Control device and control method

Abstract

The goal is to be able to distinguish between fluctuations in observed quantities that can be resolved by control operations and those that cannot be resolved by control operations, and to perform appropriate control operations accordingly. [Solution] The system comprises an operation determination unit that determines the operation and the amount of operation for a process, a disturbance fluctuation detection unit that detects short-term fluctuations in the estimated value of the first disturbance, and an operation execution determination unit that determines whether or not to execute the operation determined by the operation determination unit based on the detection results of the disturbance fluctuation detection unit. If the disturbance fluctuation detection unit detects short-term fluctuations in the estimated value of the first disturbance, the operation execution determination unit excludes a predetermined operation from the operations determined by the operation determination unit from execution.

Description

[Technical Field]

[0001] The present invention relates to a control device and a control method, and more particularly to a control device and a control method that can distinguish between fluctuations in observed quantities that are expected to be resolved by a control operation and fluctuations in observed quantities that are not expected to be resolved by a control operation, and perform an appropriate control operation. [Background technology]

[0002] In recent years, blast furnace processes have favored the use of inexpensive iron ore and operation under low reducing agent ratio conditions. However, using inexpensive, low-quality raw materials and operating under low reducing agent ratio conditions can easily lead to instability in furnace conditions, resulting in fluctuations in production volume and a decrease in furnace temperature. Therefore, the need for stabilizing blast furnace conditions is increasing.

[0003] In particular, various types of disturbances occur, mainly caused by the charging materials such as ore and coke, resulting in fluctuations in blast furnace operation. To achieve stable blast furnace operation, it is necessary to suppress fluctuations in operational indicators caused by these disturbances.

[0004] In other words, blast furnace operation involves various disturbances, including changes in the properties of the charged raw materials, and it is necessary to continue stable operation in the presence of these disturbances. More specifically, it is necessary to maintain stable operation by keeping specific indicators to be managed, such as the amount of iron produced and the furnace heat index, within the control range.

[0005] For example, disturbance estimation observers are known that estimate disturbances superimposed on observed pig iron production (see Non-Patent Literature 1). In addition, model predictive control is used in process control during blast furnace operation (see Non-Patent Literature 2).

[0006] However, various disturbances are superimposed on the measured value of Pmax (pig tapping), a representative indicator, and using this measured value directly to perform control operations may actually exacerbate operational fluctuations. Furthermore, since Pmax exhibits different time variation patterns depending on the cause of the disturbance, it is necessary to identify the type of disturbance according to the time variation pattern and then appropriately select the operations to be performed.

[0007] Patent Document 1 proposes a molten iron temperature control method that can detect future decreases in molten iron temperature that are difficult to predict using physical models and propose actions to increase the molten iron temperature. According to this method, the rate of change of the estimated error of process variables is calculated from the error between the physical model and actual values, and the presence or absence of unburned pulverized coal is determined. Then, the determination of whether or not to perform the action to increase the molten iron temperature is made by combining the unburned pulverized coal generation determination process and the ventilation abnormality determination process.

[0008] Furthermore, Patent Document 2 proposes a method for simultaneously estimating parameters and disturbances in a control model, which allows for simultaneous and more accurate estimation of disturbances and parameters. According to this method, parameters (unknown steady-state values) and disturbances within the control model are simultaneously estimated. Then, by applying a bandstop filter (notch filter) to the time series of the disturbance estimates, the relevant frequency components are cut out and reflected in the control evaluation function. [Prior art documents] [Patent Documents]

[0009] [Patent Document 1] Patent No. 7522999 [Patent Document 2] Japanese Patent Publication No. 2016-181247 [Non-patent literature]

[0010] [Non-Patent Document 1] Ikeda, Fujisaki: Multivariable System Control (Corona Publishing Co.) [Non-Patent Document 2] J. Maciejowski: Model Predictive Control (Tokyo Denki University Press) Summary of the Invention Problem to be Solved by the Invention

[0011] However, for example, when the measured value of Pmax fluctuates in a short cycle, it is difficult to eliminate the fluctuation through control operations when considering time delays (dead times) such as the time delay of control operations, the descent time of the charge in the furnace, and the residence time of molten iron in the lower part of the furnace. In other words, even if an operation is performed to suppress fluctuations in Pmax caused by disturbances that fluctuate in a short cycle, Pmax will fluctuate before the effect of the performed operation appears, and as a result, there is a risk that the fluctuation of Pmax will be exacerbated.

[0012] Additionally, for example, a sharp rise in Pmax will subsequently lead to a decrease in furnace heat, so when a sharp rise in Pmax is detected, operations that have a heat-reducing effect should be refrained from. Similarly, a sharp drop in Pmax will subsequently lead to an increase in furnace heat, so when a sharp drop in Pmax is detected, operations that have a heat-increasing effect should be refrained from. In this way, if the implementation timing of a control operation corresponding to short-term changes such as a sharp rise or sharp drop in the measured Pmax value is incorrect, the operation may反而 exacerbate fluctuations. However, when short-term changes such as a sharp rise or sharp drop occur, it is generally difficult to implement a control operation at an appropriate timing.

[0013] In other words, in order to stably continue operation in the presence of disturbances, it is necessary to properly distinguish between fluctuations in indicators that can be expected to be eliminated by control operations and fluctuations that cannot be expected to be eliminated by control operations, prevent unnecessary operational actions, and appropriately implement fluctuation suppression actions for fluctuations that can be expected to be eliminated by control operations. That is, it is necessary to appropriately switch whether to implement control operations according to the type of indicator fluctuation.

[0014] In contrast, Patent Document 1 only performs simple statistical processing that merely calculates the "time change rate" for the estimated disturbance value, so it cannot determine whether to execute an action in accordance with the periodic fluctuation of the estimated disturbance value. Further, Patent Document 2 merely applies a band-stop filter (notch filter) to disturbance in a control model, cuts the frequency component, and reflects the result in a control evaluation function, and cannot identify the type of change in the estimated disturbance value and execute an appropriate operational action according to the type.

[0015] One aspect of the present invention aims to implement a technology that enables identification of fluctuations in an observed variable that are expected to be resolved by a control operation and fluctuations in an observed variable that are not expected to be resolved by a control operation, thereby allowing execution of an appropriate control operation. [Means for Solving the Problems]

[0016] To solve the above problems, a control device according to one aspect of the present invention is a control device that controls a process, wherein the process includes a first process that outputs a first observable quantity observable with a relatively fast time response and a second process that outputs a second observable quantity observable with a relatively slow time response, and the first observable quantity output in response to an operation is input to a disturbance estimation observer to obtain an estimated first disturbance value superimposed on the first observable quantity, and the second observable quantity output in response to an operation is input to a disturbance estimation observer to obtain an estimated second disturbance value superimposed on the second observable quantity The system includes an operation determination unit that determines an operation and an operation amount for the process based on an estimated value of the first disturbance and an estimated value of the state of the process obtained by inputting the first and second observed quantities to a disturbance estimation observer; a disturbance fluctuation detection unit that detects short-term fluctuations in the estimated value of the first disturbance; and an operation execution determination unit that determines whether or not to execute the operation determined by the operation determination unit based on the detection result of the disturbance fluctuation detection unit. If the disturbance fluctuation detection unit detects a short-term fluctuation in the estimated value of the first disturbance, the operation execution determination unit excludes a predetermined operation from the operations determined by the operation determination unit.

[0017] To solve the above problems, a control method according to one aspect of the present invention is a control method for which a process is the target of control, wherein the process includes a first process that outputs a first observable quantity observable with a relatively fast time response and a second process that outputs a second observable quantity observable with a relatively slow time response, and the first observable quantity output in response to an operation is input to a disturbance estimation observer to obtain an estimated first disturbance value which is superimposed on the first observable quantity, and the second observable quantity output in response to an operation is input to a disturbance estimation observer to obtain an estimated second disturbance value which is superimposed on the second observable quantity, and the first The operation includes an operation determination step that determines an operation and an operation amount for the process based on an estimated value of the state of the process obtained by inputting the observed quantity and the second observed quantity into a disturbance estimation observer; a disturbance fluctuation detection step that detects short-term fluctuations in the estimated value of the first disturbance; and an operation execution determination step that determines whether or not to execute the operation determined in the operation determination step based on the detection result of the disturbance fluctuation detection step. If the disturbance fluctuation detection step detects a short-term fluctuation in the estimated value of the first disturbance, the operation execution determination step excludes a predetermined operation from the operations determined in the operation determination step from execution. [Effects of the Invention]

[0018] According to one aspect of the present invention, it is possible to distinguish between fluctuations in observed quantities that are expected to be resolved by control operations and fluctuations in observed quantities that are not expected to be resolved by control operations, and to perform appropriate control operations. [Brief explanation of the drawing]

