An aluminum electrolysis process parameter optimization method based on a double-space coordination strategy
The aluminum electrolysis process parameter optimization method based on the dual-space collaborative strategy uses Ward hierarchical clustering and Gaussian kernel function to calculate density values. Combining density ranking and dual-objective optimization, it solves the problems of the honey badger algorithm's lack of diversity and convergence in multimodal problems in aluminum electrolysis. It realizes the identification and retention of multiple sets of process parameter combinations, and improves the flexibility and efficiency of the aluminum electrolysis process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZUNYI ALUMINUM
- Filing Date
- 2026-06-02
- Publication Date
- 2026-07-31
AI Technical Summary
Existing multi-objective optimization methods based on the honey badger algorithm suffer from several problems in aluminum electrolysis, including maintaining a single diversity, lack of adaptability in the search strategy, and difficulty in balancing the diversity of the decision space and the objective space. These problems lead to solution set collapse in the optimization of aluminum electrolysis process parameters, making it impossible to provide multiple sets of equivalent process parameter combinations and limiting the flexibility of engineering applications.
A dual-space collaborative strategy is adopted. The decision space is divided by Ward hierarchical clustering, the individual density value is calculated by Gaussian kernel function and a local penalty mechanism is introduced. The search stage is dynamically selected by density ranking. A dual-objective optimization problem is constructed to perform non-dominated ranking, so as to achieve collaborative optimization of the diversity and convergence of the decision space and the objective space.
It effectively identifies and retains multiple equivalent Pareto solution sets, outputs multiple sets of optional aluminum electrolysis process parameter combinations, improves the flexibility of engineering applications, avoids pseudo-diversity and premature convergence, and provides a variety of process parameter selection schemes.
Smart Images
Figure CN122490134A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent control technology for aluminum electrolysis, and specifically to a method for optimizing aluminum electrolysis process parameters based on a dual-space collaborative strategy. Background Technology
[0002] The aluminum electrolysis process is characterized by high energy consumption, strong coupling, nonlinearity, and multimodal behavior. Optimizing process parameters (such as cell voltage and electrolysis temperature) is key to improving current efficiency and reducing energy consumption and carbon emissions. However, different combinations of process parameters may produce similar optimization target values (i.e., multimodal characteristics), making traditional single-objective optimization methods difficult to apply.
[0003] Swarm intelligence optimization algorithms, by simulating the behavior of biological groups to perform global searches, have shown significant advantages in solving complex optimization problems. Among them, the Honey Badger Algorithm (HBA) simulates the foraging behavior of honey badgers, effectively avoiding getting trapped in local optima by balancing the mining phase (global exploration) and the honey phase (local exploitation). In recent years, researchers have attempted to apply HBA to multi-objective optimization, integrating strategies such as non-dominated sorting, crowding distance, and dynamic updates of external archives to enable HBA to solve multi-objective optimization problems.
[0004] However, existing multi-objective optimization methods based on HBA still have the following shortcomings: (1) Diversity is maintained in a single way: Most methods assess population diversity based on the crowding distance of the target space and do not explicitly model the decision space. When the optimization problem has multimodal characteristics (i.e., multiple different combinations of decision variables correspond to similar target values), the algorithm is prone to collapse to a single equivalent Pareto solution set (PS), which cannot provide multiple sets of equivalent process parameters and limits the flexibility of engineering applications; (2) The search strategy lacks adaptability: In existing methods, the switching between exploration and development stages often depends on a fixed random threshold and cannot adaptively adjust the search behavior according to the distribution density of individuals in the decision space. This leads to low efficiency of the algorithm in exploring multimodal regions and difficulty in covering multiple equivalent solution sets at the same time; (3) The diversity of the decision space and the target space is difficult to balance: Traditional non-dominated sorting only focuses on the convergence and distribution of the target space. The diversity of solutions in the decision space is easily lost in the iteration process, while simply increasing the diversity of the decision space may sacrifice the convergence accuracy of the target space. There is a lack of an effective mechanism that can balance the diversity of the two spaces at the same time.
