Rotary hearth furnace operation parameter optimization method and system fusing XGBoost and PSO

Abstract

The invention belongs to the technical field of industrial automation, and particularly discloses a rotary hearth furnace operation parameter optimization method and system fusing XGBoost and PSO, and the method comprises the following steps: obtaining a mathematical model of rotary hearth furnace operation parameter optimization through defining decision variables, constraint conditions and a target function based on a rotary hearth furnace process principle and an energy-mass conversion relation; the method comprises the following steps: acquiring actual production data of a rotary hearth furnace industrial site, preprocessing the actual production data, and establishing a mapping relation between material and energy input and material output and carbon emission indexes based on an XGBoost proxy model; and in the process of iteratively optimizing the operation parameters by using the PSO algorithm, calculating a target function by using the trained XGBoost proxy model, thereby obtaining a near-optimal solution of the optimization model. According to the technical scheme, the agent model and the swarm intelligence algorithm are fused and integrated, so that the guiding effect on industrial field operation parameter setting is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial automation and relates to a rotary hearth furnace process parameter optimization method and system integrating XGBoost (eXtreme Gradient Boosting) and PSO (Particulate Swarm Optimization). Background Art

[0002] In the steel industry, steel dust and organic solid waste are characterized by large output, many types, great differences in properties, and low resource utilization rate. Improving the steel waste treatment process and co-disposal of steel dust and organic solid waste are important ways to promote the clean and low-carbon development of the steel industry.

[0003] As a process for treating metallurgical waste, the rotary hearth furnace process has the advantages of strong raw material applicability, high metallization rate and high resource recovery rate. However, it also has problems such as low energy utilization efficiency and high investment. In addition, the increased carbon input also increases carbon emissions.

[0004] The parameters such as the proportion of ingredients, temperature and pressure in the rotary hearth furnace process will have a direct impact on important indicators such as the quality of the finished product and the metallization rate. Appropriate parameters can effectively improve the efficiency of the rotary hearth furnace process recycling and treatment, and the process parameters need to be set according to the changes in raw materials. However, at present, the setting of parameters often uses constant values ​​in the process, and the adjustment depends on manual experience, which is difficult to meet the needs of intelligent production and also limits the improvement of its recycling efficiency. Therefore, further research and improvement are still needed.

[0005] Due to the large variety of raw materials and their different properties in the rotary hearth furnace process, and the black box characteristics of the process mechanism, it is difficult to directly establish a mechanism model related to the rotary hearth furnace process through mathematical model methods. In addition, the production conditions on site are harsh, manual operations are frequent, and the degree of digitization is low, which makes it difficult to meet the needs of intelligent production. Therefore, it is of great significance to establish a data-driven model of the rotary hearth furnace process and optimize its process parameters to realize its intelligent and digital production, improve energy utilization efficiency and resource recovery efficiency, and reduce carbon emissions. Summary of the invention

[0006] The object of the present invention is to provide a rotary hearth furnace operating parameter optimization method and system integrating XGBoost and PSO, so as to achieve optimal adjustment of process parameters in industrial sites.

[0007] In order to achieve the above object, the basic scheme of the present invention is: a rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO, comprising the following steps:

[0008] Based on the process principle of rotary hearth furnace and its energy-quality conversion relationship, the decision variables, constraints and objective functions are defined to obtain the operation parameter optimization model;

[0009] The actual production data of the rotary hearth furnace industrial site was collected and preprocessed, and a proxy model based on XGBoost was established to obtain the mapping relationship between material and energy input and material output and carbon emission indicators during the operation of the rotary hearth furnace;

[0010] The objective function value is calculated based on the XGBoost model, and the PSO algorithm is used to iteratively optimize the operating parameters to obtain the near-optimal solution of the rotary hearth furnace operating parameter optimization model.

