Final cooling temperature prediction method in power station steel plate production process

Abstract

The invention discloses a final cooling temperature prediction method in a power station steel plate production process, belongs to the technical field of power station steel plate production and manufacturing, and aims to solve the problems of poor precision and large error of an existing final cooling temperature prediction method for a steel plate in a cooling process. The method comprises the following steps: in a power station steel plate cooling process, collecting data of steel plate parameters in the cooling process, and preprocessing the data; taking data of parameters of the steel plate in the cooling process as input data of the XGBoost prediction model, and taking the final cooling temperature of the steel plate as output data of the XGBoost prediction model; performing optimization adjustment on hyper-parameters of the XGBoost prediction model by adopting a Bayesian law, constructing a BO-XGBoost-based prediction model of the final cooling temperature of the power station steel plate, and obtaining a BO-XGBoost prediction model; and a BO-XGBoost prediction model is adopted to predict the final cooling temperature of the power station steel plate in real time. The method is used for monitoring the steel plate in the cooling process.

Description

Technical Field

[0001] The invention relates to a method for predicting the final cooling temperature in a power station steel plate production process, and belongs to the technical field of power station steel plate production and manufacturing. Background Art

[0002] In recent years, as the economy has shifted from a high-speed development stage to a high-quality development stage, the steel industry has entered a stage of structural adjustment and transformation. High quality, low cost, environmental friendliness and good stability are issues that the steel industry urgently needs to solve. Steel companies are stepping up their efforts to innovate technology, focus on developing green and intelligent new technologies, new equipment and new products, and enhance their core competitiveness.

[0003] Steel plates are important raw materials required for the manufacture of power station boilers and pressure vessels. In the current steel plate post-rolling cooling control system, the energy balance equation is generally used to describe the various heat exchange phenomena of steel plates during the cooling process. Through continuous optimization of model construction and parameter prediction, the control cooling mechanism model has played an important role in the cooling process of steel plates. Although the heat transfer mechanism model plays an important role in the cooling process, since the steel plate involves multiple heat exchange forms in the heat exchange process, it is a multi-input and multi-output process. The heat exchange process is affected by multiple factors such as the chemical composition, thickness, speed, cooling manifold flow, water temperature, and cooling temperature of the steel plate, as well as the interference of unmeasurable intermediate variables and changes in the external environment. Therefore, it is difficult for the mechanism modeling method to accurately describe the physical and chemical changes of the steel plate during the cooling process in the process of constructing the mathematical model of the cooling process, resulting in low control accuracy of the mathematical model when the production conditions and equipment conditions change. Especially in the case of changing specifications and batches, the temperature control deviation of the first piece of steel often increases, resulting in poor stability of the steel plate product. Summary of the invention

[0004] The purpose of the present invention is to solve the problem of poor accuracy and large error in the existing method for predicting the final cooling temperature of steel plates during the cooling process, and to provide a method for predicting the final cooling temperature of power station steel plates during the production process.

[0005] The method for predicting the final cooling temperature of a power station steel plate production process according to the present invention comprises:

[0006] S1. During the cooling process of the steel plate in the power station, data of steel plate parameters during the cooling process are collected and preprocessed;

[0007] S2, using the data of the steel plate parameters during the cooling process as the input data of the XGBoost prediction model, and using the final cooling temperature of the steel plate as the output data of the XGBoost prediction model;

[0008] The Bayesian theorem is used to optimize and adjust the hyperparameters of the XGBoost prediction model, and a prediction model for the final cooling temperature of power plant steel plates based on BO-XGBoost is constructed to obtain the BO-XGBoost prediction model.

[0009] S3. Input the data of the steel plate parameters in the cooling process obtained in S1 into the BO-XGBoost prediction model to make a real-time prediction of the final cooling temperature of the power plant steel plate.

[0010] Preferably, the collected cooling process parameters in S1 include: chemical composition content, cooling start temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace temperature and final rolling temperature;

[0011] The chemical composition content includes the content of C, Cr, Mn, Mo and Nb.

