JPH0726128B2 - Blast furnace furnace heat control method - Google Patents

Description

TECHNICAL FIELD The present invention relates to a blast furnace thermal control method for stable operation of a blast furnace.

(Prior Art) As a blast furnace thermal control system that estimates the hot metal temperature in the blast furnace and manages and controls it, see, for example, JP-A-62-270708.
There is a system disclosed in the publication.

This blast furnace thermal control system includes data input means for taking in data from various sensors installed in the blast furnace at a predetermined timing, and unloading speed, pressure loss, shaft pressure, shaft temperature, based on the data from the sensors. A means for creating various data indicating the condition of the blast furnace such as the temperature of the fixed sonde, the gas utilization rate, the temperature of the furnace mouth sonde, etc., a means for creating authentic data by comparing the various data with the reference data thereof, Based on the storage means for temporarily storing the data, the knowledge base means for storing various knowledge bases based on the experience and results of the blast furnace, the true / false data of the storage means and the knowledge base of the knowledge base means It comprises inference means for inferring the furnace heat level and the furnace heat transition and determining the action amount for the blast furnace.

According to this system, true / false data is created from data of various sensors installed in the blast furnace, and predetermined inference as artificial intelligence is made from the data and the knowledge base based on the experience stored in the knowledge base means. Since it has been done so, the conventional experience can be fully utilized, and the capacity can be extremely small even when the system is realized by a computer. Furthermore, even if the aging of the blast furnace is changed over time, it is only necessary to change the memory contents of the knowledge base means, and it is extremely easy to change.

(Problems to be Solved by the Invention) However, the furnace heat control method by the blast furnace heat control system described above has the following problems.

Since the true and false data used for inference is obtained by comparing the size of the reference data and the sensor data, if the sensor data fluctuates near the reference data, the true and false data values will be completely opposite and the furnace heat change Leading to a decrease in the inference accuracy of.

Further, since the reference data itself is a fixed value, it is not possible to cope with fluctuations (level changes) in the average value of the sensor data due to changes in the raw material properties in the furnace. From this point as well, the inference accuracy of the furnace heat change is reduced.

Since there is no means for selecting the operation action to be executed, the operation action performed by this blast furnace thermal control system is fixed (blast moisture), and it cannot be said that the optimum operation action was performed.

Since the operation action amount is determined by referring to the action table in the knowledge base, it is difficult to determine the action amount at a fine level.

The present invention has been made to solve the above-mentioned problems (1), and predicts the furnace heat change with high accuracy and quantitatively, and executes the optimum operation action with the optimum control amount for this furnace heat change. It is an object of the present invention to obtain a blast furnace heat control method that can be performed.

(Means for Solving the Problems) A blast furnace heat control method according to the present invention captures at least one blast furnace data indicating a blast furnace internal condition, and compares the blast furnace data with a reference value to obtain A first step of outputting a first prediction result that predicts a short-term change is provided, and the reference value is changing every moment according to the blast furnace data in a past predetermined period, and indicates at least the in-reactor state of the blast furnace. A second step of taking in one blast furnace data and hot metal temperature, outputting a second prediction result for predicting a long-term change in furnace heat by adding the influence of operation action to these data; The furnace heat level of each of the first and second prediction results is obtained by the membership function, and the total furnace heat level is determined by fuzzy reasoning from these furnace heat levels and the first rule determined in advance. Third to seek
And the operation data indicating the total furnace heat level and the current operation status are inferred by collating them with the second rule set in advance so that the operation action to be performed for the stabilization of the furnace heat can be performed. A fourth step of selecting an execution operation action, and a fifth step of obtaining an operation action amount, which is a control amount of the execution operation action, by a function peculiar to the execution operation action with the total furnace heat level as an argument. And a sixth step of executing the execution operation action according to the operation action amount.

(Operation) In the first step of the present invention, since the reference value is changed every moment according to the blast furnace data in the past predetermined period, it is possible to deal with the level change of the blast furnace data.

Further, in the third step, the furnace heat level of each of the first and second prediction results is obtained by a membership function, and from the predetermined first rule and each furnace heat level, by fuzzy reasoning, Since the total furnace heat level is sought, the quantitative total furnace heat level is calculated by adding the first prediction result that predicts the short-term furnace heat change to the second prediction result that predicts the long-term furnace heat change. be able to.

(Example) A. Furnace wall temperature predicting means (A-1) First reason for decrease in furnace heat The following are possible causes of decrease in furnace heat in the blast furnace.

