A method, apparatus, equipment and storage medium for predicting converter steel output
By training a neural network model to predict the steel output of the converter, the problem of inaccurate prediction of the steel output in the converter process was solved, and precise control of the alloy addition amount was achieved, thereby reducing costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG IRON & STEEL CO LTD
- Filing Date
- 2023-08-16
- Publication Date
- 2026-06-30
AI Technical Summary
The lack of a mature method for predicting steel output in the converter process in the existing technology leads to inaccurate alloy addition, which increases alloy consumption and cost.
By obtaining converter production process data and corresponding steel output from historical production data, a specified type of neural network is trained to obtain a converter steel output prediction model, and this model is used to predict real-time converter process data.
It enables accurate prediction of converter steel output, reduces the amount of alloy added during the alloying process, and reduces costs.
Smart Images
Figure CN116842852B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of iron and steel smelting, and in particular to a method for predicting converter steel output. This invention also relates to a method, apparatus, equipment and computer-readable storage medium for predicting converter steel output. Background Technology
[0002] In the steelmaking process, alloying in the converter process is a crucial step in ensuring that the product composition meets requirements. The amount of alloy added in the converter process needs to be determined based on the steel output of the converter process. However, there is a lack of mature methods for predicting the steel output of the converter process in related technologies, resulting in poor accuracy of the determined output. Under such circumstances, in order to ensure the minimum requirements of each element composition in the converter process, it is necessary to appropriately increase the amount of alloy added, thereby increasing the alloy consumption in the converter process and increasing costs.
[0003] Therefore, how to provide a solution to the above-mentioned technical problems is a problem that needs to be solved by those skilled in the art. Summary of the Invention
[0004] The purpose of this invention is to provide a method for predicting converter steel output. By using a trained converter steel output prediction model to predict the converter steel output, an accurate converter steel output can be obtained, which helps to reduce the amount of alloy added during the alloying process in the converter and lowers costs. Another purpose of this invention is to provide a method, apparatus, equipment, and computer-readable storage medium for predicting converter steel output. By using a trained converter steel output prediction model to predict the converter steel output, an accurate converter steel output can be obtained, which helps to reduce the amount of alloy added during the alloying process in the converter and lowers costs.
[0005] To solve the above-mentioned technical problems, the present invention provides a method for predicting converter steel output, comprising:
[0006] Several sets of converter production process data and their corresponding converter steel output were obtained from historical production data as training datasets.
[0007] The converter production process data in the training dataset is used as input, and the converter steel output corresponding to the input converter production process data is used as output. The specified type of neural network is trained to obtain a converter steel output prediction model.
[0008] The converter production process data in real time is used as input, and the converter output prediction model is used to predict the converter output.
[0009] Preferably, the step of using converter production process data from the training dataset as input and the converter tapping amount corresponding to the input converter production process data as output to train a specified type of neural network to obtain a converter tapping amount prediction model includes:
[0010] The neural network of a specified type is trained by taking a set of converter production process data from the training dataset as input and taking the converter steel output corresponding to the input converter production process data as output.
[0011] Determine whether the preset iteration termination condition has been met;
[0012] If the target is reached, the training ends and the converter steel output prediction model is obtained.
[0013] If this is not achieved, the following steps are performed: taking a set of converter production process data from the training dataset as input and the converter steel output corresponding to the input converter production process data as output, to train a neural network of a specified type.
[0014] Preferably, obtaining several sets of converter production process data and their corresponding converter steel output from historical production data as a training dataset includes:
[0015] Several sets of converter production process data and their corresponding converter steel output were obtained from historical production data, and training datasets and validation datasets were divided.
[0016] The determination of whether the preset iteration termination condition has been met includes:
[0017] The accuracy of the neural network's current converter steel output prediction is verified using the verification dataset to determine whether it meets the preset standard.
[0018] If the conditions are met, it is determined that the preset iteration termination condition has been met;
[0019] If the conditions are not met, it is determined that the preset iteration termination condition has not been met.
[0020] Preferably, verifying whether the prediction accuracy of the current converter steel output of the neural network meets the preset standard using the verification dataset includes:
[0021] The verification dataset is used to verify whether the prediction accuracy of the current converter steel output of the neural network meets the preset standards of root mean square error and / or correlation coefficient.