[0019] [Figure 1] This diagram illustrates the overview of the blast furnace process. [Figure 2] This figure shows the relationship between the blast furnace process and the manipulated and observed quantities. [Figure 3] This diagram illustrates the overview of operational inputs in blast furnace operation. [Figure 4]This block diagram shows an example configuration of a control device that controls the blast furnace process. [Figure 5] This figure shows an example of a frequency spectrum obtained by FFT processing. [Figure 6] This figure illustrates an example of how manipulated variables in a blast furnace process change over time. [Figure 7] This figure illustrates the change in Pmax over time observed when the operation shown in Figure 6 is performed. [Figure 8] This is a flowchart illustrating an example of a control process. [Figure 9] This flowchart illustrates an example of the disturbance detection process. [Figure 10] This flowchart illustrates an example of the process flow for determining whether an operation should be executed. [Figure 11] This is a flowchart illustrating an example of the process for periodic gender correspondence processing. [Figure 12] This graph shows the change in Pmax, which is obtained as an observable quantity corresponding to process P1. [Figure 13] This graph shows the change in molten iron temperature, which is obtained as an observed quantity corresponding to process P2. [Figure 14] This graph shows the changes in the input parameters for the blast furnace process. [Figure 15] This figure compares the accuracy of disturbance estimation when a long-period disturbance is superimposed on Pmax. [Figure 16] This is a block diagram illustrating the physical configuration of a computer used as a control device. [Modes for carrying out the invention]

[0020] Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings. First, an overview of the blast furnace process will be given.

[0021] Figure 1 is a diagram illustrating the blast furnace process. In the blast furnace process, as shown in Figure 1, sintered ore and coke are charged alternately into the blast furnace 101 from the top of the furnace, and hot air and reducing agents such as pulverized coal are blown in from the tuyeres at the bottom of the furnace. The pulverized coal and coke are gasified by this hot air, and high-temperature reducing gases such as carbon monoxide and hydrogen are blown up into the furnace, melting the sintered ore and removing oxygen. The molten iron comes into contact with the carbon in the coke and is reduced, becoming molten iron containing slightly less than 5% carbon, which accumulates in the molten metal reservoir at the bottom of the furnace.

[0022] This molten iron is removed from a tap located on the side of the furnace bottom and transported to the next steelmaking process. Additionally, gases such as coke oven gas (COG) or natural gas (NG) may be blown in through the tuyeres (or other tuyeres).

[0023] Furthermore, in blast furnace operation, control indicators are set, and the time progression of these indicators is monitored. By performing operational operations on the blast furnace process with corresponding control units, the furnace conditions are controlled.

[0024] Figure 2 shows the relationship between the blast furnace process and the manipulated variable u and the observed variable y. As shown in Figure 2, the blast furnace process takes the manipulated variable u as input and outputs the observed variable y. The state of the blast furnace process is denoted as x.

[0025] The manipulated quantity u is, for example, a value related to the airflow rate, oxygen enrichment rate, and pulverized coal amount (pulverized coal injection amount PCI or pulverized coal ratio PCR), such as the airflow conditions at the tuyeres.

[0026] Observed quantities y include molten iron temperature, gas composition (e.g., CO concentration, CO2 concentration, N2 concentration, H2 concentration), gas utilization rate (e.g., CO utilization rate, H2 utilization rate), carbon solution loss (CSL), CO utilization rate, H2 utilization rate, iron tapping volume, molten iron Si, and molten iron temperature. Of these observed quantities y, molten iron temperature, molten iron Si, and iron tapping volume are often considered management indicators for which target values ​​have been set.

[0027] In recent years, blast furnace processes have seen a shift towards using inexpensive iron ore and operating under low reducing agent ratio conditions, leading to a growing need to stabilize blast furnace conditions.

[0028] However, operating with inexpensive, low-quality raw materials or low-reducing agents can easily lead to unstable furnace conditions, causing fluctuations in production volume and a decrease in furnace heat. For example, it is known that various types of disturbances can occur due to charging materials such as iron ore and coke, causing fluctuations in blast furnace operation. Therefore, in order to achieve stable blast furnace operation, it is necessary to operate while taking these disturbances into consideration.

[0029] Furthermore, in blast furnace operation, various operations are performed on the blast furnace, such as adding materials and changing conditions. In order to optimize blast furnace operation, control is carried out by observing the changes resulting from these operations as observed quantities and using these observed quantities as feedback to adjust the quantities of operations to be performed next (also called control operations).

[0030] (Control of input and process overview) Figure 3 illustrates the control of operational inputs and the process overview in blast furnace operation. As shown in Figure 3, in the control of operational inputs in blast furnace operation, process control performs disturbance estimation and state estimation using a disturbance estimation observer, and calculates the estimated disturbance d1^, estimated disturbance ζ1^, estimated disturbance d2^, and estimated state of the blast furnace process x^. Here, x^, d1^, ζ1^, and d2^ are notated with a caret (^) above x and d, respectively. Letters with carets will be notated similarly from now on.

[0031] Based on the calculated disturbance estimates d1^, d2^, and x^, the blast furnace process state estimate is used to determine the manipulated variable u through model predictive control, and an operational input is applied to the blast furnace process. The manipulated variable u may include at least one of the following: airflow rate, oxygen enrichment rate, or pulverized coal injection rate.

[0032] The operation corresponding to the manipulated variable u affects the blast furnace process. The blast furnace process includes a fast-responding process P1 and a slow-responding process P2 to operations and disturbances. Process P1 is a process in which the time taken to obtain a response to operations and disturbances is relatively short, while process P2 is a process in which the time taken to obtain a response to operations and disturbances is relatively long.

[0033] Process P1 alters factors such as the amount of iron tapped, gas composition, carbon solution loss (CSL), CO utilization rate, H2 utilization rate, and gas utilization rate. The amount of iron tapped is also referred to as Pmax. Process P2 alters factors such as molten iron Si and molten iron temperature. Process P1 corresponds to chemical reaction processes such as the reduction reaction of ore and the consumption reaction of coke, while process P2 roughly corresponds to the heat transfer and heat transfer phenomena of the reaction heat generated in process P1.

[0034] Observable quantities related to indicators that change due to process P1 (e.g., iron tapping rate, gas utilization rate, CSL) can be observed relatively soon after the operation or disturbance occurs, and such observable quantities can be said to be observable with a relatively fast time response to the operation or disturbance. On the other hand, observable quantities related to indicators that change due to process P2 (e.g., molten iron temperature) can be observed relatively soon after the operation or disturbance occurs, and such observable quantities can be said to be observable with a relatively slow time response to the operation or disturbance.

[0035] The values ​​of indicators such as Pmax, gas utilization rate, CSL, and molten iron temperature, which change due to processes P1 and P2, are fed back into process control as observed quantities. That is, based on the observed quantities, disturbance and state estimation is performed using a disturbance estimation observer, and estimated values ​​of disturbances and blast furnace states are recalculated.

[0036] The disturbance estimates d1^ and ζ1^ are calculated by inputting the observed variable corresponding to process P1 (e.g., the value of Pmax) into the disturbance estimation observer. The disturbance estimate d2^ is calculated by inputting the observed variable corresponding to process P2 (e.g., the value of molten iron temperature) into the disturbance estimation observer. The blast furnace process state estimate x^ is calculated by merging the state estimate x1^, which is calculated by inputting the observed variable corresponding to process P1 (e.g., the value of Pmax) into the disturbance estimation observer, with the state estimate x2^, which is calculated by inputting the observed variable corresponding to process P2 (e.g., the value of molten iron temperature) into the disturbance estimation observer.

[0037] Then, based on the calculated disturbance estimates d1^, ζ1^, d2^, and blast furnace process state estimate x^, a new manipulated variable u is determined by model predictive control, and the operation input is performed. In this way, the changes resulting from the operation are observed as observed variables, and the manipulated variable for the next operation is adjusted by feeding back these observed variables.

[0038] In this way, the blast furnace process is controlled so that a predetermined indicator is kept close to a preset target value. For example, the blast furnace process is controlled so that the value of a predetermined indicator is always between a preset upper limit and lower limit of the control range.

[0039] However, if, for example, the measured value of Pmax fluctuates with a short period, it is difficult to eliminate the fluctuations by control operations, considering the time delays (dead time) of control operations, the time it takes for the charge to descend into the furnace, and the residence time of the molten iron at the bottom of the furnace. In other words, even if operations are performed to suppress fluctuations in Pmax caused by disturbances that fluctuate with a short period, Pmax may fluctuate again before the effects of the operations become apparent, potentially exacerbating the fluctuations in Pmax.