[0005] Therefore, how to improve the diversity of solution sets of multi-objective optimization algorithms under multi-modal problems while ensuring convergence is a technical problem that urgently needs to be solved in the optimization of aluminum electrolysis process parameters. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention aims to provide a method for optimizing aluminum electrolysis process parameters based on a dual-space collaborative strategy. This method can maintain diversity in both the decision space and the target space, and obtain multiple equivalent Pareto solution sets while ensuring convergence, thus obtaining multiple sets of optional process parameter combinations.
[0007] The basic solution provided by this invention is a method for optimizing aluminum electrolysis process parameters based on a dual-space synergistic strategy, comprising the following steps: S1: Initialize the population and external files of the honey badger algorithm, where each individual in the population represents a set of aluminum electrolysis process parameters; S2: Estimate the decision space density of the current population. Divide the decision space into multiple regions through hierarchical clustering. Calculate the density value of each individual in its region using the Gaussian kernel function. Introduce a local penalty mechanism to adjust the density value. Use the adjusted density value as the basis for evaluating population diversity. S3: Sort the individuals in the population in ascending order according to the adjusted density value obtained in S2 to obtain the density ranking of each individual. Calculate the selection factor of each individual based on the density ranking, and dynamically determine whether the individual enters the exploration or development stage according to the relationship between the selection factor and the preset threshold. Generate the offspring population according to the position update formula of the honey badger algorithm, wherein the current optimal individual position used in the position update formula of the honey badger algorithm is obtained from the external archive. S4: Merge the parent population with the offspring population generated in S3. For each individual in the merged population, calculate its density value in the decision space and its crowding distance in the objective space. Construct a bi-objective optimization problem with the goal of minimizing the density value and maximizing the crowding distance. Use this bi-objective optimization problem to perform non-dominated sorting on the merged population and sort them in descending order of crowding distance within the same Pareto front. Select a new generation of population from this population and store the non-dominated solutions in the new generation of population in the external archive. S5: Repeat S2 to S4 until the termination condition is met, and then output the Pareto optimal solution set stored in the external file. The solution set corresponds to multiple equivalent combinations of aluminum electrolysis process parameters in the decision space.
[0008] Furthermore, S2 includes: S21: The Ward hierarchical clustering method is used to determine the number of clusters by minimizing the intra-cluster variance. Divide the decision space into For each cluster, the clustering distance metric used is Euclidean distance, for two individuals and The distance calculation formula is as follows:
[0009] in, As a dimension of the decision space, and Individuals and In the Coordinates in dimensional space; S22: In each cluster Internally, the Gaussian kernel function is used to calculate individual... density value :
[0010] in, For two individuals and The Euclidean distance between them For clusters The standard deviation of all individuals within the range; S23: Introduce a local penalty mechanism to adjust the density value according to the cluster size, and obtain the adjusted density value. :
[0011] in, For the current cluster The number of individuals in This represents the total number of individuals in the population.
[0012] Furthermore, S3 includes: S31: The adjusted density value obtained in S2 Sort in ascending order to obtain density ranking. ,in, Indicates minimum density. This indicates the highest density. Population size; S32: Calculate each individual Selection factors :
[0013] in, A random number uniformly distributed in the range [0,1]. S33: If Once an individual enters the exploration phase, the new individual location is generated using the honey badger algorithm's location update formula for the mining phase. :
[0014] like Once an individual enters the development phase, the honey badger algorithm's honey stage position update formula is used to generate a new individual position. :
[0015] in, The current optimal individual position is obtained from the external archive. The parameters governing the honey badger's ability to obtain food adjust the individual's tracking intensity towards prey. The density factor characterizes the intensity of prey odor and is inversely proportional to the square of the distance between individuals. An adaptive decreasing factor that decreases with the number of iterations. For the current individual With the optimal individual The Euclidean distance between them , , A random number uniformly distributed in the range [0,1]. A marker used to control the search direction.
[0016] Furthermore, the adaptive reduction factor is calculated using the following formula:
[0017] in, As a preset constant, This represents the current iteration number. The maximum number of iterations, a flag controlling the search direction. Randomly generated as 1 or -1.
[0018] Furthermore, in S4: Density values in the decision space The calculation formula is:
[0019] in, For two individuals and The Euclidean distance between them For clusters The standard deviation of all individuals within the range; The formula for calculating the crowding distance in the target space is:
[0020] in, For the target number, for the th A target value is set, and all individuals in the population are sorted in ascending order according to that target value. and Individuals The target value of two adjacent individuals in this sequence, and These are the maximum and minimum values of the target, respectively. The bi-objective optimization problem is expressed as:
[0021] in, Minimize to enhance decision space diversity. Minimize to enhance the diversity of the target space.
