Wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and main kitchen optimization algorithm

Abstract

The invention discloses a wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and a main kitchen optimization algorithm, and relates to the field of comprehensive energy, and the method comprises the steps: obtaining wind-solar output data, and carrying out the clustering processing of the wind-solar output data through employing a dyeing clustering algorithm of an ant colony path-finding mechanism, and obtaining a wind-solar combined output combination set; based on the wind-solar combined output combination set and a pre-obtained electrolytic cell response feature set, an electrolytic cell combination operation mode is constructed, and an electrolytic cell power distribution model is constructed according to the electrolytic cell combination operation mode in combination with the output fluctuation rate and the output coincidence rate; and constructing a capacity configuration multi-target optimization model according to a preset hydrogen production condition in combination with the electrolytic cell power distribution model, and solving the capacity configuration multi-target optimization model by using a multi-target main-kitchen optimization algorithm with a hybrid dominant relationship to obtain a capacity configuration optimal solution. According to the method, the adaptability to complex resource distribution and engineering constraints is remarkably improved through dyeing clustering.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of comprehensive energy, in particular to a wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm. BACKGROUND

[0002] With the transformation of global energy structure to low carbonization, renewable energy hydrogen production technology as an important path to realize large-scale production of green hydrogen has become a research hotspot in the field of energy; wind energy and solar energy have significant spatio-temporal volatility and intermittency characteristics, and their output characteristics are coupled with the dynamic response demand of water electrolysis hydrogen production system (such as alkaline electrolyzer AEL and proton exchange membrane electrolyzer PEMEL), which leads to low efficiency, short equipment life and insufficient economy of hydrogen production system. Therefore, how to coordinate the volatility of wind-solar resources and the operation characteristics of electrolyzer through scientific and reasonable capacity configuration strategy has become a key technical challenge to improve the economy and reliability of green hydrogen system.

[0003] In the modeling of wind-solar joint output scene, traditional clustering methods (such as K-means and fuzzy C-means) depend on preset cluster number and initial center, which is difficult to adapt to complex and variable resource distribution and engineering constraints, and is easy to lead to scene division deviation, which further affects the accuracy of subsequent capacity optimization results. At the same time, existing researches focus on single electrolyzer operation mode, lack of in-depth exploration of multi-electrolyzer collaborative operation mechanism, especially the flexible combination strategy for the difference of AEL and PEMEL dynamic response (such as start-up time and load adjustment range), which limits the efficient response ability of the system to wind-solar fluctuation. In addition, the power distribution method of hydrogen production system is mostly based on static rules or single objective optimization, which fails to comprehensively consider the dynamic balance between output fluctuation suppression, equipment simultaneous rate optimization and operation cost, resulting in low renewable energy utilization rate and intensified equipment wear.

[0004] In the aspect of capacity configuration model construction and solution, existing researches usually take hydrogen production cost or renewable energy utilization rate as a single optimization target, which is difficult to realize multi-objective collaborative optimization. Traditional multi-objective optimization algorithms (such as NSGA-II) are easy to fall into local optimum when dealing with high-dimensional nonlinear constraints, and the solving efficiency is difficult to meet the demand of large-scale system planning. In addition, existing methods mostly use pure data-driven or mechanism model independent analysis, which fails to effectively integrate wind-solar output statistical characteristics and dynamic operation mechanism of hydrogen production system, resulting in insufficient robustness of configuration scheme under complex fluctuation scene.

[0005] In summary, the current wind and light hydrogen production system still has significant technical bottlenecks in scene generation, electrolytic cell operation mode design, power distribution strategy and capacity optimization method, etc. It is urgent to break through the parameter dependence limitation of traditional clustering algorithm, to construct a flexible cooperative operation mechanism of multiple electrolytic cells, to develop a power distribution method considering fluctuation suppression and equipment efficiency, and to establish a mechanism and data hybrid driven multi-objective optimization framework, in order to realize the comprehensive improvement of the economy, reliability and renewable energy consumption capacity of green hydrogen system.

[0006] At present, there is no effective solution to the problems in the related art. SUMMARY

[0007] In view of the problems in the related art, the present application proposes a wind and light hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm to overcome the above technical problems existing in the prior art.

[0008] To this end, the specific technical solutions adopted by the present application are as follows:

[0009] A wind and light hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm, the method comprising the following steps:

[0010] S1, obtaining wind and light output data, and using the dyeing clustering algorithm of the ant colony path finding mechanism to cluster the wind and light output data to obtain a wind and light combined output combination set;

[0011] S2, based on the wind and light combined output combination set and the pre-acquired electrolytic cell response feature set, constructing an electrolytic cell combined operation mode, and according to the electrolytic cell combined operation mode, constructing an electrolytic cell power distribution model combined with output fluctuation rate and output simultaneous rate;

[0012] S3, constructing a capacity configuration multi-objective optimization model according to the pre-set hydrogen production conditions combined with the electrolytic cell power distribution model, and using a multi-objective chef optimization algorithm with mixed advantage relationship to solve the capacity configuration multi-objective optimization model to obtain an optimal solution of capacity configuration.

[0013] Further, in order to avoid the dependence on pre-set cluster number and initial center, similar data points can be automatically guided to gather under the clustering effect of the dyeing clustering algorithm of the ant colony path finding mechanism, avoiding the dependence on prior parameters such as cluster number, obtaining wind and light output data, and using the dyeing clustering algorithm of the ant colony path finding mechanism to cluster the wind and light output data to obtain a wind and light combined output combination set, comprising:

[0014] S11, obtaining wind and light output data, and using the wind and light output data and a pre-set ant colony base to construct an ant colony generation model to generate the number of ants through the ant colony generation model;

[0015] S12, constructing a neighborhood matrix based on the wind and light output data, calculating the nearest neighbor density by using the neighborhood matrix, constructing a path searching model by using the reliability score and the inverse S-shaped function according to the nearest neighbor density, and performing ant colony path selection based on the path searching model;

[0016] S13, generating an ant colony termination mechanism through a preset rule, constructing a path searching termination model according to the ant colony termination mechanism, and performing ant colony path searching termination by using the path searching termination model;

[0017] S14, performing dye clustering on the ant colony path when the ant colony path searching is terminated, and obtaining a wind and light combined output combination set.

[0018] Further, in order to provide basic support for the path and decision, the spatial correlation between data points is described under the construction of the neighborhood matrix, and the density relationship is effectively represented and the ant colony is helped to select the path. The neighborhood matrix is constructed based on the wind and light output data, the nearest neighbor density is calculated by using the neighborhood matrix, the path searching model is constructed by using the reliability score and the inverse S-shaped function according to the nearest neighbor density, and the ant colony path selection is performed based on the path searching model, which includes:

[0019] S121, constructing a neighborhood matrix by using the wind and light output data according to the preset forward neighborhood and reverse neighborhood, and calculating the nearest neighbor density based on the neighborhood matrix;

[0020] S122, calculating the path reliability score based on the reliability score and the inverse S-shaped function in combination with the nearest neighbor density and the pheromone obtained in advance, and updating the pheromone by using the path reliability score;

[0021] S123, performing probability distribution by using roulette according to the path reliability score, generating a path searching model based on the probability distribution result, and performing ant colony path selection by using the path searching model.

[0022] Further, in order to prevent invalid extension of the path, the ant colony cannot search endlessly under the termination of the ant colony termination mechanism, and the path searching is terminated once sufficient data information is collected. The ant colony termination mechanism is generated through a preset rule, the path searching termination model is constructed according to the ant colony termination mechanism, and the ant colony path searching termination is performed by using the path searching termination model, which includes:

[0023] S131, generating an ant colony global cross termination mechanism by using Gaussian distribution based on a preset neighborhood point distance threshold;

[0024] S132, generating an ant colony local dilemma termination mechanism according to a preset access frequency threshold in combination with a preset access rule;

[0025] S133, constructing a path searching termination model by using the ant colony global cross termination mechanism and the ant colony local dilemma termination mechanism, and performing ant colony path searching termination by using the path searching termination model.