[0040] Furthermore, for example, a sharp rise in Pmax can lead to a subsequent decrease in furnace temperature; therefore, when a sharp rise in Pmax is detected, operations that reduce heat should be avoided. Similarly, a sharp drop in Pmax can lead to a subsequent increase in furnace temperature; therefore, when a sharp drop in Pmax is detected, operations that increase heat should be avoided. In this way, control operations that respond to short-term changes such as sharp rises and falls in measured Pmax values ​​can actually exacerbate fluctuations.

[0041] In this embodiment, as shown in Figure 3, the model predictive control uses estimated values ​​d1^, d2^, and x^, in addition to estimated value d 1notch ^, and the estimated value ζ1^ are provided. Here, the estimated value d 1notch ^ is obtained by applying a notch filter to the estimated value d1^. The estimated value ζ1 is obtained by inputting the observed value of Pmax into a disturbance estimation observer, which is configured with a disturbance model different from the one used to calculate the estimated value d1^.

[0042] (Control to respond to short-term fluctuations in disturbances) If a short-term fluctuation in the estimate of the first disturbance is detected, a predetermined operation from the determined operations will be excluded from execution.

[0043] In model predictive control, the continuously input estimated values ​​d1^ are stored as time-series data, and it is determined whether the estimated value d1^ has changed significantly based on the most recently input estimated value d1^. If it is determined that the estimated value d1^ has changed significantly, it means that a short-term fluctuation in the estimated value d1^ has been detected. As will be explained in detail later, model predictive control detects short-term fluctuations in the estimated value d1^ by calculating the standard deviation and regression line of the time-series data, and excludes (skips) some or all of the operations determined according to the statistical properties of the detected short-term fluctuation. For example, by setting the value of the change from the previous value of the operation variable for a given operation to 0, that operation is skipped.

[0044] (Control to respond to periodic fluctuations of disturbances) If no short-term fluctuations are detected in the estimate of the first disturbance, the operation and manipulated quantities are determined based on the estimate of the second disturbance, the estimate of the process state, and one of the estimates of the first disturbance, the third disturbance, and the fourth disturbance, and the estimate of the third disturbance (d 1notch ^) is obtained by filtering the fluctuations of the estimated value of the first disturbance to remove periodic fluctuations corresponding to a predetermined frequency, and the estimated value of the fourth disturbance (ζ1^) is obtained by inputting the first observation into a disturbance estimation observer that is set to a different disturbance model than the disturbance model set in the disturbance estimation observer when estimating the estimated value of the first disturbance.

[0045] (Control to respond to short-periodic fluctuations of disturbances) Furthermore, in model predictive control, the continuously input estimated value d1^ is stored as time-series data, periodic fluctuations in the estimated value d1^ are detected, and if the detected fluctuation is due to a short period, the estimated value d is filtered using a notch filter to remove the periodic fluctuations in the estimated value d1^. 1notch The manipulated variables are determined using ^. That is, if the estimated value d1^ changes periodically with a short period, the estimated disturbance d2^, the estimated state of the blast furnace process x^, and the filtered estimated disturbance d 1notch The control variable u is determined based on ^.

[0046] (Control to respond to long-term periodic fluctuations of disturbances) Furthermore, in model predictive control, the continuously input estimated value d1^ is stored as time-series data, periodic fluctuations in the estimated value d1^ are detected, and if the detected fluctuation is due to a long period, the manipulated variable is determined using the estimated disturbance ζ1^ obtained by inputting the observed value of Pmax to a disturbance estimation observer with a different disturbance model set up. In other words, if the estimated value d1^ changes periodically over a long period, the manipulated variable u is determined based on the estimated disturbance d2^, the estimated state of the blast furnace process x^, and the estimated disturbance ζ1^.

[0047] If the periodic fluctuations in the estimated disturbance are due to long periods, it is considered possible to actively eliminate the effects of the disturbance through control operations. However, the disturbance model set in the disturbance estimation observer when estimating the estimated value d1^ is a constant-value disturbance model, and therefore is not suitable as a disturbance model to represent periodic fluctuations.

[0048] Therefore, in this embodiment, a more suitable disturbance model is adopted when estimating periodically fluctuating disturbances. That is, in this embodiment, a constant-value disturbance model and a disturbance model capable of representing periodic fluctuations are combined to estimate periodically fluctuating disturbances.

[0049] (A combination of a constant-value disturbance model and a disturbance model capable of representing periodic fluctuations) While a constant-value disturbance model is typically used as the disturbance model set in the disturbance estimation observer, in this embodiment, a disturbance model combining a constant-value disturbance model and a disturbance model capable of representing periodic fluctuations is set in the disturbance estimation observer.

[0050] For example, suppose the estimated value d1^ fluctuates periodically over a long period, and the peak frequency of the fluctuation is frequency f (=1 / T, where T is the period). In this case, a disturbance model capable of representing the periodic fluctuation is generated using the angular frequency ω (=2πf). This disturbance model is expressed by equation (1).

number

number

number

[0051] The above continuous-time differential equations are discretized in time with a sampling period Δt. For example, in the case of time discretization using zero-order hold, the disturbance model is obtained by the discrete-time models of equations (4) and (5).

number

number

number

[0052] (Control device configuration) Figure 4 is a block diagram showing an example configuration of a control device for controlling a blast furnace process. The control device 300 shown in Figure 4 is a device that controls the blast furnace process by observing operational indicators. As shown in Figure 4, the control device 300 has a disturbance fluctuation detection unit 310, an operation decision unit 330, and an operation execution decision unit 340. The control device 300 may further include an estimation unit that outputs estimated values ​​of disturbances and states through calculation processing related to a disturbance observer, and a filter processing unit that performs calculation processing related to a notch filter.

[0053] (Disturbance detection unit) The disturbance detection unit 310 detects short-term fluctuations in the estimated value of the first disturbance. The disturbance detection unit 310 periodically acquires the estimated value of the disturbance d1^ from an estimation unit (not shown). The disturbance detection unit 310 generates time-series data of the continuously acquired estimated value of the disturbance d1^ and detects short-term fluctuations in the estimated value of the disturbance d1^ based on the most recently acquired estimated value of the disturbance d1^.

[0054] The disturbance detection unit 310 includes a standard deviation calculation unit 311, a regression line calculation unit 312, an FFT processing unit 313, and a frequency setting unit 314.

[0055] (Standard deviation calculation part) The standard deviation calculation unit 311 calculates the standard deviation of the time series data of the disturbance estimate d1^. For example, the time series data is considered to be a series of estimates d1^ for one hour. The standard deviation calculation unit 311 further determines whether the calculated standard deviation is greater than a preset threshold. The determination result is supplied to the operation execution determination unit 340.

[0056] If the calculated standard deviation is determined to be greater than a predetermined threshold, it indicates that a short-term fluctuation in the estimate of the first disturbance has been detected, and the above determination result shows the statistical properties of the short-term fluctuation.

[0057] (Regression Line Calculation Unit) The regression line calculation unit 312 calculates a regression line relating to the time series data of the disturbance estimate d1^. The time series data used here may be longer in duration than the time series data processed by the standard deviation calculation unit 311. For example, two consecutive estimates of d1^ over two hours may be used as the time series data.

[0058] The regression line may be calculated using, for example, the least squares method. The regression line calculation unit 312 compares the current value of the disturbance estimate d1^ obtained from the estimation unit (not shown) with the current value (theoretical value) of the disturbance estimate d1^ shown by the calculated regression line.

[0059] The regression line calculation unit 312 calculates the difference between the current value and the theoretical value, compares this difference with the upper and lower thresholds, and determines whether the difference exceeds the upper threshold or falls below the lower threshold. The determination result is supplied to the operation execution determination unit 340.

[0060] If the difference exceeds the upper threshold, or if the difference falls below the lower threshold, a short-term fluctuation in the estimate of the first disturbance is detected, and the above determination result indicates the statistical properties of the short-term fluctuation.

[0061] In this manner, the disturbance detection unit 310 calculates the standard deviation and regression line for the time series data of the estimated value of the first disturbance, detects short-term fluctuations of the estimated value of the first disturbance based on the standard deviation and regression line, and identifies the statistical properties of the detected short-term fluctuations.

[0062] (FFT processing unit) The FFT processing unit 313 performs an FFT operation on the time series data of the disturbance estimate d1^. The FFT processing unit 313 determines whether the power of frequencies within a predetermined range in the frequency spectrum obtained by the FFT operation is greater than a threshold. The frequencies within the predetermined range may be, for example, relatively high frequencies.

