Process optimization method based on self-organization characteristics of non-ferrous metallurgy system

Abstract

The invention relates to the technical field of non-ferrous metallurgy, and particularly discloses a process optimization method based on self-organization characteristics of a non-ferrous metallurgy system, which discovers a self-organization phenomenon from industrial data, excavates a self-organization rule, and analyzes the self-organization characteristics in a material conversion process in the metallurgy system to optimize the self-organization characteristic of the non-ferrous metallurgy system. According to the method, the system operation state and the coupling relation between associated units are reasonably described, key indexes of the technological process are accurately estimated based on different operation states, and finally a set of operation optimization scheme is formed, so that consumption of industrial impurity removal displacers is reduced, method support is provided for dynamic-ordered and collaborative-continuous operation of the process, and the industrial impurity removal efficiency is improved. The problems that the traditional nonferrous metallurgy reaction process adopting a self-organizing cascade collaborative optimization method highly depends on historical production data, historical production samples have noise, are incomplete or are not high in quality, and then the accuracy and reliability of the model are affected are solved.

Description

Technical Field

[0001] The present application relates to the technical field of nonferrous metallurgy, and specifically discloses a process optimization method based on the self-organizing characteristics of a nonferrous metallurgical system. Background Art

[0002] Nonferrous metals usually exist in the form of alloys and need to be refined through specific process technologies. When refining nonferrous metals, the quality and purity of nonferrous metal products can be effectively improved by setting up a purification process, which protects the safety and stable operation of equipment. In the purification process of nonferrous metals, complex, open, irreversible, and far-from-equilibrium nonferrous metal metallurgical systems, such as the purification process in the hydrometallurgical zinc smelting system, have inherent self-organization.

[0003] In view of the process optimization of the self-organizing nonferrous metallurgical system, a new method for the optimization of associated unit processes applied in the manufacturing industry has been introduced in the prior art: the self-organizing cascade collaborative optimization method. Based on the theory of self-organizing criticality, it combines the system's mechanism knowledge with production data to construct a coupling relationship model of the reactor. Through the online estimation model of reaction efficiency, the reaction participation of each unit process is analyzed to obtain its actual operating efficiency. Finally, based on the coupling relationship between the associated unit processes and the cascade characteristics within the cascade system and the unit processes, a global optimization method for the system is constructed. The results show that this method can achieve global optimization of the cascade system, improve economic benefits compared to existing production conditions, and ensure the stability of the system.

[0004] However, in the process of optimizing the associated unit processes, the optimization model and optimization strategy of the self-organizing cascade collaborative optimization method are highly dependent on historical production data in highly complex or uncertain systems. After the state evolves, the category to which the current operating state belongs is obtained by evaluating the similarity between the process variables and the historical samples. These historical samples may be noisy, incomplete, or of low quality, and when there is a problem of low data quality, the accuracy and reliability of the model may be affected.

[0005] The present invention provides a process optimization method based on the self-organizing characteristics of the nonferrous metallurgical system to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to solve the problem that the traditional non-ferrous metallurgical reaction process using a self-organizing cascade collaborative optimization method is highly dependent on historical production data, and the historical production samples may be noisy, incomplete or of low quality, thereby affecting the accuracy and reliability of the model.

[0007] In order to achieve the above object, the basic scheme of the present invention provides a process optimization method based on the self-organizing characteristics of the nonferrous metallurgical system, comprising the following steps: Step A1: determine the associated reactors in the cascade system; Step A2: collecting process variables of all reactors in step A1, obtaining graph structure data between historical data and process variables through processing, and calculating reaction efficiency from historical data; Step A3: Based on the power law characteristics of the associated reactor, a power law distribution curve is calculated and fitted, and a critical value of the reaction efficiency is determined based on the fitted curve. The critical value is pre-divided into operation status labels of corresponding quantity types, and the calculated reaction efficiency is marked corresponding to the process variable collection time; Step A4: construct an operation status classification model, and use the process variables and the corresponding operation status labels as input to train the model; Step A5: Establish a reaction efficiency estimation model, combine the graph structure data, use the process variables divided according to the operation state labels as input training models, and obtain a reaction efficiency estimation sub-model 1 adapted to each operation state label; Step A6: Collect the process variables of the associated reactor in real time and input them into the operation status classification model to obtain the real-time operation status label, input the process variables and the corresponding real-time operation status label into the reaction efficiency estimation sub-model to calculate and obtain the estimated reaction efficiency, with the final impurity ion concentration at the reactor outlet meeting the standard and the lowest overall impurity removal and replacement agent addition amount as the goal, compare the estimated reaction efficiency through the particle optimization algorithm to obtain the optimal impurity removal rate of the main impurity ions at the reactor outlet, determine the optimal impurity removal and replacement agent addition amount, and complete the operation optimization.

[0008] Furthermore, during the process of performing step A1, it also includes determining the remaining reactors in the cascade system; During step A5, the reaction efficiency estimation model is trained based on the process variables of the remaining reactors as input to obtain a second reaction efficiency estimation sub-model adapted to each reactor; The process of performing step A6 also includes real-time collection of process variables of the remaining reactors, inputting the process variables of the remaining reactors into each reaction efficiency estimation sub-model 2 to calculate the estimated reaction efficiency of each reactor, with the goal of achieving the impurity ion concentration at the final reactor outlet and the lowest overall impurity removal and replacement agent addition amount, and comparing the estimated reaction efficiencies through the particle optimization algorithm to obtain the optimal impurity removal rate of the main impurity ions at the outlet of each reactor, determine the optimal impurity removal and replacement agent addition amount, and complete the operation optimization.