Hot air temperature (gas flow) for adjusting the hot metal temperature and hot metal amount blown from the tuyere of the blast furnace is usually blown into the central part of the furnace. However, a large amount of gas flow may suddenly flow around the inside of the furnace due to the conditions of raw material charging, distribution of charged materials, and the like.
As a result, the endothermic reaction of FeO + C → Fe + CO is promoted and the furnace heat is reduced.

By the way, when a large amount of gas flow flows into the periphery of the furnace, Na,
The in-furnace deposits such as K and Pb and the stagnant layer separate and fall off the wall, causing the temperature of the furnace wall to rise rapidly. If this sudden temperature rise is detected, the furnace heat drop can be predicted.

(A-2) Second Reason for Lowering Furnace Heat In addition, the following may be considered as a cause of lowering furnace heat of the blast furnace.

When the unloading rate in the blast furnace increases for the same reason as in (A-1), the so-called raw mine descent lowers the cohesive zone level in the blast furnace, resulting in a decrease in furnace heat.

By the way, when the cohesive zone level is lowered, the temperature of the furnace wall in the relevant portion is also rapidly lowered. If this rapid temperature drop is detected, the furnace heat drop can be predicted.

(A-3) Third Reason for Decreasing Furnace Heat Further, the following factors can be considered as one of the causes of decreasing the furnace heat of the blast furnace.

Hot air temperature (gas flow) for adjusting the hot metal temperature and hot metal amount blown from the tuyere of the blast furnace is usually blown into the central part of the furnace. However, due to the same reason as (A-1) and (A-2), a part of the gas flow may flow to the inner peripheral portion of the furnace. If this state continues for a long time, the heat dissipation of the gas flow from the furnace wall of the blast furnace becomes larger than that during normal operation, and as a result, the furnace heat decreases.

By the way, when a part of the gas flow steadily flows to the inner peripheral portion of the furnace, the temperature of the furnace wall gradually rises. By detecting such a gradual temperature rise for a relatively long time, a decrease in furnace heat can be predicted.

(A-4) First to Third Prediction Means FIGS. 1A and 1B are a side sectional view and a plan sectional view, respectively, showing the arrangement of the inner wall thermometer used in the embodiment of the present invention. is there. As shown in FIG. 2A, the inner wall thermometer 3 has seven pieces in the height direction of the blast furnace 1 (three pieces on the back, two pieces on the abdomen, and the bosh section 2).
4) in the circumferential direction of the blast furnace 1 as shown in FIG. That is, a total of 28 inner wall thermometers 3 are installed in 7 levels in 4 directions.

The inner wall thermometer is, for example, the actual applicant's 57-81531,
The one disclosed in Japanese Utility Model Publication No. 59-16816 may be used, and FIG. 2 is a conceptual diagram showing the inner wall thermometer (hereinafter referred to as "FM sensor") disclosed in the latter.

In the figure, reference numeral 4 denotes a sheath-type temperature measuring element having two conducting wires 5 embedded in parallel insulatively and having a temperature-sensing portion 6 on the front end side. The temperature-sensitive parts 6 are arranged in parallel so as to be arranged at different parts in the length direction, and the sheath type dummy rod 7 is connected to the tip of the temperature-sensitive part 6 to align the leading edge. The sheath type dummy rod 7 has two conductive wires 5 embedded in parallel in an insulating manner, and has substantially the same thermal conductivity as the sheath type temperature measuring element 4. The FM sensor 3 keeps the sheath type temperature measuring element 4 out of contact with the insulating material 8,
It is formed by being housed in the sheath tube 9.

FIG. 3 is an explanatory diagram of installation of the FM sensor 3. In the figure, 10 to 13 are furnace walls of the blast furnace, 10 is a brick, 11 is a stave, 12 is a stamp, and 13 is an iron skin. As shown in the figure, the FM sensor 3 is installed inside the furnace wall by welding to the packing 14 and the welded portion 15. In addition, 16 is a filler, 17 is a milk injection port, and the part where the filler 16 is poured.

Because of the installation and structure of the FM sensor 3 described here, the FM sensor 3 itself may be eroded as well as the erosion of the furnace wall, and the sheath-type temperature measuring element 4 may be exposed in the furnace near the furnace wall.
Actually, it means that the "temperature inside the furnace near the furnace wall" is measured together with the "temperature inside the furnace wall". Hereinafter, the concept including both is described as “furnace wall temperature”. As described above, the FM sensor 3 has more measurement points than conventional thermometers such as sheath thermocouples, satisfies the quick temperature measurement response, and is capable of continuous temperature measurement for a long period of time. Are being improved, durability is improved, and workability is improved.