[0063] Figure 5 shows an example of a frequency spectrum obtained by FFT processing. In this figure, the horizontal axis represents frequency or frequency (times / hour), and the vertical axis represents power, showing the frequency spectrum of the time-series data of the estimated disturbance d1^.

[0064] The FFT processing unit 313 compares the power corresponding to relatively high frequencies (relatively short periods) with a threshold. For example, in the frequency spectrum shown in Figure 5, the power of frequencies corresponding to frequencies of 0.4 times / hour or more is compared with a threshold (e.g., 80).

[0065] In this example, the power corresponding to a frequency of 0.5 times / hour exceeds the threshold. This means that a disturbance with a short period of approximately 2 hours is superimposed on the observed quantity (Pmax), and in this case, the power at frequencies within a predetermined range is determined to be greater than the threshold. The determination result is supplied to the frequency setting unit 314 and the operation determination unit 330.

[0066] Furthermore, the FFT processing unit 313 compares the power corresponding to relatively low frequencies (relatively long periods) with a threshold. For example, in the frequency spectrum shown in Figure 5, the power of frequencies corresponding to frequencies less than 0.4 times / hour is compared with a threshold (e.g., 80).

[0067] In this example, the power at frequencies corresponding to a frequency of less than 0.4 times / hour does not exceed the threshold. This means that disturbances with long periods are not superimposed on the observed quantity (Pmax), and in this case, it is determined that the power at frequencies within a predetermined range is not greater than the threshold. The determination result is supplied to the frequency setting unit 314 and the operation determination unit 330.

[0068] (Frequency setting section) The frequency setting unit 314 compares the power of frequencies within a predetermined range (relatively short period) in the frequency spectrum obtained by the FFT processing unit with a threshold value, and sets the frequencies corresponding to power exceeding the threshold value as the period of periodic fluctuations to be removed in the filter.

[0069] The frequency setting unit 314 sets a frequency (a frequency to be removed by filtering) in a notch filter (not shown), for example. If the FFT processing unit 313 determines that the power corresponding to a frequency of 0.4 times / hour or more exceeds a threshold, it sets the frequency corresponding to the power exceeding the threshold in the notch filter as the frequency to be removed. In the example in Figure 5, a frequency of 0.5 times / hour is set in the notch filter.

[0070] Furthermore, the frequency setting unit 314 compares the power of frequencies within a predetermined range (relatively long period) in the frequency spectrum obtained by the processing of the FFT processing unit with a threshold, and sets the frequency corresponding to the power exceeding the threshold as the frequency of periodic fluctuations related to the disturbance model set in the disturbance estimation observer.

[0071] (Operation decision unit) The operation determination unit 330 determines the operation and manipulated quantities for the process based on the estimated state of the process obtained by inputting the first and second observed quantities to the disturbance estimation observer. In other words, the operation determination unit 330 is a functional block corresponding to the model predictive control shown in Figure 3.

[0072] The operation determination unit 330 determines the operations to be input to the blast furnace process and their corresponding amounts by taking estimated values ​​of the blast furnace process state and estimated values ​​of disturbances as input and outputting control quantities corresponding to predetermined operations. The operation determination unit 330 may, for example, calculate control quantities indicating increases or decreases in the airflow rate, increases or decreases in the pulverized coal injection rate, etc., by predetermined model calculations.

[0073] Furthermore, the operation determination unit 330, in accordance with the determination result of the FFT processing unit 313, sets estimated values ​​of the disturbance to be used as input (disturbance superimposed on the observed quantity corresponding to process P1), estimated value d1^, estimated value d 1notch Use either ^ or the estimated value ζ1^.

[0074] If the frequency power in the frequency spectrum obtained by FFT processing is determined not to be greater than the threshold, the estimated disturbance d1^ supplied by a disturbance observer (not shown) is used.

[0075] In other words, if the power corresponding to any frequency is below a threshold, the manipulated variables corresponding to various operations are output, taking the estimated disturbance d1^, estimated disturbance d2^, and estimated blast reactor process state x^ as inputs. That is, since periodic disturbances are not superimposed on the observed blast reactor process variables, the system controls the predetermined indicators to be kept close to the preset target values ​​by manipulating the manipulated variables determined by a predetermined model calculation.

[0076] If the power of frequencies within a first range is determined to be greater than the threshold in the frequency spectrum obtained by FFT processing, an estimated disturbance d is supplied by a notch filter (not shown) and the periodic fluctuations corresponding to the frequencies corresponding to the power exceeding the threshold are removed. 1notch The ^ symbol is used.

[0077] In other words, if the power corresponding to relatively high frequencies exceeds the threshold, the estimated disturbance d 1notch The system takes ^, an estimated disturbance d2^, and an estimated blast furnace process state x^ as inputs and outputs manipulated variables corresponding to various operations.

[0078] In other words, because short-term periodic disturbances are superimposed on the observed quantities of the blast reactor process, it is difficult to control the manipulated variables, which are determined by predetermined model calculations, so that the indicators remain at a predetermined target value. Even if an operation is performed to suppress fluctuations in Pmax caused by short-period disturbances, Pmax may fluctuate before the effect of the operation becomes apparent, potentially exacerbating the fluctuations in Pmax. For this reason, short-term periodic disturbances superimposed on the observed quantities of the blast reactor process are removed by a notch filter before determining the manipulated variables.

[0079] If the power of frequencies within the second range is determined to be greater than the threshold in the frequency spectrum obtained by FFT processing, the estimated disturbance ζ1^ supplied from the disturbance estimation observer, which is configured with a disturbance model that combines a constant-value disturbance model from the estimation unit (not shown) and a disturbance model capable of representing periodic fluctuations, is used.

[0080] In other words, if the power corresponding to relatively low frequencies exceeds a threshold, the estimated disturbance ζ1^, the estimated disturbance d2^, and the estimated blast furnace process state x^ are inputs, and manipulated variables corresponding to various operations are output.

[0081] In other words, since long-term periodic disturbances are superimposed on the observed quantities of the blast reactor process, we will actively remove the effects of these disturbances through manipulation. To this end, in estimating the periodically fluctuating disturbances, we will use a disturbance estimation observer with a more suitable disturbance model to estimate the disturbances and determine the manipulated quantities.

[0082] (Changes in manipulated parameters and changes in the amount of Pmax) Figure 6 illustrates an example of how manipulated quantities in a blast furnace process change over time. In Figure 6, the upper and lower graphs show the changes in airflow rate and pulverized coal injection rate (PCI), respectively. In both the upper and lower graphs, the horizontal axis represents time, and the vertical axis represents manipulated quantities, showing the changes in airflow rate and pulverized coal injection rate over time. In the example in Figure 6, the airflow rate remains constant, but the pulverized coal injection rate changes.

[0083] Figure 7 illustrates the change in Pmax over time observed when the operation shown in Figure 6 is performed. In the graph shown in Figure 7, the horizontal axis represents time, and the vertical axis represents the amount of Pmax, showing the change in the amount of Pmax over time. In the graph shown in Figure 7, line 421 shows the change in the amount of Pmax theoretically estimated based on the operation shown in Figure 6, and line 422 shows the change in the amount of Pmax actually observed.

[0084] The theoretically estimated change in the amount of Pmax can be calculated, for example, by providing a manipulated variable to a pre-defined model calculation formula. That is, if the airflow rate is kept constant and the amount of pulverized coal injected changes as shown in Figure 6, the amount of Pmax should theoretically change as shown by curve 422 in Figure 7.

[0085] As can be seen from Figure 7, the amount of Pmax actually observed is not thought to change solely in response to the change in the amount of pulverized coal blown in, as shown in Figure 6, but is also caused by other factors, and these factors constitute disturbances. In other words, the disturbance is shown by the difference between curve 421 and curve 422, and when FFT processing is performed on this disturbance, a frequency spectrum like the one shown in Figure 5 is obtained.

[0086] (Operation execution determination unit) The operation execution determination unit 340 determines whether or not to execute the operation determined by the operation decision unit based on the detection results of the disturbance fluctuation detection unit. Specifically, the operation execution determination unit 340 determines whether or not to execute the operation with the manipulated amount determined by the operation decision unit 330 based on the determination results of the standard deviation calculation unit 311 and the determination results of the regression line calculation unit 312. For example, the operation execution determination unit 340 executes a process to skip the operation by setting the value of the change amount from the previous value of the manipulated amount of a predetermined operation determined by the operation decision unit 330 to 0.

[0087] (Statistical properties of short-term fluctuations in disturbances) The operation execution determination unit 340 classifies the type of variation of the estimated disturbance d1^ into the following first to third patterns based on the determination result of the standard deviation calculation unit 311 and the determination result of the regression line calculation unit 312.

