WO2024060287A1 - Blast furnace temperature prediction method, terminal device, and storage medium - Google Patents

Abstract

The present invention relates to a blast furnace temperature prediction method, a terminal device, and a storage medium. The method comprises: collecting values of various control parameters and state parameters of a blast furnace within a historical time period; performing lag correlation analysis on all the parameters to obtain feature parameters from all the parameters; determining a sequence length of a network according to the degree of autocorrelation of a molten iron temperature in the blast furnace; determining input dimensions of recurrent units in the network according to the obtained feature parameters; on the basis of the determined sequence length and input dimensions, extracting training data from collected data to form a training set; constructing an LSTM network model on the basis of the sequence length and the input dimensions, and training the model by using the feature parameters of the blast furnace at a certain time point as an input and a blast furnace temperature at said time point as an output of the model; and predicting the blast furnace temperature by means of the trained model. The present invention improves the accuracy of existing blast furnace temperature prediction.

Description

A blast furnace temperature prediction method, terminal equipment and storage medium Technical field

The invention relates to the field of blast furnace ironmaking, and in particular to a blast furnace temperature prediction method, terminal equipment and storage medium.

Background technique

As the main upstream process of steel manufacturing, blast furnace ironmaking is an important component of the entire steel industry. It plays a very important role in both the development of the entire industry and the overall energy conservation and emission reduction of the industry. Whether the blast furnace ironmaking conditions are running smoothly is related to whether the entire ironmaking process is efficient and energy-saving. The blast furnace temperature is an important indicator for identifying the blast furnace conditions. Generally, the furnace temperature can be controlled to ensure the smooth progress of the blast furnace conditions. Establishing a reliable and accurate furnace temperature prediction model to guide blast furnace ironmaking workers in furnace temperature control is not only a theoretical research, but also has great guiding significance for the production practice of the steel industry.

The furnace temperature prediction models currently used are mainly for furnace ironmaking researchers to calculate the mechanism from the perspective of heat balance and material balance in the blast furnace production process, and thereby derive the corresponding mechanism model; or rely on the accumulation of a large amount of practical experience. Based on the generated expert knowledge, the researchers developed corresponding qualitative reasoning types. These methods are mainly derived from physical processes or experience and are interpretable, but they still cannot accurately explain the changes in furnace temperature under complex blast furnace production conditions.

Contents of the invention

In order to solve the above problems, the present invention proposes a blast furnace temperature prediction method, terminal equipment and storage medium.

The specific plans are as follows:

A method for predicting blast furnace temperature comprises the following steps:

S1: Collect the values of various control parameters and status parameters of the blast furnace during the historical time period;

S2: Perform lagged correlation analysis on all parameters to obtain characteristic parameters from all parameters;

S3: Determine the sequence length of the LSTM network based on the degree of autocorrelation of the molten iron temperature in the blast furnace; determine the input dimensions of each cycle unit in the LSTM network based on the obtained characteristic parameters;

S4: Based on the determined sequence length and input dimension, extract training data from the data collected in step S1 to form a training set;

S5: Construct an LSTM network model based on sequence length and input dimensions, take the characteristic parameters of the blast furnace at a certain moment as input, and the blast furnace temperature at that moment as the output of the model, and train the model;

S6: Predict the blast furnace temperature through the trained model.

Further, step S1 also includes preprocessing the collected data. The preprocessing includes: performing normal distribution verification on various parameters, eliminating parameters that do not conform to the normal distribution, and performing anomaly analysis on the parameters that conform to the normal distribution. Value elimination and missing value completion are performed, and finally the values of all parameters that conform to the normal distribution are standardized to make the value ranges of different parameters consistent in magnitude.

Further, the method for obtaining the characteristic parameters in step S2 is to use the principal component analysis method to map all parameters into a low-dimensional subspace through linear transformation to obtain the dimensionally reduced characteristic parameters.