[0026] Further, in order to obtain the wind and light combined output combination set, ant colony path coloring can be realized under the action of dyeing marking, thereby automatically guiding similar data points to gather, and then obtaining the wind and light combined output combination set. The ant colony path at the termination of the ant colony search is colored and clustered to obtain the wind and light combined output combination set, including:

[0027] S141, dyeing marking is performed on the ant colony path at the termination of the ant colony search based on a preset dyeing condition, and a dyeing marking set is obtained;

[0028] S142, dyeing covering is performed on the dyeing marking set by using a majority voting mechanism, and the ant colony path is colored and clustered based on the dyeing covering result to obtain the wind and light combined output combination set.

[0029] Further, in order to realize reasonable power distribution, more electrolytic cells can be preferentially ensured to be in a rated or efficient state under the action of a preset operation switching rule, the length of unstable operation is reduced, the service life of the equipment is prolonged, and an electrolytic cell combination operation mode is constructed based on the wind and light combined output combination set and a pre-acquired electrolytic cell response feature set. An electrolytic cell power distribution model is constructed according to the electrolytic cell combination operation mode in combination with output fluctuation rate and output simultaneous rate, including:

[0030] S21, the operation state of the electrolytic cell is divided by using the wind and light combined output combination set and the pre-acquired electrolytic cell response feature set, and the electrolytic cell combination operation mode is constructed according to the division result in combination with the preset operation switching rule;

[0031] S22, the average fluctuation rate of power generation output and the output simultaneous rate are calculated based on the electrolytic cell combination operation mode, and the electrolytic cell power distribution model is constructed through the average fluctuation rate of power generation output and the output simultaneous rate.

[0032] Further, in order to take into account the response capability and cost advantage of the electrolytic cell, the capacity utilization rate can be improved while reducing the output fluctuation and shutdown risk under the reasonable distribution of the electrolytic cell power distribution model, and the electrolytic cell power distribution model includes:

[0033] When the total power demand of the electrolytic cell is equal to the sum of the rated power of all electrolytic cells, all electrolytic cells are operated at rated power, and the output simultaneous rate is improved;

[0034] When the total power demand of the electrolytic cell is lower than the rated power of all electrolytic cells, a specific alkaline electrolytic cell is preferentially started, and the average fluctuation rate of power generation output is reduced;

[0035] When the total power demand of the electrolytic cell is between the rated power of all electrolytic cells, the power distribution ratio of the proton exchange membrane electrolytic cell and the alkaline electrolytic cell is dynamically adjusted according to the real-time fluctuation rate.

[0036] Further, in order to realize multi-objective collaborative optimization, the capacity configuration multi-objective optimization model can be constructed to realize the multi-objective collaborative optimization of low hydrogen production cost, high hydrogen production efficiency and sustainable hydrogen production. The capacity configuration multi-objective optimization model is constructed according to the preset hydrogen production condition combined with the electrolytic cell power distribution model, and the multi-objective chef optimization algorithm with mixed dominance relationship is used to solve the capacity configuration multi-objective optimization model to obtain the capacity configuration optimal solution, including:

[0037] S31, based on the electrolytic cell power distribution model, the hydrogen production efficiency is obtained, and the hydrogen production efficiency is used to construct the capacity configuration multi-objective optimization model of the wind-solar coupled hydrogen production system combined with the preset hydrogen production cost;

[0038] S32, according to the preset hydrogen production parameter, the hydrogen production equipment constraint and the hydrogen production operation constraint are generated, and the constraint condition of the capacity configuration multi-objective optimization model is constructed through the hydrogen production equipment constraint and the hydrogen production operation constraint;

[0039] S33, the multi-objective chef optimization algorithm with mixed dominance relationship is used to solve the capacity configuration multi-objective optimization model to obtain the capacity configuration optimal solution.

[0040] Further, in order to effectively fuse the wind-solar output statistical characteristics and the dynamic operation mechanism of the hydrogen production system, the robustness of the configuration method in complex fluctuation scenarios can be further improved under the dual constraints of hydrogen production equipment constraints and hydrogen production operation constraints. The hydrogen production equipment constraints include: electric energy storage constraint, thermal energy storage constraint and hydrogen storage balance constraint;

[0041] The hydrogen production operation constraints include: electric load constraint, thermal load constraint and hydrogen load constraint.

[0042] Further, in order to obtain the capacity configuration optimal solution set, the operation fluctuation and wind and light abandonment phenomenon can be effectively reduced, and the renewable energy utilization rate and the overall performance of the hydrogen production system can be improved under the optimization of the multi-objective chef optimization algorithm. The multi-objective chef optimization algorithm with mixed dominance relationship is used to solve the capacity configuration multi-objective optimization model to obtain the capacity configuration optimal solution, including:

[0043] S331, initialize the population size, mixed dominance threshold and maximum iteration number of the chef instructor and apprentice subpopulation in the multi-objective chef optimization algorithm, and calculate the objective function value of each individual through the capacity configuration multi-objective optimization model;

[0044] S332, use fast non-dominated sorting and crowding distance to select individuals according to the objective function value of each individual to obtain the initialization archive set;

[0045] S333, based on the master-slave strategy, the population is constantly updated using the initialization archive set, and the archive set is iteratively updated combined with the selected dominance relationship to obtain the capacity configuration optimal solution set.

[0046] The beneficial effects of the present application are:

[0047] 1、The ant colony path mechanism dyeing clustering algorithm can dynamically adjust the number of ants and spatial density measurement, effectively guide the ants to distribute uniformly in the data set, avoid the problem of easily falling into local optimum in the traditional clustering algorithm, accurately depict the relationship between points through forward and reverse neighborhood matrix, provide scientific basis for path selection combined with nearest neighbor density, improve the rationality of the path and the accuracy of clustering, and introduce global crossover and local dilemma termination mechanism to effectively prevent path invalid extension and local trap, ensure the convergence and stability of the algorithm, finally realize dynamic expansion of clustering labels by using path dyeing strategy combined with majority voting mechanism, so that the clustering result has good continuity and robustness, and the ant colony path mechanism dyeing clustering algorithm is applied to scene clustering, which is more robust and continuous compared with common clustering algorithms.

[0048] 2、The power distribution model of electrolytic cells is constructed by combining output fluctuation rate and output simultaneous rate to realize the power distribution method of multiple electrolytic cell combined operation, the efficient and stable operation of the system is realized by dynamically coordinating the operation of AEL and PEMEL electrolytic cells, the low-cost advantage of AEL and the rapid response ability of PEMEL are fully utilized, the output fluctuation is effectively suppressed, the output simultaneous rate and the installed capacity utilization rate are improved, the operation cost and the shutdown risk are reduced, the adaptability of the hydrogen production system to power demand fluctuation is enhanced, and the overall economy and reliability are significantly improved.

[0049] 3、The multi-objective chef optimization algorithm comprehensively considers the hydrogen production cost, fluctuation rate and renewable energy utilization rate, and provides an efficient system capacity planning scheme, which effectively reduces the system operation fluctuation and wind and light curtailment phenomenon, improves the renewable energy utilization rate and the overall performance of the hydrogen production system. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Figure 1 is a flow chart of a wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm according to an embodiment of the present application;

[0052] Figure 2is a dyeing clustering algorithm flow chart of an ant colony path finding mechanism in a wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm according to an embodiment of the application. DETAILED DESCRIPTION

[0053] To further illustrate the embodiments, the present application provides drawings which are part of the disclosure of the present application and mainly serve to illustrate the embodiments and can be used to explain the operating principle of the embodiments in conjunction with the relevant description of the specification. Those skilled in the art should be able to understand other possible implementations and advantages of the present application by referring to these contents.

