Method for controlling the molten iron temperature of a blast furnace and method for producing molten iron.

Abstract

To provide a blast furnace molten iron temperature control method that can suppress a decrease in the accuracy of molten iron temperature control even when a large amount of inferior raw materials are used, when the moisture content of inferior raw materials increases, or when the function of the means for observing the moisture content of coke is impaired. [Solution] The molten iron temperature control method for a blast furnace according to the present invention includes the steps of: selecting an action model corresponding to the observation means to be used from among a plurality of action models trained by machine learning for each of two or more observation means that observe one observation quantity, using image data obtained by imaging one or more historical data of observation quantities indicating the operating state of a blast furnace process or the operating state of equipment as input data, and an operation quantity of the process or equipment determined by an operator based on the historical data as output data; and controlling the state of the process or equipment according to the operation quantity obtained by inputting image data to the selected action model, thereby controlling the molten iron temperature.

Description

Technical Field

[0004] , , , , , , , , ,

[0005] , , , , ,

[0001] The present invention relates to a method for controlling the molten iron temperature in a blast furnace and a method for producing molten iron.

Background Art

[0002] In the blast furnace process in the ironmaking industry, the molten iron temperature is an important management index. In particular, in recent years, blast furnace operations have been carried out under conditions of low coke ratio and high pulverized coal ratio in order to pursue rationalization of raw fuel costs. Since the furnace condition is likely to become unstable, there is a great need to reduce the variation in molten iron temperature. However, in the blast furnace process, various reactions such as gasification of coke, reduction and melting of ore occur. In addition, since the operation is carried out with the furnace filled with solids, the heat capacity of the entire process is large and the time constant of the response to an action is long. Therefore, in order to reduce the variation in molten iron temperature, a control rule for molten iron temperature that appropriately considers the complex dynamics of the blast furnace is required.

[0003] From such a background, a method for controlling molten iron temperature (see Patent Document 1) has been proposed in which operation data is imaged and machine-learned, and the operation amount of an operator is determined using the machine learning result. Specifically, Patent Document 1 describes a method for controlling molten iron temperature in which image data obtained by imaging history data of one or more observation amounts indicating the operation state of a process or the operation state of equipment is used as input data, and a machine learning model in which the operation amount of a process or equipment determined by an operator based on the history data is used as output data is used to determine the operation amount of the process or equipment.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the method described in Patent Document 1, the accuracy of molten iron temperature control may decrease when large quantities of inferior raw materials are used or when the moisture content of inferior raw materials increases. Here, in this specification, inferior raw materials are defined as coke containing a large amount of moisture that is left directly in the field (hereinafter referred to as yard coke). In particular, in the current situation where steel mills are coke-tight due to the rising price of high-quality coal, the weight ratio of inferior raw materials used, such as yard coke containing a large amount of moisture that is left directly in the field, and the moisture content of inferior raw materials may increase. In this case, the amount of moisture circulating through repeated evaporation and condensation in the upper part of the blast furnace increases, which slows down the heating in the upper part of the furnace, causing the molten iron temperature to tend to decrease and making it difficult to control the molten iron temperature. Similarly, if the function of the means for observing the moisture content of coke, which is relatively difficult to measure continuously, is impaired, the accuracy of molten iron temperature control may also decrease.

[0006] The present invention has been made to solve the above problems, and its objective is to provide a blast furnace molten iron temperature control method that can suppress a decrease in the accuracy of molten iron temperature control even when a large amount of inferior raw material is used, when the moisture content of the inferior raw material increases, or when the function of the means for observing the moisture content of coke is impaired. Another objective of the present invention is to provide a method for producing molten iron that can produce molten iron with a high yield. [Means for solving the problem]

[0007] The present invention provides a method for controlling the molten iron temperature of a blast furnace, comprising the steps of: selecting an action model corresponding to the observation means to be used from among a plurality of action models trained by machine learning for each of two or more observation means that observe one observation quantity, using image data obtained by imaging one or more historical data of observation quantities indicating the operating state of a blast furnace process or the operating state of equipment as input data, and an operation quantity of the process or equipment determined by an operator based on the historical data as output data; and controlling the state of the process or equipment according to the operation quantity obtained by inputting the image data to the selected action model, thereby controlling the molten iron temperature.

[0008] The two or more observation means may observe the moisture content of the coke charged into the blast furnace as the observed quantity.

[0009] The observed quantities may include at least one of the following: molten iron temperature, tuyeres embedded temperature, coke ratio, moisture content of coke, and weight ratio of yard coke to dry-extinguished coke. The manipulated quantities may include at least one of the following, which are manipulated quantities for controlling the molten iron temperature of the blast furnace: pulverized coal ratio, blown air moisture content, and coke ratio.

[0010] The image data is a two-dimensional image in which the time axis is placed on one axis and the history data is placed on the other axis, and the history data of the image data is preferably an image obtained by associating color and / or intensity of color with the numerical value of each history data.