[0011] The working principle and beneficial effects of this basic solution are as follows: This technical solution establishes a data-driven model of the rotary hearth furnace process and optimizes its process parameters through the rotary hearth furnace principle and its process input and output, merges and integrates the agent model and swarm intelligence algorithm, optimizes and adjusts the industrial site process parameters, so as to achieve a guiding role in the setting of industrial site operation parameters, which is of great significance to improving energy utilization efficiency and resource recovery efficiency and reducing carbon emissions.

[0012] Furthermore, based on the process principle of the rotary hearth furnace and its energy-quality conversion relationship, the method of defining decision variables, constraints and objective functions to obtain the operating parameter optimization model is as follows:

[0013] The operating parameters include material operating parameters and energy operating parameters. The material operating parameters are the ratio of each raw material silo, and the energy operating parameters are the set temperatures of the heating section and each reduction section in the rotary hearth furnace process. The raw material ratio vector x and the set temperature vector θ are used as controllable decision vectors. Let x = {x i |1≤i≤I},θ={θ j |1≤j≤J}, where x i represents the ratio of the i-th silo; θ j represents the jth set temperature parameter; I represents the number of silos; J represents the number of set temperature parameters;

[0014] After determining the decision variables, according to the energy-quality conversion principle in the rotary hearth furnace process, three types of indicators, namely carbon emission index f1(x), metal recovery rate f2(x) and proportion cost f3(x), are selected as the optimization objectives of the model, and the objective function min f is constructed:

[0015] minf=(f1,f2,f3)

[0016] f1=XGB 1 (x,θ)

[0017] f2=XGB 2 (x,θ)

[0018]

[0019] Among them, the f1 and f2 indicators need to be obtained through the XGBoost proxy model output. 1 (x,θ),XGB 2 (x,θ) represents the corresponding carbon emission index and metal recovery rate XGBoost proxy model; Indicates the value in the last allocation scheme.

[0020] Build relevant optimization models and conduct optimization research on rotary hearth furnace process parameters.

[0021] Furthermore, the constraints include:

[0022] Element ratio constraint: After the batching is completed, the upper and lower limits of the sum of the mass fraction of each element are constrained;

[0023] Bin discharge constraints: After batching is completed, the upper and lower limits of the total discharge of each bin are set;

[0024] Total proportion constraint: the total proportion and upper and lower limit constraints of all internal material silos;

[0025] Temperature setting constraints: upper and lower limit constraints on the temperature values ​​set in the heating section and each reduction section;

[0026] Construct constraint functions from constraint conditions:

[0027]

[0028] Where w(x) represents the element mass fraction vector; w L and w U Respectively represent the lower and upper limits of the elements; calculate the mass fraction of each element in each batch, and add up the mass fractions of the same elements to obtain the sum of the mass fractions of each element; Q represents the total batching amount; Qx represents the batching amount of the corresponding batching silo; x L and x U They represent the lower and upper limits of the silo material level respectively. t(θ) represents the set temperature; t L and t U They represent the lower limit and upper limit of the set temperature respectively; the conversion relationship between the set temperature and the actual temperature is obtained based on experience.

[0029] After completing the construction of the objective function, the constraints of the optimization model are constructed according to the process requirements of the rotary hearth furnace, which is conducive to model solution.

[0030] Furthermore, the method for collecting actual production data of the rotary hearth furnace industrial site and preprocessing it is as follows:

[0031] Based on the box plot principle, the actual production data is preprocessed:

[0032] IQR=Q3-Q1

[0033] Among them, IQR represents the interquartile range; Q3 represents the upper quartile of the data set; Q1 represents the lower quartile of the data set;

[0034] The normal data set in the actual production data set is represented as:

[0035] D * ={(Q1-1.5×IQR)≤d i ≤(Q3+1.5×IQR)|d i ∈D,i=1,2,…n}

[0036] Among them, D * represents a normal data set; d i represents the i-th sample data; D represents the entire data set; n represents the total number of all samples in the data set;

[0037] All data within the range will be considered normal data, and data outside the range will be eliminated as abnormal data.