[0088] Pattern 1: The standard deviation of the estimated disturbance d1^ is below the threshold, and the difference between the current value and the theoretical value of the estimated disturbance d1^ is below the lower threshold.

[0089] Pattern 2: The standard deviation of the estimated disturbance d1^ is below the threshold, and the difference between the current value and the theoretical value of the estimated disturbance d1^ exceeds the upper threshold.

[0090] Third pattern: Regardless of the difference between the current value and the theoretical value, the standard deviation of the time series data of the disturbance estimate d1^ exceeds the threshold.

[0091] If the judgment result of the standard deviation calculation unit 311 and the judgment result of the regression line calculation unit 312 are classified into the first pattern, it is considered that there is a short-term fluctuation in the estimated disturbance d1^ and that the estimated disturbance d1^ is rapidly decreasing.

[0092] In such cases, it is highly likely that the observed quantity (e.g., Pmax) superimposed with the estimated value d1^ as a disturbance is also rapidly decreasing. Therefore, if a change in the first pattern is detected in the estimated value d1^ of the disturbance, the operation execution determination unit 340 executes a process to skip operations that have a heat-increasing effect. In this case, for example, among the manipulated quantities determined by the operation decision unit 330, the value of the change from the previous value of the manipulated quantity of an operation with a heat-increasing effect is overwritten to 0 and output, and various operations are executed.

[0093] If the judgment result of the standard deviation calculation unit 311 and the judgment result of the regression line calculation unit 312 are classified into the second pattern, it is considered that there is a short-term fluctuation in the estimated disturbance d1^ and that the estimated disturbance d1^ is rising sharply.

[0094] In such cases, it is highly likely that the observed quantity (e.g., Pmax) superimposed as a disturbance by the estimated value d1^ is also rapidly increasing. Therefore, if a second pattern of change is detected in the disturbance estimate d1^, the operation execution determination unit 340 executes a process to skip operations that have a heat-reducing effect. In this case, for example, among the manipulated quantities determined by the operation decision unit 330, the value of the change from the previous value of the manipulated quantity for operations with a heat-reducing effect is overwritten to 0 and output, and various operations are executed.

[0095] If the result of the standard deviation calculation unit 311 is classified into the third pattern, it is considered that there are short-term fluctuations in the estimated disturbance d1^, and that the estimated disturbance d1^ is fluctuating wildly.

[0096] In such cases, the cause of the disturbance is complex, and operations performed under such circumstances may反而 promote fluctuations. For this reason, when it is detected that the third pattern change has occurred in the estimated disturbance value d1^, the operation execution determination unit 340 executes a process of skipping operations with a heating effect and operations with a cooling effect. In this case, for example, among the operation amounts determined by the operation determination unit 330, the change amounts from the previous value of the operation amount of the operation having a heating effect and the operation amount of the operation having a cooling effect are rewritten to 0 and output, and various operations are executed.

[0097] As described above, the operation execution determination unit 340 classifies short-term fluctuations in the estimated value of the first disturbance into fluctuations in which the estimated disturbance value sharply decreases, fluctuations in which it sharply increases, or fluctuations in which it fluctuates sharply up and down based on statistical properties, and excludes the operation with a heating effect and / or the operation with a cooling effect from being executed depending on which category the short-term fluctuation of the estimated value of the first disturbance falls into.

[0098] (Notch filter) Specifically, the notch filter (continuous time) is designed, for example, by equation (7). Here, s is the Laplace operator.

Math

[0099] g notch (≦1) the suppression gain (depth of the notch) is set.

[0100] Q notch (>1) the suppression frequency band (width of the notch) is set.

[0101] ω notch the suppression angular frequency (=2πf notch ) is set. Here, f notch is the suppression frequency, and the frequency determined by the FFT processing unit 313 to have power equal to or greater than a threshold is set.

[0102] When performing disturbance estimation using a disturbance estimation observer, and employing a discrete-time version of the state-space model that is time-discretized by the sampling period, it is necessary to use a discrete-time version of the notch filter that is also time-discretized by the sampling period.

[0103] We will show the formulation when discretized with sampling time Δt using a bilinear transform. Continuous-time notch filter F notch When (s) is time-discretized, a discrete-time state-space model is obtained by equations (8) and (9) below, where n is the discrete-time step and x notch is the state variable within the filter, u notch is the input variable to the filter, y notch represents the output variables from the filter, respectively. That is, u notch corresponds to the input d1^ to the notch filter, and y notch The output d from the notch filter is 1notch Corresponds to ^.

number

number

number

number

number

number

number

number

number

[0104] (Control process flow) Next, the flow of control processing by the control device 300 will be explained. Figure 8 is a flowchart illustrating an example of control processing.

[0105] In step S21, the disturbance observer estimates the state and disturbance from the observed quantities of process P1.

[0106] At this time, by inputting the observed quantity corresponding to process P1 (for example, the value of Pmax) into the disturbance estimation observer, the estimated disturbance d1^ and the estimated state x1^ are calculated, thereby estimating the state and the disturbance.

[0107] In step S22, the disturbance observer estimates the state and disturbance from the observed quantities of process P2.

[0108] At this time, by inputting the observed quantity corresponding to process P2 (for example, the value of the molten iron temperature) into the disturbance estimation observer, the estimated disturbance value d2^ and the estimated state value x2^ are calculated, thereby estimating the state and the disturbance.

[0109] In step S23, the disturbance detection unit 310 performs disturbance detection processing. This detects whether there are any short-term fluctuations in the estimated disturbance value d1^, along with the statistical properties of those fluctuations.

[0110] Now, with reference to Figure 9, we will explain the details of the disturbance detection process in step S23 of Figure 8. Figure 9 is a flowchart illustrating an example of the flow of the disturbance detection process.

[0111] In step S41, the disturbance detection unit 310 acquires time-series data of the estimated disturbance d1^.

[0112] In step S42, the standard deviation calculation unit 311 calculates the standard deviation of the time series data of the disturbance estimate d1^. For example, the time series data is assumed to be a series of estimates d1^ over one hour.

[0113] In step S43, the standard deviation calculation unit 311 determines whether the standard deviation calculated in step S42 is greater than a preset threshold. If it is determined in step S43 that the standard deviation is greater than the threshold, the process in step S44 is executed.

[0114] In step S44, the disturbance detection unit 310 outputs data showing a third pattern as the statistical properties of the short-term fluctuations of the estimated disturbance d1^.

[0115] If it is determined in step S43 that the standard deviation is not greater than the threshold, the process in step S45 is executed.

[0116] In step S45, the regression line calculation unit 312 calculates a regression line relating to the time series data of the disturbance estimate d1^. The time series data here may be longer in duration than the time series data processed by the standard deviation calculation unit 311. For example, the time series data may consist of two consecutive estimates d1^ over two hours. The regression line may be calculated using, for example, the least squares method.

[0117] The regression line calculation unit 312 compares the current value of the disturbance estimate d1^ obtained by the process in step S21 with the current value (theoretical value) of the disturbance estimate d1^ shown by the regression line calculated in the process in step S45.

[0118] In step S46, the regression line calculation unit 312 calculates the difference between the current value and the theoretical value and determines whether the difference falls below the lower threshold. If it is determined in step S46 that the difference between the current value and the theoretical value falls below the lower threshold, the process in step S48 is executed.

[0119] In step S48, the disturbance detection unit 310 outputs data showing a first pattern as the statistical properties of the short-term fluctuations of the estimated disturbance d1^.

[0120] On the other hand, if it is determined in step S46 that the difference between the current value and the theoretical value did not fall below the lower threshold, the process in step S47 is executed.

[0121] In step S47, the regression line calculation unit 312 calculates the difference between the current value and the theoretical value and determines whether the difference exceeds the upper threshold. If it is determined in step S47 that the difference between the current value and the theoretical value exceeds the upper threshold, the process in step S49 is executed.

[0122] In step S49, the disturbance detection unit 310 outputs data showing a second pattern as the statistical properties of the short-term fluctuations of the estimated disturbance d1^.

[0123] On the other hand, if it is determined in step S47 that the difference between the current value and the theoretical value does not exceed the upper threshold, the process in step S50 is executed.

[0124] In step S50, the disturbance detection unit 310 outputs data indicating no short-term fluctuations in the estimated disturbance value d1^.

[0125] The disturbance detection process is executed in this manner.

[0126] Returning to the explanation of Figure 8, in step S24, the disturbance detection unit 310 determines whether or not a short-term fluctuation in the estimated disturbance value d1^ has been detected as a result of the processing in step S23. That is, if no data indicating no short-term fluctuation is output as a result of the processing in step S23, it is determined in step S24 that a short-term fluctuation has been detected. If it is determined in step S24 that a short-term fluctuation has been detected, the processing in step S25 is executed.