[0011] The method for producing molten iron according to the present invention includes the step of producing molten iron while controlling the molten iron temperature using the molten iron temperature control method for a blast furnace according to the present invention. [Effects of the Invention]

[0012] According to the blast furnace molten iron temperature control method of the present invention, even when a large amount of inferior raw materials are used, when the moisture content of the inferior raw materials increases, or when the function of the means for observing the moisture content of coke is impaired, it is possible to suppress a decrease in the accuracy of molten iron temperature control. Furthermore, according to the molten iron manufacturing method of the present invention, molten iron can be manufactured with a high yield. [Brief explanation of the drawing]

[0013] [Figure 1] Figure 1 is a block diagram showing the configuration of a learning model generation device, which is one embodiment of the present invention. [Figure 2] Figure 2 shows an example of a method for measuring the moisture content of coke. [Figure 3] Figure 3 is a flowchart showing the flow of the learning model generation process, which is one embodiment of the present invention. [Figure 4]Figure 4 is a flowchart showing the flow of a molten iron temperature control process, which is one embodiment of the present invention. [Modes for carrying out the invention]

[0014] The configuration and operation of a learning model generation device, which is one embodiment of the present invention, will be described below with reference to the drawings.

[0015] 〔composition〕 Figure 1 is a block diagram showing the configuration of a learning model generation device, which is one embodiment of the present invention. As shown in Figure 1, the learning model generation device 1, which is one embodiment of the present invention, is a device that generates an action model by machine learning that determines the amount of operation of a process or equipment that controls the molten iron temperature of a blast furnace by an operator. The learning model generation device 1 is composed of a well-known information processing device equipped with a processor, memory, etc., and the processor functions as a model learning unit 11 and a model output unit 12 by executing a computer program. The functions of each of these units will be described later.

[0016] The learning model generation device 1 is connected to a history data database (history data DB) 2 and a learning model database (learning model DB) 3 in a data-readable format. The history data DB 2 stores the learning data used when generating the behavioral model described above. Specifically, the history data DB 2 stores learning data that is associated with image data of one or more historical quantities that indicate the operating state of the controlled process or the operating state of the controlled equipment, and the process or equipment operation quantities determined by the operator based on the history data. Furthermore, learning data is prepared for each of the two or more observation means that observe a single observed quantity.

[0017] Here, the observed quantities include at least one of the following: molten iron temperature, tuyeres embedded temperature, coke ratio, moisture content of coke, and the weight ratio of yard coke to dry-fired coke. The coke ratio refers to the weight ratio of coke to the amount of iron converted to molten iron contained in the iron ore charged from the top of the furnace. The weight ratio of yard coke to dry-fired coke refers to the weight ratio of coke placed directly in the field to coke cooled in a dry coke quenching (CDQ) system. Furthermore, if the observed quantity is the moisture content of the coke charged into the blast furnace, two or more observation means can be exemplified: an infrared moisture meter 22 that measures the moisture content of coke transported by a belt conveyor 21 as shown in Figure 2(a), and a neutron moisture meter 24 equipped with a neutron source 24a and a neutron detector 24b that measures the moisture content of coke in a coke hopper 23 as shown in Figure 2(b). Furthermore, the controllable variables include at least one of the following: pulverized coal ratio, blown air moisture content, and coke ratio, which are controllable variables for controlling the molten iron temperature of the blast furnace.

[0018] Furthermore, the image data is presented as a two-dimensional image data set with historical data of observed quantities, where one axis represents the time axis and the other axis represents the observed quantity, and one or more observed quantities are placed in a direction opposite to the time axis. The historical display of each observed quantity may be a line diagram like a normal trend graph, or it may be a so-called contour plot obtained by associating the numerical value of each historical data with color and / or intensity of color. In addition, the image data may also include historical data of process or instrument manipulation quantities. Furthermore, it is advisable to normalize the historical data of observed quantities and manipulation quantities before creating the image data in order to unify the value range.

[0019] The learning model DB3 stores data for behavioral models generated for each observation method using the learning data for each observation method. When new learning data is added to the history data DB2, the learning model generation device 1 should update the behavioral models stored in the learning model DB3 by using the newly added learning data as additional learning data to further train the behavioral models.

[0020] The learning model generation device 1 having such a configuration generates, by executing the learning model generation process shown below, an action model for an operator to determine the operation amount of a process or equipment by machine learning for each observation means. Hereinafter, the operation of the learning model generation device 1 when executing the learning model generation process will be described with reference to FIG. 3.

[0021] 〔Learning model generation process〕 FIG. 3 is a flowchart showing the flow of the learning model generation process according to an embodiment of the present invention. The flowchart shown in FIG. 3 starts at the timing when an execution command for the learning model generation process is input to the learning model generation device 1, and the learning model generation process proceeds to the process of step S1. The learning model generation process is executed for the learning data of each observation means stored in the history data DB2.

