Multi-objective collaborative prediction-based converter endpoint control system, method, device and medium

Abstract

This application discloses a multi-objective collaborative prediction converter endpoint control system, method, equipment, and medium, mainly relating to the field of converter endpoint control technology, to solve the problems of low accuracy in existing manual experience-based judgments and functional limitations in automatic control judgments. It includes: a data acquisition module, which determines the calculated endpoint data for the current furnace corresponding to the furnace entry conditions; a data processing module, which obtains the predicted endpoint data for the current furnace corresponding to the historical furnace endpoint data set; a recurrent neural network processing module, used to determine the output endpoint data using the calculated endpoint data, the predicted endpoint data, and a trained preset recurrent neural network algorithm; and a symmetric connected neural network calculation module, used to obtain the final predicted endpoint data using a trained preset symmetric connected neural network algorithm and the output endpoint data.

Description

Technical Field

[0001] This application relates to the field of converter endpoint control technology, and in particular to a multi-objective collaborative prediction converter endpoint control system, method, equipment and medium. Background Technology

[0002] Converter (steelmaking) endpoint control is an operational technique that controls the duration of the converter steelmaking process to ensure that the temperature and composition of the molten steel meet the requirements at the end of the blowing process.

[0003] The specific objectives of converter end-point control are: (1) the carbon content of the molten steel should reach the target range required for the steel grade being produced; (2) the phosphorus and sulfur content in the steel should be lower than the lower limit required by the specifications; (3) the tapping temperature should ensure the smooth progress of the next process; and (4) the molten steel should have suitable oxidizing properties. End-point control is essentially the control of the converter blowing process. The quality of end-point control is related to steelmaking productivity, metal yield, production cost, and steel quality, and therefore it is a very important link in the converter steelmaking process.

[0004] Existing converter steelmaking endpoint control can be broadly categorized into two types: empirical control and automatic control. Empirical control refers to manual control of the endpoint (final carbon) under conventional blowing conditions, utilizing standard monitoring methods. Empirical control commonly employs two operational methods: "carbon reduction" and "carbon increase," with frequently used auxiliary methods including temperature and carbon determination and rapid analysis of samples taken before the furnace. Because control relies primarily on manual judgment, the accuracy of endpoint prediction is low, and fluctuations are significant. Automatic control of the converter furnace opening has limitations such as furnace gas overflow and air mixing, low furnace gas analysis accuracy leading to reduced model prediction accuracy, and the lack of endpoint temperature prediction functionality, requiring further exploration and research. Summary of the Invention

[0005] To address the aforementioned shortcomings of existing technologies, this application provides a converter endpoint control system, method, equipment, and medium for multi-objective collaborative prediction, in order to solve the problems of low accuracy in existing manual experience-based judgments and functional limitations in automatic control judgments.

[0006] In a first aspect, this application provides a multi-objective collaborative prediction converter endpoint control system, comprising: a data acquisition module, used to acquire the furnace inlet conditions and preset calculation formulas for the current furnace, and determine the calculated endpoint data for the current furnace corresponding to the furnace inlet conditions; determine the trained preset recurrent neural network algorithm and trained preset symmetric connection neural network algorithm corresponding to the preset calculation formulas; a data processing module, used to acquire a set of historical furnace inlet conditions within a historical time period, and determine the cluster set to which the furnace inlet conditions for the current furnace belong in the set of historical furnace inlet conditions through a preset clustering algorithm; and obtain the predicted endpoint data for the current furnace corresponding to the set of historical furnace endpoint data based on the cluster set and the set of historical furnace endpoint data within the historical time period; a recurrent neural network processing module, used to determine the output endpoint data through the calculated endpoint data for the current furnace, the predicted endpoint data for the current furnace, and the trained preset recurrent neural network algorithm; and a symmetric connection neural network calculation module, used to obtain the final predicted endpoint data through the trained preset symmetric connection neural network algorithm and the output endpoint data.