[0038] Pre-process the actual production data of the rotary hearth furnace industrial site for subsequent use.

[0039] Furthermore, an XGBoost-based proxy model was established to obtain the mapping relationship between material and energy input and material output and carbon emission indicators during the operation of the rotary hearth furnace. The specific method is as follows:

[0040] The characteristic attributes of actual production data are divided into input raw materials, output products and operating conditions. The characteristic recursive elimination and cross-validation RFECV method is used to screen and sort the characteristics of actual production data to obtain the importance priority of input characteristics: operating conditions are optimal, energy input is second optimal, and material input is third optimal;

[0041] Feed the input features into the XGBoost model for training:

[0042]

[0043] in, Represents the output carbon emission index and the predicted value of metal recovery rate; x is the raw material ratio vector, θ is the set temperature vector; XGB 1 、XGB 2 XGBoost proxy model representing carbon emission index and metal recovery rate.

[0044] According to the importance of different feature values, the proxy model can be improved in a targeted manner and the model can be trained for easy use.

[0045] Furthermore, the particle swarm optimization algorithm is used to optimize the process parameters of the rotary hearth furnace based on the process parameter optimization model and the XGBoost proxy model. The specific steps are as follows:

[0046] 1) Initialize a group of particles of size POP_SIZE and obtain their random positions (X, θ) and velocities V;

[0047] 2) Obtain the carbon emission index f1 and metal recovery rate f2 corresponding to the current position of the raw material ratio and the set temperature through the XGBoost proxy model, and evaluate the current fitness fit(x,θ) of each particle;

[0048] 3) After obtaining the fitness, compare fit(x,θ) with the individual historical optimal value fit(pbest), and update the individual historical optimal position pbest;

[0049] 4) Compare fit(x,θ) with the population historical optimal value fit(gbest), and update the population historical optimal position gbest;

[0050] 5) Update particle position and velocity according to the update formula;

[0051] 6) If the end condition is not met, that is, the fitness fit(x,θ) is not good enough or the number of iterations does not reach the maximum number of iterations G max , then return to step 2), otherwise output the result.

[0052] Further, S1, particle swarm initialization:

[0053] Define the values ​​of decision variables such as the raw material ratio and set temperature of each silo as the position vector of the particle group x, θ represent the position of each particle in the particle swarm, i.e., the value of the raw material ratio and the set temperature; I represents the number of silos; J represents the number of set temperature parameters; m represents the iteration number, where m = 1; the speed vector of the raw material ratio and the set temperature each time the particle swarm is updated v represents the speed of each particle in the particle swarm; the position and speed are initialized, the particle swarm size POP_SIZE, the number of iterations MAX_ITER, the inertia factor w, the learning factors c1 and c2, c1 represents the acceleration weight that makes the particle tend to the individual optimal position; c2 represents the acceleration weight that makes the particle tend to the global optimal position;

[0054] S2, obtain the current position of raw material ratio and set temperature through XGBoost proxy model Calculate the current fitness fit(x,θ) of the particle based on the corresponding carbon emission index f1 and metal recovery rate f2:

[0055] fit(x,θ) = αf1 + βf2 + γf3 + ε

[0056] That is

[0057] Convert to

[0058] where α, β, γ represent the weights of the corresponding objective function; ε represents the penalty term corresponding to the constraint condition; represents the prediction result of the carbon emission index; represents the prediction result of the metal recovery rate; represents the ratio of the previous batching plan, x i represents the ratio of the i-th bin;

[0059] S3. Compare fit(x,θ) with the individual historical optimal value fit(pbest) and the population historical optimal value fit(gbest) respectively, and update the individual historical optimal position pbest and the population historical optimal position gbest. If fit(x,θ) < fit(pbest), then update pbest = (x,θ); if fit(x,θ) < fit(gbest), then update gbest = fit(x,θ);

[0060] Update the positions of each decision variable respectively and the velocity where x m and θ m represent the positions of each particle in the m-th generation particle swarm, that is, the raw material ratio and the set temperature; v m represents the velocity of each particle in the m-th generation particle swarm; the update formula is as follows:

[0061] V i = w * V i + c1 * rand * (pbest i - (X,θ) i ) + c2 * rand * (gbest - (X,θ) i )

[0062] (X,θ) i = (X,θ) i + V i

[0063] If the maximum number of iterations is reached, output the result; otherwise, return to step S1.