[0022] Preferably, several sets of converter production process data and their corresponding converter steel output are obtained from historical production data, and training and validation datasets are divided, including:
[0023] Several sets of converter production process data and their corresponding converter steel output were obtained from historical production data;
[0024] For each data point of converter production process data and its corresponding converter steel output obtained from historical production data, it is determined whether the value of the data exceeds the preset normal range of the data category to which the data belongs.
[0025] If the number exceeds the limit, the data will be removed.
[0026] If the limit is not exceeded, the data will be retained.
[0027] Based on all the retained data, a training dataset and a validation dataset are created.
[0028] Preferably, the specified type of neural network is a radial basis function neural network.
[0029] Preferably, the converter production process data includes at least:
[0030]
[0031] Where M is the total amount of scrap steel added, in wt n R represents the amount of scrap steel of type n added. n Let r be the proportion of type n scrap steel added. n Let be the water yield of the nth type of scrap steel.
[0032] To solve the above-mentioned technical problems, the present invention also provides a converter steel output prediction device, comprising:
[0033] The acquisition module is used to acquire several sets of converter production process data and their corresponding converter steel output from historical production data as training datasets.
[0034] The model training module is used to take the converter production process data in the training dataset as input, and the converter steel output corresponding to the input converter production process data as output, to train a neural network of a specified type to obtain a converter steel output prediction model.
[0035] The prediction module is used to take the real-time converter production process data as input and use the converter steel output prediction model to predict the converter steel output.
[0036] To solve the above-mentioned technical problems, the present invention also provides a converter steel output prediction device, comprising:
[0037] Memory, used to store computer programs;
[0038] A processor is used to execute the computer program to implement the steps of the converter steel output prediction method as described above.
[0039] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the converter steel output prediction method described above.
[0040] This invention provides a method for predicting converter steel output. Considering that historical production data, including converter production process data and corresponding converter steel output, can comprehensively reflect the impact of converter production process data on converter steel output, this application can train a specified type of neural network based on historical production data and corresponding converter steel output to obtain a converter steel output prediction model. Then, using real-time converter production process data as input, the converter steel output prediction model is used to predict the converter steel output, thus obtaining an accurate converter steel output. This helps to reduce the amount of alloy added during the alloying process in the converter, thereby reducing costs.
[0041] The present invention also provides a converter steel output prediction device, equipment and computer-readable storage medium, which have the same beneficial effects as the converter steel output prediction method described above. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the relevant technologies and the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 A flowchart illustrating a converter steel output prediction method provided by the present invention;
[0044] Figure 2 A flowchart illustrating another converter steel output prediction method provided by the present invention;
[0045] Figure 3 A schematic diagram of the structure of a converter steel output prediction device provided by the present invention;
[0046] Figure 4 This is a schematic diagram of a converter steel output prediction device provided by the present invention. Detailed Implementation
[0047] The core of this invention is to provide a method for predicting converter steel output. By using a trained converter steel output prediction model to predict the converter steel output, an accurate converter steel output can be obtained, which helps to reduce the amount of alloy added during the alloying process in the converter and lowers costs. Another core aspect of this invention is to provide a method, apparatus, equipment, and computer-readable storage medium for predicting converter steel output. By using a trained converter steel output prediction model to predict the converter steel output, an accurate converter steel output can be obtained, which helps to reduce the amount of alloy added during the alloying process in the converter and lowers costs.
[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0049] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a converter steel output prediction method provided by the present invention. The converter steel output prediction method includes:
[0050] S101: Obtain several sets of converter production process data and their corresponding converter steel output from historical production data as training datasets;
[0051] Specifically, considering the technical problems mentioned above, and considering that the converter production process data and corresponding converter steel output contained in the historical production data can comprehensively reflect the impact of the converter production process data on the converter steel output, and that the neural network itself has learning and reasoning capabilities, this application intends to train the neural network using the converter production process data and corresponding converter steel output from the historical production data, so as to obtain a converter steel output prediction model and predict the steel output of the converter process. Therefore, firstly, several sets of converter production process data and corresponding converter steel output can be obtained from the historical production data as training datasets to serve as the data basis for subsequent steps.
