Blast furnace thermal state prediction method for optimizing ELM based on improved RSO algorithm

Abstract

The invention is applicable to the technical field of metallurgy, and particularly relates to a blast furnace thermal state prediction method for optimizing ELM based on an improved RSO algorithm, and the method comprises the following steps: obtaining blast furnace production sample data; preprocessing the blast furnace production sample data to obtain preprocessed data; a blast furnace molten iron Si content prediction model is constructed, parameter identification is conducted on the blast furnace molten iron Si content prediction model through an improved RSO algorithm, and the blast furnace molten iron Si content prediction model adopts an ELM neural network model; and importing the preprocessed data into the blast furnace molten iron Si content prediction model, and outputting a state prediction result. According to the method, the heat state of the blast furnace is predicted by predicting the Si of the molten iron of the blast furnace, and the heat state of the blast furnace is accurately and rapidly predicted under the precondition that the stable operation of the process is met, so that important theoretical guidance and technical support are provided for regulation and control of field operators, and good conditions are created for stabilizing the steel quality and the production process.

Description

Technical Field

[0001] The invention belongs to the technical field of metallurgy, and in particular relates to a blast furnace thermal state prediction method based on an improved RSO algorithm to optimize ELM. Background Art

[0002] Large-scale blast furnace ironmaking is a complex production process with high dimensions, strong hysteresis, nonlinearity and multi-noise. Under conditions of high temperature, high pressure and high dust, it is often accompanied by complex physical and chemical reactions, and has characteristics such as large time lag, strong coupling and nonlinearity. The thermal state of the blast furnace is one of the important factors affecting blast furnace smelting. Monitoring and prediction of it is the focus of the industry. Its control level determines important indicators such as molten iron quality, fuel consumption, production efficiency and blast furnace operation status. Therefore, it is of great practical significance to characterize the thermal state of the blast furnace based on the characteristic variables in the blast furnace production process. Since it is very difficult to directly measure the physical temperature inside the blast furnace, the thermal state of the blast furnace cannot be characterized by directly obtaining its physical temperature. Scholars have found through a large number of studies that there is a positive correlation between the Si content of molten iron and temperature. Therefore, most of the blast furnace thermal state is characterized by predicting the Si content of molten iron. However, due to the characteristics of high temperature, closedness, and complex and changeable working conditions of blast furnaces, it is very difficult to predict the Si content in molten iron. Currently, smelting control mostly relies on the manual experience of the furnace master. This control method is blind and therefore cannot achieve optimal production of the blast furnace.

[0003] At present, the existing prediction methods of Si content in molten iron cannot meet the actual production needs well, which has led to the rapid development of algorithm models that characterize the thermal state of blast furnaces by predicting Si content in molten iron. Generally, existing research is divided into two categories: one is the classification problem of predicting the actual change trend of Si content in molten iron, and the other is the regression problem of predicting the actual production value of Si content in molten iron. In the actual application process, most of the prediction models of Si content in molten iron are based on intelligent optimization algorithms, neural networks and support vector machines. Compared with traditional prediction models, their prediction effects have been improved to a certain extent. However, these models still have certain limitations and are far from accurately reflecting the inherent change laws of furnace temperature. Summary of the invention

[0004] The purpose of the present invention is to provide a blast furnace thermal state prediction method based on an improved RSO algorithm to optimize ELM, aiming to solve the problem that the prediction method provided by the prior art has limitations and cannot accurately reflect the inherent change law of the furnace temperature.

[0005] The present invention is implemented as follows: a blast furnace thermal state prediction method based on an improved RSO algorithm to optimize ELM, the method comprising:

[0006] Obtain blast furnace production sample data;

[0007] Preprocessing the blast furnace production sample data to obtain preprocessed data;

[0008] A prediction model for Si content in molten iron of a blast furnace is constructed, and parameters of the prediction model for Si content in molten iron of a blast furnace are identified by improving the RSO algorithm, wherein the prediction model for Si content in molten iron of a blast furnace adopts an ELM neural network model;

[0009] The preprocessed data is imported into the blast furnace molten iron Si content prediction model and the state prediction results are output.