[0064] The PSO algorithm is used to update the particle position and speed to optimize the process parameters.

[0065] The present invention also provides a rotary hearth furnace operating parameter optimization system integrating XGBoost and PSO, comprising a data acquisition module and a processing module, wherein the data processing module is used to obtain the principle of the rotary hearth furnace and its process input and output, as well as actual production data of the industrial site;

[0066] The input end of the processing module is connected to the output end of the data acquisition module, and the processing module executes the method of the present invention to optimize the process parameters of the rotary hearth furnace.

[0067] This system uses data acquisition modules and processing modules to fuse and integrate machine learning models and optimization algorithms to provide guidance for adjusting process parameters on industrial sites. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic flow chart of a rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO according to the present invention. DETAILED DESCRIPTION

[0069] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0070] In the description of the present invention, it is necessary to understand that the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0071] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0072] The present invention discloses a rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO, which integrates the machine learning model and the optimization algorithm to achieve the optimization adjustment of industrial field process parameters. Figure 1 As shown in the figure, the rotary hearth furnace process parameter optimization method integrating XGBoost and PSO includes the following steps:

[0073] Based on the process principle of rotary hearth furnace and its energy-quality conversion relationship, the decision variables, constraints and objective functions are defined to obtain the operation parameter optimization model;

[0074] The actual production data of the rotary hearth furnace industrial site was collected and preprocessed, and a proxy model based on XGBoost was established to obtain the mapping relationship between material and energy input and material output and carbon emission indicators during the operation of the rotary hearth furnace;

[0075] The objective function value is calculated based on the XGBoost model, and the PSO algorithm is used to iteratively optimize the operating parameters to obtain the near-optimal solution of the rotary hearth furnace operating parameter optimization model.

[0076] In a preferred embodiment of the present invention, based on the process principle of the rotary hearth furnace and its energy-quality conversion relationship, a method for defining decision variables, constraints and objective functions to obtain an operating parameter optimization model is as follows:

[0077] The operating parameters include material operating parameters and energy operating parameters. The material operating parameters are the ratio of each raw material silo, and the energy operating parameters are the set temperatures of the heating section and each reduction section in the rotary hearth furnace process. The raw material ratio vector x and the set temperature vector θ are used as controllable decision vectors. Let x = {x i |1≤i≤I},θ={θ j |1≤j≤J}, where x i represents the ratio of the i-th silo; θ j represents the jth set temperature parameter; I represents the number of silos; J represents the number of set temperature parameters;

[0078] After determining the decision variables, according to the energy-quality conversion principle in the rotary hearth furnace process, three types of indicators, namely carbon emission index f1(x), metal recovery rate f2(x) and proportion cost f3(x), are selected as the optimization objectives of the model, and the objective function min f is constructed:

[0079] minf=(f1,f2,f3)

[0080] f1=XGB 1 (x,θ)

[0081]

[0082] Among them, the f1 and f2 indicators need to be obtained through the XGBoost proxy model output. 1 (x,θ),XGB 2 (x,θ) represents the corresponding carbon emission index and metal recovery rate XGBoost proxy model; Indicates the value in the last allocation scheme.