[0009] Furthermore, in step A2, the calculated historical data include impurity ion consumption, input amount of impurity removal replacement agent and corresponding impurity ion concentration change, and the reaction efficiency is the ratio of the theoretical amount of impurity removal replacement agent consumed by the main impurity ions to the actual amount.

[0010] Further, in step A2, the reaction efficiency is the ratio of the theoretical amount of impurity removal agent consumed by the main impurity ions to the actual amount, the cascade system is a cobalt removal purification subsystem of the hydrometallurgical zinc smelting system, cobalt ions are the main impurity ions at the reactor outlet, and the reaction efficiency γ1 of the associated reactor is calculated as follows: ; In the formula, is the theoretical amount of impurity removal agent consumed by the associated reactor to remove the remaining copper ions. is the theoretical amount of impurity removal agent consumed by the associated reactor to remove cobalt ions, is the actual amount of impurity removal agent added to the associated reactor, is the amount of impurity removal agent added to the associated reactor to remove hydrogen ions. is the amount of impurity removal agent consumed by the associated reactor to remove nickel ions; The reaction efficiency of the i-th reactor among the remaining reactors is γ i The formula is as follows: ; In the formula, is the theoretical amount of impurity removal agent consumed by the ith reactor to remove the remaining cobalt ions, is the actual amount of impurity removal agent added to the ith reactor, is the amount of impurity removal agent added to the ith reactor to remove hydrogen ions, is the amount of impurity removal agent added to the i-th reactor for removing nickel ions, i=2,3,….

[0011] Further, in step A3, the calculation formula for fitting the power law distribution curve is as follows: ; ; ; ; In the formula, d represents the distance from the point to the line, and D represents the sum of the squares of the distances from the point to the line; In the formula, X = lnc I , Y = lnf(c I ), lnf(c I )=ab lnc I , c I represents the average concentration of the main impurity ion concentration of the previous process in each cycle k, f(c I ) represents the number of operation state switching corresponding to the average concentration of main impurity ions in each previous process, and a and b are linear fitting coefficients; Wherein, m1, m2 and m3 represent the critical values ​​of reaction efficiency under different operating conditions.

[0012] Further, in step A4, the training process of the running status classification model is as follows: The process variables of the associated reactor and the corresponding operating status labels are input as the training set, and the process variables are used as samples; Autonomous sampling of the training set, randomly selecting samples and features; Decision trees are constructed recursively to partition the data according to the selected features until the number of decision trees reaches a preset value, thus forming a random forest. Establish a regression sub-decision tree and train it with the goal of minimizing the loss function; The training ends when the loss function reaches the minimum, the voting results of all decision trees are integrated, and the category with the most votes is selected as the final prediction result.

[0013] Furthermore, in step A4, the loss function of the established regression sub-decision tree is as follows: ; In the formula, x i represents the feature vector of the i-th sample, y i is the label value of the i-th sample, j represents the feature dimension, the j-th feature is selected as the segmentation criterion among all features, s represents the threshold used for segmentation on the j-th feature, I1 and I2 represent the two subsets divided according to feature j and segmentation point s, and when x is satisfied i,j The sample index set with ≤s is I1 (j,s), otherwise it is I2 (j,s), x i,j represents the jth eigenvalue of the i-th sample, d1 and d2 represent the optimal prediction values ​​selected from the two subsets respectively, and L represents the loss function obtained by selecting the optimal (j, s) and the optimal (d1, d2).

[0014] Furthermore, in step A5, the reaction efficiency estimation model includes a plurality of graph convolution modules and Transformer modules stacked in sequence, and a fully connected layer is established.

[0015] Further, in step A5, the reaction efficiency estimation model processes the input data as follows: The graph convolution module receives the graph structure data and encodes the graph structure data to obtain node features; The extracted node features are input into the Transformer for position encoding to introduce time information. The correlation between different time steps is calculated through the self-attention mechanism to obtain the spatiotemporal features of the nodes, and finally sent to the fully connected layer for integration to obtain the reaction efficiency prediction results.

[0016] The principle and effect of this scheme are: 1. Compared with the prior art, the present invention discovers self-organization phenomena and explores self-organization laws from industrial data. By analyzing the self-organization characteristics of the material transformation process in the metallurgical system, the coupling relationship between the system operation state and the related units is reasonably described. Based on different operation states, the key indicators of the process flow are accurately estimated, and finally a set of operation optimization schemes are formed to reduce the consumption of industrial impurity removal and replacement agents, providing method support for the dynamic, orderly, coordinated and continuous operation of the process.

