A difficult-to-push focus warning method based on RNN model

Abstract

The present invention discloses a method for early warning of difficult coke pushing based on an RNN model, which relates to the technical field of bearing seat production and comprises the following steps: S1: data preprocessing of operating parameters and state parameters of the coke oven production process; S2: mechanism analysis combined with correlation analysis to select the main influencing factors of the coke pushing current; S3: principal component analysis is used to reduce the dimension of the data, and the reduced data is used as the input of the RNN network. The output result is denormalized to obtain the predicted value of the coke pushing current; S4: combining historical data and expert experience to set the coke pushing current warning threshold to achieve a difficult coke pushing warning; by using the principal component analysis method to reduce the coupling between the input data of the RNN network model, the PCA method is used to improve the prediction accuracy and prediction efficiency of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of coke ovens, and in particular to a coke pushing difficulty early warning method based on an RNN model. Background Art

[0002] Most existing research on early warning of difficult coke pushing conditions starts with historical data of coke pushing current. This data is processed using appropriate data-driven methods, and artificial neural network technology is used to predict the pushing current. Based on expert experience, the pushing current threshold is set to achieve early warning of difficult coke pushing during the coke oven production process.

[0003] The coking process has a complex mechanism and is affected by numerous factors. Existing research on early warning of difficult coke pushing lacks research on the mechanism of the coke oven production process. Most studies only consider the historical data of the coke oven production process when selecting the input of the prediction model, which reduces the rationality of the model. At the same time, most studies do not consider the input dimension of the model when dealing with the problems of model calculation accuracy and efficiency, resulting in the model's calculation efficiency and prediction accuracy failing to meet industrial production requirements. Summary of the Invention

[0004] In order to overcome the above-mentioned technical problems, the purpose of the present invention is to provide a difficult-to-push focus warning method based on an RNN model, so as to solve the problem that the input dimension of the model is not considered in the existing technology for difficult-to-push focus warning, resulting in the model's computing efficiency and prediction accuracy failing to meet industrial production requirements.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for early warning of difficult-to-push coke based on an RNN model, the automatic blanking production method comprises the following steps:

[0007] S1: Data preprocessing of the operating parameters and state parameters of the coke oven production process.

[0008] S2: Combine mechanism analysis with correlation analysis to select the main influencing factors of the pushing current.

[0009] S3: Use principal component analysis to reduce the data dimension, and use the reduced data as the input of the RNN network. Denormalize the output result to obtain the predicted value of the push current.

[0010] S4: Combine historical data and expert experience to set the push current warning threshold to achieve difficult push warning.

[0011] As a further solution of the present invention: step S1 further includes the following steps:

[0012] A1: Select the parameter data of the coke pushing process within one year and use the data interpolation method to process the abnormal data caused by sensor failure and downtime.

[0013] A2: Use the Min-Max normalization method to normalize the data after abnormal data processing. The normalization formula is as follows:

[0014]

[0015] Among them, x N is the normalized expected value; x min is the minimum value in the sample data; x max is the maximum value in the sample data.

[0016] As a further solution of the present invention: step S2 further includes the following steps:

[0017] B1: The mechanism of the coke pushing process is analyzed and expert experience is taken into consideration. The main factors affecting the coke pushing current are the total number of coal charging revolutions, coal leveling current, coking time, actual coal hopper ratio, coke oven temperature, coal quality, and coal moisture content.

[0018] B2: Perform Pearson correlation analysis on the pre-processed focus pushing process parameter data. Pearson correlation analysis can be replaced by Spearman correlation analysis and Kendall correlation analysis. Set the correlation threshold ρ = 0.55. Combined with the mechanism analysis considering expert experience in B1, parameters with correlation greater than ρ and a large impact on the focus pushing current are selected as the objects of PCA processing. The calculation formula of Pearson correlation coefficient is as follows:

[0019]

[0020] Where x i and y i are the variable to be calculated and the i-th sampling value of the carbonization chamber pushing current respectively; is the mean value of the variable to be calculated; It is the average value of the coke pushing current in the carbonization chamber.