[0052] The converter production process data and the corresponding converter steel output can be automatically retrieved from the secondary production system and alloy inspection system.
[0053] Specifically, the converter tapping amount in the training dataset can be the actual measured accurate tapping amount, which helps to improve the prediction accuracy of the final trained model.
[0054] The quantity of converter production process data and corresponding converter steel output in the training dataset can be set independently, and this embodiment of the invention does not impose any limitations on it.
[0055] S102: Take the converter production process data in the training dataset as input, and the converter steel output corresponding to the input converter production process data as output, train the specified type of neural network to obtain the converter steel output prediction model.
[0056] Specifically, after obtaining the aforementioned training dataset, the converter production process data in the training dataset can be used as input, and the converter steel output corresponding to the input converter production process data can be used as output to train a specified type of neural network to obtain a converter steel output prediction model.
[0057] The specified type of neural network can be of various types, and this embodiment of the invention does not limit it.
[0058] S103: Using real-time converter production process data as input, the converter output is predicted using a converter output prediction model.
[0059] Specifically, since the neural network has learned the correlation between converter production process data and converter steel output during the training process, the resulting converter steel output prediction model has reasoning ability. Therefore, by taking the real-time converter production process data as input, the converter steel output prediction model can be used to predict the converter steel output with high accuracy.
[0060] It is worth mentioning that the steel output predicted in this embodiment of the invention is the steel output before the alloying process in the converter process. That is, the steel output obtained based on the prediction model in this embodiment of the invention can be used to set the amount of various alloys added in the alloying process of the converter process.
[0061] This invention provides a method for predicting converter steel output. Considering that historical production data, including converter production process data and corresponding converter steel output, can comprehensively reflect the impact of converter production process data on converter steel output, this application can train a specified type of neural network based on historical production data and corresponding converter steel output to obtain a converter steel output prediction model. Then, using real-time converter production process data as input, the converter steel output prediction model is used to predict the converter steel output, thus obtaining an accurate converter steel output. This helps to reduce the amount of alloy added during the alloying process in the converter, thereby reducing costs.
[0062] Based on the above embodiments:
[0063] In a preferred embodiment, the converter production process data in the training dataset is used as input, and the converter tapping amount corresponding to the input converter production process data is used as output. A specified type of neural network is trained to obtain a converter tapping amount prediction model, including:
[0064] The neural network of a specified type is trained by taking a set of converter production process data from the training dataset as input and the converter steel output corresponding to the input converter production process data as output.
[0065] Determine whether the preset iteration termination condition has been met;
[0066] If the target is reached, the training ends and the converter steel output prediction model is obtained.
[0067] If this is not achieved, the process involves taking a set of converter production process data from the training dataset as input and the converter steel output corresponding to the input converter production process data as output, and training the specified type of neural network.
[0068] Specifically, the neural network can be repeatedly trained through iterative training, and a preset iteration termination condition can be set. After each training session, it can be determined whether the preset iteration termination condition has been met. Once the condition is met, the training can be terminated and a converter steel output prediction model can be obtained, which can ensure that the trained converter steel output prediction model has high prediction accuracy.
[0069] As a preferred embodiment, several sets of converter production process data and their corresponding converter steel output are obtained from historical production data as a training dataset, including:
[0070] Several sets of converter production process data and their corresponding converter steel output were obtained from historical production data, and training datasets and validation datasets were divided.
[0071] Determining whether the preset iteration termination condition has been met includes:
[0072] The accuracy of the neural network's current converter steel output prediction is verified by using a validation dataset to determine whether it meets the preset standard.
[0073] If the conditions are met, the preset iteration termination condition is determined to have been met.
[0074] If the conditions are not met, it is determined that the preset iteration termination condition has not been met.
[0075] Specifically, considering that the converter production process data and the corresponding converter steel output in the historical production data can be used to verify the model prediction accuracy, in this embodiment of the invention, several sets of converter production process data and the corresponding converter steel output can be obtained in advance from the historical production data, and training datasets and verification datasets can be divided. Then, the prediction accuracy of the current converter steel output of the neural network can be verified by the verification dataset to determine whether the iteration should be terminated. This has high convenience and can better ensure the model prediction accuracy.