[0017] 2. Compared with the prior art, the present invention constructs an operation status classification model based on the random forest algorithm, and constructs a reaction efficiency estimation model based on the operation status classification model. By estimating the reaction efficiency of each reactor online, and then setting the optimization target as the lowest reaction material addition amount and the outlet impurity concentration meeting the standard based on the particle optimization algorithm, combined with the reaction efficiency estimation results of each reactor, the impurity removal task of each reactor is reallocated to obtain the optimized reaction material addition amount, which solves the problem that the traditional non-ferrous metallurgical reaction process using a self-organizing cascade collaborative optimization method is highly dependent on historical production data, and the historical production samples may be noisy, incomplete or of low quality, thereby affecting the accuracy and reliability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A schematic diagram of the purification and impurity removal process proposed in the embodiment of the present application is shown; Figure 2 A schematic diagram showing the relationship between process variables proposed in the embodiment of the present application; Figure 3 A flow chart of a process optimization method based on the self-organizing characteristics of a nonferrous metallurgical system proposed in an embodiment of the present application is shown; Figure 4 A power law distribution curve fitting effect diagram proposed in an embodiment of the present application is shown; Figure 5 A schematic diagram showing the range of impurity removal rates of the five reactors in the stage II process proposed in the embodiment of the present application; Figure 6 The optimal setting curve diagram of impurity removal rate of each reactor proposed in the embodiment of the present application is shown; Figure 7A comparison chart of the amount of impurity removal and replacement agent added before and after optimization proposed in the examples of the present application is shown. DETAILED DESCRIPTION

[0020] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0021] The following is an example of the process optimization method based on the self-organizing characteristics of nonferrous metallurgical systems provided by the present invention being applied to the purification and impurity removal process of a zinc smelter, specifically applied to the cascade system of the hydrometallurgical zinc smelting system: the cobalt removal purification subsystem.

[0022] like Figure 1 As shown, the cascade system realizes the purification of the main impurity ions at the reactor outlet through the purification and impurity removal process, which is divided into the first stage impurity removal process and the second stage impurity removal process which are carried out in sequence. The equipment includes 7 reactors and 2 concentrators which are connected in sequence and arranged in a stepped manner, and the entrance of each continuous reactor is located at the top. Specifically, the first stage impurity removal process includes the first stage reactor No. 1, the second stage reactor No. 2 and the first stage concentrator which are connected in series in sequence. The second stage impurity removal process includes the second stage reactor No. 1, the second stage reactor No. 2...the second stage reactor No. 5 and the second stage concentrator which are connected in series in sequence, and the second stage reactor No. 1 is connected in series with the first stage concentrator.

[0023] The cascade system is open to the outside world, and its internal operating state is affected by temperature, air pressure, reactant feed rate, etc. Impurity ions overflowing from stage I impurity removal process to stage II impurity removal process mainly affect the internal reaction of stage II reactor No. 1, and stage II reactor No. 1 is associated with two consecutive process steps of stage I impurity removal process and stage II impurity removal process. Therefore, in this embodiment, stage II reactor No. 1 is used as the associated reactor, in which cobalt ions are the main impurity ions at the outlet of each reactor.

[0024] Since the purification and impurity removal process is a process in which the concentration of impurity ions is gradually reduced, according to the method provided by the present invention, the amount of impurity removal agent added to each reactor in the entire process is used as the final optimization target. By reasonably setting the impurity removal percentage of each reactor, the input of impurity removal agent is reduced, thereby saving process raw materials and improving process efficiency. Figure 3 As shown, the process optimization method based on the self-organizing characteristics of the nonferrous metallurgical system specifically includes the following steps: Step A1: Determine the associated reactors in the cascade system. In this embodiment, it is also necessary to determine the remaining reactors in the same process step.

[0025] Specifically, the No. 1 reactor of Section II is selected as the associated reactor, and the remaining No. 2 reactor of Section II to No. 5 reactor of Section II in the impurity removal process of Section II are selected.

[0026] Step A2: The process variables of all reactors in step A1 are collected in real time, and the graph structure data between the historical data and the process variables is obtained through processing, and the reaction efficiency of each reactor is calculated based on the historical data.

[0027] The collection of process variables is carried out at equal time intervals of minutes. The collected process variables include impurity ion concentration, temperature, flow rate, liquid level, potential, and feeder material quantity. The collected process variables are used as historical process variables and are historical data. In this embodiment, all process variables collected in this cycle are sorted as a group of data to be preprocessed with a cycle of 120 minutes. After the following preprocessing operations such as missing value screening, feature scaling, and calculation of mean variance data standardization, historical data are obtained. The obtained historical data include impurity ion consumption, the input amount of impurity removal substitution agent, and the corresponding impurity ion concentration change.

[0028] The missing value screening process for process variables includes: deleting the missing values ​​of cobalt in the copper removal overflow, and obtaining the index of the row without missing values; The characteristic scaling process for process variables includes: scaling the units of characteristic variable concentration and potential for subsequent calculations; The process of calculating the mean variance and standardizing the process variables includes: The mean represents the average value of a set of data and is defined as: , μ is the sample mean, N is the total number of samples, x i is the ith data point of a process variable.

[0029] Variance measures the degree of dispersion of data and is defined as: , σ 2 Represents the population variance.

[0030] Data standardization uses Z-score standardization. The mean of the standardized data is 0 and the variance is 1. The calculation formula is: ; Through these calculations, the central tendency and distribution of the data can be obtained, providing high-quality data input for data analysis and machine learning.

[0031] For the calculation of graph structure data between process variables, it is necessary to first determine the relationship between process variables, such as Figure 2 As shown, Figure 2 The specific matters of the process variables collected are shown in Table 1 below.