[0022] Furthermore, the external archive is updated as follows: the non-dominated solutions in the new generation population selected in S4 are compared with the existing solutions in the external archive. If the new solution is not dominated by any solution in the external archive, the new solution is added to the external archive, and the solutions dominated by the new solution in the external archive are deleted. When the number of solutions in the external archive exceeds the preset capacity, pruning is performed according to the crowding distance, and solutions with larger crowding distances are retained.
[0023] Furthermore, the termination condition is: reaching the preset maximum number of iterations, or the change in the optimal target value of the population for a consecutive preset number of generations is less than a preset threshold.
[0024] Furthermore, the aluminum electrolysis process parameters include electrolysis temperature, electrolyte level, aluminum level, cell voltage, and aluminum fluoride feed rate.
[0025] The principle and beneficial effects of this invention are as follows: 1. Construct a decision space density estimation method to achieve active identification and preservation of multimodal solution sets: Traditional multi-objective optimization algorithms (such as NSGA-II) and existing multi-objective optimization methods based on the honey badger algorithm maintain population diversity through crowding distance in the objective space, but do not explicitly model the decision space. When the optimization problem has multimodal characteristics (i.e., multiple different combinations of decision variables correspond to similar objective values), these methods are prone to collapse into a single equivalent Pareto solution set (PS) in the decision space, resulting in only one set of process parameters being output, which cannot meet the actual production requirement of "flexibly selecting parameters according to raw materials and tank conditions". To address this problem, this invention first employs Ward hierarchical clustering to adaptively partition the decision space of the current population. By minimizing the intra-cluster variance, the number of clusters is automatically determined, dividing the decision space into several clusters. Based on this, a Gaussian kernel function is used to calculate the density value of each individual within its cluster. A local penalty mechanism is introduced to adjust the density value according to the cluster size (the density of individuals in large clusters is suppressed, while the density of individuals in small clusters is relatively increased). This mechanism enables the invention to proactively identify the boundary regions of multiple equivalent Pareto solution sets (PS) and distinguish different modes. During the iteration process, individuals from different clusters are retained to avoid the suppression of small clusters (sparse regions) by large clusters (dense regions). The adjusted density value is used as a direct basis for evaluating population diversity, rather than relying solely on target space information.
[0026] 2. Constructing a non-dominated ranking system with dual-space density balance to achieve synergistic optimization of decision space diversity and target space convergence: In traditional non-dominated ranking systems, the quality of an individual is determined based on its Pareto dominance in the target space, and the crowding distance is also calculated based on the objective function value. This mechanism completely ignores the distribution information of the decision space, leading to a potential loss of diversity in the decision space while pursuing target space convergence. This invention constructs a non-dominated ranking system with dual-space density balance to achieve synergistic optimization of decision space diversity and target space convergence. The opposite of the target space crowding distance As two optimization objectives, construct a bi-objective optimization problem. In each generation of selection, the algorithm performs a non-dominated sort of the bi-objective: minimizing... This encourages the algorithm to retain individuals in sparse regions of the decision space, thereby maintaining multimodal coverage; minimizing In essence, it maximizes the crowding distance to maintain the distribution of the target space. Through non-dominated sorting, the algorithm naturally achieves a diversity balance between the two spaces: an individual that is sparse in the decision space but clustered in the target space, or vice versa, may be retained; only individuals that are dense in both spaces are eliminated. Compared with existing methods, this mechanism avoids the pseudo-diversity situation of "large crowding distance in the target space but collapse of the decision space," and does not require manual setting of balancing weights, but is entirely adaptively adjusted by the Pareto dominance relationship, thus preserving as much multimodal information in the decision space as possible while ensuring the convergence accuracy of the target space.
[0027] 3. Adaptive Search Stage Selection Based on Density Ranking Improves Search Efficiency: In existing multi-objective optimization methods based on the honey badger algorithm, the switching between exploration and development stages relies entirely on a fixed random threshold independent of individual states. This invention further dynamically calculates selection factors based on the density ranking of individuals in the decision space, making individuals in sparse areas more inclined to explore and individuals in dense areas more inclined to develop. This strategy further improves the efficiency of multimodal optimization.