[0021] As a further embodiment of the present invention: and The calculation formula is as follows:

[0022]

[0023] As a further solution of the present invention: step S3 further includes the following steps:

[0024] C1: Construct the matrix M with the selected parameters;

[0025] C2: Calculate the covariance and correlation coefficient r of the parameter x(i) ij , the calculation formula is as follows:

[0026]

[0027]

[0028] Among them, cov(x i ,x j ) is the covariance between the i-th parameter and the j-th parameter, n is the number of parameter samples, r ij is the correlation coefficient between the two parameters;

[0029] C3: Calculate the value of r ij The eigenvalue λ of the correlation coefficient matrix R i and the corresponding eigenvectors

[0030]

[0031] C4: Set the principal component approximation threshold θ and calculate the variance contribution rate β of each component i and cumulative contribution rate γ v , select some parameters with cumulative contribution rate greater than 85% as the input of RNN neural network, and the calculation formula is as follows;

[0032]

[0033]

[0034] where β i is the variance contribution rate of the i-th parameter, γ v is the cumulative variance contribution rate of the first v parameters, and the IMF component is calculated based on β i In descending order, when γ v When ≥θ, PCA ends, and at this time, the previous parameter matrix is ​​used as the input of the focus current prediction model.

[0035] Beneficial effects of the present invention:

[0036] 1. In the present invention, the data interpolation method and the Pearson correlation coefficient method are used to screen the influencing factors of blast furnace gas generation. Compared with the existing technical solutions, the mechanism of the coke oven production process is fully considered.

[0037] 2. In the present invention, the principal component analysis method is used to reduce the coupling between the input data of the RNN network model. Compared with the existing technical solutions, the PCA method improves the prediction accuracy and prediction efficiency of the model.

[0038] 3. In the present invention, the difficulty in pushing coke in the coke oven is judged based on historical data and expert experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be further described below with reference to the accompanying drawings.

[0040] Figure 1 It is a process flow diagram of the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making any creative efforts shall fall within the scope of protection of the present invention.

[0042] like Figure 1 As shown, a difficult-to-push coke warning method based on an RNN model, the automatic blanking production method includes the following steps:

[0043] S1: Data preprocessing of the operating parameters and state parameters of the coke oven production process;

[0044] The specific steps include:

[0045] A1: Select the parameter data of the coke pushing process within one year and use the data interpolation method to process the abnormal data caused by sensor failure and downtime.

[0046] A2: Use the Min-Max normalization method to normalize the data after abnormal data processing. The normalization formula is as follows:

[0047]

[0048] Among them, x N is the normalized expected value; x min is the minimum value in the sample data; x max is the maximum value in the sample data;

[0049] S2: Mechanism analysis combined with correlation analysis to select the main influencing factors of the pushing current;

[0050] The specific steps include:

[0051] B1: The mechanism of the coke pushing process is analyzed and expert experience is taken into consideration. The main factors affecting the coke pushing current are the total number of coal charging revolutions, coal leveling current, coking time, actual coal hopper ratio, coke oven temperature, coal quality, and coal moisture content.

[0052] B2: Perform Pearson correlation analysis on the pre-processed focus pushing process parameter data. Pearson correlation analysis can be replaced by Spearman correlation analysis and Kendall correlation analysis. Set the correlation threshold ρ = 0.55. Combined with the mechanism analysis considering expert experience in B1, parameters with correlation greater than ρ and a large impact on the focus pushing current are selected as the objects of PCA processing. The calculation formula of Pearson correlation coefficient is as follows:

[0053]

[0054] Where x i and y i are the variables to be calculated and the i-th sampling value of the carbonization chamber pushing current respectively; is the mean value of the variable to be calculated; is the average value of the coke pushing current in the carbonization chamber;

[0055] and The calculation formula is as follows:

[0056]

[0057] S3: Use principal component analysis to reduce the data dimension, and use the reduced data as the input of the RNN network. Denormalize the output result to obtain the predicted value of the push current;

[0058] The specific steps include:

[0059] C1: Construct the matrix M with the selected parameters;

[0060] C2: Calculate the covariance and correlation coefficient r of the parameter x(i) ij , the calculation formula is as follows:

[0061]

[0062]

[0063] Among them, cov(x i ,x j ) is the covariance between the i-th parameter and the j-th parameter, n is the number of parameter samples, r ij is the correlation coefficient between the two parameters;

[0064] C3: Calculate the value of r ij The eigenvalue λ of the correlation coefficient matrix R i and the corresponding eigenvectors

[0065]