[0054] According to an embodiment of the present application, a wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm is provided.

[0055] The present application will be further described in conjunction with the drawings and specific embodiments. As shown in the drawings, the wind-solar hydrogen production system capacity configuration method based on dynamic ant colony clustering and chef optimization algorithm according to an embodiment of the present application includes the following steps: Figure 1

[0056] S1, obtaining wind-solar output data and performing clustering processing on the wind-solar output data by using a dyeing clustering algorithm of an ant colony path finding mechanism to obtain a wind-solar combined output combination set.

[0057] Specifically, obtaining wind-solar output data and performing clustering processing on the wind-solar output data by using a dyeing clustering algorithm of an ant colony path finding mechanism to obtain a wind-solar combined output combination set includes:

[0058] S11, obtaining wind-solar output data and constructing an ant colony generation model by using the wind-solar output data and a preset ant colony base number, and generating the number of ants through the ant colony generation model;

[0059] S12, constructing a neighborhood matrix based on the wind-solar output data, calculating the nearest neighbor density by using the neighborhood matrix, constructing a path finding model by using the reliability score and the inverse S-shaped function according to the nearest neighbor density, and performing ant colony path selection based on the path finding model.

[0060] Specifically, constructing a neighborhood matrix based on the wind-solar output data, calculating the nearest neighbor density by using the neighborhood matrix, constructing a path finding model by using the reliability score and the inverse S-shaped function according to the nearest neighbor density, and performing ant colony path selection based on the path finding model includes:

[0061] S121, constructing a neighborhood matrix by using the wind-solar output data according to the preset forward neighborhood and inverse neighborhood, and calculating the nearest neighbor density based on the neighborhood matrix;

[0062] ​S122, calculate the path reliability score based on the reliability score, the reverse S-shaped function, the nearest neighbor density and the pre-acquired pheromone, and update the pheromone using the path reliability score;

[0063] S123, perform probability distribution using roulette based on the path reliability score, generate a path finding model based on the probability distribution result, and perform ant colony path selection using the path finding model; the spatial correlation between data points can be described under the construction of the neighborhood matrix, and the density relationship is effectively represented and the ant colony is helped to select the path.

[0064] S13, generate an ant colony termination mechanism by a preset rule, and construct a path finding termination model according to the ant colony termination mechanism, and perform ant colony path finding termination using the path finding termination model.

[0065] Specifically, generating an ant colony termination mechanism by a preset rule, and constructing a path finding termination model according to the ant colony termination mechanism, and performing ant colony path finding termination using the path finding termination model includes:

[0066] S131, generate an ant colony global cross termination mechanism using Gaussian distribution based on a preset neighbor point distance threshold;

[0067] S132, generate an ant colony local dilemma termination mechanism according to a preset access frequency threshold and in combination with a preset access rule;

[0068] S133, construct a path finding termination model by the ant colony global cross termination mechanism and the ant colony local dilemma termination mechanism, and perform ant colony path finding termination using the path finding termination model; the ant colony can be prevented from searching endlessly under the termination of the ant colony termination mechanism, and the path finding is terminated once sufficient data information is collected.

[0069] S14, perform dye clustering on the ant colony path at the time of ant colony path finding termination to obtain a wind and light joint output combination set; similar data points can be automatically guided to gather under the clustering of the dye clustering algorithm of the ant colony path finding mechanism, and the dependence on prior parameters such as cluster number is avoided.

[0070] Specifically, performing dye clustering on the ant colony path at the time of ant colony path finding termination to obtain a wind and light joint output combination set includes:

[0071] S141, dye mark the ant colony path at the time of ant colony path finding termination based on a preset dye condition to obtain a dye mark set;

[0072] S142, perform dye coverage using a majority voting mechanism according to the dye mark set, perform dye clustering on the ant colony path based on the dye coverage result, and obtain a wind and light joint output combination set; the ant colony path can be dyed under the action of the dye mark, thereby automatically guiding similar data points to gather, and further obtaining the wind and light joint output combination set.

[0073] Specifically, the dyeing clustering algorithm based on ant colony routing mechanism is adopted to process the wind power and photovoltaic output data (i.e. wind and light output data), so as to obtain a typical scene set of wind and light combined output (i.e. a wind and light combined output combination set). The characteristics of pheromone guided path selection in the foraging process of ants in nature are introduced into the clustering process, and the clustering attribution of data points is represented by dyeing conditions; the pheromone concentration and color propagation mechanism are dynamically adjusted in the clustering process, so as to automatically guide similar data points to gather and avoid the dependence on prior parameters such as cluster number.

[0074] Specifically, the initial number of ants is generated according to the size m of the data set and the set ant colony base β (the value range is 0 to 0.5). By reasonably setting the size of the ant colony, it is ensured that the algorithm has sufficient individual distribution in different density areas. The routing model guides the path selection of the ant colony in the data set by constructing multiple spatial density relationship measurement methods. First, the forward and reverse neighborhoods are defined to represent the nearest neighbor relationship between points, and then the corresponding neighborhood matrix is constructed to represent the global relationship between points. The "nearest neighbor density" is introduced as a means to characterize local spatial density, which provides a basis for path preference for ants. In the decision-making process, the ant colony refers to the pheromone value, density information and relationship measurement group, determines the candidate target through reliability score, and adjusts the path repetition phenomenon combined with the reverse S-shaped function. Finally, the roulette strategy is adopted to realize the probabilistic jump.

[0075] To prevent invalid extension of the path, the model sets two termination mechanisms: global intersection and local dilemma. Global intersection refers to the case where the distance between adjacent points in the path of the ant colony is abnormally large, indicating that it has crossed the reasonable density range; the dynamic threshold is set by Gaussian distribution to realize the judgment. Local dilemma monitors the frequency of repeated access of ants to a certain point in the path, and if the frequency exceeds the specified number, it is determined that the ant has fallen into a local density trap, and the path is terminated.

[0076] When the ant completes the routing, the points covered by its path will participate in the dyeing clustering process. Initially, all data points are not dyed (the label is -1), and when the ant path is terminated, the clustering label is marked according to the dyeing condition in the path. If all points in the path are not dyed, a new cluster is generated; if it contains dyed points, the most frequent label is selected as the new cluster label through the majority voting mechanism, so as to realize cluster expansion.

[0077] Specifically, the ant colony routing mechanism-based dyeing clustering algorithm includes four models, specifically, an ant colony generation model, a routing model, a routing termination model, and a dyeing clustering. First, a certain number of ants are generated. Then, routing is performed according to the serial number thereof. The routing model includes decision-making of the ant colony and updating of pheromone, and data density correlation is represented by pheromone. Next, two basic conditions are set for the ant colony to stop routing. When the conditions are triggered, the ant colony stops spatial routing. Finally, the dyeing clustering algorithm is implemented with the support of the routing data, and a clustering structure is output. The above models are executed in sequence, and an ordered closed loop is constructed. The clustering labels of the data set are updated each time the iteration is performed until the ant colony is completed.

[0078] For the ant colony generation model, according to a given wind power output data set, a certain number of ants are generated by using the ant colony generation model. The calculation formula of the number of generated ants is as follows:

[0079]

[0080] In the formula, ant n represents the number of ants; represents the ant colony base, which is usually set in the interval (0, 0.5); m represents the size of the data set.

[0081] The ant colony routing mechanism-based dyeing clustering algorithm divides clustering into several rounds, ant n represents the number of ants in each round. If the number of rounds in which the algorithm is executed is r, then a total of N ants are required to be executed in sequence. The calculation formula of the total number of ants of the ant colony is as follows:

[0082] N = ant n · r;

[0083] In the formula, N represents the total number of ants of the ant colony; and r represents the number of rounds in which the ant colony routing mechanism-based dyeing clustering algorithm needs to be executed.

[0084] For the ant colony routing model, several basic measurement models are first proposed. These measurement models describe the spatial correlation between points, effectively represent the density relationship, and provide basic support for path and decision-making.