[0007] Furthermore, the data acquisition module includes a data acquisition unit, which is used to acquire the furnace entry conditions for the current furnace through a preset furnace data acquisition interface; and to acquire preset calculation formulas through a preset formula editing interface.

[0008] Furthermore, the data processing module includes an algorithm update unit, which is used to randomly select a backup clustering algorithm from a preset clustering algorithm set; obtain the set of historical furnace entry conditions and the set of historical furnace endpoint data within a preset detection period to train the backup clustering algorithm; and update the backup clustering algorithm to the preset clustering algorithm with the preset detection period as the update node.

[0009] Furthermore, the system also includes a hazard alarm module, which imports the final predicted endpoint data into a preset threshold detection program so that when the predicted endpoint data exceeds the preset threshold, an alarm task is generated to a preset maintenance terminal, and an alarm device is triggered at the same time.

[0010] Secondly, this application provides a multi-objective collaborative prediction method for converter endpoint control. The method includes: obtaining the current furnace inlet conditions and a preset calculation formula; determining the current furnace inlet calculation endpoint data corresponding to the current furnace inlet conditions; determining the trained preset recurrent neural network algorithm and the trained preset symmetric connection neural network algorithm corresponding to the preset calculation formula; obtaining a set of historical furnace inlet conditions within a historical time period; determining the cluster set to which the current furnace inlet conditions belong in the set of historical furnace inlet conditions using a preset clustering algorithm; obtaining the current furnace predicted endpoint data corresponding to the historical furnace endpoint data set based on the cluster set and the set of historical furnace endpoint data within a historical time period; determining the output endpoint data using the current furnace inlet calculation endpoint data, the current furnace predicted endpoint data, and the trained preset recurrent neural network algorithm; and obtaining the final predicted endpoint data using the trained preset symmetric connection neural network algorithm and the output endpoint data.

[0011] Furthermore, the furnace entry conditions and preset calculation formulas for this furnace batch are obtained, specifically including: obtaining the furnace entry conditions for this furnace batch through the preset furnace batch data acquisition interface; and obtaining the preset calculation formulas through the preset formula editing interface.

[0012] Furthermore, the method also includes: randomly selecting a backup clustering algorithm from a preset clustering algorithm set; obtaining a set of historical furnace entry conditions and a set of historical furnace endpoint data within a preset detection period to train the backup clustering algorithm; and updating the backup clustering algorithm to the preset clustering algorithm using the preset detection period as the update node.

[0013] Furthermore, the method also includes: importing the final predicted endpoint data into a preset threshold detection program, so that when the predicted endpoint data exceeds the preset threshold, an alarm task is generated to a preset maintenance terminal, and an alarm device is triggered at the same time.

[0014] Thirdly, this application provides a converter endpoint control device with multi-objective collaborative prediction. The device includes: a processor; and a memory storing executable code thereon. When the executable code is executed, the processor performs a converter endpoint control method with multi-objective collaborative prediction as described above.

[0015] Fourthly, this application provides a non-volatile computer storage medium storing computer instructions thereon, which, when executed, implement a multi-objective collaborative prediction converter endpoint control method as described above.

[0016] Those skilled in the art will understand that this application has at least the following beneficial effects:

[0017] This application acquires preset calculation formulas in real time through a data acquisition module, enabling flexible processing of the furnace entry conditions for this furnace batch. This ensures that the calculation endpoint data for this furnace batch is not limited by time and can obtain more accurate endpoint data by updating the preset calculation formula. To prevent updates to the preset calculation formula from affecting the accuracy of the neural network algorithm, the neural network algorithm is automatically updated to the corresponding trained neural network algorithm after the preset calculation formula is obtained. This realizes the calculation and application of recurrent neural networks combined with scenario requirements. The data processing module obtains the predicted endpoint data for this furnace batch corresponding to the historical furnace batch endpoint data set, achieving effective and accurate acquisition of the predicted endpoint data for this furnace batch by combining the furnace entry conditions for this furnace batch and the historical furnace batch endpoint data set. In addition, the furnace entry conditions for this furnace batch in this application can include various functional conditions (e.g., temperature), reducing functional limitations. Attached Figure Description

[0018] The following description refers to some embodiments of this disclosure, in which:

[0019] Figure 1 This is a schematic diagram of the internal structure of a converter endpoint control system for multi-objective collaborative prediction provided in an embodiment of this application.

