A data-driven approach to improving pellet compressive strength

Abstract

The present invention discloses a method for improving the compressive strength of pellets based on data-driven methods, and relates to the technical field of pellet production. The temperature parameters of each process section in the pellet production cooling process are obtained, the importance of each variable in the original data set is analyzed, and the variables that affect the temperature of the blast drying section, the temperature of the exhaust drying section, the temperature of the preheating section 1, the temperature of the preheating section 2, the temperature of the rotary kiln head, the temperature in the rotary kiln, the temperature at the kiln tail of the rotary kiln, the temperature of the ring cooling section 1, the temperature of the ring cooling section 2, the temperature of the ring cooling section 3, and the temperature of the ring cooling section 4 are selected to establish a pellet quality prediction model based on compressive strength; a data-driven model is derived based on a support vector machine proposed by statistical learning theory, and a genetic optimization algorithm is used to solve the problem. The pellet compressive strength is used as the optimization target, and the thermal parameters are used as decision variables to obtain the optimal thermal parameters. The wind temperature is controlled in the optimal state with the optimal thermal parameters as the target to ensure the stable progress of the entire production process, thereby improving and enhancing the compressive strength of the pellets.

Description

Technical Field

[0001] The present invention relates to the technical field of pellet production, and in particular to a method for improving the compressive strength of pellets based on data driving. Background Art

[0002] With the rapid development of the steel industry, the proportion of pellets used in blast furnace ironmaking has continued to increase, leading to the widespread application of grate-rotary kiln production systems. Determining appropriate thermal parameters is crucial to the success of the grate-rotary kiln process. These parameters not only influence the entire pelletizing reaction process but also, in turn, affect coal combustion and directly impact the compressive strength of the pellets. The pellet production process is complex, with various process and operating parameters coupled to each other. This creates uncertainties and makes it difficult to accurately assess pellet quality using mathematical models. Therefore, it is imperative to develop a thermal parameter optimization model and, based on this model, optimize the air temperature to ensure smooth operation of the entire production process, thereby enhancing and improving the compressive strength of the pellets while reducing coal consumption. Summary of the Invention

[0003] The purpose of the present invention is to provide a data-driven method for improving the compressive strength of pellets, so as to solve the problem that in the existing pellet production process, it is difficult to accurately control the temperature due to the complex process, and thus it is difficult to ensure the quality of the pellets and their compressive strength. Through modeling analysis and precise control of wind temperature, the purpose of improving the compressive strength of the pellets and reducing coal consumption can be achieved.

[0004] To solve the above technical problems, the present invention adopts the following technical solution: a data-driven method for improving the compressive strength of pellets, characterized by comprising the following steps:

[0005] S1. Obtain the thermal parameters of the pellet production process and preprocess the data. The preprocessing includes data cleaning and transformation, eliminating abnormal data, and obtaining the original data set;

[0006] S2. Analyze the importance of each variable in the original data set, delete the variables and their corresponding data that have no obvious effect on the compressive strength of the pellets, and establish a prediction model based on the 11 variables and their corresponding data, including the temperature of the blast drying section, the temperature of the exhaust drying section, the temperature of the preheating section 1, the temperature of the preheating section 2, the temperature of the rotary kiln head, the temperature of the rotary kiln middle, the temperature of the rotary kiln tail, the temperature of the ring cooling section 1, the temperature of the ring cooling section 2, the temperature of the ring cooling section 3, and the temperature of the ring cooling section 4;

[0007] S3, dividing the data processed in step S2 into a training data set and a test data set, applying an SVM model with an adaptively selected RBF kernel function, and building a prediction model based on the training set samples;

[0008] S4. Apply the SVM model with adaptively selected RBF kernel function, test the prediction model based on the test set samples, and detect the accuracy of the SVM model. The SVM model with the thermal parameters in the pellet production process as input and the pellet compressive strength as output is as follows:

[0009] y=f(x1,x2,...,x 11 )

[0010] x1 is the temperature of the blast drying section, x2 is the temperature of the exhaust drying section, x3 is the temperature of the preheating section 1, x4 is the temperature of the preheating section 2, x5 is the temperature of the rotary kiln head, x6 is the temperature of the rotary kiln middle, x7 is the temperature of the rotary kiln tail, x8 is the temperature of the ring cooling section 1, x9 is the temperature of the ring cooling section 2, x 10 The temperature of the three ring cooling sections, x 11 It is the temperature of the 4th stage of ring cooling;

[0011] S5. From the SVM model established in step S4, it can be seen that each set of thermal parameters corresponds to a pellet strength value. Given the expected pellet compressive strength value y * , so that the square value of the difference between the compressive strength of the pellets obtained by the neural network model under different thermal parameters is minimized, and the thermal parameters are continuously optimized through the optimization process, so that the thermal parameter value obtained when infinitely approaching the expected target value is the optimal solution, and each temperature parameter is regulated according to the optimal solution.