[0085] A preset forward neighborhood and a reverse neighborhood are defined, specifically, a forward neighborhood is defined, that is, taking point p as the center, point p' is the nearest neighbor serial number starting from point p, denoted as Δ(p, p'); and a reverse neighborhood is defined, that is, taking point p as the center and taking point p' as a reference point, point p is the nearest neighbor serial number starting from point p', denoted as Δ'(p, p').

[0086] Δ(p, p') = {p' | p' is the nearest neighbor of p}; The relation measure group is used to describe the spatial density relation between points. Unlike traditional density calculation, the relation measure group does not need quantitative methods such as distance, but only needs to evaluate the density through the nearest neighbor serial number, effectively simplifying the correlation calculation.

[0087] The forward (reverse) neighborhood matrix describes the forward (reverse) neighborhood relation between any two points in the point set, and the matrix proportion is m x m.

[0088] Taking the forward neighborhood matrix as an example, the matrix is as follows:

[0089]

[0090] Taking the reverse neighborhood matrix as an example, the matrix is as follows:

[0091]

[0092] The neighborhood matrix applies the concept of the relation measure group to the entire data set to represent the data correlation. Accordingly, it has two attributes: the forward neighborhood matrix and the reverse neighborhood matrix are both asymmetric matrices; the forward neighborhood matrix and the reverse neighborhood matrix are both transpose matrices. The two attributes prove the inherent characteristics of the two types of matrices. Due to the difference in the main angle of calculation, the two types of matrices are asymmetric matrices. In addition, due to the exchange of the center point, the two types of matrices are transpose matrices.

[0093] The nearest neighbor density is the average value of the ratio of the relation measure group formed by p and the nearest neighbor p' with p as the center and k as the nearest neighbor parameter. According to the neighborhood matrix, the number of forward neighborhoods Δ and reverse neighborhoods ▽ can be effectively obtained. The expression of the nearest neighbor density is as follows:

[0094]

[0095] In the formula, NNDp represents the nearest neighbor density; p i represents the ith adjacent data point.

[0096] The nearest neighbor density combines the concept of the relation measure group to describe the density correlation of points, and can effectively reflect the correlation information of points.

[0097] The pheromone constitutes the selection basis of the ant colony in the path finding process, and represents the preference of the ant colony in the path finding problem; the pheromone is the attribute of the point; in the initial stage of the algorithm, the pheromone of the data point is initialized to 1; only when the ant reaches this point, the pheromone is updated; the pheromone update formula is as follows:

[0098]

[0099] In the formula, is the updated pheromone; is the original pheromone value; PRS(p i ) is the reliability score of the path of the ant to this point. The higher the reliability, the stronger the pheromone laid by the ant, which will also attract more ants in the future.

[0100] The ant colony decision is the result of the current ant selecting a feasible point target within a certain range (Φ) as the next path. The decision is affected by multiple factors to establish a reasonable spatial path. The basic mode of ant colony decision is to regard each data object within the Φ range as a potential candidate point and calculate the path reliability score of each point. The reliability value consists of two parts. First, the influencing factors of reliability calculation, such as pheromone, relationship measurement group, etc., these factors have a guiding effect on the high-density area of the ant colony, making it easier to reach high-density points. Second, due to the attractiveness problem of high-density areas, the ant colony is prone to repeatedly visit two or more data points, resulting in a large deviation in pheromone between a single data point and other points. To avoid this problem, a reverse S-shaped access control factor is introduced.

[0101]

[0102] where PRS is the path reliability score, which is the specific value of the path reliability of points p and p'; PRS ||p,p'|| is the path reliability relationship of points p and p'; the reliability score of p' is composed of a reliability influencing factor (factor 1) and an access control factor (factor 2); PathReliability represents the reliability influencing factor; ReverseSigmoid represents the reverse S-shaped access control factor; phe p′ represents the pheromone value of point p'; d represents the distance; nor represents the normalization process; fo represents the frequency of access; where the reliability influencing factor is affected by the basic measurement factors, including the pheromone of p', the relationship measurement group, the distance, and the nearest neighbor density value of p', which collectively guide the ant colony to develop towards high-density space.

[0103] In addition, during the decision-making process, the ant may repeatedly select two data points. This makes the pheromone of a single data point very strong. To avoid this phenomenon, factor 2 is introduced to adjust the path reliability score between the two points appropriately, which can appropriately reduce the path reliability. By proposing a reverse S-shaped function, the value range is constrained within the [0, 1] interval, and its shape is opposite to that of the S-shaped function.

[0104] After the current ant obtains the reliability scores of all neighbors, it will make a final decision. Ideally, all candidate points should have a chance to be selected by this ant, not just the highest scoring point. Therefore, the algorithm is based on the probability of point number allocation and uses roulette to make the final decision. The calculation formula of the roulette probability is:

[0105]

[0106] where RNA(0,1) represents the roulette probability.

[0107] The score of each object is related to the probability of being selected. The greater the reliability score of p', the greater the probability of roulette, i.e. the greater the probability of being selected by the ant. At the same time, however, the ant can also select a point with a lower probability, which will cause the ant colony to move in a sparse direction. This reverse route, i.e. the phenomenon contrary to the original algorithm's expectation, is not an error, but a mechanism to prompt the ant colony to jump out of the current high-density range. Finally, when the ant confirms the next target on the path, it will move to the point and update the pheromone value of the point.

[0108] For the termination model of the ant colony path search, the ant colony cannot search endlessly, and can end its operation once sufficient data information is collected. The path search termination model is established so that the ant can check whether the condition is triggered to end the path search each time it reaches a new data object in the path search process. The model contains two sub-conditions, and the triggering mechanism is composed of an OR relationship. Condition 1 is global crossing. The motivation of this case is to prevent the ant colony from crossing a large space, causing two points far apart to be connected together. When any two points on the path are far apart, they have a low probability of belonging to the same cluster. If the ant connects them, it is easy to cause color pollution in the clustering step. Condition 2 is local dilemma. The motivation of this condition is to avoid excessive selection of the same point. At this time, whether the path rotation phenomenon occurs or not, as long as the threshold is exceeded, the algorithm considers that the local information collection of the ant has reached saturation, and immediately terminates the operation to save resources. In summary, the algorithm expects the ant to run in a certain area without the need to repeatedly visit any point.

[0109] For the global crossing problem, if the distance between two adjacent points on the ant path is too far, it means that the global crossing problem is caused, and this condition is triggered at this time. Therefore, it is necessary to limit the crawling of the ant within a certain density range. When the ant moves from p i to p j , if the following formula is satisfied, the condition is triggered. The condition formula of the global crossing problem is:

[0110]

[0111] where ant μ represents the range of neighbors selected by the ant, which is set as a variable and subject to the variation trend of Gaussian distribution; μ represents the median in the nearest neighbor curve; σ represents the parameter of the normal distribution curve; ant num represents the number of neighbors of the ant; lim upperIndicates the upper limit; lim lower Indicates the upper limit.

[0112] When combined with dye clustering, the pathfinding range of the ant colony can be divided into the following two parts: one is to stimulate external expansion. Within this range, as the number of ants in the ant colony increases, the pathfinding range continues to expand, radiating to a wider data space for path exploration and information collection. Subsequently, the algorithm obtains non-local information and performs color clustering; the second is to repair local errors. Within this range, as the number of ants in the ant colony increases, the pathfinding range continues to shrink to correct the misjudgments in the early large range until the dye clustering stops oscillating and gradually converges.