[0020] Figure 2 This is a flowchart of a converter endpoint control method based on multi-objective collaborative prediction provided in an embodiment of this application.

[0021] Figure 3 This is a schematic diagram of the internal structure of a converter endpoint control device for multi-objective collaborative prediction provided in an embodiment of this application. Detailed Implementation

[0022] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.

[0023] It should also be noted that 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 process, method, article, or apparatus. Unless otherwise specified, 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.

[0024] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0025] Figure 1 This application provides a multi-objective collaborative prediction converter endpoint control system as an embodiment. For example... Figure 1 As shown in the embodiments of this application, the system mainly includes:

[0026] The system obtains the furnace entry conditions and preset calculation formulas for this furnace through the data acquisition module 110, determines the calculation endpoint data for this furnace corresponding to the furnace entry conditions, and determines the trained preset recurrent neural network algorithm and trained preset symmetric connection neural network algorithm corresponding to the preset calculation formula.

[0027] It should be noted that the data acquisition module 110 can be any feasible device or apparatus capable of acquiring data and determining the corresponding neural network algorithm (a pre-trained preset recurrent neural network algorithm and a pre-trained preset symmetric connection neural network algorithm). The furnace charging conditions for this furnace run (and the furnace charging conditions mentioned in this application) can specifically include the molten iron temperature, molten iron composition (e.g., the content of Si, Mn, P, and S), the amount of scrap steel and molten iron added, the amount of slag-forming material and alloy added (e.g., the content of lime, dolomite, ore, etc.), and oxygen consumption. The preset calculation formula is a formula determined by those skilled in the art based on actual conditions to calculate the endpoint data for this furnace run; this application does not limit the specific content of the preset calculation formula itself. The endpoint data for this furnace run (and the endpoint data mentioned in this application) can specifically include multiple objectives such as endpoint carbon content, endpoint temperature, endpoint slag composition, and endpoint molten pool level height.

[0028] The process of obtaining the training data for the pre-trained pre-defined recurrent neural network algorithm and the pre-trained pre-defined symmetric connection neural network algorithm corresponding to the pre-defined calculation formula can be as follows: Obtain the set of historical furnace entry conditions and the set of historical furnace endpoint data within a pre-defined historical time period; obtain the set of calculation furnace entry conditions within the pre-defined historical time period using the pre-defined calculation formula; and then train the pre-defined recurrent neural network algorithm and the pre-defined symmetric connection neural network algorithm to obtain the pre-trained pre-defined recurrent neural network algorithm and the pre-trained pre-defined symmetric connection neural network algorithm. The specific training process of the neural network algorithm is an existing process, and this application does not limit it.

[0029] In addition, the specific scheme for obtaining the furnace entry conditions and preset calculation formula in this application can be as follows: the data acquisition unit 111 in the data acquisition module 110 obtains the furnace entry conditions for this furnace through the preset furnace data acquisition interface; and obtains the preset calculation formula through the preset formula editing interface.

[0030] It should be noted that the data acquisition unit 111 can be any feasible device or apparatus capable of acquiring data or formulas through a preset interface.

[0031] The system obtains the set of historical furnace entry conditions within a historical time period through the data processing module 120, and determines the cluster set to which the entry conditions of the current furnace belong in the set of historical furnace entry conditions through a preset clustering algorithm; based on the cluster set and the set of historical furnace endpoint data within the historical time period, the system obtains the predicted endpoint data of the current furnace corresponding to the set of historical furnace endpoint data.

[0032] It should be noted that the historical time period can be any feasible time period. The data processing module 120 can be any feasible device or apparatus capable of calling algorithms to perform data processing.

