JP2020105606A - Converter blowing control device, converter blowing control method, and program - Google Patents

Abstract

To improve accuracy of predicting a phosphorus concentration in molten steel at a time point before the start of blowing.SOLUTION: A converter blowing control device includes prediction value calculation means and correction term calculation means. The prediction value calculation means calculates a predicted value of phosphorus concentration in molten steel by adding, a first correction term which is learned on the basis of actual performance of predicted data including a predicted value and an actual value of the phosphorus concentration in the molten steel in the past blowing process in the converter, to the function for associating the molten steel data for the molten steel subjected to a blowing process in a converter and the auxiliary raw material data of the auxiliary raw material supplied to the converter with the phosphorus concentration of the molten steel in the blowing process in the converter. The correction term calculation means constructs a multivariate state space model representing the first correction term and a second correction term relating to a carbon concentration in the molten steel, and calculates the first correction term by applying a Kalman filter having a difference between the predicted value and the actual value included in the predicted actual performance data as an observation value to the state space model.SELECTED DRAWING: Figure 1

Description

The present invention relates to a converter blowing control device, a converter blowing control method, and a program.

In converter blowing, blowing control combining static control and dynamic control is performed in order to bring the molten steel component concentration (for example, carbon concentration) and the molten steel temperature at the time of blowing stop to a target value. In the static control, before starting the blowing, a mathematical model based on a mass balance and a heat balance is used to determine the amount of injected oxygen and the amount of various auxiliary materials to be introduced to achieve the above target. In the dynamic control, the same calculation is executed based on the molten steel component concentration and the molten steel temperature measured by using the sublance during blowing, and the blown oxygen amount and the input amounts of various auxiliary raw materials are corrected.

Here, in Patent Document 1, the exhaust gas component and the exhaust gas flow rate during converter blowing are periodically measured, and blowing is performed using a dephosphorization rate constant estimated based on these measured values and operating conditions. A technique for sequentially estimating the phosphorus concentration in molten steel is described. By changing the operating condition according to the result of such an estimation, the accuracy of controlling the phosphorus concentration in the molten steel at the time of blowing can be improved.

Patent Document 2 describes a technique for appropriately controlling the phosphorus concentration in molten steel at the time of blowing stop using a lime input amount calculation formula or a function representing a lime input amount calculation procedure. The function includes terms representing the molten steel temperature, the phosphorus concentration in the molten steel, the carbon concentration in the molten steel, and other operating factors, and a learning term updated for each charge of blowing, and by sequentially updating the learning term, It is possible to accurately calculate an appropriate lime input amount and appropriately control the phosphorus concentration in molten steel.

JP, 2013-23696, A JP-A-2000-309817

In recent years, in the primary refining process including converter blowing, there is an increasing need to suppress the amount of slag generated. One of the main sources of slag is a solvent solution such as quick lime that is added in the early stage of blowing. The solvent is a source of CaO used for the dephosphorization reaction as shown below. Therefore, if the phosphorus concentration in the molten steel can be accurately estimated at the initial stage of blowing, the amount of slag generated can be suppressed to the minimum by adding a necessary and sufficient amount of the medium-melting material.
3(CaO)+5(FeO)+2[P]=(3CaO·P ₂ O ₅ )+5[Fe]
*() indicates inside the slag, and [] indicates inside the hot metal.

However, since the technique described in the above-mentioned Patent Document 1 sequentially estimates the phosphorus concentration in the molten steel by using the measured values of the exhaust gas component and the exhaust gas flow rate after the start of the blowing, it is possible to obtain It cannot be used to optimize the input amount. Further, the technique described in the above-mentioned Patent Document 2 is premised on that an error corresponding to the learning term similarly occurs in each charge. That is, when the error value varies in each charge, it is difficult to accurately control the lime input amount and the molten steel phosphorus concentration even with this technique.

Therefore, an object of the present invention is to provide a converter blowing control device, a converter blowing control method, and a program capable of improving the prediction accuracy of the phosphorus concentration in molten steel before the start of blowing. ..

According to an aspect of the present invention, the hot metal data relating to the hot metal blown in the converter and the auxiliary raw material data relating to the auxiliary raw materials fed into the converter are related to the phosphorus concentration in the molten steel during the blowing process in the converter. The phosphorus concentration in molten steel is added to the function by adding the first correction term learned based on the first predicted actual data including the predicted value and the actual value of the phosphorus concentration in molten steel during the past blowing process in the converter. By constructing a multivariate state space model expressing the first correction term and the second correction term and applying a Kalman filter to the state space model. The second correction term is a correction term added to a function that relates the hot metal data and the auxiliary raw material data to the carbon concentration in the molten steel during the blowing process in the converter. , The Kalman filter is learned based on the second predicted actual data including the predicted value and the actual value of the carbon concentration in the molten steel during the past blowing process in the converter, and the Kalman filter uses the molten steel medium phosphorus contained in the first predicted actual data. Provided is a converter blowing control device that uses, as an observed value, a difference between a predicted value of a concentration and an actual value, and a difference between a predicted value of a carbon concentration in molten steel and an actual value included in the second predicted actual data.
According to the above configuration, the learning of the correction term for calculating the predicted value of the phosphorus concentration in the molten steel can be sequentially executed, and the influence of the essential process variation included in the correction term and the noise can be distinguished. You can Further, by expressing the correction term with a multivariate state space model, the prediction accuracy of the correction term is improved. Therefore, it is possible to improve the prediction accuracy of the phosphorus concentration in the molten steel before the start of blowing.