[0012] Preferably, the specific method for preprocessing data in S1 includes:

[0013] S1-1. Select the input parameters and output parameters of the model according to the production process;

[0014] S1-2, use the Laida criterion to eliminate outliers in the parameters;

[0015] The elimination method is:

[0016]

[0017] Among them, x i represents the i-th sample, represents the average value of all samples, S x represents the standard deviation of all samples;

[0018]

[0019] Where i = 1, 2, ..., n, n represents the total number of samples;

[0020] S1-3, the parameter data were standardized using the Z-score method;

[0021]

[0022] Among them, q j represents the value of the power station steel plate data in the jth column without standard processing, q' j represents the value after normalization, μ j represents the average value of the jth column of the power plant steel plate data, σ j represents the standard deviation of the jth column of the power plant steel plate data, and m represents the total number of columns of the power plant steel plate data;

[0023] S1-4. Divide the data set, divide 80% of the data into a training set, and divide 20% of the data into a test set.

[0024] Preferably, the input parameters and output parameters of the selection model described in S1-1 specifically include:

[0025] The chemical composition content, cooling start temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace exit temperature and final rolling temperature are taken as the input parameters of the model;

[0026] The final cooling temperature is taken as the output parameter of the model.

[0027] Preferably, the XGBoost prediction model in S2 includes:

[0028] S2-1. Calculate the objective function based on the training data set and loss function;

[0029] Training set Data set T = {(x1,y1),(x2,y2),...,(x n ,y n )}, loss function Calculate the objective function Obj:

[0030]

[0031] Where k represents the kth CART tree, k = 1, 2, ..., K, K represents the total number of CART trees, y i Represents sample x i The true value of Represents sample x i The predicted value of

[0032] Represents the sample x of the t-1th iteration i The predicted value, f k (x i ) represents the predicted output of the kth CART tree, f t (x i ) represents the sample x of the tth iteration i The predicted output of

[0033] S2-2, introduce regularization into the objective function;

[0034]

[0035] Where Ω(f) represents the introduced regularization function, T represents the number of leaf nodes, γ represents the number of post-pruning of leaf nodes, l2 represents the 2-norm of the leaf node weight vector, λ represents the l2 regularization term, and w jrepresents the weight of the j-th leaf node, and j represents the j-th leaf node;

[0036] S2-3. Calculate the model objective function of the tth iteration:

[0037]

[0038] Among them, Ω(f k ) represents the regularization function of the k-th CART tree;

[0039] S2-4. Calculate the overall objective function of the model:

[0040]

[0041] S2-5. Expand the overall objective function obtained in S2-4 using the second-order Taylor expansion method:

[0042]

[0043] Among them, g i represents the first-order partial derivative after Taylor expansion h i represents the second-order partial derivative after Taylor expansion

[0044]

[0045] S2-6. The iteration of the tree model is transformed into the iteration of the leaf nodes of the tree, and the objective function is transformed into:

[0046]

[0047] S2-7, omit the constant term and simplify the objective function to:

[0048]

[0049] S2-8. All samples x belonging to the jth leaf node i The sample set I that is assigned to a leaf node j In this case, all samples are grouped according to leaf nodes, and finally we get:

[0050]

[0051] S2-9, Order Get the optimal solution of the objective function And the optimal solution Obj:

[0052]

[0053] Preferably, the specific method of optimizing and adjusting the hyperparameters of the XGBoost prediction model by using the Bayesian theorem in S2 to construct a prediction model for the final cooling temperature of a power plant steel plate based on BO-XGBoost includes:

[0054] Calculate the prior function f(x) and the acquisition function f PI (x):

[0055] f(x)~gp(m(x),k(x,x'))

[0056]

[0057] Among them, m(x) represents the mean function, k(x,x') represents the covariance function, gp(·) represents the Gaussian regression function, and u(x) represents the mean of the objective function value.

[0058] σ(x) represents the variance of the objective function value, f(x + ) represents the optimal objective function value, ξ represents the parameter, and Φ() represents the normal distribution cumulative density function;

[0059] Compute Bayes' theorem:

[0060]

[0061] Where f represents the objective function, D 1:t represents the observed set, p(f|D 1:t ) represents the likelihood distribution of y, p(f) represents the prior probability distribution of f, and p(D 1:t ) represents the marginal likelihood distribution of the marginalization f;

[0062] Calculate the optimal parameter combination x * :

[0063]

[0064] Here, χ represents the decision space.

[0065] Preferably, obtaining the BO-XGBoost prediction model in S2 further includes:

[0066] Input the training set data obtained by S1 into the BO-XGBoost prediction model, obtain the BO-XGBoost prediction model and save it;

[0067] The test set data obtained by S1 is input into the BO-XGBoost prediction model to verify the accuracy of the BO-XGBoost prediction model.