As shown in FIG. 4, each FM sensor 3 measures the inner wall temperature of the blast furnace 1 at every predetermined sampling time Δt. Here, the inner wall temperature of the i-th FM sensor 3 at time j is Tj, i, and the inner wall temperature one sampling time Δt before time j is Tj.
_-1 ,, i, the inner wall temperature difference between Tj, i and Tj _-1 , i (difference value)
ΔTj, i is ΔTj, i = Tj, i−Tj ₋₁ , i (1). This state is shown in FIG.

The difference value ΔTj, i is multiplied by the weight wi in consideration of the height of each FM sensor 3, the circumferential direction, and the like. Further, if the difference value ΔTj, i is negative, vi = 0, and for other than that, a positive / negative parameter vi indicating vi = 1 is also multiplied to obtain a corrected difference value (positive difference) at time j. Value) Get CTj, i.

CTj, i = wi.vi · ΔTj, i (2) Next, the sum of the correction difference values CTj, i for all the FM sensors 3 is taken, and this is designated as ST1j.

Then, according to the following equation (4), if the value of the difference value sum ST1j becomes larger than a predetermined threshold value ε ₁ , it is considered that there is a rapid temperature rise, and the evaluation point 1 is given for a predetermined period.

ST1j ≧ ε ₁ (4) The above is the first prediction means based on the reason (A-1).

The second prediction means based on the reason (A-2) is as follows.

In the equation (2), the positive / negative parameter vi is set to vi = 0 when the difference value ΔTj, i is positive, and is set to vi = 1 otherwise, and then the corrected difference value CTj, all F of absolute value of i
The sum for the M sensor 3 is taken and designated as ST2j.

Then, according to the following equation (4) ′, if the value of the difference value sum ST2j based on the equation (3) ′ becomes larger than a predetermined threshold value ε _2, it is considered that there was a rapid temperature drop due to the downfall of the mine,
An evaluation score of 1 is given for a predetermined period.

The third prediction means based on the reason of ST2j ≧ ε ₂ (4) ′ C. is as follows.

The positive / negative parameter vi of the equation (2) is the same as in the first prediction means.
If the difference value ΔTj, i is negative, vi = 0, and otherwise, vi = 1. Further, a correction difference value before k sampling time (that is, Δt × k time before) at time j is CTj ₋ k, i, and a predetermined time width nΔt of this correction difference value is
The sum of the values of the moving average at time j for all FM sensors 3 is taken, and this is designated as ST3j.

Then, according to the following equation (4) ″, if the value of the moving average sum ST3j is larger than a predetermined threshold value ε _3, it is considered that the gradual temperature rise is for a long period of time, and the evaluation point 1 is given for a predetermined period.

ST3j ≧ ε ₃ (4) ″ Since each of the above-described first to third predicting means performs the difference in the furnace wall temperature (difference value), it does not depend on the absolute value of the furnace wall temperature and is accurate. Moreover, due to its good workability and good temperature measurement response, the FM sensor 3 can be placed so as to cover the entire circumference of the blast furnace, and the continuous inner wall temperature difference can be grasped. You can make accurate predictions.

(A-5) Learning of Threshold Each of the thresholds ε ^* (ε ₁ , ε ₂ , ε ₃ of the _first to third predicting means, which is a furnace wall temperature predicting means, is hereinafter collectively referred to as “ε ^* ” ^. Is used in advance.) Is obtained in advance so as to obtain the optimum prediction of furnace heat drop or furnace heat rise. However, the initially optimal threshold value ε ^* may not be optimal during blast furnace operation due to changes in various conditions such as the production plan and raw material conditions. Therefore, it is necessary to change the threshold value ε ^* to an optimum value during blast furnace operation.
The threshold value ε ^* is learned as follows.

When measuring the furnace wall temperature of the blast furnace, the temperature measuring point of the FM sensor 3 which is the furnace wall thermometer changes with the erosion of the furnace wall, as described in (A-4). Therefore, not only the measured value of the furnace wall temperature itself, but also the peak value (maximum value) of the furnace wall temperature difference, that is, the sensor sensitivity, may change according to the erosion of the furnace wall. Since the first to third predicting means make predictions based on the furnace wall temperature difference, the prediction accuracy also fluctuates in such a case.