[0012] A further technical solution is to analyze the importance of each variable in the original data set in step S2. The specific process is as follows: use feature selection based on decision tree to determine the importance of each variable to the target by comparing the feature importance attributes; based on the obtained relevant feature parameters, sort them according to feature importance and select the top 11 parameters.

[0013] A further technical solution is that the temperature range of the blast drying section is 130-200°C, the temperature range of the exhaust drying section is 330-380°C, the temperature range of preheating section 1 is 500-650°C, the temperature range of preheating section 2 is 900-1050°C, the temperature range of the rotary kiln head is 950-1150°C, the temperature range of the rotary kiln is 1100-1250°C, the temperature range of the rotary kiln tail is 950-1050°C, the temperature range of the ring cooling section 1 is 850-950°C, the temperature range of the ring cooling section 2 is 550-650°C, the temperature range of the ring cooling section 3 is 130-200°C, and the temperature range of the ring cooling section 4 is 50-150°C.

[0014] A further technical solution is that the specific process of step S5 is as follows:

[0015] S5-1. Establish a single-objective model to express the optimal solution for pellet compressive strength:

[0016] minJ=(f(x1,x2,...,x 11 )-y * ) 2

[0017]

[0018] Among them, x1 is the temperature of the blast drying section, x2 is the temperature of the exhaust drying section, x3 is the temperature of the preheating section 1, x4 is the temperature of the preheating section 2, x5 is the temperature of the rotary kiln head, x6 is the temperature of the rotary kiln middle, x7 is the temperature of the rotary kiln tail, x8 is the temperature of the ring cooling section 1, x9 is the temperature of the ring cooling section 2, x 10 The temperature of the three ring cooling sections and x 11 It is the temperature of the 4th stage of ring cooling;

[0019] S5-2, using genetic algorithm to solve the single objective model, by adjusting the blast drying section temperature x1, exhaust drying section temperature x2, preheating section 1 temperature x3, preheating section 2 temperature x4, rotary kiln head temperature x5, rotary kiln middle temperature x6, rotary kiln tail temperature x7, ring cooling section 1 temperature x8, ring cooling section 2 temperature x9, ring cooling section 3 temperature x 10 , Ring cooling 4-stage temperature x 11 The values ​​of these 11 controllable variables are adjusted to achieve the best fitness value and obtain the optimal solution for the compressive strength of the pellets.

[0020] A further technical solution is to use a genetic algorithm to solve the single-objective model as described in step S5-2. The specific process is as follows:

[0021] Step 1: Initialize the population, set the maximum number of evolutionary iterations T, randomly generate M individuals as the initial population P(0), set the crossover rate P_0, and set the mutation rate P_v;

[0022] Step 2: Determine the decision variables and constraints, that is, determine the individual phenotype and the solution space of the problem;

[0023] Step 3: Establish an optimization model and determine the type of objective function (whether to maximize or minimize the objective function) and its mathematical description form or quantification method;

[0024] Step 4: Select the chromosome encoding method to determine the individual's genotype and the search space of the genetic algorithm;

[0025] Step 5: Calculate the fitness of each individual in the population P(t) according to the objective function;

[0026] Step 6: Apply the selection operator to the population. The purpose of selection is to pass optimized individuals directly to the next generation or to generate new individuals through pairing and crossover and then pass them to the next generation. The selection operation is based on the fitness evaluation of individuals in the population.