[0033] The above-mentioned method of obtaining the predicted endpoint data for the current furnace based on the cluster set and the historical furnace endpoint data set within the historical time period can be specifically described as follows: by using the text similarity calculation formula (e.g., the Manhattan distance calculation method), the historical furnace entry conditions with the highest similarity to the current furnace entry conditions in the cluster set are obtained, and then the historical furnace endpoint data corresponding to the historical furnace entry conditions with the highest similarity are obtained from the historical furnace endpoint data set as the predicted endpoint data for the current furnace.

[0034] In addition, in order to ensure that the preset clustering algorithm has high accuracy and avoid the time gap caused by training the preset clustering algorithm, this application can train a backup clustering algorithm to replace the preset clustering algorithm in a timely manner, thereby maintaining a high clustering accuracy.

[0035] The specific process can be as follows: the algorithm update unit 121 in the data processing module 120 randomly selects a backup clustering algorithm from the preset clustering algorithm set; obtains the set of historical furnace entry conditions and the set of historical furnace endpoint data within the preset detection period to train the backup clustering algorithm; and updates the backup clustering algorithm to the preset clustering algorithm with the preset detection period as the update node.

[0036] The recurrent neural network processing module 130 in the system determines the output endpoint data by calculating the endpoint data of this furnace, predicting the endpoint data of this furnace, and using the trained preset recurrent neural network algorithm.

[0037] It should be noted that the recurrent neural network processing module 130 can be any feasible device or apparatus capable of calling a preset recurrent neural network algorithm for data processing. The specific implementation process of the algorithm can be implemented by existing methods or devices, and this application does not limit it in this regard.

[0038] The symmetric connected neural network calculation module 140 in the system obtains the final predicted endpoint data through the trained preset symmetric connected neural network algorithm and the output endpoint data.

[0039] It should be noted that the symmetric connected neural network computing module 140 can be any feasible device or apparatus capable of calling a preset symmetric connected neural network algorithm for data processing. The specific implementation process of the algorithm can be implemented by existing methods or devices, and this application does not limit it in this regard.

[0040] In addition, to avoid data anomalies, this application can use a threshold detection method for alarm processing. The specific process is as follows: the system imports the final predicted endpoint data into the preset threshold detection program through the danger alarm module 150, so that when the predicted endpoint data exceeds the preset threshold, an alarm task is generated to the preset maintenance terminal, and the alarm device is triggered at the same time.

[0041] It should be noted that the danger alarm module 150 can be any feasible device or apparatus capable of calling a preset threshold detection program to perform data detection, generate and transmit tasks, and trigger alarm devices.

[0042] In addition, the preset threshold detection program includes preset thresholds corresponding to the predicted endpoint data. The program can compare the predicted endpoint data with the preset thresholds to detect whether the preset thresholds are exceeded.

[0043] In addition, embodiments of this application also provide a multi-objective collaborative prediction method for converter endpoint control, such as... Figure 2 As shown in the embodiments of this application, the method mainly includes the following steps:

[0044] Step 210: Obtain the furnace entry conditions and preset calculation formulas for this furnace, determine the calculation endpoint data for this furnace corresponding to the furnace entry conditions, and determine the trained preset recurrent neural network algorithm and trained preset symmetric connection neural network algorithm corresponding to the preset calculation formulas.

[0045] Specifically, obtaining the furnace entry conditions and preset calculation formulas for this furnace run can be achieved by: obtaining the furnace entry conditions for this furnace run through the preset furnace run data acquisition interface; and obtaining the preset calculation formulas through the preset formula editing interface.

[0046] Step 220: Obtain the set of historical furnace entry conditions within the historical time period. Using a preset clustering algorithm, determine the cluster set to which the entry conditions of this furnace belong in the set of historical furnace entry conditions. Based on the cluster set and the set of historical furnace endpoint data within the historical time period, obtain the predicted endpoint data of this furnace corresponding to the set of historical furnace endpoint data.

[0047] In order to ensure that the preset clustering algorithm has high accuracy and avoid the time gap caused by training the preset clustering algorithm, this application can train a backup clustering algorithm to replace the preset clustering algorithm in a timely manner, thereby maintaining a high clustering accuracy.