In order to solve this problem, the following processing is performed. First of all, climbing reverse from the current time t to the time width h _S past,
Average value of each sum ST1j to ST3j in this section h _S ▲
▼ to ▲ ▼ and standard deviations σ _{1 to} σ ₃ are obtained, and thresholds ε _{1 to} ε ₃ are determined by the following equations (5) to (7).

ε ₁ = a ₁・ ▲ ▼ + b ₁・ σ ₁ …… (5) ε ₂ = a ₂・ ▲ ▼ + b ₂・ σ ₂ …… (6) ε ₃ = a ₃・ ▲ ▼ + b ₃・ σ ₃ … (7) Note that a _{1 to} a ₃ and b _{1 to} b ₃ are coefficients.

As described above, the threshold value ε ^* is automatically determined based on the changes in the sampling data in each of the first to third predicting means in a timely manner. Prediction can always be performed with high accuracy regardless of changes in sensor sensitivity due to changes.

(A-6) Comprehensive prediction When the evaluation point 1 is given by any of the above-mentioned first to third predicting means, 1 is output, otherwise 0 is output as the prediction result I _FM of the furnace wall temperature predicting means. It

Note that the following abnormal value correction processing may be performed. That is, when predicting by the first to third predicting means, T
When j, i (Tj _-1 , i) satisfies the following equations (8) to (10), Tj, i-Tj _-1 , i> ΔT (8) Tj, i> Tmax ...... (9 ) Tj, i <Tmin (10) (However, ΔT, Tmax, Tmin are predetermined threshold values.) Tj, i is replaced with Tj _-1 , i to correct the abnormal value. Thereby, the prediction accuracy can be further improved.

B. Sollos C amount predicting means Solros C (solution loss carbon) amount (kg / t-p) is calculated for each sampling time Δt based on analysis of furnace top gas components by gas chromatograph, blowing conditions, raw material charging conditions, etc. To do. Here, if the amount of sol loss C at time j is xj and the amount of sol loss C before k sampling time (that is, Δt × k time) before time j is xj ₋ k, the predetermined time width nΔt, lΔt
Each moving average x _M , xl _M of (l> n) is Can be calculated by

The moving averages x _M and xl _M based on the equations (11a) and (11b) are calculated for each sampling time Δt, and Ix = x _M −xl _M …… (12) Fourth prediction by the following equation (12) The prediction result Ix by the means is obtained. The increase in the amount of Sollos C indicates that the Sollos reaction, which is an endothermic reaction, is promoted. Therefore, if Ix> 0, the furnace heat decreases and Ix <
If 0, it can be said that the furnace heat rise is predicted.

When obtaining the moving average of the amount of sol loss C, the instantaneous value of the amount of sol loss C may cause an abnormal value due to noise or the like, so the following abnormal value correction processing may be performed. That is, assuming that the amount of sol loss C at time j is xj and the amount of sol loss C before sampling time Δt is xj ₋₁ , the absolute value Δxj of the difference value of the sol loss C amount is (| xj−xj ₋₁ |).
An abnormal value is found by comparing this Δxj with a preset threshold value εz, and when an abnormal value is detected, the previous moving average value is replaced with the current moving average value. By applying this method, more accurate prediction becomes possible.

C. Gas Chloro N ₂ Prediction Method Further, the nitrogen content (%) in the furnace top gas detected by gas chromatography (hereinafter, referred to as “gas black N ₂ content”) has a strong negative correlation with the Solros C content. , Instead of increasing the amount of Solros C, by increasing and decreasing the amount of gas chromatography N ₂ ,
Blast furnace heat change can be predicted.

As a result, the amount of gas black N _{2 at the} current time j is
Let yj be the amount of gas chromatogram N ₂ before k sampling time (that is, Δt × k time) before time j, and yj ₋ k, each movement of the predetermined time width nΔt, lΔt (l> n) at the current time j. The average y _M , yl _M is Can be calculated by

Y _M and yl _M based on the equations (13a) and (13b) are calculated for each sampling time Δt, and Iy = y _M −yl _M …… (14) Obtain the prediction result Iy. That is, Iy>
If 0, the furnace heat rise is predicted, and if Iy <0, the furnace heat fall is predicted.

In addition, the abnormal value correction processing used in the Sollos C predicting means is
It is of course possible to apply it to the gas chromatograph N ₂ predicting means.

Since the two prediction means described in BC are based on the moving average for each sampling time Δt, it can be said that the prediction can be obtained quickly and the prediction accuracy is sufficiently reliable.