According to another aspect of the present invention, the hot metal data relating to the hot metal blown in the converter, and the auxiliary raw material data relating to the auxiliary raw material to be fed into the converter are set to the phosphorus concentration in the molten steel during the blowing process in the converter. By adding the first correction term learned based on the first predicted actual data including the predicted value and the actual value of the phosphorus concentration in the molten steel during the past blowing process in the converter to the associating function, A predictive value calculation step of calculating a predicted value of the concentration, a multivariate state space model expressing the first correction term and the second correction term, and applying a Kalman filter to the state space model The second correction term is a correction term added to a function that relates the hot metal data and the auxiliary raw material data to the carbon concentration in the molten steel during the blowing process in the converter. Yes, the Kalman filter is learned based on the second predicted actual result data including the predicted value and the actual value of the carbon concentration in the molten steel during the past blowing process in the converter, and the Kalman filter includes the molten steel contained in the first predicted actual data. Provided is a converter blowing control method in which the difference between the predicted value and the actual value of the phosphorus concentration and the difference between the predicted value and the actual value of the carbon concentration in molten steel included in the second predicted actual data are observed values. ..

According to still another aspect of the present invention, the hot metal data relating to the hot metal blown in the converter, and the auxiliary raw material data relating to the auxiliary raw materials to be fed into the converter are used as the phosphorus concentration in the molten steel during the blowing treatment in the converter. By adding the first correction term learned based on the first predicted actual result data including the predicted value and the actual value of the phosphorus concentration in the molten steel during the past blowing process in the converter to the function relating to By constructing a predictive value calculating means for calculating a predictive value of phosphorus concentration and a multivariate state space model expressing the first correction term and the second correction term, and applying a Kalman filter to the state space model. And a correction term added to a function relating the hot metal data and the auxiliary raw material data to the carbon concentration in the molten steel during the blowing process in the converter. That is, the learning is performed based on the second predicted actual data including the predicted value and the actual value of the carbon concentration in the molten steel during the past blowing process in the converter, and the Kalman filter includes the molten steel included in the first predicted actual data. A computer is used as a converter blowing control device that uses the difference between the predicted value and the actual value of the medium phosphorus concentration and the difference between the predicted value and the actual value of the molten steel carbon concentration included in the second predicted actual data as the observed value. A program is provided to make it work.

It is a figure which shows the schematic structure of the refining equipment containing the converter blowing control apparatus which concerns on one Embodiment of this invention. It is a flow chart which shows roughly the process of the converter blowing control method concerning one embodiment of the present invention. 3 is an example of a time chart when executing the method shown in FIG. 2. 6 is a graph for explaining the effect of the state space model in the embodiment of the present invention. 6 is a graph for explaining the effect of the state space model in the embodiment of the present invention. 6 is a graph for explaining the effect of the state space model in the embodiment of the present invention. 6 is a graph for explaining the effect of the state space model in the embodiment of the present invention. It is a graph for explaining the optimization of the amount of solvent injection in one embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this specification and the drawings, constituent elements having substantially the same functional configuration are designated by the same reference numerals, and duplicate description is omitted.

In one embodiment of the present invention to be described below, in the hot metal blowing process (converter blowing process) in a converter, before the start of the blowing process, a model formula is used to melt phosphorus in molten steel during the blowing process. Calculate the predicted concentration. Here, in the present specification, the blowing process means a period from the start of the blowing process to the end (blowing stop) of the blowing process, and the predicted value of the phosphorus concentration in the molten steel is arbitrary during this period. It is calculated for the time point. Specifically, for example, a predicted value of the phosphorus concentration in the molten steel at the time of the intermediate sublance measurement described later may be calculated, or a predicted value of the phosphorus concentration in the molten steel at the time of blowing stop may be calculated.

Further, in the process of calculating the predicted value of the phosphorus concentration in the molten steel, the actual value of the phosphorus concentration in the molten steel in the past blowing process in the converter is referred to, but this actual value is also arbitrary in the blowing process. What was measured at the time of can be used. Therefore, for example, if the actual value of the phosphorus concentration in the molten steel at the time of the intermediate sublance measurement is available, the actual value of the phosphorus concentration in the molten steel at the time of blowing is not necessarily required. Therefore, one embodiment of the present invention described below can be used even when a so-called direct tap is adopted, which is used for tapping without measuring the molten steel component concentration or the molten steel temperature at the time of blowing.