[0083] In a preferred embodiment of the present invention, the constraints include:

[0084] Element ratio constraint: After the batching is completed, the upper and lower limits of the sum of the mass fraction of each element are constrained;

[0085] Bin discharge constraints: After batching is completed, the upper and lower limits of the total discharge of each bin are set;

[0086] Total proportion constraint: the total proportion and upper and lower limit constraints of all internal material silos;

[0087] Temperature setting constraints: upper and lower limit constraints on the temperature values ​​set in the heating section and each reduction section;

[0088] Construct constraint functions from constraint conditions:

[0089]

[0090] Where w(x) represents the element mass fraction vector; w L and w U Respectively represent the lower and upper limits of the elements; calculate the mass fraction of each element in each batch, and add up the mass fractions of the same elements to obtain the sum of the mass fractions of each element; Q represents the total batching amount; Qx represents the batching amount of the corresponding batching silo; x L and x U They represent the lower and upper limits of the silo material level respectively. t(θ) represents the set temperature; t L and t U They represent the lower limit and upper limit of the set temperature respectively; the conversion relationship between the set temperature and the actual temperature is obtained based on experience to obtain the objective function and constraint function, and the batching ratio and operating parameters are used as decision variables to optimize the process parameters of the rotary hearth furnace through the particle swarm optimization (PSO) algorithm.

[0091] In a preferred embodiment of the present invention, the method for collecting actual production data of the rotary hearth furnace industrial site and preprocessing the data is as follows:

[0092] The actual production data of the rotary hearth furnace process is read from the PLC of the industrial site. The items included in the data are shown in Table 1:

[0093] Table 1 Industrial site data collection

[0094]

[0095] Collect actual production data of rotary hearth furnace process. Industrial data generally has the problem of poor data quality, small data magnitude, incomplete data, and large data noise. Therefore, it is necessary to perform data preprocessing and feature engineering operations for subsequent model learning and training. Based on the box plot principle, the actual production data is preprocessed:

[0096] IQR=Q3-Q1

[0097] Among them, IQR represents the interquartile range; Q3 represents the upper quartile of the data set; Q1 represents the lower quartile of the data set;

[0098] The normal data set in the actual production data set is represented as:

[0099] D * ={(Q1-1.5×IQR)≤d i ≤(Q3+1.5×IQR)|d i ∈D,i=1,2,…n}

[0100] Among them, D * represents a normal data set; d i represents the i-th sample data; D represents the entire data set; n represents the total number of all samples in the data set;

[0101] All data within the range will be considered normal data, and data outside the range will be eliminated as abnormal data.

[0102] In a preferred embodiment of the present invention, an XGBoost-based proxy model is established to obtain the mapping relationship between material and energy input and material output and carbon emission indicators during the operation of the rotary hearth furnace. The specific method is:

[0103] After data preprocessing, feature engineering is performed on the data set to prevent the proxy model from overfitting or underfitting, which may result in poor prediction performance. The characteristic attributes of the actual production data are divided into input raw materials, output products, and operating conditions. The selected data and their types are shown in Table 2.

[0104] Table 2 Selection of data features

[0105]

[0106] The Recursive Feature Elimination and Cross Validation (RFECV) method is used to screen and sort the features of the actual production data (the principle is based on multiple training and cross validation, and the importance of features is rated through multiple training comparison models. After feature rating, the optimal number of features is selected through cross validation). The importance priority of input features is obtained: operating conditions (gas temperature and pressure) are optimal, energy input (air and gas consumption) is second optimal, and material input (ingredient ratio) is third optimal. According to the importance of different feature values, the proxy model can be improved in a targeted manner.

[0107] Feed the input features into the XGBoost model for training:

[0108]

[0109] in, Represents the output carbon emission index and the predicted value of metal recovery rate; x is the raw material ratio vector, θ is the set temperature vector; XGB 1 、XGB 2 XGBoost proxy model representing carbon emission index and metal recovery rate.