[0113] The local dilemma is the second control condition, which is used to limit the ant colony from crawling repeatedly in a local area. The algorithm only needs to record the path of the ant colony and check whether the repetition rate of the point is higher than the threshold ν. Suppose the ant moves to step j If any point p in the path appears more than ν times, the algorithm considers that the ant colony is in a local dilemma and the ants stop searching for the path. The expression of the ant colony path is:

[0114]

[0115] Where, Ant i A list of paths containing most of the repeatable data points, represents the sequence number of the ant at step j and the cluster label of its corresponding point id. The conditional expression of the local dilemma is:

[0116] count(p id )>v;

[0117] Function count counts the same points p id ,If the threshold is exceeded, the algorithm defaults to having ,enough repetitions in the local region and stops iterating.

[0118] The coloring clustering algorithm clusters by coloring. At the beginning of the algorithm, all points are marked with a color of -1, which means they are not colored. When an ant dies, the data point is colored according to the information left by the ant. During the path finding process, the model guides the ants to detect local points, so the clustering module tends to The data in are grouped into the same cluster.

[0119] In the ant colony path formula, for Any one of If S represents a single category and M represents multiple categories, then its distribution has the following three cases:

[0120] 1. All points on the path are unvisited.

[0121] 2、 The path contains unvisited points and visited points, where the visited points come from one or more clusters.

[0122] 3、 All points on the path are visited, or they originate from one cluster or belong to multiple clusters.

[0123] In the dyeing clustering, voting is used to distinguish the cluster to which the mixed points in the path belong. That is, the color with the highest frequency covers the other colors, and the covering behavior is the expansion process of the cluster in the cluster. The expression of covering is:

[0124]

[0125] In the formula, freq label represents the frequency of the highest frequency label; others represents other labels; label represents the covering label. The principle of cluster expansion is given in the formula; if all points in the crawling path are not dyed, that is, their labels are -1, then all points in are assigned to a new cluster, and the label is increased by 1. Otherwise, the label with the highest frequency is selected to cover the other labels, realizing the expansion of the cluster.

[0126] As shown in Figure 2 , first generate the number of ants related to the ant colony base, and then start serial traversal. Each ant respectively performs path finding and decision checking whether the predetermined stop condition is reached. If so, the loop terminates; if not, the current ant starts foraging, and the related variables are initialized. Subsequently, the ant decides to visit the next point to be reached. If the ant does not die in the path finding process, it updates the pheromone and other related information, and continues to select the next destination. If the ant dies, dyeing clustering occurs, and after processing the current ant, the cluster information is updated. Finally, the algorithm continues to iterate until the traversal of the ants ends.

[0127] S2, based on the wind-solar combined output combination set and the pre-acquired electrolytic cell response feature set, construct an electrolytic cell combination operation mode, and construct an electrolytic cell power distribution model according to the electrolytic cell combination operation mode combined with the output fluctuation rate and the output simultaneous rate.

[0128] Specifically, based on the wind-solar combined output combination set and the pre-acquired electrolytic cell response feature set, construct an electrolytic cell combination operation mode, and construct an electrolytic cell power distribution model according to the electrolytic cell combination operation mode combined with the output fluctuation rate and the output simultaneous rate.

[0129] S21, dividing the running state of the electrolyzer by using the wind and light combined output combination set and the pre-acquired electrolyzer response characteristic set, and constructing an electrolyzer combined running mode according to the division result and a pre-set running switching rule;

[0130] S22, calculating the average fluctuation rate of power generation output and the simultaneous rate of output based on the electrolyzer combined running mode, and constructing an electrolyzer power distribution model through the average fluctuation rate of power generation output and the simultaneous rate of output; under the action of the pre-set running switching rule, more electrolyzers can be preferentially ensured to be in the rated or efficient state, the unstable running time is reduced, and the equipment life is prolonged.

[0131] Specifically, the electrolyzer power distribution model comprises:

[0132] When the total power demand of the electrolyzer is equal to the sum of the rated power of all electrolyzers, all electrolyzers run at the rated power, and the simultaneous rate of output is improved;

[0133] When the total power demand of the electrolyzer is lower than the rated power of all electrolyzers, a specific alkaline electrolyzer is preferentially started, and the average fluctuation rate of power generation output is reduced;

[0134] When the total power demand of the electrolyzer is between the rated power of all electrolyzers, the power distribution ratio of the proton exchange membrane electrolyzer and the alkaline electrolyzer is dynamically adjusted according to the real-time fluctuation rate; under the reasonable distribution of the electrolyzer power distribution model, the capacity utilization rate is improved, and the output fluctuation and shutdown risk are reduced.

[0135] Specifically, for the dynamic response characteristics (i.e., the pre-acquired electrolyzer response characteristic set) of AEL-PEMEL (alkaline electrolyzer-proton exchange membrane electrolyzer), a plurality of flexible modes of electrolyzer combined running (i.e., the electrolyzer combined running mode) are proposed, and a power distribution method considering the output fluctuation and the simultaneous rate (i.e., the electrolyzer power distribution model) is proposed.

[0136] The alkaline electrolyzer (AEL) has low cost but poor dynamic response, and the proton exchange membrane electrolyzer (PEMEL) has high adaptability but high cost. In order to coordinate the advantages and disadvantages of the two, the system divides the running state of the electrolyzer into five categories: rated, efficient, unstable, standby and shutdown, of which the first three states can produce hydrogen. Unstable running has a long-term impact on the life, so the switching strategy needs to be optimized. The standby state maintains the temperature and pressure, which is convenient for rapid start; the shutdown state needs to meet the start or shutdown constraints. Through reasonable power distribution, more electrolyzers can be preferentially ensured to be in the rated or efficient state, the unstable running time is reduced, the equipment life is prolonged, and the hydrogen production efficiency is improved. The electric power of the i-th electrolyzer and the electrolyzer array at time t is as follows:

[0137]

[0138] In the formula, represents the electric power of the i-th electrolytic cell at time t; represents the standby power of the i-th electrolyzer; represents the power of the i-th electrolyzer with the maximum hydrogen production efficiency; represents the rated power of the i-th electrolytic cell; represents the power of the i-th electrolytic cell in an unstable operating state; P t Array represents the electrical power of the electrolyzer array; Indicates the rated operating state of the electrolytic cell. When the i-th electrolytic cell is in the rated operating state at time t otherwise Indicates the efficient operation state of the electrolyzer; Indicates the unstable operating state of the electrolytic cell; Indicates the standby state of the electrolytic cell; N EL Represents the total amount of a single electrolyzer in the electrolyzer array.

[0139] The average fluctuation rate of power generation output during period t is calculated as:

[0140]

[0141] Where, γ ave,t represents the average fluctuation rate of power generation output; T represents the total number of time periods; P t max Indicates the maximum power during the period t; P t min Indicates the minimum power value during the period t; P WP Indicates the rated power of the wind and solar power station.

[0142] Output simultaneous rate S OS Calculated as:

[0143]

[0144] Where, P Σ Indicates the total output of wind and solar power; E AEL Indicates the rated power of AEL; E PEMEL Indicates the rated power of PEMEL.

[0145] When the total power demand is equal to the sum of the rated power of both, they are operated at rated power, improving the power output while the rate; when the total power demand is low, the small-capacity AEL is started first, reducing its frequent adjustment and reducing the fluctuation rate; when the demand is between the two, the power distribution is dynamically adjusted according to the real-time fluctuation rate, and when the fluctuation is large, the PEMEL bears more, and when the fluctuation is small, the fixed proportion is restored, ensuring the efficient operation of AEL. The overall strategy takes into account the response ability of PEMEL and the cost advantage of AEL, improving the capacity utilization rate while reducing system fluctuation and production risk.

[0146] For the flexible mode of multiple electrolytic cell combination operation, the electrolytic cell array can convert wind and light into hydrogen energy to meet the demand of green hydrogen production and renewable energy adaptation. There is no big difference in hydrogen production between AEL and PEMEL electrolytic cells. AEL is low in cost, but slow in dynamic response. When the renewable energy power generation fluctuates, PEMEL has good adaptability to the power fluctuation of renewable energy, but the construction cost is high, which reduces the income. In order to improve the utilization ability of renewable energy and the cost-effectiveness of hydrogen production, a flexible mode of multiple electrolytic cell combined operation is proposed.