[0027] Step 7: Apply the crossover operator to the population and perform a crossover operation on the two groups of chromosome individuals with a certain crossover probability;

[0028] Step 8: Apply the mutation operator to the population to change the gene values ​​at certain loci of the individual strings in the population; after the population P(t) undergoes selection, crossover, and mutation operations, the next generation population P(t+1) is obtained;

[0029] Step 9: If t = T, the individual with the maximum fitness obtained in the evolution process is output as the optimal solution and the calculation is terminated;

[0030] Here, the maximum number of evolutionary iterations T is set to 100, the number of individuals in the initial population P(0) is 40, the crossover rate P_0 is set to 0.9, the mutation rate P_v is set to 0.1, and the number of populations per generation is 1.

[0031] A further technical solution is that the specific process of detecting the accuracy of the SVM model in step S4 is as follows: Assume that y is the predicted value obtained from the model, y i is the actual value measured at the sensor, and y m are the average of the predicted and observed values, respectively, and N is the number of test data points;

[0032] Calculate the mean absolute percentage error MAPE respectively, the specific formula is:

[0033]

[0034] To obtain the score deviation FB, the specific formula is:

[0035]

[0036] Calculate the root mean square error RMSE, the specific formula is:

[0037]

[0038] To obtain the normalized mean square error NMSE, the specific formula is:

[0039]

[0040] Calculate the coefficient of determination R 2 , the specific formula is:

[0041]

[0042] The accuracy of the prediction results is judged based on the above calculation results.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows: by acquiring the thermal parameters in the pellet production process and then analyzing the importance of each variable in the original data set, the blast drying section temperature, exhaust drying section temperature, preheating section 1 temperature, preheating section 2 temperature, rotary kiln head temperature, rotary kiln middle temperature, rotary kiln tail temperature, ring cooling section 1 temperature, ring cooling section 2 temperature, ring cooling section 3 temperature, ring cooling section 4 temperature and pellet compressive strength that affect the pellet compressive strength are selected to establish a prediction model to avoid data redundancy; a data-driven model is obtained by adaptively selecting an SVM model with an RBF kernel function and solving it using a genetic optimization algorithm, with pellet compressive strength as the optimization target and thermal parameters as decision variables to obtain optimal thermal parameters, and then with the optimal thermal parameters as the target to control the wind temperature in the optimal state and ensure the stable progress of the entire production process, thereby improving and enhancing the pellet compressive strength, improving production quality and reducing coal consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Flowchart of the present invention.

[0045] Figure 2 Flowchart of the genetic algorithm in the present invention.

[0046] Figure 3 This is the curve diagram of the SVM model prediction results of the pellet compressive strength.

[0047] Figure 4 This is the comparison curve between the actual data of pellet compressive strength and the genetic algorithm optimization results. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] like Figure 1 As shown, a data-driven method for improving the compressive strength of pellets includes the following steps:

[0050] S1. Obtain the production data of a pelletizing plant, which is a daily sample from November 4, 2021 to November 4, 2022, including the temperature of the blast drying section, the temperature of the exhaust drying section, the temperature of preheating section 1, the temperature of preheating section 2, the temperature of the rotary kiln head, the temperature in the middle of the rotary kiln, the temperature of the rotary kiln tail, the temperature of ring cooling section 1, the temperature of ring cooling section 2, the temperature of ring cooling section 3, the temperature of ring cooling section 4, the thickness of the cloth, the wind speed of each fan, the compressive strength of the pellets and other thermal parameters related to pellet production and pellet quality parameters.

[0051] Preprocess the data, which includes data cleaning, transformation, and elimination of abnormal data to obtain the original data set.

[0052] S2. It is necessary to delete the data that has no obvious effect on the compressive strength of the pellets to solve the problem of data redundancy affecting the calculation speed and accuracy in the subsequent calculation process. This is done by gradually analyzing the importance of each variable.

[0053] Analyze the importance of each variable in the original data set. The characteristics of importance will affect the accuracy of the model and cause deviations. Use feature selection based on decision trees to determine the importance of each variable to the target by comparing the feature importance attributes. Based on the relevant feature parameters that have been obtained, sort the features according to their importance and select the top-ranked features. Delete the variables that have no obvious effect on the compressive strength of the pellets and their corresponding data, and obtain the 11 variables and their corresponding data, including the temperature of the blast drying section, the temperature of the exhaust drying section, the temperature of the preheating section 1, the temperature of the preheating section 2, the temperature of the rotary kiln head, the temperature of the rotary kiln in the kiln, the temperature of the rotary kiln tail, the temperature of the ring cooling section 1, the temperature of the ring cooling section 2, the temperature of the ring cooling section 3 and the temperature of the ring cooling section 4, to establish a prediction model.