[0076] Of course, in addition to this specific preset iteration termination condition, the preset iteration termination condition can also be of other types, and the embodiments of the present invention are not limited here.
[0077] As a preferred embodiment, verifying whether the prediction accuracy of the current converter steel output by the neural network meets the preset standard by verifying the dataset includes:
[0078] The accuracy of the neural network's prediction of the current converter steel output is verified by using a validation dataset to determine whether it meets the preset standards for root mean square error and / or correlation coefficient.
[0079] Specifically, to evaluate the accuracy of the model's predictions, the root mean square error (RMSE) and correlation coefficient, the most widely used performance evaluation metrics, were selected as evaluation indicators. The RMSE characterizes the deviation between the predicted value and the actual value; a smaller value indicates that the predicted value is closer to the actual value, indicating a better model prediction effect. The correlation coefficient characterizes the ratio of the regression sum of squares to the total sum of squares, with a value ranging from [0, 1]. A value closer to 1 indicates a better model prediction effect. Table 1 shows the evaluation of the predicted model values and validation values. The predicted model values are the steel output predicted by the model, while the validation values are the historical actual steel outputs in the validation dataset corresponding to the data input to the predicted model.
[0080] Table 1 Evaluation of Predicted Model Values and Validation Values
[0081] Evaluation Criteria Evaluation value Root Mean Square Error (RMSE) 0.0326 R² (R²) 0.9635
[0082] As shown in Table 1, the predicted values of each alloy have good predictive performance compared with the measured values.
[0083] Of course, in addition to root mean square error and correlation coefficient, the preset standard can be of other types, which are not limited in this embodiment of the invention.
[0084] As a preferred embodiment, several sets of converter production process data and their corresponding converter steel output are obtained from historical production data, and training and validation datasets are divided, including:
[0085] Several sets of converter production process data and their corresponding converter steel output were obtained from historical production data;
[0086] For each data point of converter production process data and its corresponding converter steel output obtained from historical production data, it is determined whether the value of the data exceeds the preset normal range of the data category to which the data belongs.
[0087] If the number exceeds the limit, the data will be removed.
[0088] If the limit is not exceeded, the data will be retained.
[0089] Based on all the retained data, a training dataset and a validation dataset are created.
[0090] Specifically, considering that historical production data may also contain erroneous or abnormal data, it is possible to clean the data obtained from historical production data, that is, to remove the abnormal data. Therefore, in this embodiment of the invention, a preset normal range can be set for each type of data in the converter production process data and its corresponding converter steel output. Then, for each data to be judged, it is determined whether the value of the data exceeds the preset normal range of the data category in which the data belongs, thereby completing the cleaning work of the obtained data.
[0091] The preset normal range can be set in various ways; for example, the preset normal range for each type of data can be:
[0092] (α(X max -X min ),β(X max -X min ));
[0093] Among them, X max X represents the maximum value of a class of data obtained from historical production data. min Let α be the minimum value of a class of data obtained from historical production data, and let β be the first coefficient and β be the second coefficient. The first coefficient is less than the second coefficient and both are positive fractions not greater than 1.
[0094] Of course, in addition to this specific form, the preset normal range can also be of many other types, and the embodiments of the present invention are not limited here.
[0095] As a preferred embodiment, the specified type of neural network is a radial basis function neural network.
[0096] Specifically, radial basis function (RBF) neural networks have advantages such as fast computation speed and high accuracy.
[0097] Among them, the prediction model based on radial basis function neural network can have a three-layer structure, and the number of nodes in the middle layer can be determined by the mean square error between the predicted value and the measured value.
[0098] Of course, in addition to this specific type, the neural network of the specified type can also be other types, and this embodiment of the invention does not limit it here.
[0099] As a preferred embodiment, the converter production process data includes at least:
[0100]
[0101] Where M is the total amount of scrap steel added, in wt n R represents the amount of scrap steel of type n added. n Let r be the proportion of type n scrap steel added. n Let be the water yield of the nth type of scrap steel.