Further, in step S3, the autocorrelation function ACF and the partial autocorrelation function PACF are used to jointly determine the molten iron temperatures at the first t moments that are most relevant to the current molten iron temperature. The sequence length is the number of molten iron temperatures that are most relevant to the current molten iron temperature. t.

A blast furnace temperature prediction terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above embodiments of the present invention are implemented. steps of the method.

A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the method described above in the embodiment of the present invention are implemented.

The present invention adopts the above technical solution to improve the accuracy of prediction of the existing blast furnace temperature, and solves the problem that the existing system does not fully accurately simulate the complex production process in the blast furnace when inferring the furnace temperature through the heat balance process or based on empirical rules. The result is not as expected.

Description of the drawings

Figure 1 shows a flow chart of Embodiment 1 of the present invention.

Figure 2 is a comparison diagram between the model prediction results and the actual results in this embodiment.

Detailed ways

To further explain various embodiments, the present invention provides drawings. These drawings are part of the disclosure of the present invention, and are mainly used to illustrate the embodiments, and can be used to explain the operating principles of the embodiments in conjunction with the relevant descriptions in the specification. With reference to these contents, those of ordinary skill in the art will be able to understand other possible implementations and advantages of the present invention.

The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

Example 1:

An embodiment of the present invention provides a blast furnace temperature prediction method, as shown in Figure 1. The method includes the following steps:

S1: Collect the values of various control parameters and status parameters of the blast furnace during the historical time period.

In this embodiment, the control parameters include blast furnace hot air temperature, oxygen enrichment rate, etc., and the state parameters include top temperature, top pressure, etc.

Further, this embodiment also includes preprocessing the collected data. The preprocessing includes: performing normal distribution verification on various parameters, eliminating parameters that do not conform to the normal distribution, and performing verification on the parameters that conform to the normal distribution. Outlier elimination and missing value completion are performed, and finally the values of all parameters that conform to the normal distribution are standardized to make the value ranges of different parameters consistent in order of magnitude to speed up the network learning process.

S2: Perform lagged correlation analysis on all parameters to filter out characteristic parameters from all parameters.

The specific lag correlation analysis algorithm is:

Mutual information entropy is used to measure the dimensionless statistic (generally unit is bit) that a random variable The degree of reduction in uncertainty in Y can therefore be used as a measure of the degree of correlation between X and Y.

Principal component analysis (PCA) is used to map data to a dimensionality reduction method in a low-dimensional subspace through linear transformation, while preventing information loss as much as possible. Subtract the respective average values from each dimensional feature, calculate the covariance matrix, and obtain the eigenvalues and eigenvectors of the covariance matrix. Sort the eigenvalues from large to small, and select the eigenvectors corresponding to the largest k eigenvalues. Finally, the original data is converted into a new space constructed by k feature vectors. In the new space, each principal component is a linear combination of the original variables, and the principal components are uncorrelated with each other. The number of principal components is much less than the number of original variables, and most of the information of the original variables is retained.

The specific process of obtaining characteristic parameters is as follows:

After performing lag correlation analysis on all parameter sequences, the parameter lag sequence set {X ₁ ,...,X _n } is obtained. The principal component analysis method PCA is used to reduce the dimensionality of the original n input feature sequences to obtain a new feature input sequence {X ₁ ',...,X _k '}, how many characteristic parameters are obtained, then how many input dimensions are there.

S3: Determine the sequence length of the LSTM network based on the degree of autocorrelation of the molten iron temperature in the blast furnace; determine the input dimensions of each cycle unit in the LSTM network based on the obtained characteristic parameters.

Since the molten iron temperature is continuous, that is, the value at a certain moment has a certain correlation with the values at the previous moments, through autocorrelation analysis, the molten iron temperature at the previous t moments that is most correlated with the current molten iron temperature can be determined, thereby determining the input length t of the network.

