JP6624212B2 - Blast furnace heat prediction device and blast furnace heat prediction method - Google Patents

Description

  The present invention relates to a blast furnace heat prediction device and a blast furnace heat prediction method for predicting hot metal temperature of a blast furnace as blast furnace heat.

  It is important to stabilize the hot metal temperature of the blast furnace from the viewpoint of reducing the fuel consumption rate and stabilizing the Si content of the hot metal. The hot metal temperature is controlled based on the blast temperature, blast moisture, the amount of pulverized coal injected, the amount of oxygen enrichment, etc., which are the blast conditions from the tuyere. For example, when the humidification is reduced, the hot metal temperature increases later. However, usually, there is a time delay of two hours or more from the change of the blowing condition to the change of the hot metal temperature. For this reason, if the blowing conditions are changed based only on the hot metal temperature at the time of the operation, excessive operation or reverse operation occurs, and the fluctuation of the hot metal temperature increases. Therefore, in order to stabilize the hot metal temperature, the hot metal temperature several hours ahead, specifically, accurately predicts the hot metal temperature at least two hours or more ahead where the change in the blowing conditions is reflected in the hot metal temperature, based on the prediction result It is necessary to control the blowing conditions.

  For the prediction of hot metal temperature, many methods have been proposed for estimating hot metal temperature based on the similarity between current operation data and past operation data using time series data of multiple items past the prediction timing. . For example, Patent Literature 1 describes a method of searching for a case similar to a past process state from a process time-series database. The method described in Patent Literature 1 uses operation data in which the number of variables is reduced by converting a plurality of process variable values into independent component variable values, and uses a similarity between current operation data and past operation data. Predicts the future state of the process based on

JP 2005-135010 A

  However, in order to ensure the accuracy of hot metal temperature prediction using the past operation data, operation data similar to the operation data from the prediction timing to a predetermined time before is included to some extent in the past operation data. There is a need. For this reason, when the current operation data is in a “sparse” region in the data space of the past operation data, the prediction accuracy of the hot metal temperature of the blast furnace decreases. Specifically, the accuracy of predicting the hot metal temperature in an unsteady state, such as an abnormal drop in the hot metal temperature, which is originally desired to be predicted, decreases because there are few similar cases in the past.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a blast furnace furnace heat prediction device and a blast furnace furnace heat prediction method capable of always maintaining the prediction accuracy of hot metal temperature in a high state. .

  The blast furnace furnace heat prediction device according to the present invention, for each actual value of the actual process data including information on the actual value of the operating conditions of the blast furnace and the actual value of the hot metal temperature when operating the blast furnace under the operating conditions, A data expansion unit that extracts data up to a predetermined time before and creates a performance data set; a data difference value calculation unit that calculates a time change amount of each performance value using the performance data set; A data difference value developing unit that extracts a time change amount data of each actual value for a predetermined time to create a time change amount data set; and a time change of the actual value of a plurality of operating conditions in the time change amount data set. For the amount, the similarity calculation unit that calculates the similarity to the time change of the operating conditions of the blast furnace at the time of prediction of the hot metal temperature, and the time change of the actual value of the operating condition in the time change data set and the class Using the similarity calculated by the temperature calculation unit, a furnace heat prediction formula for creating a prediction formula of the time change of the hot metal temperature representing the relationship between the time change of the operating conditions of the blast furnace and the time change of the hot metal temperature. A furnace heat prediction for predicting a time change amount of the hot metal temperature at the prediction time by substituting the time change amount of the operating condition of the blast furnace at the prediction time into the prediction formula prepared by the preparation unit and the furnace heat prediction formula preparation unit. And a unit.

  The blast furnace furnace heat prediction device according to the present invention, in the above invention, wherein the operating conditions include blast sensible heat, tuyere tip combustion heat, solution reaction heat, furnace top gas sensible heat, blast moisture decomposition heat, furnace body heat dissipation. , Pulverized coal combustion heat, pulverized coal decomposition heat, slag sensible heat, charged raw material sensible heat, and hot metal sensible heat.

  In the blast furnace furnace heat prediction device according to the present invention, in the above invention, the furnace heat prediction unit considers a time of the hot metal temperature in consideration of a time delay of the time change of the hot metal temperature with respect to a time change of the amount of heat included in the operating conditions. It is characterized by predicting a change amount.

