dimension, d is the number of dimensions, r2 is a random number in [0,1], is the new growth vector, For the i ter The growth vector at the iteration; In the first state of chemotaxis analysis, the growth direction is guided to the most nutrient-rich location based on the preset probability mechanism. The mathematical model of the mycelium's growth direction toward the current optimal solution or the randomly selected solution is: Where, is a random number uniformly distributed within the interval [0,1], r3 is a random number within [0,1], ζ is a parameter between 0 and 1, and β is a randomly selected value. is the growth direction of the τth mycelium; at the same time, based on the fitness value, the development operator is associated with the fitness value of each solution. The higher the fitness value, the faster the convergence speed; the mycelium growth update definition expression according to the first state is: Where η τ is the growth step of mycelium under the current development behavior; In the second state of the chemotaxis behavior analysis, the expression of the second state is: Where, is a random number in the interval [0,1], The growth direction of the τth hypha is towards the individual randomly selected from the population Location, For the i ter At the +1th iteration, the τth individual is in The location of the dimension, For the At the same time, by adding the exploration step, the mycelium continues to explore in the search space; For the i ter At the +1th iteration, the τth individual is in Dimensional location; The expression for switching between the first and second states in the chemotaxis behavior assay is: Where r 10 、r 11 、r 12 are random numbers in different intervals [0,1], is the adjustment factor, r 13 、r 14 is a random number in the interval [0,1]; By normalizing the fitness value of each solution, the hyphae in the worst position will try to apply the development operator and move towards the nutrient-rich area; the hyphae in the nutrient-rich area will continue to explore; When the normalized fitness value is less than the defined exploration rate, the exploration operation is performed, otherwise the development operation is performed.

13. The method for optimizing long-term and short-term cloud energy storage configuration considering the leasing market according to claim 11, characterized in that: During the hyphae branching and spore germination stages, the search space is explored in multiple directions to escape from the local optimal solution. The memory sharing update mechanism is used to update the dimensions of the individual during iteration, including: In the first mode of hyphae branching stage, two solutions are randomly selected from the current population, and the growth direction of the new hyphae is determined according to the difference between the two solutions; In the second mode during the hyphae branching phase, the growth direction of new hyphae is determined by the difference between the current optimal solution and a randomly selected solution from the current population; According to the first and second patterns in the hyphae branching stage, the growth direction of new hyphae is comprehensively determined; During the hyphal branching stage, the growth rate of hyphae generated by lateral branches is calculated as follows: Where, E L is the growth rate of hyphae generated by lateral branches, r 16 and r 17 is a random number in different intervals [0,1], M is the number of mycelium, f σ is the fitness value of the σth solution, f τ is the fitness value of the τth solution; Introducing the growth rate into the comprehensively determined growth direction of the new hyphae, the expression for the adjusted growth direction of the new hyphae is: Where r2, r3, r 15 are random numbers in different intervals [0,1], For the Random values ​​generated in dimensions, d is the number of dimensions, is the growth direction of new hyphae in the first mode, is the growth direction of new hyphae in the second mode; For the i ter At the τth iteration, the τth individual The location of the dimension, For the i ter At the +1th iteration, the τth individual is in Dimensional location; In the spore germination stage, new solutions are randomly generated in the search space, and the exploration and development are balanced based on the mean of the current optimal solution and the random solution. The expression of the spore germination process is: Where, X g is a randomly selected value between -1 and 1, r2 and r5 are random numbers in different intervals [0,1], i ter is the current iteration number, is the maximum number of iterations, For the i ter At the τth iteration, the τth individual The location of the dimension, For the i ter At the τth iteration, the τth individual The location of the dimension, For the i ter At the +1th iteration, the τth individual is in Dimensional location; For the The optimal individual of dimension; is randomly selected from the population The position of the dimension, E is the factor used to adjust the exploration and exploitation operators during the optimization process; A random threshold is preset to update hyphae branching and spore germination on some dimensions. For dimensions that meet the conditions, the original position is retained. Otherwise, the dimensions are updated based on the memory sharing update mechanism.

Images