(System configuration)
FIG. 1 is a diagram showing a schematic configuration of refining equipment including a converter blowing control device according to an embodiment of the present invention. As shown in FIG. 1, the refining equipment 1 includes a converter equipment 10, a measurement control device 20, and a converter blowing control device 30. Of these, the converter equipment 10 includes a converter 11, an upper blowing lance 12, and a charging device 13. In the converter equipment 10, the decarburization process of the primary refining is performed by the oxygen gas supplied to the hot metal 111 by the upper blowing lance 12 inserted from the furnace opening of the converter 11. The hot metal 111 that has been decarburized is sent to the next step as molten steel 112. Further, in the decarburization treatment, phosphorus and silicon in the hot metal 111 also react with the oxygen gas 121 or the auxiliary raw material contained in the slag 113, and are taken into the slag 113 and stabilized. The charging device 13 charges the auxiliary material 131 including the quick lime forming the slag 113 into the converter 11. When the auxiliary material 131 is powder, it is also possible to blow it together with the oxygen gas 121 using the upper blowing lance 12.

The measurement control device 20 executes various measurements regarding refining processing in the converter equipment 10 and controls refining processing. Specifically, the measurement control device 20 includes a sublance 21, an oxygen supply device 22, and an auxiliary material charging control device 23. The sub-lance 21 is inserted together with the upper blowing lance 12 from the furnace opening of the converter 11, and the measuring device provided at the tip is immersed in the molten steel 112 at a predetermined timing during the decarburization process, so that the carbon concentration and the phosphorus concentration can be reduced. The component concentration of the molten steel 112 containing, the temperature of the molten steel 112 (hereinafter, also referred to as molten steel temperature), etc. are measured. The measurement using the sublance 21 during blowing is referred to as an intermediate sublance measurement in this specification. The result of the sublance measurement is transmitted to the converter blowing control device 30. The oxygen supply device 22 supplies the oxygen gas 121 to the upper blowing lance 12. The flow rate of the oxygen gas 121 supplied can be adjusted. The auxiliary material charging control device 23 controls the charging of the auxiliary material 131 by the charging device 13. Specifically, the auxiliary raw material input control device 23 controls the timing and the input amount of the auxiliary raw material 131. The above-described operations of the oxygen supply device 22 and the auxiliary raw material charging control device 23 are both executed according to a control signal received from the converter blowing control device 30.

The converter blowing control device 30 includes a communication unit 31, a calculation unit 32, a storage unit 33, and an input/output unit 34. The communication unit 31 is various communication devices that communicate with each element of the measurement control device 20 in a wired or wireless manner. The communication unit 31 receives a measurement result obtained by the measurement control device 20 and transmits a control signal to the measurement control device 20. To do. The calculation unit 32 is a calculation device that executes various calculations according to a program, and is configured by, for example, a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Only Memory). The program is stored in the ROM or the storage unit 33. In the converter blowing control device 30 described above, the calculation unit 32 functions as a molten steel phosphorus concentration prediction unit 321, a correction term calculation unit 322, and a medium-fluid material amount correction unit 323 by operating according to a program. The storage unit 33 is a storage capable of storing various data, and stores hot metal/auxiliary raw material data 331, predicted performance data 332, target data 333, and parameters 334. These data are stored as initial data, for example, and are updated at any time according to the result of the calculation in the calculation unit 32. The input/output unit 34 includes an output device such as a display or a printer and an input device such as a keyboard, a mouse, or a touch panel. The output device outputs, for example, a value such as the phosphorus concentration in molten steel during blowing, which is predicted by the phosphorus concentration in molten steel prediction unit 321. The input device acquires, for example, an instruction input regarding the control executed by the medium-melting material amount correction unit 323.

In the converter blowing control device 30 described above, the hot metal/auxiliary raw material data 331 stored in the storage unit 33 is the hot metal data relating to the hot metal 111 blown in the converter 11, and the secondary hot metal data supplied to the converter 11. And auxiliary material data regarding the material 131. The hot metal data includes, for example, the initial hot metal weight for each charge, the concentrations of hot metal components (carbon, silicon, phosphorus, manganese, etc.), hot metal temperature, hot metal ratio, and the like. In addition, the auxiliary material data includes the components of the auxiliary material 131, the input amount for each charge, and the like. As described above, in the present embodiment, since the predicted value of the phosphorus concentration in the molten steel during the blowing process is calculated before the start of the blowing process, the hot metal data and the auxiliary raw material data are newly predicted to be the phosphorus concentration in the molten steel. Contains the planned value in the charge for which the value is calculated. The predicted actual data 332 includes predicted values and actual values (first predicted actual data) of phosphorus concentration in molten steel during past blowing treatment in the converter 11, and predicted values and actual values (second values) of carbon concentration in molten steel. Predicted actual data) and the predicted value and actual value of molten steel temperature (third predicted actual data). The target data 333 includes target values such as the component concentration and temperature in the hot metal 111 (or molten steel 112) at the time of the intermediate sublance measurement or at the time of blowing stop, for example. The parameter 334 includes a parameter of a model formula for calculating a predicted value of the phosphorus concentration in molten steel described later.