[0068] Advantages of the present invention: The present invention proposes a method for predicting the final cooling temperature of a power plant steel plate production process, and proposes a BO-XGBoost model to predict the final cooling temperature of a power plant steel plate. Compared with the DNN model and the BO-GBDT model, the BO-XGBoost model has the highest prediction accuracy and the smallest prediction error. The training set MAE, RMSE and R 2 The MAE, RMSE and R of the test set are 1.47℃, 2.03℃ and 0.9947℃ respectively. 2 They are 1.97℃, 2.91℃ and 0.9886℃ respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flowchart of the method for predicting the final cooling temperature of the power station steel plate production process according to the present invention;

[0070] Figure 2 This is a schematic diagram comparing the final cooling temperature model prediction value and the actual value. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0073] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0074] Embodiment 1:

[0075] Combine the following Figure 1 This embodiment is described. The method for predicting the final cooling temperature of a power plant steel plate production process in this embodiment includes:

[0076] S1. During the cooling process of the steel plate in the power station, data of steel plate parameters during the cooling process are collected and preprocessed;

[0077] S2, using the data of the steel plate parameters during the cooling process as the input data of the XGBoost prediction model, and using the final cooling temperature of the steel plate as the output data of the XGBoost prediction model;

[0078] The Bayesian theorem is used to optimize and adjust the hyperparameters of the XGBoost prediction model, and a prediction model for the final cooling temperature of power plant steel plates based on BO-XGBoost is constructed to obtain the BO-XGBoost prediction model.

[0079] S3. Input the data of the steel plate parameters in the cooling process obtained in S1 into the BO-XGBoost prediction model to make a real-time prediction of the final cooling temperature of the power plant steel plate.

[0080] Furthermore, the cooling process parameters collected in S1 include: chemical composition content, cooling start temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace temperature and final rolling temperature;

[0081] The chemical composition content includes the content of C, Cr, Mn, Mo and Nb.

[0082] Furthermore, the specific method of preprocessing the data in S1 includes:

[0083] S1-1. Select the input parameters and output parameters of the model according to the production process;

[0084] S1-2, use the Laida criterion to eliminate outliers in the parameters;

[0085] The elimination method is:

[0086]

[0087] Among them, x i represents the i-th sample, represents the average value of all samples, S x represents the standard deviation of all samples;

[0088]

[0089] Where i = 1, 2, ..., n, n represents the total number of samples;

[0090] S1-3, the parameter data were standardized using the Z-score method;

[0091]

[0092] Among them, q j represents the value of the power station steel plate data in the jth column without standard processing, q' j represents the value after normalization, μ j represents the average value of the jth column of the power plant steel plate data, σ j represents the standard deviation of the jth column of the power plant steel plate data, and m represents the total number of columns of the power plant steel plate data;

[0093] S1-4. Divide the data set, divide 80% of the data into a training set, and divide 20% of the data into a test set.

[0094] Furthermore, the input parameters and output parameters of the selection model described in S1-1 specifically include:

[0095] The chemical composition content, cooling start temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace exit temperature and final rolling temperature are taken as the input parameters of the model;

[0096] The final cooling temperature is taken as the output parameter of the model.

[0097] Furthermore, the XGBoost prediction model in S2 includes:

[0098] S2-1. Calculate the objective function based on the training data set and loss function;

[0099] Training set Data set T = {(x1,y1),(x2,y2),...,(x n ,y n )}, loss function Calculate the objective function Obj:

[0100]

[0101] Where k represents the kth CART tree, k = 1, 2, ..., K, K represents the total number of CART trees, y i Represents sample x i The true value of Represents sample x i The predicted value of Represents the sample x of the t-1th iteration i The predicted value, f k (x i ) represents the predicted output of the kth CART tree, f t (x i ) represents the sample x of the tth iteration i The predicted output of

[0102] S2-2, introduce regularization into the objective function;

[0103]

[0104] Where Ω(f) represents the introduced regularization function, T represents the number of leaf nodes, γ represents the number of post-pruning of leaf nodes, l2 represents the 2-norm of the leaf node weight vector, λ represents the l2 regularization term, and w j represents the weight of the j-th leaf node, and j represents the j-th leaf node;

[0105] S2-3. Calculate the model objective function of the tth iteration:

[0106]

[0107] Among them, Ω(f k ) represents the regularization function of the k-th CART tree;

[0108] S2-4. Calculate the overall objective function of the model:

[0109]

[0110] S2-5. Expand the overall objective function obtained in S2-4 using the second-order Taylor expansion method:

[0111]

[0112] Among them, g i represents the first-order partial derivative after Taylor expansion h i represents the second-order partial derivative after Taylor expansion

[0113]

[0114] S2-6. The iteration of the tree model is transformed into the iteration of the leaf nodes of the tree, and the objective function is transformed into:

[0115]

[0116] S2-7, omit the constant term and simplify the objective function to:

[0117]

[0118] S2-8. All samples x belonging to the jth leaf node i The sample set I that is assigned to a leaf node j In this case, all samples are grouped according to leaf nodes, and finally we get:

[0119]

[0120] S2-9, Order Get the optimal solution of the objective function And the optimal solution Obj:

[0121]

[0122] Furthermore, the specific method of optimizing and adjusting the hyperparameters of the XGBoost prediction model by using the Bayesian theorem in S2 and constructing a prediction model for the final cooling temperature of power plant steel plates based on BO-XGBoost includes:

[0123] Calculate the prior function f(x) and the acquisition function f PI (x):

[0124] f(x)~gp(m(x),k(x,x'))

[0125]

[0126] Among them, m(x) represents the mean function, k(x,x') represents the covariance function, gp(·) represents the Gaussian regression function, u(x) represents the mean of the objective function value, σ(x) represents the variance of the objective function value, and f(x + ) represents the optimal objective function value, ξ represents the parameter, and Φ() represents the normal distribution cumulative density function;

[0127] Compute Bayes' theorem:

[0128]

[0129] Where f represents the objective function, D 1:t represents the observed set, p(f|D 1:t ) represents the likelihood distribution of y, p(f) represents the prior probability distribution of f, and p(D 1:t ) represents the marginal likelihood distribution of the marginalization f;

[0130] Calculate the optimal parameter combination x * :

[0131]

[0132] Here, χ represents the decision space.

[0133] Furthermore, obtaining the BO-XGBoost prediction model in S2 also includes:

[0134] Input the training set data obtained by S1 into the BO-XGBoost prediction model, obtain the BO-XGBoost prediction model and save it;

[0135] The test set data obtained by S1 is input into the BO-XGBoost prediction model to verify the accuracy of the BO-XGBoost prediction model.

[0136] In the present invention, the method for predicting the final cooling temperature of a power station steel plate production process of the present invention comprises the following steps:

[0137] Step 1: Data collection and preprocessing of hot rolled plate cooling process;

[0138] Step 1.1: Take 15 parameters including chemical composition content (C, Cr, Mn, Mo, Nb), cooling temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace temperature, and final rolling temperature as the input of the model, and the final cooling temperature as the output of the model;

[0139] Step 1.2: Use the Pauta criterion to remove outliers. After removal, a total of 31,971 data samples remain.

[0140]

[0141]

[0142] Where: and S y are the mean and standard deviation of all samples respectively; L is the number of samples; y i is the i-th sample;

[0143] Step 1.3: Use the Z-score method to standardize the power station steel plate data;

[0144]

[0145] Where: x i is the value of the unstandardized power plant steel plate data in the i-th column; x i ' is the standardized value; μ i is the average value of the data in the i-th column of the power station steel plate data; σ i is the standard deviation of the data in the i-th column of the power station steel plate data;

[0146] Step 1.4: Randomly divide the data set into 80% of the data as the training set and 20% of the data as the test set.

[0147] Step 2: Construct a prediction model for the final cooling temperature of power plant steel plates based on XGBoost;

[0148] Step 2.1: Based on the training data set T = {(x1,y1),(x2,y2),...,(x n ,y n )}、Loss function Calculate the objective function:

[0149]

[0150] Where: i is the number of samples; k is the kth tree; y i and The samples x arei The true value and predicted value of f k (x i ) is the predicted output of the kth CART tree.

[0151] Step 2.2: Introduce regularization into the objective function:

[0152]

[0153] Where: γT is the number of leaf nodes; l2 is the 2-norm of the leaf node weight vector.