[0110] In a preferred embodiment of the present invention, a particle swarm optimization algorithm is used to optimize the process parameters of a rotary hearth furnace based on a process parameter optimization model and an XGBoost proxy model. The specific steps are as follows:

[0111] 1) Initialize a group of particles of size POP_SIZE and obtain their random positions (X, θ) and velocities V;

[0112] 2) Obtain the carbon emission index f1 and metal recovery rate f2 corresponding to the current position of the raw material ratio and the set temperature through the XGBoost proxy model, and evaluate the current fitness fit(x,θ) of each particle;

[0113] 3) After obtaining the fitness, compare fit(x,θ) with the individual historical optimal value fit(pbest), and update the individual historical optimal position pbest;

[0114] 4) Compare fit(x,θ) with the population historical optimal value fit(gbest), and update the population historical optimal position gbest;

[0115] 5) Update particle position and velocity according to the update formula;

[0116] 6) If the end condition is not met, that is, the fitness fit(x,θ) is not good enough or the number of iterations does not reach the maximum number of iterations G max , then return to step 2), otherwise output the result.

[0117] In a preferred embodiment of the present invention,

[0118] S1, particle swarm initialization:

[0119] Define the values ​​of decision variables such as the raw material ratio and set temperature of each silo as the position vector of the particle group x, θ represent the position of each particle in the particle swarm, i.e., the value of the raw material ratio and the set temperature; I represents the number of silos; J represents the number of set temperature parameters; m represents the iteration number, where m = 1; the speed vector of the raw material ratio and the set temperature each time the particle swarm is updated v represents the speed of each particle in the particle swarm; the position and speed are initialized, the particle swarm size POP_SIZE, the number of iterations MAX_ITER, the inertia factor w, the learning factors c1 and c2, c1 represents the acceleration weight that makes the particle tend to the individual optimal position; c2 represents the acceleration weight that makes the particle tend to the global optimal position;

[0120] S2, obtain the current position of raw material ratio and set temperature through XGBoost proxy model The corresponding carbon emission index f1 and metal recovery rate f2 are used to calculate the current fitness of the particle fit(x,θ):

[0121] fit(x,θ)=αf1+βf2+γf3+ε

[0122] Right now

[0123] Convert to

[0124] Among them, α, β, γ represent the weights of the corresponding objective function; ε represents the penalty term corresponding to the constraint condition; Indicates the prediction result of carbon emission index; It represents the predicted result of metal recovery rate; Indicates the ratio of the previous batching scheme, x i represents the ratio of the i-th silo;

[0125] S3. Compare fit(x,θ) with the individual historical optimal value fit(pbest) and the population historical optimal value fit(gbest) respectively, and update the individual historical optimal position pbest and the population historical optimal position gbest. If fit(x,θ) < fit(pbest), then update pbest = (x,θ); if fit(x,θ) < fit(gbest), then update gbest = fit(x,θ).

[0126] Update the positions of each decision variable respectively and the velocity where x m and θ m represent the positions of each particle in the m-th generation particle swarm respectively, that is, the raw material ratio and the set temperature; v m represents the velocity of each particle in the m-th generation particle swarm; the update formula is as follows:

[0127] V i = w * V i + c1 * rand * (pbest i - (X,θ) i ) + c2 * rand * (gbest - (X,θ) i )

[0128] (X,θ) i = (X,θ) i + V i

[0129] If the maximum number of iterations is reached, output the result; otherwise, return to step S1.

[0130] The present invention also provides a rotary hearth furnace operation parameter optimization system integrating XGBoost and PSO, including a data acquisition module and a processing module. The data processing module is used to obtain the rotary hearth furnace principle, its process inputs and outputs, and the actual production data of the industrial site.

[0131] The input end of the processing module is electrically connected to the output end of the data acquisition module. The processing module executes the method of the present invention to optimize the rotary hearth furnace process parameters. This system uses the data acquisition module and the processing module to integrate and combine the machine learning model and the optimization algorithm to realize the guiding function for the adjustment of the industrial site process parameters.

[0132] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0133] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO, characterized in that: The steps include: Based on the process principle of rotary hearth furnace and its energy-quality conversion relationship, the decision variables, constraints and objective functions are defined to obtain the operation parameter optimization model; The actual production data of the rotary hearth furnace industrial site was collected and preprocessed, and a proxy model based on XGBoost was established to obtain the mapping relationship between material and energy input and material output and carbon emission indicators during the operation of the rotary hearth furnace; The objective function value is calculated based on the XGBoost model, and the PSO algorithm is used to iteratively optimize the operating parameters to obtain the near-optimal solution of the rotary hearth furnace operating parameter optimization model.