In the calculation unit 32, the molten steel in-concentration phosphorus concentration prediction unit 321 which is the predicted value calculation unit in the present embodiment, based on the molten pig iron/auxiliary raw material data 331 and the predicted performance data 332 read from the storage unit 33, Calculate the predicted value of phosphorus concentration in molten steel. Specifically, the molten steel phosphorus concentration predicting unit 321 adds a correction term β _p (first correction term), which will be described later, to a function that associates the hot metal/auxiliary raw material data 331 with the molten steel phosphorus concentration during the blowing process. The predicted value of phosphorus concentration in molten steel is calculated by. As will be described later, the correction term calculation unit 322 constructs a multivariate state space model expressing the correction term β _p used for calculating the predicted value of the phosphorus concentration in the molten steel, and a Kalman filter is applied to this state space model. The correction term β _p is calculated by applying. In addition, in the computing unit 32, the medium-melting-material-amount correcting unit 323, which is the input amount correcting unit in the present embodiment, uses the auxiliary raw material 131 based on the predicted value of the phosphorus concentration in molten steel calculated by the phosphorus concentration in molten steel prediction unit 321. The amount of CaO-containing auxiliary material contained in the above, specifically, the amount of the medium-melting material such as quick lime is corrected from the planned value included in the hot metal/auxiliary material data 331 to a more appropriate value.

(Outline of process)
FIG. 2 is a flowchart schematically showing steps of the converter blowing control method according to the embodiment of the present invention. The illustrated process is executed before the start of the blowing process in the converter 11. In the illustrated example, first, the molten steel phosphorus concentration predicting unit 321 of the converter blowing control device 30 reads the hot metal/auxiliary raw material data 331 and the predicted performance data 332 from the storage unit 33 (steps S11 and S12). Here, the correction term calculation unit 322 calculates the correction term β _p for predicting the phosphorus concentration in the molten steel based on the read predicted performance data 332 (step S13). Next, the molten steel phosphorus concentration predicting unit 321 calculates a predicted value of the molten steel phosphorus concentration using the calculated correction term β _p (step S14). Next, the medium-melting material amount correction unit 323 determines the medium-melting material based on the target value of the phosphorus concentration in molten steel contained in the target data 333 and the predicted value of the phosphorus concentration in molten steel calculated by the molten steel phosphorus concentration predicting unit 321. The input amount of is corrected (step S15). The control signal including the corrected charging amount is transmitted to the auxiliary material charging control device 23 via the communication unit 31. After that, in the initial stage of the blowing process, the auxiliary material injection control device 23 injects a medium-soluble material such as quick lime as the auxiliary material 131 according to the corrected input amount.

FIG. 3 is an example of a time chart when the processing shown in FIG. 2 is executed. In the illustrated example, the predicted value of the phosphorus concentration in the molten steel at the time of the intermediate sublance measurement at the n-th charge (n=1, 2,...) Is calculated, and the phosphorus concentration in the molten steel is actually measured at the intermediate sublance measurement. The value is retrieved. For the carbon concentration in molten steel and the molten steel temperature, the predicted value is calculated and the measured value is acquired in the same manner. When the intermediate sublance measurement (step S21) is performed in the nth charge, the measured values such as the phosphorus concentration in the molten steel are acquired, and the series of steps shown in FIG. 2 is performed for the n+1th charge. It will be possible. Specifically, the correction term calculation unit 322 calculates the correction term β _p (step S22), and the molten steel in-mol phosphorus concentration prediction unit 321 uses the calculated correction term β _p for the intermediate sublance measurement in the (n+1)th charge. A predicted value of the phosphorus concentration in the molten steel at that time is calculated (step S23).

Next, the medium-melting-material-amount correcting unit 323 corrects the amount of the medium-melting material charged (step S24), so that the appropriate amount of the medium-melting material in the (n+1)th charge is set. The series of steps may be executed so that the above step S24 is completed before the introduction of the solvent solution in the (n+1)th charge. Therefore, in the illustrated example, the processes of step S22 and subsequent steps are started immediately after the intermediate sublance measurement (step S21) in the nth charge, but in these processes, the blowing process of the nth charge is completed. May be started from. Alternatively, the processes up to step S24 may be finished before the nth charge blowing process is finished.

Hereinafter, the model formula and the correction term used to calculate the predicted value of the phosphorus concentration in the molten steel in this embodiment, and the optimization of the amount of the medium-melting material input based on the predicted value will be described more specifically.

(Model formula)
In the present embodiment, the temporal change of the phosphorus concentration in molten steel (hereinafter, also referred to as [P](%)) is represented by the primary reaction equation of the following equation (1). Note that [P] _ini (%) represents the initial value of [P] (phosphorus concentration in the hot metal), and k (sec ⁻¹ ) represents the dephosphorization rate constant.

From Expression (1), [P] after t seconds from the start of the blowing process is expressed by the following Expression (2). However, the dephosphorization rate constant k is assumed to be represented by a multiple regression equation with the operating factor X _j included in the hot metal/sub-material data 331 as an explanatory variable, as shown in equation (3). Note that α _j represents a regression coefficient.

The above equation (2) is an example of a function that associates the hot metal/auxiliary raw material data 331 with the phosphorus concentration [P] in the molten steel during the blowing process. In the present embodiment, the accuracy of the predicted value of the phosphorus concentration in molten steel [P] is improved by adding the correction term (learning term) β _p to this function as shown in the following equation (4). The correction term β _p is learned based on the predicted value and the actual value (first predicted actual data) of the phosphorus concentration in molten steel [P].

Here, as a formula widely known in the refining field of molten steel and the like, there is a Healy phosphorus distribution formula represented by formula (5) below. In the formula (5), Lp is the distribution ratio, (%P) is the phosphorus concentration in the slag, [P] is the phosphorus concentration in the molten steel, T is the temperature, (%T.Fe), (%CaO) is in the slag. T. Fe, CaO concentrations, a, b, c, d are constants.