[0154] Step 2.3: Calculate the model objective function at the tth iteration:

[0155]

[0156] Step 2.4: Calculate the overall objective function of the model:

[0157]

[0158] Step 2.5: The objective function adopts the second-order Taylor expansion:

[0159]

[0160] Where: g i and h i They are the first-order partial derivatives after Taylor expansion. and the second-order partial derivative

[0161] Step 2.6: Convert the iteration of the tree model into the iteration of the leaf nodes of the tree, and the objective function is transformed:

[0162]

[0163] Step 2.7: The regularization term involving t-1 trees is omitted as a constant, and the objective function is simplified:

[0164]

[0165] Step 2.8: All samples x belonging to the jth leaf node i The sample set I that is assigned to a leaf node j In this case, all samples are grouped according to leaf nodes, and finally we get:

[0166]

[0167] Where: w j is the weight of leaf node j; and They respectively represent the cumulative sum of the first-order partial derivatives and the cumulative sum of the second-order partial derivatives of the samples contained in the leaf node j.

[0168] Step 2.9: Finally, let The optimal solution point and optimal solution of the objective function are as follows:

[0169]

[0170] Step 3: Use Bayesian optimization to adjust the hyperparameters of the XGBoost model and build a prediction model for the final cooling temperature of power plant steel plates based on BO-XGBoost;

[0171] Step 3.1: Calculate the prior function and acquisition function:

[0172] f(x)~gp(m(x),k(x,x'))

[0173]

[0174] Where: m(x) is the mean function, m(x) = E[f(x)]. For simplicity, the mean function m(x) = 0 is usually set; k(x, x') is the covariance function, k(x, x') = E[(f(x)-m(x))(f(x')-m(x'))]; u(x) and σ(x) are the mean and variance of the objective function value respectively; f(x + ) is the optimal objective function value; ξ is the parameter; Φ(*) is the normal distribution cumulative density function.

[0175] Step 3.2: Calculate Bayes' law:

[0176]

[0177] Where: f is the objective function; D 1:t ={(x1,y1),(x2,y2),...,(x t ,y t )} is the observed set; x t is the decision vector; y t is the observed value; p(D 1:t |f) is the likelihood distribution of y; p(f) is the prior probability distribution of f; p(D 1:t ) is the marginal likelihood distribution of the marginalization f.

[0178] Step 3.3: Calculate the optimal parameter combination. The final BO-XGBoost model parameters are shown in Table 1:

[0179]

[0180] Where: x *is the optimal parameter combination; χ is the decision space; f(x) is the objective function.

[0181] Table 1

[0182] XGBoost Parameters Parameter meaning Search Space Optimal parameter values learning_rate Learning Rate (0.1,1) 0.3583 n_estimators Number of decision trees (0,1000) 1000 max_depth Maximum tree depth (1,10) 3 min_child_weight Minimum leaf node sample weight and (1,10) 3

[0183] Step 4: Train and save the power plant steel plate final cooling temperature prediction model based on BO-XGBoost;

[0184] Step 4.1: Input the training set data obtained in step 1 into the BO-XGBoost model to obtain the BO-XGBoost prediction model and save it;

[0185] Step 4.2: Input the test set data obtained in step 1 into the BO-XGBoost model obtained in step 4.1, and use the mean absolute error (MAE), root mean square absolute error (RMSE) and determination coefficient (R2) to verify that the prediction accuracy is qualified. Otherwise, repeat steps 3.1 to 4.1 until the requirements are met. By comparing the DNN model, BO-GBDT model, and XGBoost, the prediction results are as follows: Figure 2 As shown, it is shown that the BO-XGBoost model in the present invention has the highest prediction result.

[0186] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A method for predicting the final cooling temperature of a power plant steel plate production process, characterized in that: It includes: S1. During the cooling process of the steel plate in the power station, collect the data of the steel plate parameters during the cooling process and pre-process the data; S2, using the data of the steel plate parameters during the cooling process as the input data of the XGBoost prediction model, and using the final cooling temperature of the steel plate as the output data of the XGBoost prediction model; The Bayesian theorem is used to optimize and adjust the hyperparameters of the XGBoost prediction model, and a prediction model for the final cooling temperature of power plant steel plates based on BO-XGBoost is constructed to obtain the BO-XGBoost prediction model. S3. Input the data of the steel plate parameters in the cooling process obtained in S1 into the BO-XGBoost prediction model to make a real-time prediction of the final cooling temperature of the power plant steel plate.

2. The method for predicting the final cooling temperature of a power plant steel plate production process according to claim 1, characterized in that: S1 The collected cooling process parameters include: chemical composition content, cooling start temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace temperature and final rolling temperature; The chemical composition content includes the content of C, Cr, Mn, Mo and Nb.