[0102] Specifically, considering that the total amount of scrap steel added in the converter production process data has a significant impact on the steel output, and that related technologies usually treat all types of scrap steel as one category and count the total amount of scrap steel added to calculate the steel output, but the added scrap steel has many different types, and different types of scrap steel contribute differently to the steel output, this embodiment of the invention can classify and count the contribution of each type of added scrap steel. Specifically, the contribution of each type of scrap steel can be the product of the amount added, the proportion added, and the water content of each type of scrap steel, which can improve the prediction accuracy of the steel output.
[0103] It is worth mentioning that, in addition to the total amount of scrap steel added, the converter production process data can also include other types of data, such as the amount of molten iron charged, the composition of molten iron (carbon content, silicon content, manganese content, phosphorus content, sulfur content), the final composition of molten steel (carbon content, silicon content, manganese content, phosphorus content, sulfur content), the amount of cold materials added (lime / limestone added, pellets added, etc.), the total iron content in the slag, and the amount of slag, etc. The embodiments of the present invention are not limited here.
[0104] Specifically, for a better explanation of the embodiments of the present invention, please refer to... Figure 2 , Figure 2 The following is a flowchart illustrating another converter steel output prediction method provided by the present invention, and a specific embodiment of the present invention is provided below:
[0105] In the production of Q345qD steel at a steel plant, the production staff previously calculated the alloy addition amount based on experience. The difference between the actual tapping amount and the experienced tapping amount ranged from -24.70 to 12.96 tons. To ensure that all heats meet the requirements, the amount of various alloys added must be increased. This not only increases the alloy consumption in the converter process but also increases the operational pressure of alloying in the refining process. Therefore, the converter tapping amount prediction model mentioned above can be used to predict the tapping amount and guide the setting of alloy addition amounts. This can improve the stability of the elemental content of the converter steel and reduce the operational intensity of alloying in the refining process.
[0106] Of the 959 heats of data obtained after cleaning, 60% (575 heats) was selected for model training, and 40% (384 heats) was used for model validation and optimization to improve prediction performance. To verify the effectiveness of the method in actual converter production, the output of 20 heats was statistically analyzed, and the differences between the converter output determined under the original process and the converter output using the method of this invention were compared. Specific data are shown in Table 2.
[0107] Table 2 Comparison of converter steel output between the original process and the method of this invention.
[0108]
[0109]
[0110] As can be seen from the data comparison in Table 2, under the original process, the difference between the predicted steel output and the actual steel output ranged from -12.0 to 9.7t, while the difference between the predicted steel output and the actual steel output using the method of the present invention ranged from -4.3 to 5.9t, indicating that the method of the present invention can accurately predict the converter steel output.
[0111] in, Figure 2 Compared to the above-described method embodiments, this method adds verification of prediction accuracy during actual production. Specifically, it can predict the converter steel output based on production data from the converter process. After actual steel output, the actual output can be measured, and the difference between the actual and predicted output can be used to determine if the prediction accuracy meets the standard. If it does, the input production process data and the corresponding actual output can be added to historical production data. If it does not meet the standard, the prediction model can be retrained immediately to further ensure prediction accuracy. The step of "comparing the actual and predicted output to determine if the prediction accuracy meets the standard" can be described as follows: Figure 2 As shown, it can be determined whether the absolute value of the difference between the two is less than a preset value (5 tons). Other forms are also possible, and this embodiment of the invention does not limit them here.
[0112] Please refer to Figure 3 , Figure 3 This is a schematic diagram of a converter steel output prediction device provided by the present invention. The converter steel output prediction device includes:
[0113] The acquisition module 31 is used to acquire several sets of converter production process data and their corresponding converter steel output from historical production data as training datasets.
[0114] The model training module 32 is used to take the converter production process data in the training dataset as input, and the converter steel output corresponding to the input converter production process data as output, to train a specified type of neural network to obtain a converter steel output prediction model.
[0115] The prediction module 33 is used to take the real-time converter production process data as input and use the converter steel output prediction model to predict the converter steel output.
[0116] For a description of the converter steel output prediction device provided in this embodiment of the invention, please refer to the aforementioned embodiment of the converter steel output prediction method. This embodiment of the invention will not be repeated here.
[0117] Please refer to Figure 4 , Figure 4 This is a schematic diagram of a converter steel output prediction device provided by the present invention. The converter steel output prediction device includes:
[0118] Memory 41 is used to store computer programs;
[0119] The processor 42 is used to execute computer programs to implement the steps of the converter steel output prediction method as described in the foregoing embodiments.