[0028] Therefore, compared with the prior art, this invention solves the technical problems of solution set collapse and insufficient diversity in multimodal and multi-objective optimization of aluminum electrolysis. Specifically, it can actively identify and retain multiple equivalent Pareto solution sets and output multiple sets of optional aluminum electrolysis process parameter combinations, greatly improving the flexibility of engineering applications; it achieves a dynamic balance between decision space diversity and objective space convergence, avoiding false diversity and premature convergence; in practical applications of aluminum electrolysis, it can provide a variety of different schemes, such as high temperature high molecular weight ratio and low temperature low molecular weight ratio, for on-site engineers to flexibly choose according to raw materials and cell conditions. Attached Figure Description
[0029] Figure 1This is a flowchart illustrating an embodiment of an aluminum electrolysis process parameter optimization method based on a dual-space collaborative strategy according to the present invention. Detailed Implementation
[0030] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown: This embodiment takes the optimization of process parameters of an aluminum electrolysis cell as an example. A method for optimizing aluminum electrolysis process parameters based on a dual-space collaborative strategy includes the following steps: S1. Initialization: Initialize the population and external files of the honey badger algorithm. Each individual in the population represents a set of aluminum electrolysis process parameters, including electrolysis temperature, electrolyte level, aluminum level, cell voltage and aluminum fluoride feed rate. The external files are initialized to empty and the preset capacity is set to 100.
[0031] S2, Decision space density estimation: S21: The Ward hierarchical clustering method is used to determine the number of clusters by minimizing the intra-cluster variance. Divide the decision space into For each cluster, the clustering distance metric used is Euclidean distance, for two individuals and The distance calculation formula is as follows:
[0032] in, As a dimension of the decision space, and Individuals and In the Coordinates on a dimension (corresponding to one of the process parameters); S22: In each cluster Internally, the Gaussian kernel function is used to calculate individual... density value :
[0033] in, For two individuals and The Euclidean distance between them For clusters The standard deviation of all individuals within the range; S23: Introduce a local penalty mechanism to adjust the density value according to the cluster size, and obtain the adjusted density value. :
[0034] in, For the current cluster The number of individuals in The local penalty mechanism reduces the density of individuals in large clusters, thereby encouraging the algorithm to retain individuals in small clusters (sparse regions) to maintain multimodal coverage of the decision space.
[0035] S3, Adaptive Search Based on Density Ranking: S31: The adjusted density value obtained in S2 Sort in ascending order to obtain density ranking. ,in, This indicates the lowest density (sparsest). This indicates the highest density (most concentrated). Population size; S32: Calculate each individual Selection factors :
[0036] in, A random number uniformly distributed in the range [0,1]. S33: If Once an individual enters the exploration phase, the new individual location is generated using the honey badger algorithm's location update formula for the mining phase. :
[0037] like Once an individual enters the development phase, the honey badger algorithm's honey stage position update formula is used to generate a new individual position. :
[0038] in, The optimal individual position is obtained from the external archives based on the crowding distance; the greater the crowding distance, the higher the probability of being selected. This parameter determines the honey badger's ability to obtain food; the default value is 6. It adjusts the individual's tracking intensity towards prey. The density factor characterizes the intensity of prey odor and is inversely proportional to the square of the distance between individuals. As an adaptive decreasing factor that decreases with the number of iterations, the following formula is used to calculate it in this embodiment:
[0039] in As a preset constant, This represents the current iteration number. The maximum number of iterations, For the current individual With the optimal individual The Euclidean distance between them , , A random number uniformly distributed in the range [0,1]. The flag used to control the search direction is randomly generated as 1 or -1; The offspring population is generated through the above process.