The specific implementation process is:

The autocorrelation function ACF and the partial autocorrelation function PACF are used to determine the most concerning t lagging effects of the molten iron temperature itself. ACF describes the degree of correlation between the current value of a series and its past values, while PACF finds the correlation of the residual (which remains after removing the effects already explained by previous lags) to the next lag value. Assume that the lag sequence set corresponding to the sequence X _t of a certain parameter at time t is {X _t-1 ,...,X _tm }. Calculate the mutual information entropy {I _{t- 1} ,…,I _tm }, according to the maximum mutual information entropy I _tj , the lag sequence of the lag time j of the parameter is selected as the sequence length.

S4: Based on the determined sequence length and input dimension, training data is extracted from the data collected in step S1 to form a training set.

The training data includes the value of the characteristic parameter of each sampling point within a period of time, and the temperature of the molten iron at each sampling point. The value of the characteristic parameter is used as the input of the model, and the temperature of the molten iron is used as the output of the model.

S5: Construct an LSTM network model based on the sequence length and input dimension, take the characteristic parameters of the blast furnace at a certain time (t time) as input, and the blast furnace temperature at that time (t time) as the output of the model, and train the model.

LSTM (Long Short-Term Memory) is an RNN with long memory features, which effectively solves the gradient problem that occurs during training of recurrent neural networks. The uniqueness of LSTM is that it has a storage unit dedicated to memory, which is protected by some gate neurons. These gate neurons are different from ordinary neurons in that they have two states: on and off. When they are in the on state, the connection weight is 1; otherwise, the weight is 0. The simplest LSTM contains the following three gate neurons with different functions: keep (save gate/forget gate): When the save gate is open, the contents of the storage unit memory are not cleared; when the save gate is turned off, the previously memorized contents are cleared.

LSTM introduces a gate mechanism to control the circulation and loss of features. Assume _{that the} network input at time _t is the input _feature The unit state C _t at time t is:

C _t ＝f _t *C _t-1 +i _t *C _t '

Among them, f _t is called the forgetting gate, indicating which dimensions of C _t-1 are used to calculate C _t-1 ; C _t ' represents the unit state update value, which is composed of the input data X _t and hidden node h _t-1 through a neural network layer gets.

C _t '=tanh(W _C *[h _t-1 ,X _t ]+b _C )

i _t is the input gate. Like f _t , it is also a vector whose elements are in the interval [0,1]. It is also calculated from X _t and h _t-1 .

i _t =sigmoid(W _i *[h _t-1 ,X _t ]+b _i )

Among them, W is the network weight and b is the bias term.

Finally, in order to calculate the predicted value and generate the complete input of the next time slice, the hidden layer node output h _t needs to be calculated:

o _t =sigmoid(W _o *[h _t-1 ,X _t ]+b _o )

h _t =o _t *tanh(C _t )

The long short-term memory neural network can not only remember short-term information, but also solve the problem of losing the ability to capture long-term dependencies due to gradient explosion in ordinary recurrent networks through the gate structure.

In this embodiment, the loss function of the model is set as the root mean square error RMSE, and its calculation formula is:

Through iterative training, the loss function is minimized.

S6: Predict the blast furnace temperature through the trained model.

When making predictions, the collected characteristic parameters at the current moment need to be input into the trained model, and the model predicts and outputs the blast furnace temperature corresponding to the current moment based on the input state at the previous moment.

The comparison diagram between the model prediction results and the actual results in this embodiment is shown in Figure 2. The embodiments of the present invention improve the accuracy of prediction of the existing blast furnace temperature, and solve the problem that the existing system does not fully accurately simulate the complex production process in the blast furnace when inferring the furnace temperature through the heat balance process or based on empirical rules, resulting in inconsistent results. Anticipated questions.

Example 2:

The invention also provides a blast furnace temperature prediction terminal device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the present invention is realized. The steps in the above method embodiment of the first embodiment of the invention.