The inventors of the present invention focused on the carbon concentration in molten steel (hereinafter also referred to as [C] (%)), which has no direct causal relationship based on the phosphorus distribution formula of formula (5). Since the carbon concentration [C] in the molten steel is considered to correlate with (%T.Fe), (%CaO), and the temperature T, the phosphorus concentration [P] in the molten steel is indirectly obtained from the above formula (5). Can be said to have a relationship with. Therefore, in the present embodiment, in addition to the correction term β _p (first correction term) in the model formula used to calculate the predicted value of the phosphorus concentration in molten steel, in the model formula for carbon concentration in molten steel [C] described later. By constructing a multivariate state space model expressing the correction term β _C (second correction term) and the correction term β _T (third correction term) in the model formula of the molten steel temperature T, further phosphorus in the molten steel can be obtained. The accuracy of the predicted value of the density [P] is improved.

Based on the above examination, the carbon concentration in molten steel [C] and the molten steel temperature T are expressed by multiple regression equations such as equations (6) and (7). Equations (6) and (7) are functions that relate the hot metal/auxiliary material data 331 to the carbon concentration in molten steel [C] and the molten steel temperature T during the blowing process, respectively, and the phosphorus concentration in molten steel [P] described above. Similarly to the example of 1, the correction terms (learning terms) β _C and β _T are added to improve the accuracy of the predicted value. The correction term β _C is learned based on the predicted value and the actual value (second predicted actual data) of the molten steel carbon concentration [C], and the correction term β _T is the predicted value and the actual value of the molten steel temperature T (the third predicted value). It is learned based on the predicted performance data).

When considering the continuity of charge as shown in FIG. 3 above, the correction terms β _p , β _C , and β _T in the equations (4), (6), and (7) are a kind of time series data. Can be regarded as Therefore, in the present embodiment, the correction terms β _p , β _C , and β _T are expressed by a state space model, which is one of modeling methods for time series data. State-space models can be continuous or discrete, periodic or not, univariate or multivariate, stationary or non-stationary, It is a framework of a wide range of statistical models that can be applied to various time series data, and has the feature that the increase or decrease in time series data can be decomposed into factors such as trends, seasonal fluctuations, and regression fluctuations. The correction terms β _p , β _C , and β _T used in this embodiment are considered to be multivariate and non-stationary, but the state space model can be applied as described above. On the other hand, other general auto-regressive models that are modeling methods of time series data are based on the assumption that the data to be analyzed is stationary, and thus the correction terms β _p and β that are considered to be non-stationary. It is not easy to apply it to _C and β _T (it becomes necessary to make steady by variable conversion and difference processing). In addition, it is difficult for the autoregressive model to decompose the increase and decrease in time series data.

The state space model uses two equations, a state equation and an observation equation. The state equation expresses an unmeasured quantity (state quantity), and the idea is that an observation value is obtained by an observation equation in which an observation error is added to the state quantity. When both the state equation and the observation equation are linear and it can be assumed that the observation error has a normal distribution, an algorithm called a Kalman filter that corrects the state quantity using time series data of the observation value has been established. In the present embodiment, the correction terms β _p , β _C , and β _T are used as state quantities, and the predicted values of the phosphorus concentration in molten steel [P], the molten steel temperature, and the carbon concentration in molten steel [C] during the past blowing process are used. The Kalman filter is applied with the difference from the actual value as the observed value (value obtained by adding the measurement error to the state quantity) to accurately predict the correction term β _p .

Although only the correction term β _p is used to predict the final phosphorus concentration [P] in the molten steel, the molten steel temperature T and the carbon concentration [C] in the molten steel are used to predict the phosphorus concentration [P] in the molten steel as described above. ], the correction term for the molten steel temperature T and the carbon concentration in the molten steel [C] is corrected by using a multivariate and single state space model including the correction terms β _C and β _T. by expressing beta _p, it improves the prediction accuracy of the correction term beta _p, is improved prediction accuracy of thus phosphorus concentration in the molten steel [P].

As described above, the state space model has a feature that the increase/decrease of time series data can be decomposed into elements. More specifically, in the formulation of the state space model, the concept of state quantities is used, and the observation value is the state quantity plus noise.Therefore, it is necessary to distinguish between the state quantity and noise included in the prediction error. Will be possible. Therefore, in the present embodiment, by accurately predicting the correction term β _p represented by the state space model using the Kalman filter, the prediction error corresponding to the correction term β _p is regarded as an element due to an essential process variation. It can be decomposed into noise elements. For elements that are caused by essential process variations, the relationship between the prediction error and the operation factors can be clarified by associating the variations of the elements with the specific changes of the operation factors.

(Outline of Kalman filter)
The Kalman filter is a method for sequentially estimating the state quantity inside the model from the observed value when the dynamics of the target process follows a linear state space model. In the present embodiment, it is assumed that the state space model of the correction term β _p in equation (4) is linear, so the Kalman filter can be applied. The Kalman filter targets a linear Gaussian state space model represented by the following equation (8). In addition, x _t is a state quantity vector, y _t is an observation value vector, F _t is a time-varying n×m matrix, G _t is a time-varying n×1 matrix, H _t is a time-varying n×m matrix, and R is R. ⁿ represents an n-dimensional vector space.