[0054] S3, dividing the data processed in step S2 into a training data set and a test data set, applying the SVM model with adaptively selected RBF kernel function, and building a prediction model based on 7008 groups of samples in the training set;

[0055] S4. Applying the SVM model with adaptively selected RBF kernel function, the prediction model was tested based on 1752 test set samples, and the accuracy of the RBF model was tested. The RBF model with the thermal parameters of the pellet production process as input and the pellet compressive strength as output was obtained as follows:

[0056] y=f(x1,x2,...,x 11 )

[0057] Assume that y is the predicted value obtained from the model, y i is the actual value measured at the sensor, and y m are the average of the predicted and observed values, respectively, and N is the number of test data points;

[0058] Calculate the mean absolute percentage error MAPE respectively, the specific formula is:

[0059]

[0060] To obtain the score deviation FB, the specific formula is:

[0061]

[0062] Calculate the root mean square error RMSE, the specific formula is:

[0063]

[0064] To obtain the normalized mean square error NMSE, the specific formula is:

[0065]

[0066] Calculate the coefficient of determination R 2 , the specific formula is:

[0067]

[0068] The accuracy of the prediction results is judged based on the above calculation results.

[0069] S5. From the SVM model established in step S4, it can be seen that each set of thermal parameters corresponds to a pellet strength value. Given the expected pellet compressive strength value y * , so that the square value of the difference between the compressive strength of the pellets obtained by the neural network model under different thermal parameters is minimized, and the thermal parameters are continuously optimized through the optimization process, so that the thermal parameter values ​​obtained when infinitely approaching the expected target values ​​are the optimal solution. Among them, the temperature range of the blast drying section is 130-200℃, the temperature range of the exhaust drying section is 330-380℃, the temperature range of the preheating section 1 is 500-650℃, the temperature range of the preheating section 2 is 900-1050℃, the temperature range of the rotary kiln head is 950-1150℃, the temperature range of the rotary kiln is 1100-1250℃, the temperature range of the rotary kiln tail is 950-1050℃, the temperature range of the ring cooling section 1 is 850-950℃, the temperature range of the ring cooling section 2 is 550-650℃, the temperature range of the ring cooling section 3 is 130-200℃, and the temperature range of the ring cooling section 4 is 50-150℃; specifically:

[0070] S5-1. Establish a single-objective model to express the optimal solution for pellet compressive strength:

[0071] minJ=(f(x1,x2,...,x 11 )-y * ) 2

[0072]

[0073] Among them, x1 is the temperature of the blast drying section, x2 is the temperature of the exhaust drying section, x3 is the temperature of the preheating section 1, x4 is the temperature of the preheating section 2, x5 is the temperature of the rotary kiln head, x6 is the temperature of the rotary kiln middle, x7 is the temperature of the rotary kiln tail, x8 is the temperature of the ring cooling section 1, x9 is the temperature of the ring cooling section 2, x 10 The temperature of the three ring cooling sections and x 11 It is the temperature of the 4th stage of ring cooling;

[0074] S5-2, using genetic algorithm to solve the single objective model, by adjusting the blast drying section temperature x1, exhaust drying section temperature x2, preheating section 1 temperature x3, preheating section 2 temperature x4, rotary kiln head temperature x5, rotary kiln middle temperature x6, rotary kiln tail temperature x7, ring cooling section 1 temperature x8, ring cooling section 2 temperature x9, ring cooling section 3 temperature x 10 , Ring cooling 4-stage temperature x 11 The values ​​of these 11 controllable variables are adjusted to achieve the best fitness value and obtain the optimal solution for the compressive strength of the pellets.

[0075] Genetic algorithm is used to solve the single objective model. The specific process is as follows:

[0076] Step 1: Initialize the population, set the maximum number of evolutionary iterations T, randomly generate M individuals as the initial population P(0), set the crossover rate P_0, and set the mutation rate P_v.

[0077] Step 2: Determine the decision variables and constraints, that is, determine the individual phenotype and the solution space of the problem.

[0078] Step 3: Establish an optimization model and determine the type of objective function (whether to maximize or minimize the objective function) and its mathematical description form or quantification method.