D. Charge Pitch Prediction Method It is empirically known that the number of times the raw material is charged from the furnace top per predetermined time (hereinafter referred to as "charge pitch") affects the furnace heat. That is, it is known that when the charging pitch increases, the furnace heat decreases, and when the charging pitch decreases, the furnace heat increases.

It is speculated that this is mainly due to the following reasons.

When the charging pitch increases (decreases), the endothermic sol-loss reaction is promoted (suppressed), and as a result, the furnace heat decreases (rises).

When the charging pitch increases (decreases), the amount of tapping increases (decreases), and the time from raw material charging to tapping becomes shorter (longer). As a result, the time for hot metal to exchange heat in the blast furnace is reduced (increased), and the furnace heat is reduced (increased).

Therefore, the prediction result Ip by the charging pitch prediction means is obtained from the following equation (15).

However, PITCHl ₁ : the number of raw material charges in the past l ₁ minute PITCHl ₂ : the number of raw material charges in the last l ₂ minutes (l ₁ <l ₂ ) That is, if Ip> 0, the furnace heat drop is predicted, and Ip < If 0, it means that the furnace heat rise is predicted.

It can be said that the prediction means described in A. to D. basically make a prediction by comparing one piece of blast furnace data showing the in-furnace condition of the blast furnace with the reference data. Reference data ε
^* , Xl _M , yl _M , PITCHl ₂ / l ₂ are set sequentially according to changes in blast furnace data, so the prediction methods described in A. to D. above are based on changes in sensor sensitivity or changes in raw material properties. Even if there is a change in the heat balance due to, it is possible to maintain a certain furnace heat estimation accuracy.

E. Long-term outlook forecasting means All the forecasting means described in A. to D.
Although it was a means of predicting reactor heat changes from a relatively short-term perspective, the predictive means described below are only predictive means from a long-term perspective.

Long-term forecast is based on the following equation (16).

(T + Δt ′) = a · (XN (t) −XM (t)) + b · (Z (t + Δt ′) − Z (t + Δt ′)) + y (t) (16) However, Δt ′> Δt , XN (t): average value of SolLos C amount in past N hours from time t XM (t): average value of Solloss C amount in past M hours (M >> N) from time t Z (t + Δt '): operation action (up Thermal action,
Predicted value of hot metal temperature at time (t + Δt ′) due to influence of heat-fall action (hereinafter referred to as “actual action predicted value”) Z (t + Δt ′): From actual action predicted value Z (t) at time (t + Δt ′) Average value y (t) in the past N hours: Hot metal temperature y in the past N hours from time t
It is the average value of (t).

Also, the coefficients a, b and XN (t), Z (t + Δt '), y (t)
The set time N changes every moment so as to improve the accuracy of the hot metal temperature abnormality prediction described later.

Hereinafter, the actual action predicted value Z (t) will be described in detail with reference to FIG. The actual action predicted value Z (t) is a predicted value of the hot metal temperature at time t when operating actions such as air temperature and humidity control are performed.

As shown in FIG. 6, when a heating action is performed at time 0, the response (force affecting the hot metal temperature y (t)) gradually appears, and the hot metal temperature difference tb (Z (t) -Z ( t)) rises. In this way, the predicted value of the hot metal temperature based on the change in the hot metal temperature after performing the operation action is the actual action predicted value Z (t). This actual action predicted value Z
(T) is calculated by obtaining the response of the hot metal temperature to actions such as air temperature and humidity control by a mathematical model or data analysis. As the actual action prediction value Z (t) based on the mathematical model, for example, “development of unsteady blast furnace simulation program” (“Iron and Steel” No.12.vol73S825)
(1987) page 89).

Next, a method of determining the coefficients a and b and the set time N will be described. These values a, b, N are determined so that the evaluation function J shown in the following equation (17) is minimized.

_{J = p 1 E 1 + p} 2 E 2 + p 3 E 3 + p 4 E 4 ...... (17) However, p ₁ ~p ₄ coefficients E _1: molten pig iron temperature abnormality decrease seen逃率E _2: molten iron temperature abnormality decrease false alarms Rate E ₃ : Abnormal rise in hot metal temperature overlook rate E ₄ : False alarm rate for abnormal rise in hot metal temperature. Abnormal drop in hot metal temperature (hereinafter referred to simply as "fall")
Say. ) Means that the hot metal temperature is lower than the control lower limit temperature T _CL , and the hot metal temperature abnormally rises (hereinafter, simply referred to as “rise”).
Means that the hot metal temperature exceeds the control upper limit temperature T _CH .
Further, it is assumed that the state is stable except the above-mentioned situation.