[0032] Table 1

[0033] Combining Table 1 and Figure 2 The relationships between the obtained process variables include: the concentration of impurity ions and the amount of feeder material will affect the reaction efficiency, and the change of reaction efficiency will also affect the working state of the subsequent reactor. The operating conditions (such as temperature and pH) have a direct impact on the reaction rate and the removal effect of impurity ions. The material flow rate determines the amount of reactants in each reactor, which will affect the equilibrium state of the entire process.

[0034] After determining the relationship between process variables, the graph structure data between process variables is constructed as follows: After preprocessing the data such as screening, cleaning and standardization, the relationship (edge) between variables is established by combining empirical knowledge and data mining techniques, and then converted into an adjacency matrix or adjacency list.

[0035] In the graph structure, nodes and edges are usually used to represent the relationship between process variables. A graph can be represented as: (G=(V,E) ) Where V is the set of nodes and E is the set of directed edges connecting the nodes.

[0036] In this embodiment, the obtained graph structure data is the adjacency matrix of the graph, and the expression of the adjacency matrix of the graph is as follows: ; ; In the formula, W represents the adjacency matrix, which is a square matrix of size n×n, n is the number of variable nodes, and w ij is an element of the adjacency matrix, representing the connection relationship between node i and node j, v ij is the directed edge connecting node i and node j, v ij ∈E indicates that there is a directed edge from node i to node j. In this case, w ij =1, otherwise w ij =0.

[0037] The adjacency matrix W represents the direction of information transmission between node variables and can be used to guide information aggregation in the calculation process of the graph convolutional network, thereby improving the model's ability to learn variable relationships.

[0038] In this embodiment, the ratio of the theoretical amount of impurity removal agent consumed by the main impurity ions to the actual amount is used as the reaction efficiency, and the reaction efficiency obtained by calculation describes the operating state of the reactor. Specifically, the calculation formula for the reaction efficiency of the associated reactor is as follows: ; Where γ1 is the reaction efficiency of the associated reactor, is the theoretical amount of impurity removal agent consumed by the associated reactor to remove the remaining copper ions. is the theoretical amount of impurity removal agent consumed by the associated reactor to remove cobalt ions, is the actual amount of impurity removal agent added to the associated reactor, is the amount of impurity removal agent added to the associated reactor to remove hydrogen ions. It is the amount of impurity removal agent added to the associated reactor to remove nickel ions.

[0039] The reaction efficiency calculation formula for the impurity removal stage II reactor No. 2 to stage II reactor No. 5 is as follows: ; In the formula, γ i is the reaction efficiency of the ith reactor in the current reaction process, i.e., the impurity removal II process. is the theoretical amount of impurity removal agent consumed by the ith reactor to remove the remaining cobalt ions, is the actual amount of impurity removal agent added to the ith reactor, is the amount of impurity removal agent added to the ith reactor for removing hydrogen ions, is the amount of impurity removal agent added to the i-th reactor for removing nickel ions. In this embodiment, i=2, 3, 4, 5.

[0040] Step A3: Based on the power-law characteristics of the associated reactor, the power-law distribution curve is calculated and fitted, and the critical value of the reaction efficiency is determined based on the fitted curve. The critical value is pre-divided into operation status labels of corresponding quantity types, and the calculated reaction efficiency is marked corresponding to the process variable collection time.

[0041] Specifically, the definition of the critical value is based on the degree of fit of the fitted power-law distribution curve to the original data.

[0042] For the calculation of the power-law distribution curve, the ultimate goal is to fit the power-law distribution curve. The fitting effect is based on the minimum sum of the squares of the distances from the point to the line. All possible values ​​are traversed to assign values ​​to the operating state to obtain the final fitting curve. The calculation formula for fitting the power-law distribution curve is as follows: ; ; ; ; In the formula, d represents the distance from the point to the line, and D represents the sum of the squares of the distances from the point to the line; In the formula, X = lnc I , Y = lnf(c I ), lnf(c I )=ab lnc I , c I It represents the median of the concentration of the main impurity ions in the previous process of each cycle, that is, stage I. The median of the concentration of the main impurity ions in stage I in each cycle is regarded as the average concentration of the main impurity ions in stage I of the cycle. I ) represents the number of operation state switching corresponding to the average concentration of the main impurity ions in each section I, and a and b are linear fitting coefficients. Among them, the main impurity ion concentration of the inlet section I is collected once every 2 hours, k unit time is taken as a cycle, and the number of operation state switching corresponding to the main impurity ion concentration of section I in each cycle is calculated; Wherein, m1, m2 and m3 represent the critical values ​​of reaction efficiency under different operating conditions.

[0043] Finally, based on the calculation of the power-law distribution curve, it is determined that m1=0.24, m2=0.35, and m3=0.65. The specific categories of the operating states divided by the fitted power-law distribution curve are shown in Table 2. The fitting effect of the power-law distribution curve is shown in Table 2. Figure 4 shown.

[0044] Table 2

[0045] like Figure 4 As shown, the fitting line can be approximated as a monotonically decreasing straight line, proving that the associated reactor has a power law characteristic. Figure 4 It was concluded that the higher the average concentration of the main impurity ions in stage I, the fewer times the operating state was switched. When the concentration of the main impurity ions in stage I was high, the reaction competed for a large amount of impurity removal and replacement agents, causing the operating state to remain in the third and second types of operating states, and the number of switches was less than that at other concentrations.