[0079] Step 4: Select the chromosome encoding method to determine the individual's genotype and the search space of the genetic algorithm.

[0080] Step 5: Calculate the fitness of each individual in the population P(t) based on the objective function.

[0081] Step 6: Apply the selection operator to the population. The goal of selection is to pass optimized individuals directly to the next generation or to generate new individuals through pairing and crossover, which are then passed on to the next generation. The selection operation is based on the fitness evaluation of individuals in the population.

[0082] Step 7: Apply the crossover operator to the population and perform a crossover operation on the two groups of chromosome individuals with a certain crossover probability.

[0083] Step 8: Apply the mutation operator to the population, changing the gene values ​​at certain loci of the individual strings in the population. After the selection, crossover, and mutation operations, the population P(t) is converted into the next generation population P(t+1).

[0084] Step 9: If t = T, the individual with the maximum fitness obtained in the evolution process is output as the optimal solution and the calculation is terminated.

[0085] Here, the maximum number of evolutionary iterations T is set to 100, the number of individuals in the initial population P(0) is 40, the crossover rate P_0 is set to 0.9, the mutation rate P_v is set to 0.1, and the number of populations per generation is 1.

[0086] The genetic algorithm uses the prediction results of the pellet compressive strength prediction model to optimize 11 thermal parameters, setting a target compressive strength of 2650 N / pellet. When the absolute difference between the compressive strength and the target value is close to or equal to 0, the requirement is considered met and the corresponding thermal parameter combination is output. When the difference between the compressive strength and the target value is large, there is room for improvement in the pellet production conditions. In this case, the optimized parameters given by the genetic algorithm are used to guide pellet production, increasing the compressive strength to 2650 N / pellet. After optimizing the compressive strength of the pellets using this optimization strategy, the average pellet strength can be increased by approximately 3.2%.

[0087] The case selected 40 sets of historical experimental data and brought the experimental data into the genetic optimization model. Figure 4 The optimization results of 40 pieces of data by genetic algorithm were compared with the actual pellet compressive strength. The results showed that the compressive strength of the pellets after optimization was significantly improved.

[0088] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A data-driven method for improving the compressive strength of pellets, characterized in that The steps include: S1. Obtain the thermal parameters of the pellet production process and preprocess the data. The preprocessing includes data cleaning and transformation, eliminating abnormal data, and obtaining the original data set; S2. Analyze the importance of each variable in the original data set, delete the variables and their corresponding data that have no obvious effect on the compressive strength of the pellets, and establish a prediction model based on the 11 variables and their corresponding data, including the temperature of the blast drying section, the temperature of the exhaust drying section, the temperature of the preheating section 1, the temperature of the preheating section 2, the temperature of the rotary kiln head, the temperature of the rotary kiln middle, the temperature of the rotary kiln tail, the temperature of the ring cooling section 1, the temperature of the ring cooling section 2, the temperature of the ring cooling section 3, and the temperature of the ring cooling section 4; S3, the data processed in step S2 is divided into a training data set and a test data set, and an adaptive selection is applied Kernel function Model, build a prediction model based on the training set samples; S4. Application Adaptive Selection Kernel function Model, test the prediction model based on the test set samples, and detect The accuracy of the model is obtained by taking the thermal parameters of the pellet production process as input and the pellet compressive strength as output. The model is as follows: ； is the temperature of the blast drying section, is the temperature of the exhaust drying section, To preheat 1 stage temperature, To preheat 2 temperature, is the temperature of rotary kiln head, is the temperature in the rotary kiln, is the temperature at the end of the rotary kiln, The temperature of ring cooling stage 1, Ring cooling 2 stage temperature, 3-stage ring cooling temperature, It is the temperature of ring cooling 4 stages; S5, based on the data already established in step S4 The model shows that each set of thermal parameters corresponds to a pellet strength value. Given the expected pellet compressive strength value , so that the square value of the difference between the compressive strength of the pellets obtained by the neural network model under different thermal parameters is minimized, and the thermal parameters are continuously optimized through the optimization process, so that the thermal parameter value obtained when infinitely approaching the expected target value is the optimal solution, and each temperature parameter is regulated according to the optimal solution.