2. The rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO according to claim 1, characterized in that: Based on the process principle of rotary hearth furnace and its energy-quality conversion relationship, the method of defining decision variables, constraints and objective functions to obtain the operation parameter optimization model is as follows: The operating parameters include material operating parameters and energy operating parameters. The material operating parameters are the ratio of each raw material silo, and the energy operating parameters are the set temperatures of the heating section and each reduction section in the rotary hearth furnace process. The raw material ratio vector x and the set temperature vector θ are used as controllable decision vectors. Let x = {x i |1≤i≤I}, θ={θ j |1≤j≤J}, where x i represents the ratio of the i-th silo; θ j represents the jth set temperature parameter; I represents the number of silos; J represents the number of set temperature parameters; After determining the decision variables, according to the energy-quality conversion principle in the rotary hearth furnace process, three types of indicators, namely carbon emission index f1(x), metal recovery rate f2(x) and proportion cost f3(x), are selected as the optimization objectives of the model, and the objective function min f is constructed: min f=(f1,f2,f3) f1=XGB 1 (x,θ) f2=XGB 2 (x,θ) Among them, the f1 and f2 indicators need to be obtained through the XGBoost proxy model output. 1 (x,θ),XGB 2 (x,θ) represents the corresponding carbon emission index and metal recovery rate XGBoost proxy model; Indicates the value in the last allocation scheme.

3. The rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO as claimed in claim 2, characterized in that: The constraints include: Element ratio constraint: After the batching is completed, the upper and lower limits of the sum of the mass fraction of each element are constrained; Bin discharge constraints: After batching is completed, the upper and lower limits of the total discharge of each bin are set; Total proportion constraint: the total proportion and upper and lower limit constraints of all internal material silos; Temperature setting constraints: upper and lower limit constraints on the temperature values ​​set in the heating section and each reduction section; Construct constraint functions from constraint conditions: Where w(x) represents the element mass fraction vector; w L and w U Respectively represent the lower and upper limits of the elements; calculate the mass fraction of each element in each batch, and add up the mass fractions of the same elements to obtain the sum of the mass fractions of each element; Q represents the total batching amount; Qx represents the batching amount of the corresponding batching silo; x L and x U They represent the lower and upper limits of the silo material level respectively; t(θ) represents the set temperature; t L and t U They represent the lower limit and upper limit of the set temperature respectively; the conversion relationship between the set temperature and the actual temperature is obtained based on experience.

4. The rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO according to claim 1, characterized in that: The method of collecting actual production data of rotary hearth furnace industrial site and preprocessing it is as follows: Based on the box plot principle, the actual production data is preprocessed: IQR=Q3-Q1 Among them, IQR represents the interquartile range; Q3 represents the upper quartile of the data set; Q1 represents the lower quartile of the data set; The normal data set in the actual production data set is represented as: D * ={(Q1-1.5×IQR)≤d i ≤(Q3+1.5×IQR)|d i ∈D,i=1,2,...n} Where D* represents the normal data set; d i represents the i-th sample data; D represents the entire data set; n represents the total number of all samples in the data set; All data within the range will be considered normal data, and data outside the range will be eliminated as abnormal data.

5. The rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO according to claim 1, characterized in that: An XGBoost-based proxy model was established to obtain the mapping relationship between material and energy input and material output and carbon emission indicators during the operation of the rotary hearth furnace. The specific method is as follows: The characteristic attributes of actual production data are divided into input raw materials, output products and operating conditions. The characteristic recursive elimination and cross-validation RFECV method is used to screen and sort the characteristics of actual production data to obtain the importance priority of input characteristics: operating conditions are optimal, energy input is second optimal, and material input is third optimal; Feed the input features into the XGBoost model for training: in, Represents the output carbon emission index and the predicted value of metal recovery rate; x is the raw material ratio vector, θ is the set temperature vector; XGB 1 、XGB 2 XGBoost proxy model representing carbon emission index and metal recovery rate.