[0046] According to the calculated fitted power law distribution curve, the historical operating status of the associated reactors is divided into the following four types according to the division boundaries, corresponding to four different ranges of impurity removal displacement agent utilization: The first type of operation state: the reaction efficiency is in the range of (0, m1), indicating that a very small amount of impurity removal agent added to the reactor can effectively participate in the impurity removal reaction; The second type of operation state: the reaction efficiency is within the range of (m1, m2), indicating that a small amount of impurity removal agent added to the reactor can effectively participate in the impurity removal reaction; The third type of operation state: the reaction efficiency is within the range of (m2, m3), indicating that most of the impurity removal agents added to the reactor can effectively participate in the impurity removal reaction; The fourth type of operating state: the reaction efficiency is within the range of (m3,1), indicating that almost all the impurity removal agents added to the reactor can effectively participate in the impurity removal reaction.

[0047] Step A4: Construct an operation status classification model, and use the associated reactor process variables and the corresponding operation status labels as input to train the model.

[0048] Specifically, the operation status classification model is constructed based on the random forest algorithm. The operation steps of the constructed operation status classification model are as follows: Data preparation: Prepare data sets for training and testing, which are divided into training sets and test sets. The data sets include process variables and corresponding operating status labels; Randomly select samples and features: autonomously sample the training set, randomly select a sample subset from the training set with replacement sampling, repeat the sampling multiple times, create a new dataset of the same size as the original dataset, and randomly extract the features of the original dataset to form a feature subset; Constructing decision trees: Decision trees are constructed recursively, and the data is divided according to the selected features until the number of decision trees reaches the preset value, thus forming a random forest. Then the sample subset and feature subset are used as the training set of the decision tree, and the training is carried out with the goal of minimizing the loss function. By randomly extracting features and splitting the optimal segmentation vector x j And the split point S, the loss function of the regression sub-decision tree is established as follows: ; In the formula, x i Represents the feature vector of the i-th sample, that is, the input data, y i is the label value of the i-th sample, j represents the feature dimension (index), that is, the j-th feature is selected as the segmentation criterion among all features, s represents the threshold used when segmenting on the j-th feature, I1 and I2 represent the two subsets divided according to feature j and segmentation point s, specifically: when x is satisfied i,j The sample index set with ≤s is I1 (j,s), otherwise it is I2 (j,s), x i,j represents the jth eigenvalue of the i-th sample, d1 and d2 represent the optimal prediction values ​​selected from the two subsets respectively, and L represents the loss function obtained by selecting the optimal (j, s) and the optimal (d1, d2).

[0049] Integrated decision tree: When the loss function reaches the minimum, the training ends, the voting results of all decision trees are integrated, and the category with the most votes is selected as the final prediction result.

[0050] Evaluate the model: Use the test set to evaluate the performance of the trained random forest model and calculate the accuracy.

[0051] In the present embodiment, offline training of the operating status classification model is performed only for the associated reactor: in this training, after the process variables and operating status labels of each time period of the associated reactor are input into the operating status classification model, the random forest algorithm is used to classify them according to the established critical values, and the operating status categories are divided according to the reaction efficiency range they fall into, so that they are matched to the operating status that meets the operating status label.

[0052] Step A5: Establish a reaction efficiency estimation model, combine the graph structure data, use the process variables of the associated reactors divided according to the operating status labels as input training models, and obtain a reaction efficiency estimation sub-model 1 adapted to each reactor.

[0053] In this embodiment, the reaction efficiency estimation model is further trained based on the process variables of the remaining reactors as inputs to obtain a second reaction efficiency estimation sub-model adapted to each reactor.

[0054] Specifically, the reaction efficiency estimation model is built based on the graph convolutional network (GCN) and the Transformer model, including several graph convolution modules and Transformer modules stacked in sequence, and a fully connected layer is established. The reaction efficiency estimation model is also equipped with the Huber loss function and the Adam optimizer.

[0055] The graph structured data constructed according to the interrelationships between process flows and process variables are taken as input, and an adjacency matrix is ​​generated to represent the spatial associations between process variables.

[0056] The topological structure relationship between variables is learned through GCN (graph convolution network). In each layer of graph convolution, node v i New Features It is obtained by weighted averaging the features of its neighboring nodes, expressed as: ; Where N(i) represents the node x i The set of neighbor nodes, c i is node x i The normalization constant, W (l) is the weight matrix of the lth layer, and σ is the activation function.

[0057] By stacking multiple graph convolutional layers, the network can extract higher-level node features and learn more effective graph structure information.

[0058] The Transformer module is responsible for capturing time series features, introducing time information through position encoding, and using the self-attention mechanism to calculate the correlation between different time steps to extract high-quality spatiotemporal features. The calculation formula of the self-attention mechanism of the Transformer module is as follows: ; Where Q, K, and V are projections of the input data, which are query vector, key vector, and value vector, respectively. k Represents the dimension of the key vector, which Self-Attention uses to calculate the importance of different time steps.