[0127] In step S25, the operation determination unit 330 calculates a manipulated variable corresponding to a predetermined operation, taking the estimated state value and the estimated disturbance value as input. At this time, the manipulated variable is calculated using the estimated disturbance value d1^ calculated in step S21, the estimated disturbance value d2^ calculated in step S22, and the estimated value x^ obtained by merging the estimated state value x1^ calculated in step S21 and the estimated state value x2^ calculated in step S22 as input.

[0128] In step S26, the operation execution determination unit 340 executes the operation execution determination process.

[0129] Here, we will explain the details of the operation execution determination process in step S26. Figure 10 is a flowchart illustrating an example of the flow of the operation execution determination process.

[0130] In step S61, the operation execution determination unit 340 determines the statistical properties of the short-term fluctuations (pattern of short-term changes) of the estimated disturbance d1^. At this time, the pattern of short-term changes is determined based on the processing results of steps S44 and S48 to S50, as explained with reference to Figure 9.

[0131] In step S61, if it is determined that the short-term change pattern is the first pattern, the process in step S62 is executed. In step S62, the operation execution determination unit 340 skips the operation by setting the value of the change from the previous value of the operation amount

[0132] In other words, in the first pattern, short-term fluctuations are observed in the estimated disturbance d1^, and it is considered that the estimated disturbance d1^ is rapidly decreasing, so operations that have a heat-increasing effect are excluded from consideration.

[0133] If, in step S61, it is determined that the short-term change pattern is the second pattern, the process in step S63 is executed. In step S63, the operation execution determination unit 340 skips the operation by setting the value of the change from the previous value of the operation amount of the operation amount of the operation that has a heat reduction effect, among the operation amounts determined in step S25 of Figure 8, to 0.

[0134] In other words, in the second pattern, short-term fluctuations are observed in the estimated disturbance d1^, and it is considered that the estimated disturbance d1^ is rising sharply, so operations that have a heat-reducing effect are excluded from consideration.

[0135] If, in step S61, it is determined that the short-term change pattern is the third pattern, the process in step S64 is executed. In step S64, the operation execution determination unit 340 skips the operation by setting the value of the change from the previous value of the operation amount for operations with a heat-increasing effect and operations with a heat-reducing effect, among the operation amounts determined in step S25 of Figure 8, to 0.

[0136] In other words, in the third pattern, short-term fluctuations are observed in the estimated disturbance d1^, and it is considered that the estimated disturbance d1^ is fluctuating wildly. Therefore, operations that increase heat and operations that decrease heat are excluded from consideration.

[0137] In this way, the operation execution determination process is executed.

[0138] Returning to the explanation of Figure 8, if it is determined in step S24 that no short-term fluctuations were detected, the process in step S27 is executed. That is, if the result of the process in step S23 is that data indicating no short-term fluctuations is output, it is determined in step S24 that no short-term fluctuations were detected.

[0139] In step S27, an FFT operation is performed on the time series data of the estimated disturbance d1^.

[0140] In step S28, the periodic gender matching process is executed. Here, we will explain the details of the periodic gender matching process in step S28. Figure 11 is a flowchart illustrating an example of the flow of the periodic gender matching process.

[0141] In step S81, the FFT processing unit 313 determines what period of disturbance is being superimposed.

[0142] More specifically, in step S81, the FFT processing unit 313 determines whether the power of frequencies within a first range in the frequency spectrum obtained by the FFT processing in step S27 is greater than a threshold.

[0143] The frequencies within the first range may be, for example, relatively high frequencies. As an example, the frequencies within the predetermined range may correspond to frequencies with a frequency of 0.4 times / hour or more. If the power of the frequency spectrum exceeds a threshold at frequencies corresponding to a frequency of 0.4 times / hour or more, it is determined that a short-period disturbance is superimposed. If it is determined in step S81 that a short-period disturbance is superimposed, the process in step S82 is executed.

[0144] In step S82, the frequency setting unit 314 sets the frequency (the frequency to be removed by filtering) for the notch filter. Here, the frequency with power exceeding the threshold in the frequency spectrum is set. For example, in the example in Figure 5, the frequency corresponding to 0.5 times / hour is set for the notch filter.

[0145] In step S83, filtering is performed using a notch filter. This removes short-period disturbances from the time series data of the disturbance estimate d1^, resulting in the disturbance estimate d 1notch This will result in a ^ being output.

[0146] Furthermore, in step S81, the FFT processing unit 313 determines whether the power of the frequencies within the second range in the frequency spectrum obtained by the FFT processing in step S27 is greater than a threshold.

[0147] The frequencies within the second range may be, for example, relatively low frequencies. As an example, the frequencies within the predetermined range may correspond to frequencies with a frequency of less than 0.4 times / hour. If the power of the frequency spectrum exceeds a threshold at frequencies corresponding to a frequency of less than 0.4 times / hour, it is determined that a long-period disturbance is superimposed. If it is determined in step S81 that a long-period disturbance is superimposed, the process in step S84 is executed.

[0148] In step S84, the frequency setting unit 314 specifies the angular frequency of the disturbance model to be set in the disturbance estimation observer of the estimation unit. At this time, a disturbance model capable of representing periodic fluctuations is generated using the angular frequency ω (=2πf) based on the frequency f in which the power exceeds the threshold. Then, a disturbance model (disturbance model according to equations (4) and (5)) is set in the disturbance estimation observer of the estimation unit, which combines the constant value disturbance model and the disturbance model capable of representing periodic fluctuations.

[0149] In step S85, the disturbance estimate corresponding to process P1 is recalculated. As described above, the disturbance estimation observer is configured with disturbance models given by equations (4) and (5). The observed quantity of process P1 (e.g., Pmax) is input to the disturbance estimation observer, and the disturbance estimate ζ1^ is calculated.

[0150] On the other hand, if the power of any frequency in the frequency spectrum obtained by the FFT processing in step S27 does not exceed the threshold, then in step S81 it is determined that there is no periodicity in the disturbance.

[0151] In this way, after the periodic gender correspondence processing, the estimated disturbance corresponding to process P1, which is necessary for calculating the manipulated variable that will be executed later, is supplied to the operation determination unit 330.

[0152] In other words, when short-period disturbances are superimposed, the estimated disturbance corresponding to process P1 is d 1notch ^ is assumed to be the estimated value of the disturbance corresponding to process P1 when long-period disturbances are superimposed, and d1^ is assumed to be the estimated value of the disturbance corresponding to process P1 when the disturbance is not periodic.

[0153] In this way, the periodic gender matching process is executed.

[0154] Returning to the explanation of Figure 8, in step S29, the operation determination unit 330 calculates the manipulated variable corresponding to a predetermined operation, taking the estimated state and the estimated disturbance as input.

[0155] For example, if short-period disturbances are superimposed, the estimated disturbance d obtained after processing in step S83 is 1notch The manipulated variable is calculated using the following inputs: ^, the estimated disturbance d2^ calculated in step S22, and the estimated value x^ obtained by merging the estimated state x1^ calculated in step S21 and the estimated state x2^ calculated in step S22.

[0156] Furthermore, for example, if long-period disturbances are superimposed, the manipulated variable is calculated using the estimated disturbance ζ1^ obtained after processing in step S85, the estimated disturbance d2^ calculated in step S22, and the estimated value x^ obtained by merging the estimated state x1^ calculated in step S21 and the estimated state x2^ calculated in step S22 as input.

[0157] Furthermore, for example, if the disturbance is not periodic, the manipulated variable is calculated using the estimated disturbance d1^ obtained in step S21, the estimated disturbance d2^ calculated in step S22, and the estimated value x^ obtained by merging the estimated state x1^ calculated in step S21 and the estimated state x2^ calculated in step S22 as input.

[0158] After the processing in step S26, or after the processing in step S29, in step S30, the operation amount for the next operation is output.

[0159] The control process is executed in this manner.

[0160] (Effects of the embodiment) Next, the effects of executing the control process according to this embodiment will be explained with reference to Figures 12 to 15.

[0161] (Operations related to changes in Pmax) Figure 12 is a graph showing the change in Pmax, which is obtained as an observable quantity corresponding to process P1. In Figure 12, the horizontal axis represents time, and the vertical axis represents the value of Pmax, with the change in the value of Pmax over time shown by the solid line 441.

[0162] Figure 13 is a graph showing the change in molten iron temperature, which is obtained as an observed quantity corresponding to process P2. In Figure 13, the horizontal axis represents time, and the vertical axis represents the value of molten iron temperature. The change in the value of molten iron temperature over time is shown by the solid line 442 and the dotted line 443. Note that the molten iron temperature shown in Figure 13 is not an actual observed value, but an estimated value calculated using a model formula based on the value of Pmax.