  The blast furnace furnace heat prediction method according to the present invention, for each actual value of the actual process data including the actual value of the operating conditions of the blast furnace and the actual value of the hot metal temperature when operating the blast furnace under the operating conditions, A data expansion step of extracting data up to a predetermined time before and creating a performance data set; a data difference value calculation step of calculating a time change amount of each performance value using the performance data set; A data difference value development step of extracting a time change amount data of each actual value for a predetermined time to create a time change amount data set, and a time change of the actual value of a plurality of operating conditions in the time change amount data set; A similarity calculating step of calculating a similarity to a time variation of the operating conditions of the blast furnace at the time of the prediction of the hot metal temperature; Using the change amount and the similarity calculated in the similarity calculation step, a prediction formula for a time change amount of the hot metal temperature representing a relationship between the time change amount of the operating conditions of the blast furnace and the time change amount of the hot metal temperature is created. Furnace heat prediction formula creating step, and by substituting the time change of the operating conditions of the blast furnace at the forecast time into the forecast formula created in the furnace heat forecast formula creating step, the time change of the hot metal temperature at the forecast time is calculated. And estimating the furnace heat.

  In the blast furnace furnace heat prediction method according to the present invention, in the above-mentioned invention, the operating conditions include blast sensible heat, tuyere tip combustion heat, solution reaction heat, furnace top gas sensible heat, blast moisture decomposition heat, furnace body heat dissipation. , Pulverized coal combustion heat, pulverized coal decomposition heat, slag sensible heat, charged raw material sensible heat, and hot metal sensible heat.

  In the blast furnace furnace heat prediction method according to the present invention, in the above-described invention, the furnace heat prediction step includes a time period of the hot metal temperature in consideration of a delay time of the time change of the hot metal temperature with respect to a time change of the heat amount included in the operating conditions. The method includes a step of predicting a change amount.

  According to the blast furnace heat prediction apparatus and the blast furnace heat prediction method according to the present invention, not the similarity between the absolute value of the operation condition of the prediction target and the absolute value of the past operation condition, but the time change of the operation condition of the prediction target Since the similarity between the amount and the time change amount of the past operating conditions is calculated, even if the absolute value of the operating condition to be predicted is in a sparse region in the data space of the absolute value of the past operating condition. It is possible to extract past operating conditions that are similar to the time variation of the operating conditions to be predicted, and to always maintain high accuracy in predicting the hot metal temperature.

FIG. 1 is a block diagram illustrating a configuration of a blast furnace heat prediction system according to an embodiment of the present invention. FIG. 2 is a flowchart showing a flow of a blast furnace heat prediction process according to an embodiment of the present invention. FIG. 3 is a diagram illustrating an example of a time variation data set. FIG. 4 is a diagram showing the actual value and the predicted value of the time change amount of the hot metal temperature.

  Hereinafter, the configuration of a blast furnace heat prediction system according to an embodiment of the present invention will be described with reference to the drawings.

[Configuration of blast furnace heat prediction system]
FIG. 1 is a block diagram illustrating a configuration of a blast furnace heat prediction system according to an embodiment of the present invention. As shown in FIG. 1, a blast furnace furnace heat prediction system 1 according to one embodiment of the present invention includes an input device 2 operated by an operator when the operator inputs various operation inputs, and a hot metal temperature (blast furnace furnace heat). Output device 3 that outputs prediction results, etc., and stores actual process data including information on actual values of operating conditions of blast furnace system A and actual values of hot metal temperature when operating blast furnace system A under the operating conditions. The operation result database 4 and the blast furnace heat prediction device 5 are provided as main components.

  The blast furnace heat prediction device 5 is configured by an information processing device such as a personal computer, and the internal arithmetic processing device executes a computer program, whereby a calorie calculation unit 51, a data development unit 52, a data difference value calculation unit 53, It functions as a data difference value developing unit 54, a similarity calculating unit 55, a furnace heat prediction formula creating unit 56, and a furnace heat predicting unit 57. The functions of these units will be described later.