E _{1 to} E ₄ are determined by the following equations (18) to (21).

However, DG: number of passes of decline (decision prediction hits) DE1: number of misses of decline (negation of decline prediction) DE2: number of false alarms of false decline (wrong forecast of decline) UG: pass of decline (rising of rise forecast Number of hits UE1: Number of missed climbs (failed to forecast rises) UE2: Number of false rises (wrong forecasts of rises).

The above six determinations are made when the operation action is performed based on the prediction (t + Δt ′) of the present embodiment (that is,
When the decrease (increase) is predicted, the heat up (heat down) action is performed, and when the stable prediction is performed, no action is performed), and the determination as shown in Table 1 is made.

In addition, the operation action is the prediction of this embodiment (t + Δ
It is performed regardless of t '), and the determination in the case where the operation action is not performed with respect to the prediction of the fall and rise in this embodiment is as shown in Table 2.

In addition, the operation action is the prediction of this embodiment (t + Δ
When the operation action is an action that is contrary to the prediction (t + Δt ′) of the present embodiment, that is, it is performed regardless of t ′), that is, even though the decrease (rise) is predicted, the heat fall (rise) is performed. The judgment when the action "heat" is performed is as shown in Table 3.

Furthermore, the operation action is the prediction of this embodiment (t + Δ
It is performed regardless of t ′), and the determination when the operation action is performed at the time of predicting the stable state in this embodiment is as shown in Table 4.

Based on the judgments in Tables 1 to 4 above, (18) to (2
The prediction failure rates E _{1 to} E _{4 in the} past predetermined period are calculated from the equation (1). Then, the coefficients a and b and the set time N are determined so that the evaluation function J obtained from the equation (17) is minimized. The determinations in Tables 1 to 4 described above are performed at a predetermined time interval l '.
Each time, it is performed using the predicted value (t) within the time width 1'and the actual hot metal temperature value y (t).

As a result, the coefficients a and b of the equation (16) and the set time N are changed every time l ′ so as to improve the prediction accuracy with respect to the presence or absence of abnormal decrease or abnormal rise of the furnace heat in the time width l ′. Therefore, it can be expected that the blast furnace furnace heat prediction according to the present embodiment can accurately predict whether or not the furnace heat will abnormally rise or fall in the time width l ′.

F. Estimation of total reactor heat level (F-1) Calculation of reactor heat evaluation points Prediction results obtained by each of the prediction means A. to E. I _FM (prediction means
A.), Ix (prediction means B.), Iy (prediction means C.), Ip (prediction means D.), (t + Δt ′) (prediction means E.),
By comparing with the membership functions M _{1 to} M ₅ shown in FIGS. 7 to 11, the furnace heat evaluation points P _{1 to} P ₅ are obtained. For example,
I _FM = 1, Ix = 4, Iy = -0.35, Ip = 0.55,
When (t + Δt ′) = 1477.5 is obtained, the results shown in FIGS.
As shown by the broken line in Fig. 11, P ₁ (evaluation point of I _FM ) is PB = 1.0 P ₂ (evaluation point of Ix) is P S = 0.5 PM = 0.5 P ₃ (evaluation point of Iy) is NB = 0.25 For M = 0.75 P ₄ (evaluation point of Ip), PB = 0.75 M = 0.25 P ₅ (evaluation point of (t + Δt ′)) is NS = 0.75 M = 0.25.

In addition, in FIG. 7 to FIG. 11, NB, NM, NS, M, PS, PM, PB
Are furnace heat evaluation parameters, and each parameter means NB ... fairly low NM ... low NS ... rather low M ... stable PS ... rather high PM ... high PB ... quite high. These parameters have a fitness of 0 to 1. Further, the shapes of the membership functions M _{1 to} M ₅ shown in FIGS. 7 to 11 can be appropriately changed depending on the difference in the blast furnace and the control temperature.

(F-2) Determination of intermediate conclusion The furnace heat evaluation points P _{1 to} P ₅ obtained in this way are compared with the IF-THEN type furnace heat evaluation rule created in advance, and if successful, the intermediate conclusion CI is obtained. obtain. The verification succeeds when all the positive values of the furnace heat evaluation parameters (NB, NM, NS, M, PS, PM, PB) at the furnace heat evaluation points P _{1 to} P ₅ shown in the IF part (conditional part) Is the case. If the furnace heat evaluation points P _{1 to} P ₅ illustrated in (F-1) are satisfied, the collation of rules 1 to 10 shown in Table 5 is successful.