[0059] After joint feature extraction by GCN and Transformer, the data enters the fully connected layer for prediction, and the Huber loss function is used for error calculation. This loss function combines the advantages of mean square error (MSE) and absolute error (MAE) to improve the robustness to outliers, thereby preventing the model from being affected by extreme data, and finally obtaining the target prediction result. The calculation formula of the Huber loss function is: ; Where a is the error between the true value and the predicted value, and δ is a hyperparameter that determines the threshold of the error range.

[0060] In addition, the reaction efficiency estimation model also uses the Adam optimizer to update parameters and adjust weights through adaptive learning rate to improve training efficiency and convergence speed. The specific update process is as follows: ; In the formula, θ t is the parameter after the tth iteration, is the parameter to be optimized, θ t-1 is the parameter of the previous iteration, α is the learning rate, which determines the step size of each parameter update, m t is the exponentially weighted moving average of the current gradient, v t It is the exponentially weighted moving average of the current gradient square, and ε is a very small number used to prevent division by zero errors.

[0061] The entire training process is continuously iterated and optimized, so that the reaction efficiency estimation model can eventually accurately estimate the reaction efficiency prediction results of each reactor: reaction efficiency γ, providing an optimization basis for industrial production, thereby improving production efficiency and resource utilization.

[0062] As can be seen from step A4, only the associated reactor is marked with an operating status label, so the offline training of the reaction efficiency estimation model includes two parts: Based on offline training of historical data of the associated reactor, the operating state of the associated reactor is divided based on the operating state classification model, and the historical data of the associated reactor under different operating states are used as input to perform offline training for the associated reactor, so as to obtain a reaction efficiency estimation sub-model 1 for different operating state categories of the associated reactor; Based on the offline training of the No. 2 reactor of Section II to the No. 5 reactor of Section II, the historical data of the No. 2 reactor of Section II to the No. 5 reactor of Section II are respectively used as input to carry out offline training for the No. 2 reactor of Section II to the No. 5 reactor of Section II, and the reaction efficiency estimation sub-model 2 for different reactors is obtained; Specifically, the offline training of the reaction efficiency estimation model includes the following steps: After inputting the graph structure data obtained in step 2 into the reaction efficiency estimation model, the original input variables of the associated reactor obtained through the graph convolution module contain 7 features, the number of input variable features of the No. 2 reactor in section II is the same as the number of process variables of the associated reactor, the original input variables of the No. 3 reactor in section II contain 6 features, and the original input variables of the No. 4 reactor in section II and the No. 5 reactor in section II contain 5 features.

[0063] A total of 1099 sets of valid data were obtained for training and testing of the associated reactors, and 2325, 2052, 1708 and 1442 sets of valid data were obtained for training and testing of the reactor No. 2 to reactor No. 5 of section II, respectively. The reaction efficiency was calculated to label the effective data with the utilization rate of the impurity removal agent. In order to reduce the impact of outliers on model training, the mean of every 15 data of each group of 120 data was taken. After pre-processing of the data and division of the data set, the effective process data was integrated into the estimation model.

[0064] The reaction efficiency estimation model is trained with four operating states of the associated reactor. The characteristic dimension of the input vector of the GCN-TS model is 7, and the dimension of the output vector is 1. Through repeated experiments, the connection status of the graph network nodes is defined as [[1,2,3,0,1,2,3,4,6],[3,3,4,5,5,5,5,5,5]], the batch size is 32, the number of iterations is 100, the number of nodes in the graph is the characteristic dimension of the input variable, the characteristic dimension of each node is set to 8, the number of attention heads is set to 8, and the number of stacking layers is set to 5.

[0065] The reaction efficiency estimation models are trained for reactors 2 to 5 of the impurity removal stage II. The feature dimensions of the input vectors of the GCN-TS model are 7, 6, 5, and 5, respectively, and the dimensions of the output vectors are all 1. The connection of the graph network nodes is defined as [[1,2,3,0,1,2,3,6],[4,4,4,5,5,5,5,5]], [[1,1,2,0,5,2],[2,4,3,4,4,4]], [[1,1,0,4],[2,3,3,3]], and [[1,1,0,4],[2,3,3,3]]. The other hyperparameters are set the same as the associated reactors.

[0066] At this point, the nonlinear relationship between process variables and reaction efficiency is constructed based on historical data, and the offline training of the reaction efficiency estimation model is completed.

[0067] Step A6: Collect the process variables of the associated reactor in real time and input them into the operation status classification model to obtain the real-time operation status label, input the process variables and the corresponding real-time operation status label into the reaction efficiency estimation sub-model to calculate and obtain the estimated reaction efficiency, with the final reactor outlet impurity ion concentration reaching the standard and the lowest overall impurity removal and replacement agent addition amount as the goal, compare the estimated reaction efficiency through the particle optimization algorithm to obtain the optimal impurity removal rate of the reactor cobalt ion, determine the optimal impurity removal and replacement agent addition amount, and complete the operation optimization of the associated reactor.

[0068] In this embodiment, the real-time collection of process variables of the remaining reactors is also included, and the process variables of the remaining reactors are input into each reaction efficiency estimation sub-model 2 to calculate the estimated reaction efficiency of each reactor, with the goal of reaching the impurity ion concentration at the final reactor outlet and the lowest overall impurity removal and replacement agent addition amount, the particle optimization algorithm is used to compare the estimated reaction efficiencies to obtain the optimal impurity removal rate of cobalt ions in each reactor, determine the optimal impurity removal and replacement agent addition amount, and complete the operation optimization of the remaining reactors.