  In the blast furnace heat prediction system 1 having such a configuration, the blast furnace heat prediction device 5 executes the blast furnace heat prediction processing described below, thereby predicting the hot metal temperature while always maintaining a high prediction accuracy. . Hereinafter, the operation of the blast furnace furnace heat prediction device 5 when performing the blast furnace furnace heat prediction process will be described.

[Blast furnace heat prediction processing]
FIG. 2 is a flowchart showing a flow of a blast furnace heat prediction process according to an embodiment of the present invention. The flowchart shown in FIG. 2 starts when the actual process data is input to the operation result database 4, and the blast furnace furnace heat prediction process proceeds to the process of step S1. The result process data is stored in the operation result database 4 at every predetermined control cycle.

  In the process of step S1, the calorie calculating unit 51 uses the actual values of the operating conditions of the blast furnace system A input to the operational results database 4 to determine the hot metal temperature. Of combustion heat, solution reaction heat, furnace gas sensible heat, blast moisture decomposition heat, furnace body dissipated heat, pulverized coal combustion heat, pulverized coal decomposition heat, slag sensible heat, charged raw material sensible heat, and hot metal sensible heat Is calculated. These amounts of heat correspond to the operating conditions according to the present invention. Thereby, the process of step S1 is completed, and the blast furnace heat prediction process proceeds to the process of step S2.

  In the process of step S2, the data developing unit 52 extracts a plurality of actual values of the calorific value and the hot metal temperature calculated by the calorific value calculating unit 51 retroactively to the past before the first predetermined time, thereby setting the actual data set. Create Specifically, the data developing unit 52 uses the actual values of the plurality of calories calculated by the calorific value calculating unit 51 retroactively to the past before the first predetermined time as an input variable and the actual values of the hot metal temperature as output variables. Create an actual data set by extracting. Thereby, the process of step S2 is completed, and the blast furnace heat prediction process proceeds to the process of step S3.

  In the process of step S3, the data difference value calculation unit 53 calculates the difference data (the time change) of the actual values of a plurality of calories and hot metal temperature for a preset data difference time from the actual data set created by the process of step S2. Volume data). The data difference time is an arbitrary value determined by the accuracy of the prediction of the blast furnace heat thereafter. Note that the data difference value calculation unit 53 may calculate difference data after performing a smoothing process on the performance data set. Thus, the process of step S3 is completed, and the blast furnace heat prediction process proceeds to the process of step S4.

In the process of step S4, the data difference value developing unit 54 creates a time change amount data set by extracting a plurality of difference data of the actual values of the calorific value and the hot metal temperature for the set second predetermined time in the past. . Specifically, the data difference value developing unit 54 inputs the difference data of the actual values of the plurality of calories for the set second predetermined time in the past as input variables and the difference data (observation data) of the actual values of the hot metal temperature. By extracting it as an output variable, a time variation data set as shown in FIG. 3 is created. Here, in the time variation data set shown in FIG. 3, the item name of the output variable (hot metal temperature) is Y, and the item name of the M input variables (calorific value) is X _m (m = 1, 2,..., M ). The number of output variables is the N, n-th (n = 1, 2, ..., N) the value of the output variable ^y n of the value of n-th input variable is represented as _x ^{m n.} Thereby, the process of step S4 is completed, and the blast furnace heat prediction process proceeds to the process of step S5.

In the process of step S5, the similarity calculating unit 55 sets the similarity to the input variable for which the hot metal temperature is to be predicted for the set ^xn of N input variables included in the time variation data set created by the process of step S4. Calculate the degree. More specifically, first, the similarity calculating unit 55, expressed as Equation (1) below the input vector (request point) x ^r indicating the input variables of the prediction target molten iron temperature. Here, in Equation (1), X _m ^r indicates the time variation of the M heat in the prediction timing of the hot metal temperature.

Next, the similarity calculation unit 55 uses the set ^xn of N input variables included in the time variation data set to calculate the difference data of the hot metal temperature and the difference between the calorific ^values as in the following equation (2). Create a linear regression model showing the relationship to the data. The parameter b in equation _{(2), a 1, a} 2, ..., a m are calculated by least square method, is used as a distance function partial regression coefficient vector α is later shown in the following equation (3) .