[0040] S4. Population ranking and selection in a two-space density equilibrium: The parent population is merged with the generated offspring population. For each individual in the merged population, its density value in the decision space and its crowding distance in the objective space are calculated. A bi-objective optimization problem is constructed with the goal of minimizing the density value and maximizing the crowding distance. The merged population is then non-dominated and sorted in descending order of crowding distance within the same Pareto front. A new generation of population is selected from this population, and the non-dominated solutions in the new generation population are stored in the external archive. Density values in the decision space The calculation formula is:
[0041] in, For two individuals and The Euclidean distance between them For clusters The standard deviation of all individuals within the range; The formula for calculating the crowding distance in the target space is:
[0042] in, For the target number, for the th A target value is set, and all individuals in the population are sorted in ascending order according to that target value. and Individuals The target value of two adjacent individuals in this sequence, and These are the maximum and minimum values of the target, respectively. The bi-objective optimization problem is expressed as:
[0043] in, Minimize to enhance decision space diversity. Minimize to enhance the diversity of the target space; The external archive is updated as follows: the non-dominated solutions in the new generation population selected in S4 are compared with the existing solutions in the external archive. If the new solution is not dominated by any solution in the external archive, the new solution is added to the external archive, and the solutions dominated by the new solution in the external archive are deleted. When the number of solutions in the external archive exceeds the preset capacity, pruning is performed according to the crowding distance, and solutions with larger crowding distances are retained.
[0044] S5: Repeat S2 to S4 until the termination condition is met, and then output the Pareto optimal solution set stored in the external file. The solution set corresponds to multiple equivalent combinations of aluminum electrolysis process parameters in the decision space. The termination condition is: reaching the preset maximum number of iterations, or the change of the optimal target value of the population for a consecutive preset number of generations is less than a preset threshold.
[0045] It should be noted that the optimization problem of aluminum electrolysis process parameters has significant multimodal characteristics. Different combinations of process parameters (decision variables) may achieve the same optimization target values of current efficiency and original aluminum quality. Therefore, this embodiment uses hypervolume (HV) and decision space hypervolume (HVx) as evaluation indicators. The HVx index is the extension of the HV index in the decision space, representing the hypervolume formed by the solution set in the five-dimensional decision space. The larger its value, the wider the distribution of the surface solution set and the stronger the diversity of the solution set. The median of the index parameters after 50 runs of various traditional multi-objective optimization algorithms and the method of this invention is compared in Table 1.
[0046] Table 1 Comparison of various algorithms in the aluminum electrolysis model
[0047] As can be seen from Table 1, compared with the multimodal multi-objective particle swarm optimization algorithm based on ring topology (MO-Ring-PSO-SCD), the non-dominated sorting genetic algorithm II based on decomposition strategy (DN-NSGA-II), and the non-dominated sorting genetic algorithm II with elitist strategy (NSGA-II), the method of this invention has the highest HVx, that is, its solution set is more widely distributed in the five-dimensional decision space and can successfully capture multiple equivalent Pareto solution set (PS) regions.
[0048] The diversity of process parameter combinations corresponding to the solution set obtained by the method of this invention is significantly improved. For example, the maximum and minimum values of aluminum fluoride feed amount in each equivalent process parameter combination differ by tens of units. Based on the adjustment of other process parameters in the solution set, different aluminum fluoride feed amounts can be selected in actual production. It can be understood that the same applies to other process parameters. In actual production, if the selection of a certain process parameter is constrained by the real environment, then the equivalent process parameter combination obtained by this invention can meet the actual needs.
[0049] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.
Claims
1. A method for optimizing parameters of an aluminum electrolysis process based on a dual space coordination strategy, characterized in that, Includes the following steps: S1: Initialize the population and external files of the honey badger algorithm, where each individual in the population represents a set of aluminum electrolysis process parameters; S2: Estimate the decision space density of the current population. Divide the decision space into multiple regions through hierarchical clustering. Calculate the density value of each individual in its region using the Gaussian kernel function. Introduce a local penalty mechanism to adjust the density value. Use the adjusted density value as the basis for evaluating population diversity. S3: Sort the individuals in the population in ascending order according to the adjusted density value obtained in S2 to obtain the density ranking of each individual. Calculate the selection factor of each individual based on the density ranking, and dynamically determine whether the individual enters the exploration or development stage according to the relationship between the selection factor and the preset threshold. Generate the offspring population according to the position update formula of the honey badger algorithm, wherein the current optimal individual position used in the position update formula of the honey badger algorithm is obtained from the external archive. S4: Merge the parent population with the offspring population generated in S3. For each individual in the merged population, calculate its density value in the decision space and its crowding distance in the objective space. Construct a bi-objective optimization problem with the goal of minimizing the density value and maximizing the crowding distance. Use this bi-objective optimization problem to perform non-dominated sorting on the merged population and sort them in descending order of crowding distance within the same Pareto front. Select a new generation of population from this population and store the non-dominated solutions in the new generation of population in the external archive. S5: Repeat S2 to S4 until the termination condition is met, and then output the Pareto optimal solution set stored in the external file. The solution set corresponds to multiple equivalent combinations of aluminum electrolysis process parameters in the decision space.