[0147] The running state of electrolytic cell can be divided into five types, namely rated operation, efficient operation, unstable operation, standby and shutdown, as follows:

[0148]

[0149] In the formula, represents the safe running power of the i-th electrolytic cell.

[0150] First, in the rated operation, efficient operation and unstable operation state, the electrolytic cell can produce hydrogen. The unstable running state refers to the dynamic response of the electrolytic cell to the fluctuation of wind and light output. However, long-term unstable operation of the electrolytic cell will affect its service life. Second, in the standby state, the electrolytic cell stops producing hydrogen, but consumes power to maintain its temperature and pressure, which can make the electrolytic cell quickly switch to the hydrogen production state. In addition, in the shutdown state, no power is consumed, but the start and stop time constraints need to be met, as follows:

[0151]

[0152] In the formula, represents the closed state of the electrolytic cell; and are 0-1 variables, representing the on-off state of the electrolytic cell; represents the time when the electrolytic cell starts; represents the time when the electrolytic cell is closed.

[0153] Multiple electrolyzers can be combined to operate, which can expand its working range. Secondly, in the process of hydrogen production by water electrolysis, fluctuating power will cause electrolytic capacitor to produce. It will in turn cause electrode corrosion and electrolyte degradation, and have an adverse effect on the performance and service life of electrolyzer. Flexible operation mode can increase the number of electrolyzers working in high efficiency and rated state, reduce the duration of unstable operation state, and help to prolong the service life. Multiple electrolyzers are combined to operate in order to maximize the number of electrolyzers in high efficiency and rated operation state, and the remaining power is allocated to electrolyzers operating in unstable and standby state. The hydrogen production rate of electrolyzer is as follows:

[0154]

[0155] wherein, represents the hydrogen production rate of the i-th electrolyzer at time t; represents the heat value of hydrogen; represents the thermal power of the i-th electrolyzer at time t; represents the hydrogen production amount of the i-th electrolyzer at time t; η EL,el represents the efficiency of the electrolyzer.

[0156] For the power allocation method considering output fluctuation and simultaneous rate, the power allocation strategy of electrolyzer array realizes efficient and stable hydrogen production by dynamically adjusting the operation state of both. The power allocation strategy of electrolyzer array is based on the relationship between total power demand and rated power of different types of electrolyzers (AEL and PEMEL), and further optimizes the dynamic adjustment process by integrating fluctuation rate and output simultaneous rate index, to improve the stability and economy of hydrogen production system. When the total power demand P t Array is equal to the sum of rated power of AEL and PEMEL P t Array = E AEL + E PEMEL , both of which operate at rated power, maximize the utilization of installed capacity, thereby improving the output simultaneous rate S OS , ensuring that the daily maximum output is close to the upper limit of the total capacity. If the total power demand is lower than the smaller rated power P t Array ≤ min (E AEL , E PEMEL ), the small capacity electrolyzer (such as AEL) is started preferentially, which is stably operated in the high efficiency interval, and the remaining power is supplemented by the large capacity device, reducing the frequent adjustment of small capacity device, thereby reducing the average fluctuation rate γ ave,tWhen the power demand is between the sum of the two rated powers, the strategy introduces a dynamic allocation mechanism, i.e. adjusting the allocation proportion according to the real-time volatility rate; when the volatility rate is high, the PEMEL with strong adaptability is preferred to bear the fluctuation part, and the allocation proportion is adjusted from the fixed 50% to a dynamic value P AEL =k·P t Array (0<k≤0.5), the remaining power is used to smooth the power fluctuation by utilizing the fast response characteristics of PEMEL; when the volatility rate is low, the fixed proportion allocation is restored to ensure the efficient operation of AEL. At the same time, when the power demand approaches the total installed capacity, the optimization allocation makes all electrolytic cells as close to the rated power as possible, which improves the S OS to close to 100%. This strategy dynamically coordinates the low-cost advantage of AEL and the flexibility of PEMEL by predicting the power demand and volatility trend, reduces the overall volatility rate, improves the capacity utilization rate, and avoids large-scale production stoppage.

[0157] S3, constructing a capacity configuration multi-objective optimization model according to the preset hydrogen production condition combined with the electrolytic cell power allocation model, and solving the capacity configuration multi-objective optimization model by using a multi-objective chef optimization algorithm with a mixed advantage relationship to obtain an optimal solution of capacity configuration.

[0158] Specifically, the capacity configuration multi-objective optimization model is constructed according to the preset hydrogen production condition combined with the electrolytic cell power allocation model, and the capacity configuration multi-objective optimization model is solved by using a multi-objective chef optimization algorithm with a mixed advantage relationship to obtain an optimal solution of capacity configuration, which includes:

[0159] S31, obtaining the hydrogen production efficiency based on the electrolytic cell power allocation model, and constructing a wind-solar coupled hydrogen production system capacity configuration multi-objective optimization model by using the hydrogen production efficiency combined with the preset hydrogen production cost;

[0160] S32, generating hydrogen production equipment constraints and hydrogen production operation constraints according to the preset hydrogen production parameters, and constructing constraint conditions of the capacity configuration multi-objective optimization model through the hydrogen production equipment constraints and the hydrogen production operation constraints.

[0161] Specifically, the hydrogen production equipment constraints include: electrical energy storage constraints, thermal energy storage constraints, and hydrogen storage balance constraints;

[0162] The hydrogen production operation constraints include: electrical load constraints, thermal load constraints, and hydrogen load constraints; the configuration method can further improve the robustness in complex fluctuation scenarios under the double constraints of the hydrogen production equipment constraints and the hydrogen production operation constraints.

[0163] S33, solve the capacity configuration multi-objective optimization model by using the multi-objective chef optimization algorithm with mixed dominance relationship to obtain the optimal solution of capacity configuration; under the construction of the capacity configuration multi-objective optimization model, the multi-objective collaborative optimization of low hydrogen production cost, high hydrogen production efficiency and hydrogen production sustainability can be realized.

[0164] Specifically, the multi-objective chef optimization algorithm with mixed dominance relationship is used to solve the capacity configuration multi-objective optimization model to obtain the optimal solution of capacity configuration, which includes:

[0165] S331, initialize the population size of chef instructors and apprentices in the multi-objective chef optimization algorithm, the mixed dominance threshold and the maximum number of iterations, and calculate the objective function value of each individual through the capacity configuration multi-objective optimization model;

[0166] S332, use fast non-dominated sorting and crowding distance to select individuals according to the objective function value of each individual to obtain the initialization archive set;

[0167] S333, based on the master-slave strategy, constantly update the population using the initialization archive set, and iteratively update the archive set combined with the selected dominance relationship to obtain the optimal solution set of capacity configuration; under the optimization of the multi-objective chef optimization algorithm, the running fluctuation and the phenomenon of wind and light abandonment can be effectively reduced, and the renewable energy utilization rate and the overall performance of the hydrogen production system can be improved.

[0168] Specifically, a capacity configuration multi-objective optimization model of hydrogen production system considering hydrogen production cost, fluctuation rate and renewable energy utilization rate is proposed, which combines the energy flow mode of wind-solar coupled hydrogen production system to reduce the system hydrogen production cost, fluctuation rate and renewable energy utilization rate as the target, and uses the multi-objective chef optimization algorithm based on mixed dominance relationship to solve the model.

[0169] The calculation formula of the multi-objective optimization model of the hydrogen production system considering cost, efficiency and sustainability is:

[0170]

[0171] In the formula, represents the electric power of the i th electrolyzer; C EL represents the hydrogen production cost; C PN represents the penalty cost; α represents C PN the weight in the objective function; W EL represents the average value of the fluctuation; c ael and c pemel represent the construction cost coefficient of AEL and PEMEL; C EL,op represents the operating cost of the electrolyzer; L represents the service life of the electrolyzer; Δt represents the time interval; T represents the amount of time period in the planning period; cel , c on and c off respectively represent the cost coefficients of electrolyzer operation, start-up and shut-down; and represent the on-off action of electrolyzer, which is a 0-1 variable;c re represent the cost coefficient of wind and light curtailment;P t RE,PN represent the amount of wind and light curtailment.