Next, the similarity calculating unit 55 is defined as Equation (5) that indicates the distance L from the request point x ^r in the x is that the input variables space shown in the following equation (4) below. Here, since the partial regression coefficient vector α can be considered as the contribution of each input variable to the time variation of the output variable, the distance L can be said to be a weighted distance in consideration of the contribution.

Next, the similarity calculating unit 55 calculates the distance L from the request point x ^r for each of a set x ⁿ of N input variables included in the time variation data set. For details, n-th (n = 1,2, ..., N ) the distance ^{L n} from the requesting point ^{x r} of the set ^{x n} input variables is obtained by equation (6) shown below. Here, ^xn in Expression (6) is expressed as Expression (7) below. In the following description, to be expressed as Equation below summarizes the distance L ⁿ from 1~N second request point x ^r of the set x ⁿ input variables (8).

Next, the similarity calculating unit 55, a similarity W representing the closeness of the requested point x ^r is defined as Equation (9) below. Here, in Expression (9), σ (l) indicates the standard deviation of the distance 1 shown in Expression (8), and p indicates an adjustment parameter.

Next, the similarity calculating unit 55, for each of a set x ⁿ of N input variables included in the time variation data set, calculates the similarity W ⁿ from the requesting point x ^r. For details, n-th (n = 1,2, ..., N ) similarity ^{W n} from the requesting point ^{x r} of the set ^{x n} input variables can be determined from equation (10) shown below. Also, to be expressed as Equation (11) below summarizes the similarity W ⁿ from 1~N second request point x ^r of the set x ⁿ input variables. Thereby, the process of step S5 is completed, and the blast furnace heat prediction process proceeds to the process of step S6.

In the process of step S6, the furnace heat prediction equation creating unit 56, similarity W from the requesting point x ^r calculated by the processing of the set of N input variables included in the time variation data set x ⁿ and S5 ^{Using n} , a linear regression model that shows the relationship between the difference data of the hot metal temperature and the difference data of each calorific value as shown in the following Expressions (12) and (13) is created. Here, the parameter θ shown in Equation (13), obtained by the weighted least square method to weight the similarity W ^n. Result, a large amount of heat similarity W ⁿ (heat close to the required point x ^r) is large weight, low heat similarity (distant heat from the required point x ^r) is a regression equation model obtained as the weight decreases In addition, a regression model that can more accurately fit the actual value of each heat quantity near the required point can be created. The regression equation model corresponds to the prediction equation according to the present invention. Thereby, the process of step S6 is completed, and the blast furnace furnace heat prediction process proceeds to the process of step S7.

In the process at step S7, the furnace heat prediction unit 57, by substituting the input variable requirements point x ^r to regression model created by the processing in step S6, the time variation of molten iron temperature in the request point x ^r Predict. Note that the feature of the blast furnace heat prediction processing in this embodiment is to use a time change data set, and thus the prediction method is not limited to the method using the regression equation model, and the time change of each past heat quantity is not limited. Any method may be used as long as the method predicts the amount of change in the hot metal temperature over time based on the similarity between the amount and the amount of change in the amount of heat at the prediction timing. Thereby, the process of step S7 is completed, and a series of the blast furnace furnace heat prediction process ends.

  As is apparent from the above description, in the blast furnace heat prediction processing according to one embodiment of the present invention, the data development unit 52 performs the operation of the blast furnace under the actual values of the blast furnace operating conditions and the operating conditions. For each actual value of the actual process data including the information on the actual value of the hot metal temperature, data is extracted up to a first predetermined time before to create an actual data set, and the data difference value calculation unit 53 generates the actual data set. The data difference value developing unit 54 calculates the time change amount of each actual value by using the data, and extracts the data of the time change amount of each actual value for the second predetermined time to create a time change amount data set. Then, the similarity calculating unit 55 calculates the similarity to the time variation of the operating conditions of the blast furnace at the time of prediction of the hot metal temperature, with respect to the time variation of the actual values of the plurality of operating conditions in the time variation data set, The furnace heat prediction formula creation unit 56 uses the time change amount of the actual value of the operation condition in the time change amount data set and the similarity calculated by the similarity calculation unit 55 to calculate the time change amount of the operation condition of the blast furnace. A prediction formula for the time change of the hot metal temperature representing the relationship with the time change of the hot metal temperature is created, and the furnace heat prediction unit 57 uses the prediction formula created by the furnace heat prediction formula creation unit 56 to calculate the blast furnace at the prediction time. The time variation of the hot metal temperature at the prediction time is predicted by substituting the time variation of the operating conditions. This calculates not the similarity between the absolute value of the operation data to be predicted and the absolute value of past operation data, but the similarity between the time change of the operation conditions to be predicted and the time change of the past operation conditions. Therefore, even if the absolute value of the predicted operating condition is in a sparse area in the data space of the absolute value of the past operating condition, the past operating condition similar to the temporal change amount of the predicted operating condition is used. And the prediction accuracy of the hot metal temperature can always be kept high. Further, if the time change of the furnace heat can be predicted in this way, the future furnace heat itself can be predicted by adding the predicted time change amount to the current furnace heat.