[0069] Specifically, the steps include: First, the difference between the inlet and outlet impurity ion concentrations of each reactor at all times is calculated, and the ratio of the obtained difference to the inlet impurity ion concentration is calculated. Based on the statistical method, the impurity removal rate range of the five reactors in the impurity removal stage II process is obtained, as follows: Figure 5 shown.

[0070] Then, the trained reaction efficiency estimation sub-model corresponding to each reactor is used to estimate the reaction efficiency of the process variables collected in real time for each reactor input.

[0071] Finally, the estimated reaction efficiency was used to achieve the final impurity ion concentration at the reactor outlet and to minimize the overall amount of impurity removal agent added. The particle optimization algorithm was combined to obtain the optimal impurity removal rate of cobalt ions in each reactor and determine the optimal amount of impurity removal agent added.

[0072] Specifically, when the impurity removal rate of each reactor is t i When (i=1,2,3,4,5), combined with the reaction efficiency, the formula for estimating the amount of impurity removal agent added by all reactors to remove impurities is as follows: ; Where, T i is the amount of impurity removal agent added to each reactor, α is an empirical coefficient or proportional factor, M A The relative atomic mass of the main metal, M B is the relative atomic mass of the impurity metal, V is the reactor flow rate, t i is the impurity removal rate of each reactor, γ i The reaction efficiency of each reactor is estimated.

[0073] The goal of the particle optimization algorithm is to find an optimal impurity removal rate distribution so that the total impurity removal agent consumption is minimized. The goal and constraints can be expressed as: ; ; ; In the formula, c out is the set value of the impurity ion concentration at the final outlet of the impurity removal process, c outmax It represents the upper limit requirement of the impurity ion concentration at the final outlet of the impurity removal process, t imin and t imax They respectively represent the upper and lower limit requirements of the impurity removal rate of the i-th reactor in the impurity removal process.

[0074] The particle swarm optimization algorithm is a global optimization algorithm that simulates the foraging behavior of bird flocks. Each particle represents a possible solution, that is, a removal rate distribution scheme for a reactor. The particle swarm optimization algorithm adjusts the position and speed of particles through continuous iteration to make them close to the optimal solution. The process includes the following steps: Input the particle swarm and initialize the particle swarm parameters. The input data includes the particle swarm size, particle dimension, number of iterations, inertia weight, learning factor and iteration step range; Randomly initialize the position and velocity of each particle; Iteratively update the speed and position of each particle, calculate the fitness value of each particle, compare and update the individual historical optimal fitness value and position of each particle, compare and update the historical optimal fitness value and position of the group, and update other accommodations accordingly, such as inertia weight, number of iterations, etc.; Iteratively obtain the individual's best historical position, the group's best historical position, the individual's best historical fitness value, and the group's best historical fitness value; When the maximum number of iterations is reached or the minimum difference in fitness between two iterations is reached, the iteration stops, the optimal solution obtained by the iteration is output, and the optimization ends; Time intervals were randomly selected in the test set, and the optimal setting curves of the impurity removal rate of each reactor obtained using the operation optimization strategy based on the estimation of reactor reaction efficiency based on self-organization characteristic analysis are shown as follows Figure 6 shown.

[0075] The comparison results of the amount of impurity removal agent added before and after optimization are as follows Figure 7 As shown by Figure 7 It was further concluded that allocating more impurity removal tasks to associated reactors with high reaction efficiency can reduce the workload of impurity removal in subsequent reactors No. 2 to No. 5 of stage II, which is helpful to improve the operating efficiency of the process. By calculating the estimated total amount of impurity removal agent added after optimization and comparing it with the amount of impurity removal agent added in the actual production process, it can be obtained that the proposed strategy can reduce the consumption of impurity removal agent and reduce the waste of industrial raw materials.

[0076] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modifications to equivalent embodiments of equivalent changes by using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A process optimization method based on the self-organizing characteristics of nonferrous metallurgical systems, characterized in that: The steps include: Step A1: Determine the associated reactors in the cascade system; Step A2: collecting process variables of all reactors in step A1, obtaining graph structure data between historical data and process variables through processing, and calculating reaction efficiency from historical data; Step A3: Based on the power law characteristics of the associated reactor, a power law distribution curve is calculated and fitted, and a critical value of the reaction efficiency is determined based on the fitted curve. The critical value is pre-divided into operation status labels of corresponding quantity types, and the calculated reaction efficiency is marked corresponding to the process variable collection time; Step A4: construct an operation status classification model, and use the process variables and the corresponding operation status labels as input to train the model; Step A5: Establish a reaction efficiency estimation model, combine the graph structure data, use the process variables divided according to the operation state labels as input training models, and obtain a reaction efficiency estimation sub-model 1 adapted to each operation state label; Step A6: Collect the process variables of the associated reactor in real time and input them into the operation status classification model to obtain the real-time operation status label, input the process variables and the corresponding real-time operation status label into the reaction efficiency estimation sub-model to calculate and obtain the estimated reaction efficiency, with the final impurity ion concentration at the reactor outlet meeting the standard and the lowest overall impurity removal and replacement agent addition amount as the goal, compare the estimated reaction efficiency through the particle optimization algorithm to obtain the optimal impurity removal rate of the main impurity ions at the reactor outlet, determine the optimal impurity removal and replacement agent addition amount, and complete the operation optimization.

2. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 1 is characterized in that: The process of performing step A1 also includes determining the remaining reactors in the cascade system; During step A5, the reaction efficiency estimation model is trained based on the process variables of the remaining reactors as input to obtain a second reaction efficiency estimation sub-model adapted to each reactor; The process of performing step A6 also includes real-time collection of process variables of the remaining reactors, inputting the process variables of the remaining reactors into each reaction efficiency estimation sub-model 2 to calculate the estimated reaction efficiency of each reactor, with the goal of achieving the impurity ion concentration at the final reactor outlet and the lowest overall impurity removal and replacement agent addition amount, and comparing the estimated reaction efficiencies through the particle optimization algorithm to obtain the optimal impurity removal rate of the main impurity ions at the outlet of each reactor, determine the optimal impurity removal and replacement agent addition amount, and complete the operation optimization.

3. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 1 is characterized in that: In step A2, the calculated historical data include impurity ion consumption, input amount of impurity removal replacement agent and corresponding impurity ion concentration change, and the reaction efficiency is the ratio of the theoretical amount of impurity removal replacement agent consumed by the main impurity ions to the actual amount.

4. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 2 is characterized in that: In step A2, the reaction efficiency is the ratio of the theoretical amount of impurity removal agent consumed by the main impurity ions to the actual amount. The cascade system is a cobalt removal purification subsystem of the hydrometallurgical zinc smelting system. Cobalt ions are the main impurity ions at the reactor outlet. The reaction efficiency γ1 of the associated reactor is calculated as follows: ； In the formula, is the theoretical amount of impurity removal agent consumed by the associated reactor to remove the remaining copper ions. is the theoretical amount of impurity removal agent consumed by the associated reactor to remove cobalt ions, is the actual amount of impurity removal agent added to the associated reactor, is the amount of impurity removal agent added to the associated reactor to remove hydrogen ions. is the amount of impurity removal agent consumed by the associated reactor to remove nickel ions; The reaction efficiency of the i-th reactor among the remaining reactors is γ i The formula is as follows: ； In the formula, is the theoretical amount of impurity removal agent consumed by the ith reactor to remove the remaining cobalt ions, is the actual amount of impurity removal agent added to the ith reactor, is the amount of impurity removal agent added to the ith reactor to remove hydrogen ions, is the amount of impurity removal agent added to the i-th reactor for removing nickel ions, i=2,3,….

5. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 1 is characterized in that: In step A3, the calculation formula for fitting the power law distribution curve is as follows: ； ； ； ； In the formula, d represents the distance from the point to the line, and D represents the sum of the squares of the distances from the point to the line; Where, X = lnc I , Y = lnf(c I ), lnf(c I )=ab lnc I , c I represents the average concentration of the main impurity ion concentration of the previous process in each cycle k, f(c I ) represents the number of operation state switching corresponding to the average concentration of main impurity ions in each previous process, and a and b are linear fitting coefficients; Wherein, m1, m2 and m3 represent the critical values ​​of reaction efficiency under different operating conditions.

6. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 1 is characterized in that: In step A4, the training process of the running status classification model is as follows: The process variables of the associated reactor and the corresponding operating status labels are input as the training set, and the process variables are used as samples; Autonomous sampling of the training set, randomly selecting samples and features; Decision trees are constructed recursively to partition the data according to the selected features until the number of decision trees reaches a preset value, thus forming a random forest. Establish a regression sub-decision tree and train it with the goal of minimizing the loss function; The training ends when the loss function reaches the minimum, the voting results of all decision trees are integrated, and the category with the most votes is selected as the final prediction result.

7. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 6 is characterized in that: In step A4, the loss function of the established regression sub-decision tree is as follows: ； In the formula, x i represents the feature vector of the i-th sample, y i is the label value of the i-th sample, j represents the feature dimension, the j-th feature is selected as the segmentation criterion among all features, s represents the threshold used for segmentation on the j-th feature, I1 and I2 represent the two subsets divided according to feature j and segmentation point s, and when x is satisfied i,j The sample index set with ≤s is I1 (j,s), otherwise it is I2 (j,s), x i,j represents the jth eigenvalue of the i-th sample, d1 and d2 represent the optimal prediction values ​​selected from the two subsets respectively, and L represents the loss function obtained by selecting the optimal (j, s) and the optimal (d1, d2).

8. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 1 is characterized in that: In step A5, the reaction efficiency estimation model includes a plurality of graph convolution modules and Transformer modules stacked in sequence, and a fully connected layer is established.

9. The process optimization method based on the self-organizing characteristics of nonferrous metallurgical system according to claim 8, characterized in that: In step A5, the reaction efficiency estimation model processes the input data as follows: The graph convolution module receives the graph structure data and encodes the graph structure data to obtain node features; The extracted node features are input into the Transformer for position encoding to introduce time information. The correlation between different time steps is calculated through the self-attention mechanism to obtain the spatiotemporal features of the nodes, and finally sent to the fully connected layer for integration to obtain the reaction efficiency prediction results.

Images