In the above equation (8), v _t is called system noise and w _t is called observation noise. In the present embodiment, it is assumed that v _t and x _t follow a multidimensional normal distribution such as the following Expression (9). Incidentally, N (0, _{Q t)} represents the average 0, multidimensional normal distribution of variance-covariance matrix _{Q t, N (0, R} t) is zero mean, multidimensional normal distribution of variance-covariance matrix _{R t.} Hereinafter, the variance-covariance matrix of system noise Q _t, also called variance-covariance matrix of the observation noise R _t.

In the Kalman filter algorithm, the initial value x _0|0 of the estimated value of the state quantity vector and the initial value V _0|0 of the error covariance matrix of the estimated value of the state quantity vector are given in the above state space model. Above, the procedure of prediction and filtering as described below is sequentially repeated.

First, in the procedure of prediction, as shown in the following equation (10), the error variance of the estimated value x _{t−1 |t−1} of the state quantity vector and the estimated value of the state quantity vector at time (t−1) Using the covariance matrix Vt _-1|t-1 , the respective predicted values _xt|t-1 and Vt _|t-1 at time _t are calculated.

Next, in the filtering procedure, the correction value V _t|t and the Kalman gain K _t of the error covariance matrix of the estimated value of the state quantity vector at time _t are calculated as shown in the following equation (11). ..

Furthermore, using the Kalman gain K _t calculated by the above equation (11) and the observed value vector y _t at the time t, the predicted value of the state quantity vector at the time t calculated by the above equation (10) x _{t | t-1} of the correction value _{x t |} a _t, can be calculated as shown in the following equation (12).

The above equations (8) to (12) are examples of equations used in state estimation using the Kalman filter. The Kalman filter is already widely used as a state estimation method, and various specific mathematical formulas are known, not limited to the above examples. Of course, these other examples can also be applied in the present embodiment.

(When randomly walking the state quantity)
As described above, in the present embodiment, the correction terms β _p , β _C , and β _T in the equations (4), (6), and (7) for calculating the predicted value of the phosphorus concentration in the molten steel are set in the state space. The correction terms β _p , β _C , and β _T are calculated by applying a Kalman filter to this state space model by expressing the model.

Here, in equation (8), consider the setting as in equation (13) below.

This setting means that the state quantity and the observed value are multivariate, and the state transition of the state quantity is based on a random walk. In this setting, the equation (8) becomes the state space model represented by the equations (14) and (15). Note that x _t =(x _1,t , x _2,t , x _3,t ) is a state quantity representing the true value of the correction terms β _p , β _C , β _T , and y _t =(y _{1, t 1} , y _2,t , y _3,t ) are observed values corresponding to the correction terms β _p , β _C , and β _T , respectively.

In the state space model expressed by the above equations (14) and (15), the state quantity x _t (correct values of the correction terms β _p , β _C , and β _T ) is randomly walked (state quantity at the next time point). _xt is a motion randomly determined at random), and represents a situation in which the current state quantity _xt is very similar to the previous state quantity _xt-1 . Note that Q and R in the equations (14) and (15) are the prediction error actual value data of the phosphorus concentration [P] in the molten steel obtained in advance, the prediction error actual value data of the molten steel temperature T, and the molten steel medium It is a premise that the value is calculated by the maximum likelihood method or the like using the prediction error actual value data of the carbon concentration [C]. In a multivariate state space model such as equations (14) and (15), v _i,t and w _i,t are the variance-covariance matrix (for example, r ₁₁ is x ₁ (the true value of the correction term β _p . : R ₁₂ is a covariance of x ₁ (correction term β _p true value: state quantity) and x ₂ (correction term β _C true value: state quantity). It is said to follow a normal distribution with variances of Q and R, and the correlation between each variable is incorporated through this relationship. By applying the Kalman filter to the state space model expressed by the equations (14) and (15), the state quantity x _t (correction terms β _p , β) from the observed value y _t-1 in the previous charge to the current charge It becomes possible to calculate the true value of _C , β _T ) and the observed value y _t .

(Effect of state space model)
4 to 7 are graphs for explaining the effect of using the state space model in the embodiment of the present invention. In each graph, the predicted value and the actual measured value of the phosphorus concentration in molten steel [P] are the values at the time of the intermediate sublance measurement, and are therefore described as SL[P]cal and SL[P]act. The values of SL[P]cal and SL[P]act are both normalized. In FIG. 4, as Case 0, the predicted value SL[P]cal and the actually measured value when [P] is predicted using the above equation (2) that does not include any of the correction terms β _p , β _C , and β _T The relationship with the value SL[P]act is shown. In FIG. 5, as case 1, the above equation (4) including the correction term β _p of the phosphorus concentration in molten steel [P] and the equations (8) to (15) are used, and the equation (14) and the equation ( In (15), the prediction value SL[P]cal and the actual measurement value SL[P]act when [P] is predicted by introducing only the correction term β _p (ignoring the correction terms β _C and β _T ) Relationships are shown. In FIG. 6, as the case 2, the above equations (4), (6), and (8) to (15) including the correction term β _C in addition to the correction term β _p are used, and the equation (14 ) And Equation (15), the relationship between the predicted value SL[P]cal and the actual measured value SL[P]act when the correction terms β _p and β _C (correction term β _T is ignored) [P] is predicted. It is shown. In FIG. 7, as Case 3, the above equations (4), (6), (7), and (8) to (15) including all the correction terms β _p , β _C , and β _T are shown. The relationship between the predicted value SL[P]cal and the actual measured value SL[P]act when [P] is predicted using the graph is shown.