[0172] For the multi-objective chef optimization algorithm, first, the population size containing chef instructors and apprentice subgroups, hybrid dominance threshold and maximum iteration number are initialized, the target function value corresponding to each individual is calculated, individual selection is carried out by using fast non-dominated sorting and crowded distance, and the archive set is initialized. Subsequently, the master-slave strategy (chef and apprentice) is used to update the population with the archive solution, the archive is updated combined with the selected dominance relationship, and the loop is executed before the iteration reaches the maximum number. Finally, the approximate Pareto optimal solution set (i.e. the optimal solution of capacity configuration) is output.

[0173] When the installed capacity ratio of AEL is high, the cost of EL array is relatively low, but the renewable energy utilization ability and adaptability to renewable energy fluctuation are poor. It will cause significant fluctuations in hydrogen production rate. When the installed capacity ratio of PEMEL is high, the renewable energy utilization ability and hydrogen production can be effectively improved. However, the cost of PEMEL is relatively high, which will increase the construction cost of EL array and affect its cost-effectiveness. In order to realize the optimal capacity configuration of EL array AEL and PEMEL, the present application comprehensively considers the hydrogen production cost, hydrogen production efficiency and penalty cost to establish a multi-objective optimization model of hydrogen production system.

[0174] The calculation formula of the electrical energy storage constraint in the constraint condition of the multi-objective optimization model is:

[0175]

[0176] In the formula, represents the residual capacity of electrical energy storage at time t;P t EES,ch and P t EES,dis represents the charging and discharging power of electrical energy storage at time t; and represents the charging and discharging state of electrical energy storage at time t, which is a 0-1 variable; represents the maximum value of the charging and discharging power of electrical energy storage; and represents the upper and lower limits of the residual capacity of electrical energy storage.

[0177] The calculation formula of the thermal energy storage constraint is:

[0178]

[0179] where, denotes the residual capacity of the thermal storage at time t; and denotes the charge-discharge power of the thermal storage at time t; and denotes the state of charge of the thermal storage at time t, which is a 0-1 variable; denotes the maximum value of the charge-discharge power of the thermal storage; and denotes the upper and lower limits of the residual capacity of the thermal storage.

[0180] Hydrogen storage balance constraint calculation formula:

[0181]

[0182] where, denotes the residual capacity of the hydrogen storage at time t; and denotes the charge-discharge rate of the hydrogen storage at time t; denotes the maximum value of the charge-discharge rate of the hydrogen storage; denotes the maximum value of the residual capacity of the hydrogen storage.

[0183] Electric load constraint calculation formula:

[0184] P t RE +P t HFC +P t EES,dis = P t EES,ch +P t Array +P t HP +P t LD ;

[0185] where, P t RE denotes the renewable energy generation at time t; P t HFC denotes the electric power of the hydrogen fuel cell at time t; P t Array denotes the electric power of the electrolyzer array at time t; P t HP denotes the electric power of the heat pump at time t; P t LD denotes the electric load of the system at time t.

[0186] Thermal load constraint calculation formula:

[0187]

[0188] wherein, represents the total heat production of the electrolyzer array at time t; represents the thermal power of the hydrogen fuel cell at time t; represents the thermal power of the heat pump at time t; represents the thermal load of the system at time t.

[0189] Hydrogen load constraint calculation formula:

[0190]

[0191] wherein, represents the total hydrogen consumption of the electrolyzer array at time t; represents the hydrogen consumption of the hydrogen fuel cell at time t; represents the hydrogen load of the system at time t.

[0192] To solve the multi-objective model, a multi-objective chef optimization algorithm based on mixed dominance relation is used to solve the model. First, input the device parameters and electrolyzer type constraints of hydrogen fuel cell, heat pump, electricity storage, heat storage, hydrogen storage, etc., initialize the population size containing chef instructors and student subgroups, mixed dominance threshold and maximum iteration number, and set the archive capacity; randomly generate particle position and velocity, construct a candidate pool consisting of history, current and archive groups to maintain diversity. Then, calculate the objective function value of each solution, check the feasibility combining with power boundary, running logic, electrolyzer start-stop constraints, etc., and use the mixed dominance strategy to non-dominantly sort the candidate pool solutions, compare each solution with other solutions one by one using the current dominance relation, and select the non-dominated solution that is not dominated under this strategy as the current Pareto front, divide the front level and evaluate the distribution density by the crowding distance, calculate the normalized sum of the daily standard distance between each solution and the adjacent two solutions, the larger the distance, the sparser the area where the solution is located, and the solution of the front 1 is stored in the archive. When dynamically updating the archive, the dominance strategy is rotated to compare the candidate solution with the archive solution one by one, retain the non-dominated solution and control the capacity, continue to compare the new front solution with the archive solution according to the current dominance relation, and replace the dominated solution; that is, if the archive is saturated, the solution in the dense area (low crowding distance) or low dominance level is preferentially eliminated. In the population update stage, the chef instructor performs global development based on the archive solution, while the student adjusts the position through skill learning, variable replacement and local search, and only accepts new solutions that dominate the current solution. During the iteration process, if the maximum iteration number is not reached, repeat the mixed dominance sorting and updating; finally, output the Pareto front solution set in the archive that meets the mixed dominance relation, which meets the non-inferiority under the current selected dominance relation, is the "optimal compromise solution" under multiple objective functions, each solution corresponds to a specific scheduling strategy, and provides a multi-objective optimal compromise scheme.

[0193] In summary, the present application firstly adopts a dyeing clustering algorithm based on an ant colony routing mechanism to cluster typical wind-solar combined output scenarios; then, in view of the dynamic response characteristics of AEL-PEMEL, a flexible mode of combined operation of multiple electrolytic cells is proposed; at the same time, a power distribution method considering output fluctuation and simultaneity rate is proposed; finally, a multi-objective optimization model of system capacity configuration is constructed, combined with the energy flow mode of the wind-solar coupled hydrogen production system, to reduce the system hydrogen production cost, reduce the fluctuation rate and the utilization rate of renewable energy as the target, and a multi-objective chef optimization algorithm based on a hybrid dominance relationship is used to solve the model. The wind-solar scene generation technology based on the ant colony routing mechanism of the dyeing clustering method proposed in the present application breaks through the limitation of traditional clustering relying on preset parameters, dynamically optimizes the wind-solar scene division based on ant colony intelligence, significantly improves the adaptability to complex resource distribution and engineering constraints, and improves the calculation efficiency; the model through the innovative framework of mixed driving of mechanism and data not only guarantees the operation reliability of the hydrogen production system under complex fluctuation scenarios, but also realizes the efficient utilization and cost optimization of renewable energy, providing an innovative solution for the planning and design of large-scale green hydrogen production systems.