6. The rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO according to claim 1, characterized in that: The particle swarm optimization algorithm is used to optimize the process parameters of the rotary hearth furnace based on the process parameter optimization model and the XGBoost proxy model. The specific steps are as follows: 1) Initialize a group of particles of size POP_SIZE and obtain their random positions (x, θ) and velocities V; 2) Obtain the carbon emission index f1 and metal recovery rate f2 corresponding to the current position of the raw material ratio and the set temperature through the XGBoost proxy model, and evaluate the current fitness fit(x,θ) of each particle; 3) After obtaining the fitness, compare fit(x, θ) with the individual historical optimal value fit(pbest), and update the individual historical optimal position pbest; 4) Compare fit(x, θ) with the population historical optimal value fit(gbest), and update the population historical optimal position gbest; 5) Update the particle position and velocity according to the update formula; 6) If the end condition is not met, that is, the fitness fit(x,θ) is not good enough or the number of iterations does not reach the maximum number of iterations G max , then return to step 2), otherwise output the result.

7. The rotary hearth furnace operating parameter optimization method integrating XGBoost and PSO according to claim 6, characterized in that: S1. Particle swarm initialization: Define the values ​​of decision variables such as the raw material ratio and set temperature of each silo as the position vector of the particle group x, θ represent the position of each particle in the particle group, that is, the value of the raw material ratio and the set temperature; I represents the number of silos; J represents the number of set temperature parameters; m represents the iteration number, where m = 1; The speed vector of the raw material ratio and temperature setting each time the particle group is updated v represents the speed of each particle in the particle swarm; the position and speed are initialized, the particle swarm size POP_SIZE, the number of iterations MAX_ITER, the inertia factor w, the learning factors c1 and c2, c1 represents the acceleration weight that makes the particle tend to the individual optimal position; c2 represents the acceleration weight that makes the particle tend to the global optimal position; S2, obtain the current position of raw material ratio and set temperature through XGBoost proxy model The corresponding carbon emission index f1 and metal recovery rate f2 are used to calculate the current fitness of the particle fit(x,θ): fit(x, θ) = αf1 + βf2 + γf3 + ε Right now Convert to Among them, α, β, γ represent the weights of the corresponding objective function; ε represents the penalty term corresponding to the constraint condition; Indicates the prediction result of carbon emission index; It represents the predicted result of metal recovery rate; Indicates the ratio of the previous batching scheme, x i represents the ratio of the i-th silo; S3. Compare fit(x, θ) with the individual historical optimal value fit(pbest) and the population historical optimal value fit(gbest) respectively, and update the individual historical optimal position pbest and the population historical optimal position gbest. If fit(x, θ) < fit(pbest), then update pbest = (x, θ); if fit(x, θ) < fit(gbest), then update gbest = fit(x, θ); Update the position of each decision variable separately and speed Among them, x m With θ m They represent the position of each particle in the mth generation particle group, that is, the raw material ratio and the set temperature; v m Represents the speed of each particle in the mth generation particle swarm; the update formula is as follows: V i =w*V i +c1*rand*(pbest i -(x,θ)i)+c2*rand**(gbest-(x,θ) i ) (x,θ) i =(x,θ) i +V i If the maximum number of iterations is reached, output the result; otherwise, return to step S1.

8. A rotary hearth furnace operating parameter optimization system integrating XGBoost and PSO, characterized in that: It includes a data acquisition module and a processing module. The data processing module is used to obtain the principle of the rotary hearth furnace, its process inputs and outputs, as well as the actual production data of the industrial site; The input end of the processing module is connected to the output end of the data acquisition module. The processing module executes the method described in any one of claims 1-7 to optimize the process parameters of the rotary hearth furnace.

Images