In case 0 (FIG. 4), the standard deviation of the predicted value SL[P]cal is 0.858, and the average error with respect to the measured value SL[P]act is 0.101. On the other hand, in case 1 (FIG. 5), the standard deviation is 0.753 and the error average is -0.0171. In case 1, the standard deviation (variation of predicted values) is improved, and in addition, the tendency that the predicted value SL[P]cal in case 0 is higher than the measured value SL[P]act as a whole is improved. There is. In case 2 (FIG. 6), the standard deviation is 0.639, the error average is −0.0172, and the center of the distribution of the predicted value SL[P]cal is maintained close to the measured value SL[P]act. Meanwhile, the standard deviation is improved. Furthermore, in case 3 (FIG. 7 ), the standard deviation is 0.426 and the error average is 0.0236, and the standard deviation is greatly improved compared to cases 0 to 2 while the predicted value SL[P]cal The center of the distribution can be maintained close to the measured value SL[P]act.

(Adjusting the amount of solvent input)
By using the predicted value of the phosphorus concentration in molten steel [P] calculated by introducing the correction terms β _p , β _C , and β _T as described above, the CaO-containing auxiliary raw material that is input in the initial stage of the blowing process is used. Specifically, it is possible to optimize the input amount of the medium-soluble material such as quick lime. As described above, in the converter blowing, blowing control combining static control and dynamic control is performed. In the present embodiment, in static control, the input amount of a medium-soluble material such as quick lime determined in advance by using a mathematical model based on a mass balance or a heat balance includes correction terms β _p , β _C , and β _T [ Correct using the predicted value of P].

FIG. 8 is a graph for explaining the optimization of the amount of the solvent input in this embodiment. As shown by the solid line in FIG. 8, the amount of the medium-melting material input (kg/ton) per weight of the hot metal in the static control is determined in advance according to the target phosphorus concentration (%) in the molten steel. Specifically, when the target value of the phosphorus concentration in the molten steel in the current charge is P ₁ (%), the amount of the medium-solvent input determined in advance is W _CaO (kg/ton). On the other hand, when the predicted value of the phosphorus concentration in the molten steel calculated by introducing the correction term β _p is P ₂ (%) (P ₂ <P ₁ ), the medium-fluid material input amount W _CaO is maintained. , ΔP=P ₁ −P ₂ (%) causes excessive dephosphorization. In this case, it is considered that the relationship between the target phosphorus concentration in the molten steel and the amount of the medium-melting material charged per weight of the hot metal shown by the solid line in FIG. 8 is shifted to the broken line graph. Therefore, for example, as shown in the following formula (16), the amount of the solute material is corrected from W _CaO to W′ _CaO by using the solute material correction function f(ΔP). As a result, it is possible to add a necessary and sufficient amount of the medium solution material for dephosphorization, and to minimize the amount of slag generated due to the medium solution material while achieving the target value of the phosphorus concentration in the molten steel. it can.

In one embodiment of the present invention as described above, a correction for calculating the predicted value of the phosphorus concentration in molten steel [P] by introducing a state space model assuming that the state quantity randomly walks and applying a Kalman filter. The learning of the terms β _p , β _C , and β _T can be sequentially executed. Further, by applying the Kalman filter to the state space model in which the regression effect is introduced, it is possible to distinguish the influence of the essential process variation contained in the correction terms β _p , β _C , and β _T from the noise. Therefore, the prediction accuracy of the phosphorus concentration in molten steel [P] before the start of blowing can be improved. If the prediction accuracy of the phosphorus concentration in molten steel [P] is improved, it is possible to optimize the amount of CaO-containing auxiliary raw material that is introduced in the initial stage of the blowing process, while achieving the target value of phosphorus concentration in molten steel. The amount of slag generated due to the solvent material can be suppressed to the minimum.

The preferred embodiments of the present invention have been described above in detail with reference to the accompanying drawings, but the present invention is not limited to these examples. It is obvious that a person having ordinary knowledge in the technical field to which the present invention pertains can come up with various changes or modifications within the scope of the technical idea described in the claims. Of course, it is understood that these also belong to the technical scope of the present invention.

DESCRIPTION OF SYMBOLS 1... Refining equipment, 10... Converter equipment, 11... Converter, 12... Top blowing lance, 13... Input device, 20... Measurement control device, 21... Sublance, 22... Oxygen supply device, 23... Sub raw material input control device , 30... Converter blowing control device, 31... Communication unit, 32... Computing unit, 321... Phosphorus concentration in molten steel predicting unit, 322... Correction term calculating unit, 323... Solvent material amount correcting unit, 33... Storage unit, 331 ... hot metal/auxiliary material data, 332... predicted performance data, 333... target data, 334... parameter, 34... input/output section, 111... hot metal, 112... molten steel, 113... slag, 121... oxygen gas, 131... auxiliary material.