[0163] The solid line 442 shows the change in molten iron temperature when the control process according to this embodiment is performed. The dotted line 443 shows the change in molten iron temperature when the conventional control process is performed, as a comparative example.

[0164] Figure 14 is a graph showing the change in the operational quantity input to the blast furnace process. Here, the operational quantity shown is the change in the pulverized coal injection rate (PCI). In Figure 14, the horizontal axis represents time, and the vertical axis represents the PCI value, with the change in the PCI value over time shown by the solid line 444 and the dotted line 445.

[0165] The solid line 444 shows the change in PCI when the control process according to this embodiment is executed. The dotted line 445 shows the change in PCI when the conventional control process is executed, as a comparative example.

[0166] As shown in Figure 12, the solid line 441 fluctuates up and down in small increments, suggesting that the observed quantity Pmax is superimposed with short-term and short-period disturbances. When conventional control processing is performed based on Pmax as shown in Figure 12, the molten iron temperature changes as shown by the dotted line 443 in Figure 13. Also, when conventional control processing is performed based on Pmax as shown in Figure 12, PCI changes as shown by the dotted line 445 in Figure 14.

[0167] In other words, in conventional control processes, the manipulated variable is determined in response to short-term disturbances or short-period disturbances, and attempts are made to resolve changes in Pmax that cannot be expected to be resolved by control operations. As a result, when conventional control processes are executed, the PCI value rises rapidly and becomes constant, while the molten iron temperature rises above the upper limit of the control range, then continues to fall below the lower limit of the control range.

[0168] In contrast, when the control process according to this embodiment is executed, the PCI value does not rise significantly, and although a decrease in molten iron temperature is observed, it always remains between the lower limit and upper limit of the control range.

[0169] Thus, according to this embodiment, stable control operations can be performed even when short-term or short-period disturbances are superimposed on the observed quantity Pmax.

[0170] (Accuracy related to the estimation of disturbances superimposed on Pmax) Figure 15 compares the accuracy of disturbance estimation when a long-period disturbance is superimposed on Pmax. In the graph shown in Figure 15, the horizontal axis represents the calculation period, the vertical axis represents the estimated disturbance, and a curve is shown connecting the estimated disturbance values ​​calculated in each calculation.

[0171] In Figure 15, Case 1 shows the estimated value of the disturbance (estimated value d1^) obtained by a disturbance estimation observer that has set up a constant-value disturbance model.

[0172] In Figure 15, Case 2 shows the estimated disturbance (estimated value ζ1^) obtained by a disturbance estimation observer that uses a disturbance model combining a constant-value disturbance model and a disturbance model capable of representing periodic fluctuations.

[0173] Comparing Case 1 and Case 2, Case 1 exhibits a phase lag relative to the true value of the disturbance, indicating poor tracking performance. In contrast, Case 2 demonstrates good tracking performance relative to the true value of the disturbance.

[0174] Thus, compared to setting up a constant-value disturbance model, a disturbance model that combines a constant-value disturbance model with a disturbance model capable of representing periodic fluctuations allows for a more favorable estimation of periodic fluctuation disturbances. Therefore, when performing control operations using model predictive control (MPC) based on estimated disturbance values, it becomes possible to more favorably suppress future periodic fluctuations. For example, it becomes possible to suppress in advance future fluctuations in furnace heat that may be caused by fluctuations in Pmax.

[0175] Thus, according to this embodiment, stable control operations can be performed even when a long-period disturbance is superimposed on the observed quantity Pmax.

[0176] (Other embodiments) In the embodiment described above, the observed quantity corresponding to process P2, which is observable with a relatively slow time response, was input to the disturbance estimation observer. However, the observed quantity corresponding to process P2 may be an estimated value calculated by the model calculation formula, rather than an actually observed value.

[0177] A method for estimating an observable quantity corresponding to process P2 based on an observable quantity corresponding to process P1 is disclosed, for example, in Japanese Patent Publication No. 2024-137344.

[0178] Furthermore, although the above-described embodiment explained the application of the present invention to a blast furnace, the application is not limited to blast furnaces, and the present invention may also be applied to shaft furnaces used for direct reduction ironmaking, for example.

[0179] <Examples of implementation using software> The function of the control device 300 is a program that causes the device to function as a computer, and can be realized by a program that causes each block of the device to function as a computer.

[0180] Figure 16 is a block diagram illustrating the physical configuration of a computer used as a control device 300. As shown in Figure 16, the control device 300 can be configured by a computer comprising a bus 510, a processor 501, a main memory 502, an auxiliary memory 503, a communication interface 504, and an input / output interface 505. The processor 501, main memory 502, auxiliary memory 503, communication interface 504, and input / output interface 505 are connected to each other via the bus 510. An input device 506 and an output device 507 are connected to the input / output interface 505.

[0181] Processor 501 can include, for example, a CPU (Central Processing Unit), a microprocessor, a digital signal processor, a microcontroller, or a combination thereof.

[0182] For the main memory 502, for example, semiconductor RAM (random access memory) can be used.

[0183] For example, the auxiliary memory 503 may be flash memory, an HDD (Hard Disk Drive), an SSD (Solid State Drive), or a combination thereof. The auxiliary memory 503 stores a program that causes the processor 501 to execute the operations of the control device 300 described above. The processor 501 loads the program stored in the auxiliary memory 503 onto the main memory 502 and executes each instruction contained in the loaded program.

[0184] The communication interface 504 is an interface that connects to a network.

[0185] The input / output interface 505 can be, for example, a USB interface, a short-range communication interface such as infrared or Bluetooth®, or a combination thereof.

[0186] For input devices 506, for example, a keyboard, mouse, touchpad, microphone, or a combination thereof may be used. For output devices 507, for example, a display, printer, speaker, or a combination thereof may be used.

[0187] When the functions of the control device 300 are realized by a program that causes the device to function as a computer, the functions described in each of the above embodiments are realized by executing the above program using the processor 501 and the main memory 502.

[0188] The above program may be recorded on one or more computer-readable recording media, not temporary ones. These recording media may or may not be provided by the above device. In the latter case, the program may be supplied to the above device via any wired or wireless transmission medium.

[0189] Furthermore, some or all of the functions of each of the above blocks can also be realized by logic circuits. For example, an integrated circuit in which logic circuits that function as each of the above control blocks are formed is also included in the scope of the present invention.

[0190] Furthermore, some or all of the functions of each of the above blocks may operate on the above device, or they may operate on other devices (for example, an edge computer or a cloud server).

[0191] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included within the technical scope of the present invention.

[0192] 〔summary〕 A control device according to embodiment 1 of the present invention is a control device for controlling a process, wherein the process includes a first process that outputs a first observable quantity observable with a relatively fast time response and a second process that outputs a second observable quantity observable with a relatively slow time response, and the control device includes an estimated first disturbance obtained by inputting the first observable quantity output in response to an operation to a disturbance estimation observer, which is an estimated first disturbance superimposed on the first observable quantity, and an estimated second disturbance obtained by inputting the second observable quantity output in response to an operation to a disturbance estimation observer, which is an estimated second disturbance superimposed on the second observable quantity, The system includes an operation determination unit that determines an operation and an operation amount for the process based on an estimated value of the state of the process obtained by inputting the first and second observed quantities to a disturbance estimation observer; a disturbance fluctuation detection unit that detects short-term fluctuations in the estimated value of the first disturbance; and an operation execution determination unit that determines whether or not to execute the operation determined by the operation determination unit based on the detection result of the disturbance fluctuation detection unit. If the disturbance fluctuation detection unit detects a short-term fluctuation in the estimated value of the first disturbance, the operation execution determination unit excludes a predetermined operation from the operations determined by the operation determination unit.

[0193] In the control device according to embodiment 2 of the present invention, in embodiment 1 described above, if the disturbance fluctuation detection unit does not detect short-term fluctuations in the estimated value of the first disturbance, the operation determination unit determines the operation and the manipulated quantity based on the estimated value of the second disturbance, the estimated value of the state relating to the process, and any of the estimated value of the first disturbance, the estimated value of the third disturbance, and the estimated value of the fourth disturbance. The estimated value of the third disturbance is obtained by filtering the fluctuation of the estimated value of the first disturbance to remove periodic fluctuations corresponding to a predetermined frequency, and the estimated value of the fourth disturbance is obtained by inputting the first observed quantity to a disturbance estimation observer that is set to a disturbance model different from the disturbance model set to the disturbance estimation observer when estimating the estimated value of the first disturbance.