Further, as an executable solution, the blast furnace temperature prediction terminal device can be a computing device such as a desktop computer, notebook, PDA, cloud server, etc. The blast furnace temperature prediction terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above-mentioned composition structure of the blast furnace temperature prediction terminal equipment is only an example of the blast furnace temperature prediction terminal equipment and does not constitute a limitation on the blast furnace temperature prediction terminal equipment. It may include more or less than the above. components, or a combination of certain components, or different components. For example, the blast furnace temperature prediction terminal device may also include input and output devices, network access devices, buses, etc., which are not limited in the embodiment of the present invention.

Further, as an executable solution, the so-called processor can be a central processing unit (Central Processing Unit, CPU), or other general-purpose processor, digital signal processor (Digital Signal Processor, DSP), application-specific integrated circuit ( Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general processor can be a microprocessor or the processor can be any conventional processor, etc. The processor is the control center of the blast furnace temperature prediction terminal equipment and uses various interfaces and lines to connect the entire blast furnace temperature Predict various parts of the end device.

The memory may be used to store the computer program and/or module, and the processor implements the blast furnace by running or executing the computer program and/or module stored in the memory, and calling data stored in the memory. Various functions of furnace temperature prediction terminal equipment. The memory may mainly include a stored program area and a stored data area, wherein the stored program area may store an operating system and at least one application required for a function; the stored data area may store data created based on the use of the mobile phone, etc. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card , Flash Card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

The present invention also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the above method in the embodiment of the present invention are implemented.

If the module/unit integrated in the blast furnace temperature prediction terminal equipment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory) and software distribution medium, etc.

Although the invention has been specifically shown and described in connection with the preferred embodiments, it will be apparent to those skilled in the art that the invention can be modified in form and detail without departing from the spirit and scope of the invention as defined by the appended claims. Various changes are made within the scope of the present invention.

Claims (6)

1. A method for predicting blast furnace temperature, characterized in that it comprises the following steps:

   S1: Collect the values of various control parameters and status parameters of the blast furnace during the historical time period;

   S2: Perform lagged correlation analysis on all parameters to obtain characteristic parameters from all parameters;

   S3: Determine the sequence length of the LSTM network based on the degree of autocorrelation of the molten iron temperature in the blast furnace; determine the input dimensions of each cycle unit in the LSTM network based on the obtained characteristic parameters;

   S4: Based on the determined sequence length and input dimension, extract training data from the data collected in step S1 to form a training set;

   S5: Build an LSTM network model based on the sequence length and input dimension, take the characteristic parameters of the blast furnace at a certain moment as input, and the blast furnace temperature at that moment as the output of the model, and train the model;

   S6: Predict the blast furnace temperature through the trained model.
2. The blast furnace temperature prediction method according to claim 1, characterized in that: step S1 also includes preprocessing the collected data, and the preprocessing includes: performing normal distribution verification on various parameters, and eliminating those that do not conform to normality. parameters of the distribution, and perform outlier elimination and missing value completion on parameters that conform to the normal distribution. Finally, the values of all parameters that conform to the normal distribution are standardized so that the value ranges of different parameters are consistent in magnitude.
3. The blast furnace temperature prediction method according to claim 1, characterized in that: the method for obtaining the characteristic parameters in step S2 is: using the principal component analysis method and linear transformation to map all parameters into a low-dimensional subspace to obtain the reduced Dimensional characteristic parameters.
4. The blast furnace temperature prediction method according to claim 1, characterized in that: in step S3, the autocorrelation function ACF and the partial autocorrelation function PACF are used to jointly determine the molten iron temperature at the first t moments most relevant to the current molten iron temperature, and the sequence length That is, the number t of molten iron temperatures most relevant to the current molten iron temperature.
5. A blast furnace temperature prediction terminal device, characterized by: including a processor, a memory, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it realizes the following: The steps of any one of the methods described in claims 1 to 4.
6. A computer-readable storage medium stores a computer program, and is characterized in that: when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images