3. The method for predicting the final cooling temperature of a power plant steel plate production process according to claim 2, characterized in that: The specific method for preprocessing data in S1 includes: S1-1. Select the input parameters and output parameters of the model according to the production process; S1-2, use the Laida criterion to eliminate outliers in the parameters; The elimination method is: Among them, x i represents the i-th sample, x represents the average of all samples, S x represents the standard deviation of all samples; Where i = 1, 2, ..., n, n represents the total number of samples; S1-3, the parameter data were standardized using the Z-score method; Among them, q j represents the value of the power station steel plate data in the jth column without standard processing, q' j represents the value after normalization, μ j represents the average value of the jth column of the power plant steel plate data, σ j represents the standard deviation of the jth column of the power plant steel plate data, and m represents the total number of columns of the power plant steel plate data; S1-4. Divide the data set, divide 80% of the data into a training set, and divide 20% of the data into a test set.

4. The method for predicting the final cooling temperature of a power plant steel plate production process according to claim 3, characterized in that: The input parameters and output parameters of the selection model described in S1-1 specifically include: The chemical composition content, cooling temperature, cooling rate, number of headers, steel plate speed, steel plate thickness, steel plate length, water temperature, total flow, furnace temperature and final rolling temperature are taken as the input parameters of the model; The final cooling temperature is taken as the output parameter of the model.

5. The method for predicting the final cooling temperature of a power station steel plate production process according to claim 3 or 4, characterized in that: S2 The XGBoost prediction model includes: S2-1. Calculate the objective function based on the training data set and loss function; Training set Data set T = {(x1,y1),(x2,y2),...,(x n ,y n )}, loss function Calculate the objective function Obj: Where k represents the kth CART tree, k = 1, 2, ..., K, K represents the total number of CART trees, y i Represents sample x i The true value of Represents sample x i The predicted value of Represents the sample x of the t-1th iteration i The predicted value, f k (x i ) represents the predicted output of the kth CART tree, f t (x i ) represents the sample x of the tth iteration i The predicted output of S2-2, introduce regularization into the objective function; Where Ω(f) represents the introduced regularization function, T represents the number of leaf nodes, γ represents the number of post-pruning of leaf nodes, l2 represents the 2-norm of the leaf node weight vector, λ represents the l2 regularization term, and w j represents the weight of the j-th leaf node, and j represents the j-th leaf node; S2-3. Calculate the model objective function of the tth iteration: Among them, Ω(f k ) represents the regularization function of the k-th CART tree; S2-4. Calculate the overall objective function of the model: S2-5. Expand the overall objective function obtained in S2-4 using the second-order Taylor expansion method: Among them, g i represents the first-order partial derivative after Taylor expansion h i represents the second-order partial derivative after Taylor expansion S2-6. The iteration of the tree model is transformed into the iteration of the leaf nodes of the tree, and the objective function is transformed into: S2-7, omit the constant term and simplify the objective function to: S2-8. All samples x belonging to the jth leaf node i The sample set I that is assigned to a leaf node j In this case, all samples are grouped according to leaf nodes, and finally we get: S2-9, Order Get the optimal solution of the objective function And the optimal solution Obj:

6. The method for predicting the final cooling temperature of a power station steel plate production process according to claim 5, characterized in that: S2 uses the Bayesian theorem to optimize and adjust the hyperparameters of the XGBoost prediction model, and the specific method of constructing a prediction model for the final cooling temperature of the power plant steel plate based on BO-XGBoost includes: Calculate the prior function f(x) and the acquisition function f PI (x): f(x)~gp(m(x),k(x,x')) Among them, m(x) represents the mean function, k(x,x') represents the covariance function, gp(·) represents the Gaussian regression function, u(x) represents the mean of the objective function value, σ(x) represents the variance of the objective function value, and f(x + ) represents the optimal objective function value, ξ represents the parameter, and Φ() represents the normal distribution cumulative density function; Compute Bayes' theorem: Where f represents the objective function, D 1:t represents the observed set, p(f|D 1:t ) represents the likelihood distribution of y, p(f) represents the prior probability distribution of f, and p(D 1:t ) represents the marginal likelihood distribution of the marginalization f; Calculate the optimal parameter combination x * : Here, χ represents the decision space.

7. The method for predicting the final cooling temperature of a power station steel plate production process according to claim 1, characterized in that: S2 further includes obtaining the BO-XGBoost prediction model: Input the training set data obtained by S1 into the BO-XGBoost prediction model, obtain the BO-XGBoost prediction model and save it; The test set data obtained by S1 is input into the BO-XGBoost prediction model to verify the accuracy of the BO-XGBoost prediction model.

Images