[0194] In the control device according to embodiment 3 of the present invention, in embodiment 2 described above, if the power of a frequency within a first range exceeds a threshold in the frequency spectrum obtained by performing FFT processing on the time series data of the estimated disturbance, the filter removes the periodic fluctuation corresponding to the frequency corresponding to the power exceeding the threshold, thereby obtaining the estimated third disturbance.

[0195] In the control device according to aspect 4 of the present invention, in aspect 2 described above, if the power of frequencies within the second range exceeds a threshold in the frequency spectrum obtained by performing FFT processing on the time series data of the estimated disturbance, the first observed quantity is input to a disturbance estimation observer, which is set to a disturbance model combining a constant-value disturbance model and a disturbance model capable of representing periodic fluctuations, thereby obtaining the fourth estimated disturbance.

[0196] In the control device according to aspect 5 of the present invention, in any of the above aspects 1 to 4, the disturbance fluctuation detection unit calculates the standard deviation and regression line for the time series data of the estimated value of the first disturbance, detects short-term fluctuations of the estimated value of the first disturbance based on the standard deviation and regression line, identifies the statistical properties of the detected short-term fluctuations, and determines operations to be excluded from execution in accordance with the statistical properties.

[0197] In the control device according to aspect 6 of the present invention, in aspect 5 described above, the operation execution determination unit classifies the short-term fluctuations of the estimated value of the first disturbance into fluctuations in which the estimated value of the disturbance decreases sharply, fluctuations in which it increases sharply, or fluctuations in which it volatility, based on the statistical properties, and excludes operations that have a heat-increasing effect and / or operations that have a heat-reducing effect from being performed, depending on which type of fluctuation the short-term fluctuations of the estimated value of the first disturbance are classified into.

[0198] In any of the above embodiments 1 to 6, the control device according to embodiment 7 of the present invention is such that the first observed quantity is at least one of the following: gas composition, carbon solution loss, CO utilization rate, H2 utilization rate, and pig iron production rate (Pmax), and the second observed quantity is at least one of the following: molten iron Si and molten iron temperature.

[0199] In any of the above embodiments 1 to 7, the control device according to embodiment 8 of the present invention includes at least one of the following: airflow rate, oxygen enrichment rate, or pulverized coal injection rate.

[0200] In the control device according to embodiment 9 of the present invention, in any of embodiments 1 to 8 described above, the process is a blast furnace process.

[0201] A control method according to embodiment 10 of the present invention is a control method for controlling a process, wherein the process includes a first process that outputs a first observable quantity observable with a relatively fast time response and a second process that outputs a second observable quantity observable with a relatively slow time response, and the method includes: an estimated first disturbance obtained by inputting the first observable quantity output in response to an operation to a disturbance estimation observer, which is an estimated first disturbance superimposed on the first observable quantity; an estimated second disturbance obtained by inputting the second observable quantity output in response to an operation to a disturbance estimation observer, which is an estimated second disturbance superimposed on the second observable quantity; and the first observable quantity The process includes an operation determination step in which an operation and an operation amount are determined for the process based on an estimated value of the state of the process obtained by inputting the second observed quantity to a disturbance estimation observer; a disturbance fluctuation detection step in which short-term fluctuations in the estimated value of the first disturbance are detected; and an operation execution determination step in which an operation determined by the operation determination step is determined based on the detection result of the disturbance fluctuation detection step, wherein if a short-term fluctuation in the estimated value of the first disturbance is detected by the disturbance fluctuation detection step, a predetermined operation among the operations determined by the operation determination step is excluded from execution in the operation execution determination step. [Explanation of symbols]

[0202] 300 Control device 310 Disturbance detection unit 311 Standard deviation calculator 312 Regression Line Calculation Unit 313 FFT Processing Unit 314 Frequency setting section 330 Operation Determination Unit 340 Operation Execution Determination Unit

Claims

1. A control device that controls a process, The process includes a first process that outputs a first observable quantity that can be observed with a relatively fast time response, and a second process that outputs a second observable quantity that can be observed with a relatively slow time response. An estimated value of the first disturbance obtained by inputting the first observed quantity output in response to the operation into the disturbance estimation observer, wherein the estimated value of the first disturbance is superimposed on the first observed quantity. An estimated second disturbance obtained by inputting a second observed quantity, which is output in response to the operation, into a disturbance estimation observer, wherein the estimated second disturbance is superimposed on the second observed quantity, and An operation determination unit determines an operation and an operation quantity for the process based on an estimated value of the state of the process obtained by inputting the first and second observed quantities into a disturbance estimation observer, A disturbance fluctuation detection unit that detects short-term fluctuations in the estimated value of the first disturbance, The system includes an operation execution determination unit that determines whether or not to execute the operation determined by the operation determination unit based on the detection results of the disturbance fluctuation detection unit, If the disturbance fluctuation detection unit detects a short-term fluctuation in the estimated value of the first disturbance, the operation execution determination unit excludes a predetermined operation from the operations determined by the operation determination unit. Control device.

2. If the disturbance fluctuation detection unit does not detect any short-term fluctuations in the estimated value of the first disturbance, the operation determination unit determines the operation and the manipulated quantity based on the estimated value of the second disturbance, the estimated value of the state related to the process, and any of the estimated values ​​of the first disturbance, the third disturbance, and the fourth disturbance. The estimated value of the third disturbance is obtained by processing the fluctuation of the estimated value of the first disturbance with a filter that removes periodic fluctuations corresponding to a predetermined frequency. The estimate of the fourth disturbance is obtained by inputting the first observed quantity into a disturbance estimation observer that is set to a disturbance model different from the disturbance model set in the disturbance estimation observer when estimating the estimate of the first disturbance. The control device according to claim 1.

3. If, in the frequency spectrum obtained by applying FFT processing to the time series data of the estimated disturbance, the power of frequencies within the first range exceeds a threshold, The filter removes periodic fluctuations corresponding to frequencies that correspond to power exceeding the threshold, thereby obtaining an estimate of the third disturbance. The control device according to claim 2.

4. If, in the frequency spectrum obtained by performing an FFT on the time series data of the estimated disturbance, the power of frequencies within the second range exceeds a threshold, the fourth estimated disturbance can be obtained by inputting the first observed quantity into a disturbance estimation observer equipped with a disturbance model that combines a constant-value disturbance model and a disturbance model capable of representing periodic fluctuations. The control device according to claim 2.

5. The aforementioned disturbance detection unit is The standard deviation and regression line for the time series data of the estimated value of the first disturbance are calculated, and short-term fluctuations of the estimated value of the first disturbance are detected based on the standard deviation and regression line, and the statistical properties of the detected short-term fluctuations are identified. The operations that are excluded from execution are determined in accordance with the aforementioned statistical properties. The control device according to claim 1.

6. The operation execution determination unit, Based on the aforementioned statistical properties, the short-term fluctuations of the estimate of the first disturbance are classified into fluctuations in which the estimate of the disturbance decreases sharply, fluctuations in which it increases sharply, or fluctuations in which it volatility. Depending on which category the short-term fluctuations in the estimate of the first disturbance fall into, operations that increase heat and / or decrease heat are excluded from consideration. The control device according to claim 5.

7. The first observed quantities are gas composition, carbon solution loss, CO utilization rate, and H ２ At least one of the following: utilization rate and / or pig iron production (Pmax), The second observed quantity is at least one of molten iron Si and molten iron temperature. The control device according to claim 1.

8. The aforementioned operating volume includes at least one of the airflow rate, oxygen enrichment rate, or pulverized coal injection rate. The control device according to claim 1.

9. The aforementioned process is a blast furnace process. The control device according to claim 1.

10. A control method for which a process is the target of control, The process includes a first process that outputs a first observable quantity that can be observed with a relatively fast time response, and a second process that outputs a second observable quantity that can be observed with a relatively slow time response. An estimated value of the first disturbance obtained by inputting the first observed quantity output in response to the operation into the disturbance estimation observer, wherein the estimated value of the first disturbance is superimposed on the first observed quantity. An estimated second disturbance obtained by inputting a second observed quantity, which is output in response to the operation, into a disturbance estimation observer, wherein the estimated second disturbance is superimposed on the second observed quantity, and An operation determination step in which an operation and an operation quantity are determined for the process based on an estimated value of the state of the process obtained by inputting the first and second observed quantities into a disturbance estimation observer, A disturbance variation detection step for detecting short-term fluctuations in the estimated value of the first disturbance, The operation execution determination step includes determining whether or not to perform the operation determined by the operation determination step based on the detection results of the disturbance detection step, If the disturbance fluctuation detection step detects a short-term fluctuation in the estimated value of the first disturbance, the operation execution determination step excludes a predetermined operation from the operations determined in the operation determination step. Control method.

Images