[0010] Preferably, the blast furnace molten iron Si content prediction model consists of an input layer, a hidden layer and an output layer, and the decision variables are the weight between the input layer and the hidden layer and the hidden layer threshold.

[0011] Preferably, the step of performing parameter identification on the prediction model of Si content in molten iron of blast furnace by improving RSO algorithm comprises constructing a parameter identification model, wherein the objective function of the parameter identification model is the minimum root mean square error between the predicted value and the true value,

[0012]

[0013] Where J represents the objective function; f true Indicates the actual value of Si content in blast furnace hot metal; f(x i ) represents the predicted value of Si content in blast furnace hot metal of the i-th individual; x i It indicates the position of the i-th individual and represents the set of connection weights and thresholds in ELM;

[0014] The constraints of the parameter identification model are inequality constraints, including upper and lower limit constraints of each decision variable. For the ELM weight, the constraint range is between [-1, 1]; for the ELM threshold, the constraint range is between [0, 1].

[0015] Preferably, the steps for implementing the improved RSO algorithm include:

[0016] Construct an initial population consisting of N individuals, each of which represents a potential solution vector;

[0017] Calculate the individual fitness value of each individual and divide the individuals based on the individual fitness value;

[0018] The multi-group co-evolution mechanism is used to perform mutation, crossover, swarm prey pursuit and swarm attack processes on individuals;

[0019] The population is updated according to the execution results of the multi-group co-evolution mechanism, and iterations are repeated. When the iteration termination condition is reached, the global optimal solution is output.

[0020] Preferably, the step of calculating the individual fitness value of each individual and dividing the individuals based on the individual fitness value includes calculating the individual fitness value, dividing the individuals ranked in the top p% of the population as excellent individuals to form an elite sub-population, and randomly dividing the other individuals into groups of 5 individuals, and each small population does not interfere with each other.

[0021] The best individuals in the elite sub-population are mutated using DE / rand / 2 in differential evolution, which is mathematically described as:

[0022] V i =X r1 +F·(X r2 -X r3 )+F·(X r4 -X r5 )

[0023] Where V i represents the variant individual of the i-th individual; X r1 , X r2 , X r3 , X r4 and X r5 represents individuals randomly selected from excellent individuals, and r1, r2, r3, r4, r5 are not equal to i; F represents the coefficient of variation.

[0024] Preferably, a multi-group co-evolution mechanism is used to perform a crossover operation on the individual in the steps of mutation, crossover, mouse swarm prey chasing and mouse swarm attack process to generate a new individual, and its mathematical description is as follows:

[0025]

[0026] In the formula, Indicates the new individual position corresponding to the jth dimension of the i-th individual; rand j () represents a random number between [0, 1] randomly generated in the jth dimension; j rand represents a randomly selected dimension; CR represents the crossover coefficient.

[0027] Preferably, in the steps of using a multi-group co-evolution mechanism to perform mutation, crossover, swarm prey chasing and swarm attack processes on individuals, the swarm prey chasing process is mathematically described as follows:

[0028] P i =A·X i +C·(X gbest -X i )

[0029] Where, X gbest It represents the local optimal position of the current individual in the small population, where A is the exploration parameter and C is the development parameter.

[0030] Preferably, the process of the rat group attacking the prey is expressed as:

[0031]

[0032] Preferably, the step of updating the population according to the execution result of the multi-group co-evolution mechanism includes performing boundary processing on the new position of each individual, terminating the dimension exceeding the boundary on the boundary, calculating the fitness value, and updating the population based on greedy selection.