[0120] For a description of the converter steel output prediction device provided in this embodiment of the invention, please refer to the aforementioned embodiment of the converter steel output prediction method. This embodiment of the invention will not be repeated here.
[0121] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the converter steel output prediction method as described in the foregoing embodiments.
[0122] For a description of the computer-readable storage medium provided in the embodiments of the present invention, please refer to the aforementioned embodiments of the converter steel output prediction method. The embodiments of the present invention will not be repeated here.
[0123] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should also be noted that in this specification, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0124] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting converter steel output, characterized in that, include: Several sets of converter production process data and their corresponding converter steel output were obtained from historical production data; For each data point of converter production process data and its corresponding converter steel output obtained from historical production data, it is determined whether the value of the data exceeds the preset normal range of the data category to which the data belongs. If the number exceeds the limit, the data will be removed. If the limit is not exceeded, the data will be retained. Based on all the retained data, a training dataset and a validation dataset are divided. The neural network of a specified type is trained by taking a set of converter production process data from the training dataset as input and taking the converter steel output corresponding to the input converter production process data as output. Determine whether the preset iteration termination condition has been met; If the target is reached, the training ends and the converter steel output prediction model is obtained. If not achieved, the following steps are performed: taking a set of converter production process data from the training dataset as input and the converter steel output corresponding to the input converter production process data as output, and training a neural network of a specified type. The converter production process data of the real-time converter operation is used as input, and the converter steel output prediction model is used to predict the converter steel output. The determination of whether the preset iteration termination condition has been met includes: The accuracy of the neural network's current converter steel output prediction is verified using the verification dataset to determine whether it meets the preset standard. If the conditions are met, it is determined that the preset iteration termination condition has been met; If the conditions are not met, it is determined that the preset iteration termination condition has not been met. The converter production process data includes: ； Where M is the total amount of scrap steel added, in wt n R represents the amount of scrap steel of type n added. n Let r be the proportion of type n scrap steel added. n Let be the water yield of the nth type of scrap steel.
2. The converter steel output prediction method according to claim 1, characterized in that, The step of verifying whether the prediction accuracy of the current converter steel output of the neural network meets the preset standard through the verification dataset includes: The verification dataset is used to verify whether the prediction accuracy of the current converter steel output of the neural network meets the preset standards of root mean square error and / or correlation coefficient.
3. The converter steel output prediction method according to claim 1, characterized in that, The specified type of neural network is a radial basis function neural network.
4. A converter steel output prediction device, characterized in that, include: The acquisition module is used to acquire several sets of converter production process data and their corresponding converter steel output from historical production data; for each data point of converter production process data and its corresponding converter steel output acquired from historical production data, it is determined whether the value of the data exceeds the preset normal range of the data category to which the data belongs. If the number exceeds the limit, the data will be removed. If the limit is not exceeded, the data is retained; based on all retained data, the training dataset and the validation dataset are divided. The model training module is used to take a set of converter production process data in the training dataset as input, and take the converter steel output corresponding to the input converter production process data as output to train a neural network of a specified type; and determine whether the preset iteration termination condition has been reached. If the target is reached, the training ends and the converter steel output prediction model is obtained. If not achieved, the following steps are performed: taking a set of converter production process data from the training dataset as input and the converter steel output corresponding to the input converter production process data as output, and training a neural network of a specified type. The prediction module is used to take the real-time converter production process data as input and use the converter steel output prediction model to predict the converter steel output. The determination of whether the preset iteration termination condition has been met includes: The accuracy of the neural network's current converter steel output prediction is verified using the verification dataset to determine whether it meets the preset standard. If the conditions are met, it is determined that the preset iteration termination condition has been met; If the conditions are not met, it is determined that the preset iteration termination condition has not been met. The converter production process data includes: ； Where M is the total amount of scrap steel added, in wt n R represents the amount of scrap steel of type n added. n Let r be the proportion of type n scrap steel added. n Let be the water yield of the nth type of scrap steel.
5. A converter steel output prediction device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the converter steel output prediction method as described in any one of claims 1 to 3.
6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the converter steel output prediction method as described in any one of claims 1 to 3.