[0194] To sum up, by means of the technical scheme of the present application, the present application can dynamically adjust the number of ants and the spatial density measurement by the ant colony pathfinding mechanism-based coloring clustering algorithm, effectively guide the ants to be uniformly distributed in the data set, and avoid the problem of being easily trapped in local optimum in the traditional clustering algorithm; and the forward and reverse neighborhood matrix is used to accurately depict the relationship between points, and the nearest neighbor density is combined to provide a scientific basis for path selection, thereby improving the rationality of the path and the accuracy of clustering; meanwhile, the global crossover and local dilemma termination mechanism is introduced to effectively prevent invalid extension of the path and local traps, thereby ensuring the convergence and stability of the algorithm; finally, the path coloring strategy is used in combination with the majority voting mechanism to realize dynamic expansion of the clustering label, so that the clustering result has good continuity and robustness; meanwhile, the ant colony pathfinding mechanism-based coloring clustering algorithm is applied to scene clustering, and compared with the common clustering algorithm, the present application has better robustness and continuity; the present application realizes the power distribution method of combined operation of multiple electrolytic cells by constructing an electrolytic cell power distribution model by combining the output fluctuation rate and the output simultaneous rate, realizes the efficient and stable operation of the system by dynamically coordinating the operation of the AEL and the PEMEL electrolytic cells; it fully utilizes the low-cost advantage of the AEL and the fast response capability of the PEMEL, effectively suppresses the output fluctuation, improves the output simultaneous rate and the installed capacity utilization rate, reduces the operation cost and the shutdown risk, enhances the adaptability of the hydrogen production system to power demand fluctuation, and significantly improves the overall economy and reliability; the present application provides an efficient system capacity planning scheme by comprehensively considering the hydrogen production cost, the fluctuation rate and the renewable energy utilization rate by the multi-objective chef optimization algorithm, effectively reduces the system operation fluctuation and the curtailment of wind and light, and improves the renewable energy utilization rate and the overall performance of the hydrogen production system.

[0195] The above merely describes preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for capacity configuration of a wind-solar-hydrogen system based on dynamic ant colony clustering and chef optimization algorithm, characterized in that, The method comprises: S1, obtaining wind and light output data, and using an ant colony path mechanism dyeing clustering algorithm to cluster the wind and light output data to obtain a wind and light combined output combination set; S2, based on the wind and light combined output combination set and a pre-obtained electrolytic cell response feature set, constructing an electrolytic cell combined operation mode, and constructing an electrolytic cell power distribution model according to the electrolytic cell combined operation mode combined with output fluctuation rate and output simultaneous rate; S3, according to a pre-set hydrogen production condition combined with the electrolytic cell power distribution model, constructing a capacity configuration multi-objective optimization model, and using a multi-objective chef optimization algorithm with a mixed advantage relationship to solve the capacity configuration multi-objective optimization model to obtain an optimal solution of capacity configuration.

2. The method according to claim 1, wherein, The method comprises: S11, obtaining wind and light output data, and using the wind and light output data and a pre-set ant colony base to construct an ant colony generation model, generating the number of ants through the ant colony generation model; S12, constructing a neighborhood matrix based on the wind and light output data, calculating the nearest neighbor density using the neighborhood matrix, constructing a path finding model using the reliability score and the inverse S-shaped function according to the nearest neighbor density, and selecting the ant colony path based on the path finding model; S13, generating an ant colony termination mechanism through a pre-set rule, constructing a path finding termination model according to the ant colony termination mechanism, and terminating the ant colony path finding using the path finding termination model; S14, dyeing clustering the ant colony path when the ant colony path finding is terminated to obtain the wind and light combined output combination set.

3. The method according to claim 2, wherein, The method comprises: S121, constructing a neighborhood matrix using the wind and light output data according to the pre-set forward neighborhood and reverse neighborhood, and calculating the nearest neighbor density based on the neighborhood matrix; S122, calculating the path reliability score based on the reliability score and the inverse S-shaped function combined with the nearest neighbor density and the pre-obtained pheromone, and updating the pheromone using the path reliability score; S123, using roulette to allocate probability according to the path reliability score, generating a path finding model based on the probability allocation result, and selecting the ant colony path using the path finding model.

4. The method of claim 2, wherein the method is characterized by, The method comprises: S131, generating an ant colony global cross termination mechanism using Gaussian distribution based on a pre-set neighborhood point distance threshold; S132, generating an ant colony local dilemma termination mechanism according to a pre-set access frequency threshold combined with a pre-set access rule; S133, constructing a path finding termination model through the ant colony global cross termination mechanism and the ant colony local dilemma termination mechanism, and terminating the ant colony path finding using the path finding termination model.

5. The method of claim 2, wherein the method is characterized by, The method comprises: S14, dyeing clustering the ant colony path when the ant colony path finding is terminated to obtain the wind and light combined output combination set. S141, color the ant colony path when the ant colony pathfinding is terminated based on the preset coloring condition, and obtain a coloring mark set; S142, use the majority voting mechanism to perform coloring coverage according to the coloring mark set, perform coloring clustering on the ant colony path based on the coloring coverage result, and obtain a wind-solar combined output combination set.

6. The method of claim 1, wherein, The wind-solar combined output combination set and the pre-acquired electrolytic cell response feature set are used to construct an electrolytic cell combined operation mode, and an electrolytic cell power distribution model is constructed according to the electrolytic cell combined operation mode in combination with an output fluctuation rate and an output simultaneous rate, and the electrolytic cell power distribution model comprises: S21, the operation state of the electrolytic cell is divided by using the wind-solar combined output combination set and the pre-acquired electrolytic cell response feature set, and the electrolytic cell combined operation mode is constructed according to the division result in combination with a preset operation switching rule; S22, the power distribution model of the electrolytic cell is constructed by calculating the average fluctuation rate and the output simultaneous rate of the power generation output based on the electrolytic cell combined operation mode.

7. The method according to claim 6, wherein, The electrolytic cell power distribution model comprises: When the total power demand of the electrolytic cell is equal to the sum of the rated power of all electrolytic cells, all electrolytic cells operate at rated power, and the output simultaneous rate is improved; When the total power demand of the electrolytic cell is lower than the rated power of all electrolytic cells, a specific alkaline electrolytic cell is started preferentially, and the average fluctuation rate of the power generation output is reduced; When the total power demand of the electrolytic cell is between the rated power of all electrolytic cells, the power distribution ratio of the proton exchange membrane electrolytic cell and the alkaline electrolytic cell is dynamically adjusted according to the real-time fluctuation rate.

8. The method of claim 1, wherein, The capacity configuration multi-objective optimization model is constructed according to the preset hydrogen production condition in combination with the electrolytic cell power distribution model, and the capacity configuration multi-objective optimization model is solved by using a multi-objective chef optimization algorithm with a mixed advantage relationship to obtain an optimal solution of capacity configuration, and the capacity configuration multi-objective optimization model comprises: S31, the hydrogen production efficiency is obtained based on the electrolytic cell power distribution model, and the wind-solar coupled hydrogen production system capacity configuration multi-objective optimization model is constructed by using the hydrogen production efficiency in combination with a preset hydrogen production cost; S32, hydrogen production equipment constraints and hydrogen production operation constraints are generated according to preset hydrogen production parameters, and the constraint conditions of the capacity configuration multi-objective optimization model are constructed by using the hydrogen production equipment constraints and the hydrogen production operation constraints; S33, the capacity configuration multi-objective optimization model is solved by using a multi-objective chef optimization algorithm with a mixed advantage relationship to obtain an optimal solution of capacity configuration.

9. The method according to claim 8, wherein, The hydrogen production equipment constraints comprise: electrical energy storage constraints, thermal energy storage constraints, and hydrogen storage balance constraints; The hydrogen production operation constraints comprise: electrical load constraints, thermal load constraints, and hydrogen load constraints.

10. The method of claim 8, wherein the method is characterized by, The capacity configuration multi-objective optimization model is solved by using a multi-objective chef optimization algorithm with a mixed advantage relationship to obtain an optimal solution of capacity configuration, and the capacity configuration multi-objective optimization model comprises: S331, the population size, the mixed advantage threshold value, and the maximum number of iterations of the chef instructor and the apprentice subpopulation in the multi-objective chef optimization algorithm are initialized, and the objective function value of each individual is calculated by using the capacity configuration multi-objective optimization model; S332, the objective function value of each individual is selected by using fast non-dominated sorting and crowding distance to obtain an initialization archive set; S333, based on master-slave strategy using initialization file set constantly update population, and combined with the selected dominant relationship for iterative update of the file set, get the optimal solution set of capacity configuration.

Images