[0033] The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM provided by the present invention predicts the blast furnace thermal state by predicting the blast furnace molten iron Si, and accurately and quickly predicts the blast furnace thermal state under the premise of meeting the stable operation of the process, thereby providing important theoretical guidance and technical support for on-site operators to control, and creating good conditions for stabilizing steel quality and production technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flow chart of the main steps of a blast furnace thermal state prediction method based on an improved RSO algorithm to optimize ELM provided in an embodiment of the present invention;

[0035] Figure 2 A schematic diagram of a prediction model for Si content in molten iron in a blast furnace provided by an embodiment of the present invention;

[0036] Figure 3 A detailed flowchart of the steps of a blast furnace thermal state prediction method based on an improved RSO algorithm to optimize ELM provided in an embodiment of the present invention;

[0037] Figure 4 The flowchart of the improved RSO algorithm provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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.

[0039] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0040] like Figure 1 and Figure 3 As shown, it is a flow chart of the main steps of the blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM provided by an embodiment of the present invention, and the method includes:

[0041] Get sample data on blast furnace production.

[0042] In this step, the blast furnace production sample data includes raw material and fuel information, molten iron composition information, blast furnace structure parameters, process timing operation information and operating condition parameter information.

[0043] The blast furnace production sample data is preprocessed to obtain preprocessed data.

[0044] In this step, the preprocessing process includes processing missing values ​​and outliers in the blast furnace production sample data; performing 3σ screening to improve data quality; normalizing the screened data so as not to affect the analysis results; and finally performing a correlation analysis on the data, selecting features with a correlation coefficient greater than 0.2 with the Si content in blast furnace molten iron as the input to the prediction model, and the Si content in blast furnace molten iron as the output of the prediction model.

[0045] A prediction model for Si content in blast furnace iron is constructed, and parameters of the prediction model for Si content in blast furnace iron are identified by improving RSO algorithm. The prediction model for Si content in blast furnace iron adopts ELM neural network model.

[0046] In this step, the prediction model of Si content in molten iron of blast furnace adopts ELM neural network model, the structure is as follows: Figure 2 As shown in Figure 1, it consists of an input layer, a hidden layer, and an output layer. The decision variables are the weights between the input layer and the hidden layer and the hidden layer threshold.

[0047] In this step, the improved RSO algorithm is used to identify the parameters of the connection weights and thresholds in the ELM.

[0048] Construct a parameter identification model, the objective function is to minimize the root mean square error between the predicted value and the true value, the mathematical description is as follows:

[0049]

[0050] Where J represents the objective function; f true Indicates the actual value of Si content in blast furnace hot metal; f(x i ) represents the predicted value of Si content in blast furnace hot metal of the i-th individual; x i Indicates the position of the i-th individual and represents the set of connection weights and thresholds in ELM.

[0051] The constraints are mainly inequality constraints, including upper and lower limit constraints of each decision variable. For ELM weight, the constraint range is between [-1, 1]; for ELM threshold, the constraint range is between [0, 1].

[0052] like Figure 4 As shown in the figure, the improved RSO algorithm is used to identify the parameters of ELM.

[0053] Initialize the population: The population consists of N individuals, each of which represents a potential solution vector.

[0054] Individual division: Calculate the individual fitness value, divide the top p% of the individuals in the population into excellent individuals to form an elite sub-population. Since they have better location information, the idea of ​​differential evolution is introduced to enhance the information exchange between them, so as to avoid falling into the local optimum. Other individuals are randomly divided into groups of 5 individuals, and each small population does not interfere with each other to increase population diversity.

[0055] Population evolution: adopts multi-group co-evolution mechanism.

[0056] For the excellent individuals in the elite sub-population, the "DE / rand / 2" in differential evolution is used for mutation, and its mathematical description is as follows:

[0057] V i =X r1 +F·(X r2 -X r3 )+F·(X r4 -X r5 )

[0058] Where V i represents the variant individual of the i-th individual; X r1 , X r2 , X r3 , X r4 and X r5 represents individuals randomly selected from excellent individuals, and r1, r2, r3, r4, r5 are not equal to i; F represents the coefficient of variation.