[0048] The specific process can be as follows: randomly select a backup clustering algorithm from the preset clustering algorithm set; obtain the set of historical furnace entry conditions and the set of historical furnace endpoint data within the preset detection period to train the backup clustering algorithm; and update the backup clustering algorithm to the preset clustering algorithm with the preset detection period as the update node.

[0049] Step 230: Determine the output endpoint data by using the endpoint data calculated for this furnace, the endpoint data predicted for this furnace, and the trained preset recurrent neural network algorithm.

[0050] Step 240: Obtain the final predicted endpoint data by using the trained preset symmetric connection neural network algorithm and the output endpoint data.

[0051] Furthermore, to prevent data anomalies, this application can implement alarm handling using a threshold detection method. The specific process is as follows:

[0052] The final predicted endpoint data is imported into a preset threshold detection program. When the predicted endpoint data exceeds the preset threshold, an alarm task is generated and sent to a preset maintenance terminal, and the alarm device is triggered.

[0053] The above are method embodiments of this application. Based on the same inventive concept, embodiments of this application also provide a converter endpoint control device for multi-objective collaborative prediction. Figure 3 As shown, the device includes: a processor; and a memory storing executable code thereon, which, when executed, causes the processor to perform a multi-objective collaborative prediction converter endpoint control method as described in the above embodiments.

[0054] Specifically, the server obtains the furnace entry conditions and preset calculation formulas for the current furnace batch, determines the calculation endpoint data corresponding to the furnace entry conditions for the current furnace batch, determines the trained preset recurrent neural network algorithm and trained preset symmetric connection neural network algorithm corresponding to the preset calculation formula, obtains the set of furnace entry conditions for historical furnace batches within a historical time period, and determines the cluster set to which the furnace entry conditions for the current furnace batch belong in the set of furnace entry conditions for historical furnace batches through a preset clustering algorithm, obtains the predicted endpoint data for the current furnace batch corresponding to the set of endpoint data for historical furnace batches based on the cluster set and the set of endpoint data for historical furnace batches within a historical time period, determines the predicted endpoint data for the current furnace batch through the calculated endpoint data for the current furnace batch, the predicted endpoint data for the current furnace batch, and the trained preset recurrent neural network algorithm, and obtains the final predicted endpoint data through the trained preset symmetric connection neural network algorithm and the output endpoint data.

[0055] In addition, this application embodiment also provides a non-volatile computer storage medium storing executable instructions, which, when executed, implement the multi-objective collaborative prediction converter endpoint control method described above.

[0056] The technical solutions of this disclosure have been described in conjunction with the preceding embodiments. However, it will be readily understood by those skilled in the art that the scope of protection of this disclosure is not limited to these specific embodiments. Without departing from the technical principles of this disclosure, those skilled in the art can disassemble and combine the technical solutions in the above embodiments, and can also make equivalent changes or substitutions to the relevant technical features. Any changes, equivalent substitutions, improvements, etc., made within the technical concept and / or technical principles of this disclosure will fall within the scope of protection of this disclosure.