[0066] C4: Set the principal component approximation threshold θ and calculate the variance contribution rate β of each component i and cumulative contribution rate γ v , select some parameters with cumulative contribution rate greater than 85% as the input of RNN neural network, and the calculation formula is as follows;

[0067]

[0068]

[0069] where β i is the variance contribution rate of the i-th parameter, γ v is the cumulative variance contribution rate of the first v parameters, and the IMF component is calculated based on β i In descending order, when γ v When ≥θ, PCA ends, and at this time, the previous parameter matrix is ​​used as the input of the focus current prediction model;

[0070] S4: Combine historical data and expert experience to set the current warning threshold for pushing the coke, and implement early warning for difficult pushing the coke;

[0071] The above method is used to predict the peak current and the actual peak current. The peak currents at 185h, 370h, 555h, 740h and 925h are predicted respectively, and the actual peak currents at 185h, 370h, 555h, 740h and 925h are monitored, as shown in the following table:

[0072] Time (18.5 hours) 10 20 30 40 50 Predicted value (A) 203 241 265 275 190 True value (A) 204 240 265 260 188

[0073] From the data in the above table, we can see that the data interpolation method and the Pearson correlation coefficient method are used to screen the influencing factors of blast furnace gas generation to predict the peak value of coke pushing current. The results are consistent with the actual peak value data of coke pushing current, with high accuracy, meeting the requirements of industrial production.

[0074] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for early warning of difficult-to-push focus based on RNN model, characterized in that: The automatic blanking production method includes the following steps: S1: Data preprocessing of the operating parameters and state parameters of the coke oven production process; S2: Mechanism analysis combined with correlation analysis to select the main influencing factors of the pushing current; S3: Use principal component analysis to reduce the data dimension. The reduced data is used as the input of the RNN network. The output result is denormalized to obtain the predicted value of the push current. S4: Combine historical data and expert experience to set the pre-warning threshold for the current of the coke push, so as to realize the pre-warning of the coke push difficulty. Step S2 further includes the following steps: B1: Analyze the mechanism of the coke pushing process and consider expert experience. The main factors affecting the coke pushing current are the total number of revolutions of coal charging, coal leveling current, coking time, actual coal hopper ratio, coke oven temperature, coal quality, and coal moisture content. B2: Perform Pearson correlation analysis on the pre-processed focus pushing process parameter data, set the correlation threshold ρ = 0.55, and combine the mechanism analysis in B1. Parameters with a correlation greater than ρ and a large impact on the focus pushing current are selected as the objects of PCA processing. The calculation formula of the Pearson correlation coefficient is as follows: Where x i and y i are the variables to be calculated and the i-th sampling value of the carbonization chamber pushing current respectively; is the mean value of the variable to be calculated; It is the average value of the coke pushing current in the carbonization chamber.

2. The method for early warning of difficult-to-push focus based on RNN model according to claim 1, characterized in that: Step S1 further includes the following steps: A1: Select one year's worth of coke pushing process parameter data and use data interpolation to process abnormal data caused by sensor failures and downtime. A2: Use the Min-Max normalization method to normalize the data after abnormal data processing. The normalization formula is as follows: Among them, x N is the normalized expected value; x min is the minimum value in the sample data; x max is the maximum value in the sample data.

3. The method for early warning of difficult-to-push focus based on RNN model according to claim 1, characterized in that: and The calculation formula is as follows:

4. The method for early warning of difficult focus pushing based on RNN model according to claim 1, characterized in that: Step S3 further includes the following steps: C1: Construct the matrix M with the selected parameters; C2: Calculate the covariance and correlation coefficient r of the parameter x(i) ij , the calculation formula is as follows: Among them, cov(x i ,x j ) is the covariance between the i-th parameter and the j-th parameter, n is the number of parameter samples, r ij is the correlation coefficient between the two parameters; C3: Calculate the value of r ij The eigenvalue λ of the correlation coefficient matrix R i and the corresponding eigenvectors C4: Set the principal component approximation threshold θ and calculate the variance contribution rate β of each component i and cumulative contribution rate γ v , select some parameters with cumulative contribution rate greater than 85% as the input of RNN neural network, and the calculation formula is as follows; where β i is the variance contribution rate of the i-th parameter, γ v is the cumulative variance contribution rate of the first v parameters, and the IMF component is calculated based on β i In descending order, when γ v When ≥θ, PCA ends, and at this time, the previous parameter matrix is ​​used as the input of the focus current prediction model.

Images