[0059] Subsequently, a crossover operation is performed to generate new individuals, the mathematical description of which is as follows:

[0060]

[0061] In the formula, Indicates the new individual position corresponding to the jth dimension of the i-th individual; rand j () represents a random number between [0, 1] randomly generated in the jth dimension; j rand represents a randomly selected dimension; CR represents the crossover coefficient.

[0062] For each individual in a small population, the exploration parameter A and the development parameter C are updated according to the basic RSO algorithm description, and the mathematical description is as follows:

[0063]

[0064] C=2·rand()

[0065] Where R represents a random number between [1, 5]; t represents the current number of iterations; T represents the maximum number of iterations; rand() represents a random number between [0, 1].

[0066] The mathematical description of the rat-prey chasing process is as follows:

[0067] P i =A·X i +C·(X gbest -X i )

[0068] Where, X gbest Indicates the local optimal position of the current individual in the small population.

[0069] Finally, the process of rats attacking prey is defined by the following formula:

[0070]

[0071] Population update: The new position of each individual is processed at the boundary, and the dimension beyond the boundary is cut on the boundary. Then, the fitness value is calculated, and the population is updated based on greedy selection. Specifically, if the fitness value of the new individual is better than that of the old individual, the new individual is used to replace the old individual; otherwise, the individual is not updated.

[0072] If the current iteration number t is less than the maximum iteration number T, and the population is reorganized every flag generation. If mod(t,flag)=0, then the top p% of individuals in the population are reclassified as excellent individuals to form an elite population; the other individuals are randomly assigned to a small ordinary population of 5 individuals; otherwise, output the result;

[0073] Result output: Output the global optimal solution, whose decision variables represent the set of connection weights and thresholds in ELM, and whose fitness function value represents the root mean square error between the predicted value and the true value.

[0074] The preprocessed data is imported into the blast furnace molten iron Si content prediction model and the state prediction results are output.

[0075] The present invention provides a blast furnace thermal state prediction method based on an improved RSO algorithm to optimize ELM. As a single hidden layer feedforward neural network algorithm, ELM has the advantages of fast learning speed, good generalization ability and simple parameter setting. Therefore, the present invention predicts the Si content of molten iron based on ELM. However, the connection weights and thresholds in ELM are randomly generated, which greatly affects the prediction accuracy of the model, resulting in a decrease in the prediction accuracy of the model and a weakening of the generalization performance. Therefore, the present invention proposes an improved RSO algorithm to perform parameter identification on ELM to obtain a more accurate blast furnace molten iron Si prediction model, thereby achieving the purpose of predicting the thermal state of the blast furnace.

[0076] The present invention proposes a co-evolution mechanism, which divides the top p% of individuals in the population into elite populations. Since they have better position information, the present invention adopts the "DE / rand / 2" strategy to mutate them based on the idea of ​​differential evolution, and increases the diversity of the population by enhancing the information exchange between them, thereby avoiding falling into the local optimum; and for ordinary individuals, they are randomly assigned to small populations to increase the influence of the local optimal position to enhance the local search ability. At the same time, a recombination mechanism is adopted, and the population is reorganized according to a certain number of iterations, and the elite population individuals and ordinary small population individuals are re-divided, and search tasks are assigned to increase the information exchange between population individuals, aiming to avoid premature convergence, thereby better balancing the global search ability and the local search ability.

[0077] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0078] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0079] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0080] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0081] 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 protection scope of the present invention.