Claims

1. A multi-objective collaborative prediction-based converter endpoint control system, characterized in that, The system comprises: a data acquisition module, configured to acquire a current furnace charging condition and a preset calculation formula, determine current furnace calculation end point data corresponding to the current furnace charging condition, and determine a trained preset recurrent neural network algorithm and a trained preset symmetric connection neural network algorithm corresponding to the preset calculation formula; the preset calculation formula is a formula capable of calculating the current furnace calculation end point data corresponding to the current furnace charging condition; a data processing module, configured to acquire a historical furnace charging condition set in a historical time period, determine a cluster set to which the current furnace charging condition belongs in the historical furnace charging condition set by using a preset clustering algorithm, and obtain current furnace predicted end point data corresponding to a historical furnace end point data set based on the cluster set and the historical furnace end point data set in the historical time period; the obtaining of the current furnace predicted end point data corresponding to the historical furnace end point data set based on the cluster set and the historical furnace end point data set in the historical time period comprises: acquiring a historical furnace charging condition with the highest similarity to the current furnace charging condition in the cluster set by using a text similarity calculation formula, and obtaining historical furnace end point data corresponding to the historical furnace charging condition with the highest similarity in the historical furnace end point data set as the current furnace predicted end point data; a recurrent neural network processing module, configured to determine output end point data by using the current furnace calculation end point data, the current furnace predicted end point data and the trained preset recurrent neural network algorithm; a symmetric connection neural network calculation module, configured to obtain final predicted end point data by using the trained preset symmetric connection neural network algorithm and the output end point data; the data acquisition module comprises a data acquisition unit, configured to acquire the current furnace charging condition by using a preset furnace data acquisition interface; acquire the preset calculation formula by using a preset formula editing interface; the data processing module further comprises an algorithm updating unit, configured to randomly extract a standby clustering algorithm from a preset clustering algorithm set; acquire a historical furnace charging condition set and a historical furnace end point data set in a preset detection period to train the standby clustering algorithm; update the standby clustering algorithm to the preset clustering algorithm with the preset detection period as an updating node.

2. The multi-target collaborative predictive ladle endpoint control system of claim 1, wherein, The system further comprises a danger alarm module, configured to import the final predicted end point data into a preset threshold detection program, generate an alarm task to a preset maintenance terminal when the predicted end point data exceeds a preset threshold, and trigger an alarm device.

3. A multi-objective cooperative prediction-based control method for converter endpoint, characterized in that, The method comprises: acquiring a current furnace charging condition and a preset calculation formula, determining current furnace calculation end point data corresponding to the current furnace charging condition, and determining a trained preset recurrent neural network algorithm and a trained preset symmetric connection neural network algorithm corresponding to the preset calculation formula; the preset calculation formula is a formula capable of calculating the current furnace calculation end point data corresponding to the current furnace charging condition; The method comprises the following steps: obtaining a historical batch entry condition set in a historical time period, determining a cluster set to which the batch entry condition of the current batch belongs in the historical batch entry condition set through a preset clustering algorithm; obtaining the predicted end point data of the current batch corresponding to the historical batch end point data set based on the cluster set and the historical batch end point data set in the historical time period; the obtaining of the predicted end point data of the current batch corresponding to the historical batch end point data set based on the cluster set and the historical batch end point data set in the historical time period comprises: obtaining the historical batch entry condition with the highest similarity to the batch entry condition of the current batch in the cluster set through a text similarity calculation formula, and obtaining the historical batch end point data corresponding to the historical batch entry condition with the highest similarity from the historical batch end point data set as the predicted end point data of the current batch; determining the output end point data through the calculated end point data of the current batch, the predicted end point data of the current batch and the trained preset recurrent neural network algorithm; obtaining the final predicted end point data through the trained preset symmetric connection neural network algorithm and the output end point data; The method further comprises the following steps: obtaining the batch entry condition of the current batch through a preset batch data acquisition interface; obtaining the preset calculation formula through a preset formula editing interface; The method further comprises the following steps: randomly extracting a standby clustering algorithm from a preset clustering algorithm set; obtaining the historical batch entry condition set and the historical batch end point data set in a preset detection period to train the standby clustering algorithm; updating the standby clustering algorithm to the preset clustering algorithm with the preset detection period as an update node.

4. The multi-target synergistic predictive ladle endpoint control method of claim 3, wherein, The method further comprises the following steps: importing the final predicted end point data into a preset threshold detection program to generate an alarm task to a preset maintenance terminal when the predicted end point data exceeds the preset threshold, and triggering an alarm device.

5. A multi-objective cooperative prediction-based converter endpoint control apparatus, characterized by, The device comprises: a processor; and a memory having executable codes stored thereon, which, when executed, cause the processor to perform the multi-target cooperative prediction converter end point control method according to any one of claims 3-4.

6. A non-transitory computer storage medium, comprising, having computer instructions stored thereon, which, when executed, implement the multi-target cooperative prediction converter end point control method according to any one of claims 3-4.

Images