2. The method for improving the compressive strength of pellets based on data-driven method according to claim 1, characterized in that: The importance of each variable in the original data set is analyzed in step S2. The specific process is as follows: feature selection based on decision tree is used to compare the importance of features. Attributes are used to determine the importance of each variable to the target; based on the obtained relevant feature parameters, the top 11 parameters are selected according to the feature importance.

3. The method for improving the compressive strength of pellets based on data-driven methods according to claim 1, characterized in that: The temperature range of the blast drying section is 130-200°C, the temperature range of the exhaust drying section is 330-380°C, the temperature range of the preheating section 1 is 500-650°C, the temperature range of the preheating section 2 is 900-1050°C, the temperature range of the rotary kiln head is 950-1150°C, the temperature range of the rotary kiln middle is 1100-1250°C, the temperature range of the rotary kiln tail is 950-1050°C, the temperature range of the ring cooling section 1 is 850-950°C, the temperature range of the ring cooling section 2 is 550-650°C, the temperature range of the ring cooling section 3 is 130-200°C, and the temperature range of the ring cooling section 4 is 50-150°C.

4. The method for improving the compressive strength of pellets based on data-driven methods according to claim 1, characterized in that: The specific process of step S5 is as follows: S5-1. Establish a single-objective model to express the optimal solution for pellet compressive strength: ； in, is the temperature of the blast drying section, is the temperature of the exhaust drying section, To preheat 1 stage temperature, To preheat 2 temperature, is the temperature of rotary kiln head, is the temperature in the rotary kiln, is the temperature at the end of the rotary kiln, The temperature of ring cooling stage 1, Ring cooling 2 stage temperature, The temperature of the three ring cooling sections and It is the temperature of ring cooling 4 stages; S5-2, using genetic algorithm to solve the single objective model, by adjusting the temperature of the blast drying section , exhaust drying section temperature , preheat 1 stage temperature , preheat 2 temperature , rotary kiln head temperature , the temperature in the rotary kiln , rotary kiln tail temperature , ring cooling 1 stage temperature , ring cooling 2 stage temperature , ring cooling 3 stage temperature , 4-stage ring cooling temperature The values ​​of these 11 controllable variables are adjusted to achieve the best fitness value and obtain the optimal solution for the compressive strength of the pellets.

5. The method for improving the compressive strength of pellets based on data-driven method according to claim 4, characterized in that: In step S5-2, a genetic algorithm is used to solve the single-objective model. The specific process is as follows: Step 1: Initialize the population and set the maximum number of evolutionary iterations , randomly generated individuals as the initial group , set the crossover rate , set the mutation rate ; Step 2: Determine the decision variables and constraints, that is, determine the individual phenotype and the solution space of the problem; Step 3: Establish an optimization model and determine the type of objective function and its mathematical description form or quantification method; Step 4: Select the chromosome encoding method to determine the individual's genotype and the search space of the genetic algorithm; Step 5: Calculate the population based on the objective function The fitness of each individual in Step 6: Apply the selection operator to the population. The purpose of selection is to directly pass on the optimized individuals to the next generation or to generate new individuals through pairing and crossover and then pass them on to the next generation. The selection operation is based on the fitness evaluation of individuals in the population. Step 7: Apply the crossover operator to the population and perform a crossover operation on the two groups of chromosome individuals with a certain crossover probability; Step 8: Apply the mutation operator to the population to change the gene values ​​of certain loci of the individual strings in the population; After selection, crossover and mutation operations, the next generation population is obtained ; Step 9: If , the individual with the maximum fitness obtained in the evolution process is output as the optimal solution and the calculation is terminated; Here, set the maximum number of evolution iterations is 100, the initial population The number of individuals is 40, and the crossover rate is set Set the mutation rate to 0.9 is 0.1, and the population size in each generation is 1.

6. The method for improving the compressive strength of pellets based on data-driven method according to claim 4, characterized in that: The detection in step S4 The specific process of the model accuracy is as follows: Assume are the predicted values ​​obtained from the model, is the actual value measured at the sensor, and are the averages of the predicted and observed values, is the number of test data points; Calculate the mean absolute percentage error , the specific formula is: ； Calculate score deviation , the specific formula is: ； Take the root mean square error , the specific formula is: ； Calculate the normalized mean square error , the specific formula is: ； Find the coefficient of determination , the specific formula is: ； The accuracy of the prediction results is judged based on the above calculation results.

Images