DS, DB, and N in the THEN section (conclusion section) are furnace heat change parameters indicating the degree of change in furnace heat level, and other parameters are UB (large furnace heat rise) and US (small furnace heat rise). Also, the suitability of each furnace heat change parameter is determined by the value of the furnace heat evaluation parameter of the IF section, and when there are multiple furnace heat evaluation parameters in the IF section as in Rule Nos. 2 to 4 and 6 to 8. Is determined by the minimum furnace heat evaluation parameter.

In this example, there are three parameters, DS, DB, and N, for the furnace heat change parameters of the intermediate conclusion CI derived by the THEN part. As the evaluation value of these furnace heat change parameters, the maximum conformity in the same furnace heat change parameter is adopted. In other words DS = 0.
75 (Rule 9), DB = 0.5 (Rule 7, 8), N = 0.5 (Rule 1). Intermediate conclusion CI that UB = US = 0 (no hiring rules)
Will be obtained.

(F-3) Inference of total reactor heat level When the intermediate conclusion CI is obtained, the membership function B _DB (y), B of each reactor heat change parameter is obtained as shown in FIG.
_A composite area BB (y consisting of areas in which the upper part is cut with evaluation values 0.5, 0.75, 0.5, 0, 0 for _DS (y), B _N (y), B _UB (y) and B _US (y), respectively. ) (Shown with diagonal lines).

The y-coordinate y ^* of the position of the center of gravity G of the area surrounded by the combined area BB (y) and the y-axis is obtained by the following equation (22).

This y ^* is the total reactor heat level obtained by fuzzy reasoning, and y ^* = 0.5 in the example of FIG. Closer to the total furnace heat level y ^* 1.5, indicates that the downward trend of the furnace heat level is strong, as y ^* is close to -1.5, indicate that increasing tendency of the furnace heat level is strong, y ^* is 0 the vicinity If so, it indicates that the furnace heat level tends to be stable.

Thus, the total reactor heat level y ^* is a quantitative value because it is obtained by fuzzy reasoning. Also, based on the furnace heat levels of the short-term forecast prediction results I _FM , Ix, Iy, Ip described in A. to C. and the furnace heat levels of the long-term forecast prediction results (t + Δt ′) described in D. Since the furnace heat level y ^* is required, it can be said that the comprehensive furnace heat level y ^* is a highly accurate prediction result.

G. Selection of execution operation action Total furnace heat level y ^* obtained in F., current air temperature, humidity control,
By comparing the operating conditions such as pulverized coal flow rate and O / C with the knowledge base, the operation action to be executed next is selected. Knowledge base is IF-TH
It is written in EN format rules, and these rules are
It was extracted based on the know-how of the operator, and takes into consideration the constraints on the equipment of the blast furnace, the response to changes in the actual hot metal temperature, and the influence of past operation actions. An example of the rules is shown below.

IF y ^* ＞ C ₁ AND [There is no operation action in the past d hours] AND Hot metal temperature y (t) <C ₂ AND Air temperature <C ₃ THEN Air temperature lowering action IF y ^* ＞ C ₁ AND [Past d time No operation action] AND hot metal temperature y (t) <C ₂ AND wind temperature> C ₃ AND humidity control <C ₄ THEN humidity control action In this way, the total furnace heat level and the current operating conditions are taken into consideration. Since the execution operation action is selected based on the knowledge base in which the know-how of the operator is accumulated, the optimum execution operation action can be automatically selected.

H. Calculation of operation action amount When the execution operation action is selected in G., the operation action amount A, which is the control amount of the execution operation action, is calculated by the following (23).
Calculated from the formula.

A = f (y ^* ) (23) The function f (y ^* ) differs depending on the operation action performed, and can be estimated by the total reactor heat level y ^* and the actual action amount taken by the operator. FIG. 13 shows the function f in the case where the execution operation action is the air temperature lowering action.

The function f (y ^* ) can be obtained by the least-squares method based on past data, and the function f in FIG. 13 is f (y ^* ) = 0.19−0.61y ^* + 15.17 (y ^* ) ² Tasu87.32 become ^{(y *) 3 -16.5 (y} *) 4 -57.83 (y *) 5 ...... (24).