  In addition, since the input values of calorie and hot metal temperature vary greatly with time according to the operating conditions and the raw materials used, if these absolute values are used as they are, if the state at the time of prediction is a rare calorie or hot metal temperature, Since the amount of past data close to time and state is reduced, the prediction accuracy is reduced. On the other hand, even if the state at the time of prediction is a rare calorie or hot metal temperature, taking the time change amount, the probability that there is past data that is close to the state at the time of prediction, if the absolute value is used as it is , The prediction accuracy can be prevented from lowering. When data on input and output heat in the furnace is used as the input value, it is considered that a weighted linear sum of the time changes of the heat amounts appears in the change amount of the furnace heat in the future. .

  In the blast furnace furnace heat prediction processing according to one embodiment of the present invention, the operating conditions include blast sensible heat, tuyere tip combustion heat, solution reaction heat, furnace top gas sensible heat, blast moisture decomposition heat, furnace body radiation. At least two of calorie, pulverized coal combustion heat, pulverized coal decomposition heat, slag sensible heat, charged raw material sensible heat, and hot metal sensible heat are included. This makes it possible to grasp which amount of temporal change in the amount of heat contributes to the temporal change in the hot metal temperature at the timing of predicting the hot metal temperature. In addition, according to the regression equation model using only past operating conditions similar to the operating conditions to be predicted, the influence of each heat amount on the hot metal temperature can be determined by the partial regression coefficient. The hot metal temperature can be controlled by the operation. On the other hand, Patent Literature 1 discloses, as variables used for the prediction of hot metal temperature, pulverized coal injection amount, solution loss carbon, heat flow ratio, charging pitch, Si amount, hot air temperature, furnace top temperature, hot metal temperature, and the like. Has been described. However, the physical dimensions of these variables vary, and the variables are selected simply based on whether they are statistically related to the hot metal temperature. For this reason, the selected variable is not based on the physical phenomenon, and it is difficult to verify the physical validity of the variable used for prediction and to physically interpret the prediction formula.

In the present embodiment, six heat amount time changes that strongly influence the time change amount of the hot metal temperature are selected. The time for estimating the time change amount of the hot metal temperature was set to 2 hours later, and the second predetermined time when creating the time change amount data set was set to 2 hours. In addition, the delay time of the time change of the hot metal with respect to the time change of the calorie is considered (the parameter τ _i shown in Expression (14)). Note that the delay time here indicates how long the hot metal temperature has changed after the heat quantity has changed. In addition, since the data of the hot metal temperature to be used is batch data (non-periodic data) unlike the data which can be calculated at a fixed period like the calorific value, the interpolation processing was performed so as to match the calculation time of the calorific value. The calculation formula used for this prediction is shown in the following formula (14). FIG. 4 shows the prediction results of the blast furnace heat. As shown in FIG. 4, it was confirmed that the predicted value of the time change of the hot metal temperature after 2 hours almost matched the actual value of the time change of the hot metal temperature.

  As described above, the embodiment to which the invention made by the present inventors is applied has been described. However, the present invention is not limited by the description and the drawings that constitute a part of the disclosure of the present invention according to the present embodiment. That is, other embodiments, examples, operation techniques, and the like performed by those skilled in the art based on this embodiment are all included in the scope of the present invention.