[0022] In the process of step S1, the model learning unit 11 acquires the learning data of the observation means to be processed. Thereby, the process of step S1 is completed, and the learning model generation process proceeds to the process of step S2.

[0023] In the process of step S2, the model learning unit 11 uses the learning data acquired in the process of step S1 to machine-learn an action model for an operator to determine the operation amount of a process or equipment, with the image data obtained by imaging the history data of one or more observation quantities indicating the operating state of the process to be controlled or the operating state of the equipment to be controlled as input data, and the operation amount of the process or equipment determined by the operator based on the history data as output data. Here, the action model may be a convolutional neural network (CNN). Further, when the form of the image data is a two-dimensional image in which the time axis is arranged on one axis and the history data is arranged on the other axis, the convolution operation of the CNN may be performed only in the time axis direction.

[0024] Furthermore, the input data should be image data of a predetermined continuous time interval, and the output data should be the value of the manipulated variable immediately after the predetermined time interval. Also, as training data, considering that the blast furnace is a process with a long time constant for its response to changes in the manipulated variable after a predetermined time has elapsed, it is preferable to use a combination of image data and manipulated variable where one or more values ​​of predetermined observed variables (as controlled variables) after a predetermined time has elapsed are within a predetermined range centered on their respective target values. This is because data in which the controlled variable has changed appropriately as a result of the operator's actions and has been controlled to a predetermined range from the target value is used as training data. With this, the processing of step S2 is completed, and the learning model generation process proceeds to the processing of step S3.

[0025] In step S3, the model output unit 12 stores the behavioral model corresponding to the observation means being processed, which was generated in step S2, into the learning model DB3. This completes step S3, and the series of learning model generation processes are finished.

[0026] Thereafter, when controlling the molten iron temperature, as shown in Figure 4, the operator selects an action model from the action models stored in the learning model DB3 that corresponds to the observation means used to observe the observed quantity (step S11). Note that the selection of the action model may be performed automatically by the molten iron temperature control device depending on the failure or operating status of the observation means. The molten iron temperature control device then controls the state of the process or equipment according to the manipulated quantity obtained by inputting image data of the observed quantity history for a predetermined interval into the selected action model (step S12). For example, when the moisture content of the inferior raw material is decreasing, the device predicts that the molten iron temperature will rise and outputs an action such as decreasing the pulverized coal ratio or coke ratio or increasing the blown-humidified content. This makes it possible to suppress a decrease in the accuracy of molten iron temperature control even when a large amount of inferior raw material is used, when the moisture content of the inferior raw material increases, or when the function of the observation means for the moisture content of coke is impaired. As a result, molten iron can be manufactured with a high yield.

[0027] Although embodiments applying the invention made by the present inventors have been described above, the present invention is not limited by the descriptions and drawings that constitute part of the disclosure of the present invention in this embodiment. That is, all other embodiments, examples, and operational techniques made by those skilled in the art based on this embodiment are included in the scope of the present invention. [Explanation of Symbols]

[0028] 1. Learning Model Generation Device 2. Historical Data Database (Historical Data DB) 3. Learning Model Database (Learning Model DB) 11 Model Learning Section 12 Model Output Section 21 Belt conveyor 22 Infrared moisture meter 23 Coke hopper 24 Neutron moisture meter 24a Neutron source 24b Neutron detector

Claims

1. The process involves selecting an action model from among multiple action models trained on machine learning for each of two or more observation means that observe a single observed quantity, using image data obtained by visualizing one or more historical data of observed quantities that indicate the operating status of a blast furnace process or the operating status of equipment as input data, and process or equipment operation quantities determined by the operator based on the historical data as output data, and selecting an action model corresponding to the observation means to be used. The steps include controlling the molten iron temperature by controlling the state of the process or equipment according to the manipulated values ​​obtained by inputting the image data into the selected behavioral model, A method for controlling the molten iron temperature of a blast furnace, including the following.

2. The method for controlling the molten iron temperature of a blast furnace according to claim 1, wherein the two or more observation means observe the amount of water content in the coke charged into the blast furnace as the observed quantity.

3. The blast furnace molten iron temperature control method according to claim 1, wherein the observed quantity includes at least one of molten iron temperature, tuyeres embedded temperature, coke ratio, moisture content of coke, and weight ratio of yard coke to dry-extinguished coke, and the manipulated quantity includes at least one of pulverized coal ratio, blown air moisture, and coke ratio, which are manipulated quantities for controlling the molten iron temperature of the blast furnace.

4. The method for controlling the molten iron temperature of a blast furnace according to claim 1, wherein the image data is a two-dimensional image in which the time axis is arranged on one axis and the history data is arranged on the other axis, and the history data of the image data is an image obtained by associating color and / or intensity of color with the numerical value of each history data.

5. A method for producing molten iron, comprising the step of producing molten iron while controlling the molten iron temperature using a molten iron temperature control method for a blast furnace described in any one of claims 1 to 4.

Images