Claims (8)

The function of associating the hot metal data on the hot metal blown in the converter with the phosphorus content in the molten steel at the time of the blowing process in the converter, The predicted value of the phosphorus concentration in the molten steel is calculated by adding the first correction term learned based on the first predicted performance data including the predicted value and the actual value of the phosphorus concentration in the molten steel during the past blowing process. Predictive value calculating means,
Correction term calculation for calculating the first correction term by constructing a multivariate state space model expressing the first correction term and the second correction term and applying a Kalman filter to the state space model Means and
The second correction term is a correction term added to a function relating the molten pig iron data and the auxiliary raw material data to the carbon concentration in molten steel at the time of the blowing treatment in the converter, and the past blowing treatment in the converter. Learned based on the second predicted actual data including the predicted value and the actual value of the carbon concentration in molten steel at the time,
The Kalman filter uses the difference between the predicted value and the actual value of the phosphorus concentration in molten steel included in the first predicted actual data, and the predicted value and the actual value of the carbon concentration in molten steel included in the second predicted actual data. A converter blowing control device that uses the difference between the two as the observed value. The converter blowing control device according to claim 1, wherein the correction term calculation means expresses the first correction term and the second correction term by a single state space model. The first predicted actual data is the predicted value of phosphorus concentration in molten steel at the time of intermediate sublance measurement during the past blowing process in the converter, and the actual value of phosphorus concentration in molten steel acquired by the intermediate sublance measurement. Including
The second predicted performance data is a predicted value of carbon concentration in molten steel at the time of intermediate sublance measurement during the past blowing process in the converter, and an actual value of molten carbon in the intermediate sublance measured. The converter blowing control apparatus according to claim 1 or claim 2, which comprises: The state space model is a multivariate state space model that expresses a third correction term in addition to the first correction term and the second correction term,
The third correction term is a correction term added to a function that associates the molten pig iron data and the auxiliary raw material data with the molten steel temperature during the blowing process in the converter, Learned based on the third predicted actual data including the predicted value and the actual value of the molten steel temperature,
The Kalman filter uses the difference between the predicted value and the actual value of the phosphorus concentration in molten steel included in the first predicted actual data, and the predicted value and the actual value of the carbon concentration in molten steel included in the second predicted actual data. In addition to the difference, the converter of any one of claims 1 to 3, wherein the difference between the predicted value and the actual value of the molten steel temperature included in the third predicted actual data is an observed value. Blowing control device. The converter blowing control device according to any one of claims 1 to 4, wherein the correction term calculation means randomly walks the state quantity in the state space model. The injection amount correcting means for correcting the injection amount of the CaO-containing auxiliary raw material contained in the auxiliary raw material based on the predicted value of the phosphorus concentration in the molten steel, The injection amount correcting means according to claim 1, further comprising: Converter blowing control device. The function of associating the hot metal data on the hot metal blown in the converter with the phosphorus content in the molten steel at the time of the blowing process in the converter, The predicted value of the phosphorus concentration in the molten steel is calculated by adding the first correction term learned based on the first predicted performance data including the predicted value and the actual value of the phosphorus concentration in the molten steel during the past blowing process. A predicted value calculation step to
Correction term calculation for calculating the first correction term by constructing a multivariate state space model expressing the first correction term and the second correction term and applying a Kalman filter to the state space model Including process and
The second correction term is a correction term added to a function relating the molten pig iron data and the auxiliary raw material data to the carbon concentration in molten steel at the time of the blowing treatment in the converter, and the past blowing treatment in the converter. Learned based on the second predicted actual data including the predicted value and the actual value of the carbon concentration in molten steel at the time,
The Kalman filter uses the difference between the predicted value and the actual value of the phosphorus concentration in molten steel included in the first predicted actual data, and the predicted value and the actual value of the carbon concentration in molten steel included in the second predicted actual data. A converter blowing control method in which the difference between the measured values is the observed value. The function of associating the hot metal data on the hot metal blown in the converter with the phosphorus content in the molten steel at the time of the blowing process in the converter, The predicted value of the phosphorus concentration in the molten steel is calculated by adding the first correction term learned based on the first predicted performance data including the predicted value and the actual value of the phosphorus concentration in the molten steel during the past blowing process. Predictive value calculating means,
Correction term calculation for calculating the first correction term by constructing a multivariate state space model expressing the first correction term and the second correction term and applying a Kalman filter to the state space model Means and
The second correction term is a correction term added to a function relating the molten pig iron data and the auxiliary raw material data to the carbon concentration in molten steel at the time of the blowing treatment in the converter, and the past blowing treatment in the converter. Learned based on the second predicted actual data including the predicted value and the actual value of the carbon concentration in molten steel at the time,
The Kalman filter uses the difference between the predicted value and the actual value of the phosphorus concentration in molten steel included in the first predicted actual data, and the predicted value and the actual value of the carbon concentration in molten steel included in the second predicted actual data. A program that causes a computer to function as a converter blowing control device that uses the difference between the two as an observed value.

Images