1 Blast furnace heat prediction system 2 Input device 3 Output device 4 Operation results database 5 Blast furnace heat prediction device 51 Heat quantity calculation unit 52 Data development unit 53 Data difference value calculation unit 54 Data difference value development unit 55 Similarity calculation unit 56 Furnace heat Prediction formula creation unit 57 Furnace heat prediction unit A Blast furnace system

Claims (6)

For each actual value of the actual process data including the actual value of the operating conditions of the blast furnace and the actual value of the hot metal temperature at the time of operating the blast furnace under the operating conditions, the data up to the first predetermined time is extracted. A data expansion unit that creates an actual data set
A data difference value calculation unit that calculates a time change amount of each performance value using the performance data set,
A data difference value developing unit for extracting time change data of each actual value for a second predetermined time and creating a time change data set;
A similarity calculation method for calculating the similarity to the time change amount of the operating conditions of the blast furnace from the time before the first predetermined time to the predicted time point of the hot metal temperature for the time change amount of the actual values of the plurality of operating conditions in the time change amount data set. A degree calculator,
Using the time change amount of the actual value of the operation condition in the time change amount data set and the similarity calculated by the similarity calculation unit, the time change amount of the operation condition of the blast furnace and the time change amount of the hot metal temperature and Furnace heat prediction formula creation unit that creates a prediction formula of the amount of time change in hot metal temperature representing the relationship
The time variation of the hot metal temperature at the prediction time is predicted by substituting the time variation of the operating conditions of the blast furnace from the time before the first predetermined time to the prediction time into the prediction formula created by the furnace heat prediction formula creating unit. Furnace heat prediction section
A blast furnace heat prediction device comprising:   The operating conditions include blast sensible heat, tuyere tip combustion heat, solution reaction heat, furnace gas sensible heat, blast moisture decomposition heat, furnace body dissipated heat, pulverized coal combustion heat, pulverized coal decomposition heat, slag sensible heat, The blast furnace furnace heat prediction device according to claim 1, wherein at least two of the charged raw material sensible heat amount and the hot metal sensible heat are included.   The furnace heat predicting unit predicts a time change amount of the hot metal temperature in consideration of a delay time of the time change of the hot metal temperature with respect to a time change of the heat amount included in the operation condition, the time range of the hot metal temperature being characterized by the above-mentioned. Blast furnace heat prediction device. For each actual value of the actual process data including the actual value of the operating conditions of the blast furnace and the actual value of the hot metal temperature at the time of operating the blast furnace under the operating conditions, the data up to the first predetermined time is extracted. A data deployment step to create an actual data set
A data difference value calculation step of calculating a time change amount of each performance value using the performance data set,
A data difference value developing step of extracting time change data of each actual value for a second predetermined time and creating a time change data set;
A similarity calculation method for calculating the similarity to the time change amount of the operating conditions of the blast furnace from the time before the first predetermined time to the predicted time point of the hot metal temperature for the time change amount of the actual values of the plurality of operating conditions in the time change amount data set. A degree calculation step;
Using the time variation of the operating conditions in the time variation data set and the similarity calculated in the similarity calculation step, the relationship between the time variation of the operating conditions of the blast furnace and the time variation of the hot metal temperature is calculated. Furnace heat prediction formula creation step of creating a prediction formula of the time change of the hot metal temperature to represent
By substituting the time variation of the operating conditions of the blast furnace from the first predetermined time before to the prediction time into the prediction formula created in the furnace heat prediction formula creating step, the time variation of the hot metal temperature at the prediction time is predicted. Furnace heat prediction step,
A blast furnace heat prediction method comprising:   The operating conditions include blast sensible heat, tuyere tip combustion heat, solution reaction heat, furnace gas sensible heat, blast moisture decomposition heat, furnace body dissipated heat, pulverized coal combustion heat, pulverized coal decomposition heat, slag sensible heat, The blast furnace furnace heat prediction method according to claim 4, wherein at least two of the charged raw material sensible heat amount and the hot metal sensible heat are included.   The furnace heat prediction step includes a step of predicting a time change amount of the hot metal temperature in consideration of a delay time of the time change of the hot metal temperature with respect to a time change of the heat amount included in the operation condition. 6. The method for estimating blast furnace heat according to 5.

Images