2. The method for optimizing aluminum electrolysis process parameters based on a dual-space coordination strategy according to claim 1, characterized in that, S2 includes: S21: The Ward hierarchical clustering method is used to determine the number of clusters by minimizing the intra-cluster variance. Divide the decision space into For each cluster, the clustering distance metric used is Euclidean distance, for two individuals and The distance calculation formula is: in, As a dimension of the decision space, and Individuals and In the Coordinates in dimensional space; S22: In each cluster Internally, the Gaussian kernel function is used to calculate individual... density value : in, For two individuals and The Euclidean distance between them For clusters The standard deviation of all individuals within the range; S23: Introduce a local penalty mechanism to adjust the density value according to the cluster size, and obtain the adjusted density value. : in, For the current cluster The number of individuals in This represents the total number of individuals in the population.
3. The method for optimizing aluminum electrolysis process parameters based on a dual-space cooperative strategy according to claim 1, characterized in that, S3 includes: S31: The adjusted density value obtained in S2 Sort in ascending order to obtain density ranking. ,in, Indicates minimum density. This indicates the highest density. Population size; S32: Calculate each individual Selection factors : in, A random number uniformly distributed in the range [0,1]. S33: If Once an individual enters the exploration phase, the new individual location is generated using the honey badger algorithm's location update formula for the mining phase. : like Once an individual enters the development phase, the honey badger algorithm's honey stage position update formula is used to generate a new individual position. : in, The current optimal individual position is obtained from the external archive. The parameters governing the honey badger's ability to obtain food adjust the individual's tracking intensity towards prey. The density factor characterizes the intensity of prey odor and is inversely proportional to the square of the distance between individuals. An adaptive decreasing factor that decreases with the number of iterations. For the current individual With the optimal individual The Euclidean distance between them , , A random number uniformly distributed in the range [0,1]. A marker used to control the search direction.
4. The method for optimizing aluminum electrolysis process parameters based on a dual-space cooperative strategy according to claim 3, characterized in that, The adaptive reduction factor is calculated using the following formula: in, As a preset constant, This represents the current iteration number. The maximum number of iterations, a flag controlling the search direction. Randomly generated as 1 or -1.
5. The method for optimizing aluminum electrolysis process parameters based on a dual-space cooperative strategy according to claim 1, characterized in that, In S4: Density values in the decision space The calculation formula is: in, For two individuals and The Euclidean distance between them For clusters The standard deviation of all individuals within the range; The formula for calculating the crowding distance in the target space is: in, For the target number, for the th A target value is set, and all individuals in the population are sorted in ascending order according to that target value. and Individuals The target value of two adjacent individuals in this sequence, and These are the maximum and minimum values of the target, respectively. The bi-objective optimization problem is expressed as: in, Minimize to enhance decision space diversity. Minimize to enhance the diversity of the target space.
6. The method for optimizing aluminum electrolysis process parameters based on a dual-space cooperative strategy according to claim 1, characterized in that, The external archive is updated as follows: the non-dominated solutions in the new generation population selected in S4 are compared with the existing solutions in the external archive. If the new solution is not dominated by any solution in the external archive, the new solution is added to the external archive, and the solutions dominated by the new solution in the external archive are deleted. When the number of solutions in the external archive exceeds the preset capacity, pruning is performed according to the crowding distance, and solutions with larger crowding distances are retained.
7. The method for optimizing aluminum electrolysis process parameters based on a dual-space cooperative strategy according to claim 1, characterized in that, The termination condition is: reaching the preset maximum number of iterations, or the change in the optimal target value of the population for a consecutive preset number of generations is less than a preset threshold.
8. The method for optimizing aluminum electrolysis process parameters based on a dual-space cooperative strategy according to claim 1, characterized in that, The aluminum electrolysis process parameters include electrolysis temperature, electrolyte level, aluminum level, cell voltage, and aluminum fluoride feed rate.