Claims

1. Blast furnace thermal state prediction method based on improved RSO algorithm to optimize ELM, It is characterized in that The method comprises: Obtain blast furnace production sample data; Preprocessing the blast furnace production sample data to obtain preprocessed data; A prediction model for Si content in molten iron of a blast furnace is constructed, and parameters of the prediction model for Si content in molten iron of a blast furnace are identified by improving the RSO algorithm, wherein the prediction model for Si content in molten iron of a blast furnace adopts an ELM neural network model; The preprocessed data is imported into the blast furnace molten iron Si content prediction model and the state prediction results are output.

2. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 1, It is characterized in that The Si content prediction model for blast furnace molten iron consists of an input layer, a hidden layer and an output layer, and the decision variables are the weight between the input layer and the hidden layer and the hidden layer threshold.

3. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 1, It is characterized in that The step of performing parameter identification on the prediction model of Si content in molten iron of blast furnace by improving RSO algorithm comprises constructing a parameter identification model, wherein the objective function of the parameter identification model is the minimum root mean square error between the predicted value and the true value, Where J represents the objective function; f true Indicates the actual value of Si content in blast furnace hot metal; f(x i ) represents the predicted value of Si content in blast furnace hot metal of the i-th individual; x i It indicates the position of the i-th individual and represents the set of connection weights and thresholds in ELM; The constraints of the parameter identification model are inequality constraints, including upper and lower limit constraints of each decision variable. For the ELM weight, the constraint range is between [-1, 1]; for the ELM threshold, the constraint range is between [0, 1].

4. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 1, It is characterized in that The implementation steps of the improved RSO algorithm include: Construct an initial population consisting of N individuals, each of which represents a potential solution vector; Calculate the individual fitness value of each individual and divide the individuals based on the individual fitness value; The multi-group co-evolution mechanism is used to perform mutation, crossover, swarm prey pursuit and swarm attack processes on individuals; The population is updated according to the execution results of the multi-group co-evolution mechanism, and iterations are repeated. When the iteration termination condition is reached, the global optimal solution is output.

5. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 1, It is characterized in that The step of calculating the individual fitness value of each individual and dividing the individuals based on the individual fitness value includes calculating the individual fitness value, ranking the population first p％ The individuals are divided into excellent individuals to form an elite sub-population, and the other individuals are randomly divided into groups of 5, and each small population does not interfere with each other.

6. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 4, It is characterized in that The excellent individuals in the elite sub-population are mutated using DE / rand / 2 in differential evolution, which is mathematically described as: V i =X r1 +F·(X r2 -X r3 )+F·(X r4 -X r5 ) Where V i represents the variant individual of the i-th individual; X r1 , X r2 , X r3 , X r4 and X r5 represents an individual randomly selected from among the excellent individuals, and r 1 、r 2 、r 3 、r 4 、r 5 Not equal to i; F represents the coefficient of variation.

7. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 4, It is characterized in that The multi-group co-evolution mechanism is used to perform crossover operations on individuals in the steps of mutation, crossover, mouse swarm prey pursuit and mouse swarm attack to generate new individuals. The mathematical description is as follows: In the formula, Indicates the new individual position corresponding to the jth dimension of the i-th individual; rand j () represents a random number between [0, 1] randomly generated in the jth dimension; j rand represents a randomly selected dimension; CR represents the crossover coefficient.

8. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 4, It is characterized in that In the steps of using multi-group co-evolution mechanism to perform mutation, crossover, swarm prey chasing and swarm attack on individuals, the swarm prey chasing process is mathematically described as follows: P i =A·X i +C·(X gbest -X i ) Where, X gbest It represents the local optimal position of the current individual in the small population, where A is the exploration parameter and C is the development parameter.

9. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 8, It is characterized in that The process of a group of rats attacking their prey can be expressed as:

10. The blast furnace thermal state prediction method based on the improved RSO algorithm to optimize ELM according to claim 1, It is characterized in that The steps of updating the population according to the execution results of the multi-group co-evolution mechanism include boundary processing for the new position of each individual, terminating the dimension beyond the boundary on the boundary, calculating the fitness value, and updating the population based on greedy selection.

Images