As described above, since the operation action amount is obtained by the function using the total furnace heat level y ^* as an argument, the operation action amount can be finely adjusted.

I. Execution of operation action The operation action determined in G. is executed according to the operation action amount determined in H. The blast furnace heat is automatically controlled to stabilize the furnace heat. Further, by displaying each result described in A. to H. on a CRT or the like, it can be used as an action guide for the operator.

J. Flow of Blast Furnace Heat Control FIG. 14 is a flowchart showing the flow of processing when a computer is used for the blast furnace heat control method according to an embodiment of the present invention. FIG. 15 is a conceptual diagram showing a furnace heat control system to which a blast furnace heat control method according to an embodiment of the present invention is applied. The processing procedure in this embodiment will be described below with reference to FIG.

First, in step S1, as described in A. to D., short-term prospect prediction means performs short-term prospect prediction to obtain various short-term prospect prediction results.

Next, in step S2, as described in E., the long-term prospect prediction is performed by the long-term prospect prediction means based on the equation (16), and the long-term prospect prediction result is obtained.

Then, as described in F. in step S3, steps S1, S
Reactor heat evaluation points P _{1 to} P ₅ of the short-term and long-term forecast results obtained in 2 are obtained from the membership functions M _{1 to} M ₅ , and these furnace heat evaluation points P _{1 to} P ₅ and reactor heat evaluation The general reactor heat level y ^* is obtained by fuzzy reasoning from the rules.

After that, in step S4, as described in G., the execution operation action that is the optimum operation action for stabilizing the furnace heat is selected using the knowledge base in which the operation conditions are added to the total furnace heat level y ^*. To do.

Then, in step S5, as described in H., step S4
The control amount of the execution operation action selected in step 1 is determined by a function with the total furnace heat level as an argument.

Then, in step S6, as described in I., step S4
The execution operation action selected in step S5 is executed according to the operation action amount determined in step S5. Then step S1
~ S6 is repeated and the furnace heat control is continued.

K. Supplement Note that the furnace wall temperature predicting means described in A. in this embodiment uses the FM sensor as the inner wall thermometer, but a normal temperature measuring sensor (for example, a sheath thermocouple) also has a problem in terms of life. It is possible to substitute the thing. In addition, even if a stave thermometer or a brick-embedded thermometer is used, the prediction accuracy is slightly lowered due to its low reliability and low temperature response, but it can be used instead.

(Effects of the Invention) As described above, according to the present invention, in the first step, the reference value is changed every moment according to the blast furnace data in the past predetermined period, so that it is possible to cope with the level change of the blast furnace data. By doing so, high prediction accuracy can always be maintained.

Further, in the third step, the furnace heat level of each of the first and second prediction results is obtained by a membership function, and from the determined first rule and each of the furnace heat levels, by fuzzy reasoning, Highly accurate quantitative total furnace heat level that takes into account the second predicted result that predicts long-term furnace heat change, which is the first predicted result that predicts short-term furnace heat change that determines the total furnace heat level You can ask.

By inferring the total reactor heat level obtained in this way from the current operating conditions by comparing it with the second rule,
In the fourth step, to select the operational action to perform,
You can select the optimal operation action.

Then, in the fifth step, since the operation action amount of the executed operation action is obtained by the function using the total furnace heat level as an argument, the action amount can be determined at a fine level.

[Brief description of drawings]

1 (a) and 1 (b) are side sectional views, plan sectional views, FIG. 2 and FIG. 3, respectively, showing the arrangement of the FM sensor used in one embodiment of the present invention in the blast furnace wall. Conceptual diagram of FM sensor, installation explanatory diagram, Fig. 4 is a graph showing the change over time of the furnace wall temperature measured by the FM sensor, Fig. 5 is a graph showing the change over time of the difference value of the measured furnace wall temperature by the FM sensor, 6 is an explanatory view showing the situation after the heat raising action, and FIGS. 7 to 11 are graphs showing membership functions M _{1 to} M ₅ , respectively, and FIG.
The figure is a graph explaining how to determine the total reactor heat level y ^* .
FIG. 13 is a graph showing a function for determining the operation action amount, FIG. 14 is a flowchart showing a processing procedure of a blast furnace heat control method according to an embodiment of the present invention, and FIG. 15 is an embodiment of the present invention. It is a conceptual diagram of a furnace heat control system to which a certain blast furnace furnace heat control method is applied. 1 ... Blast furnace, 3 ... FM sensor

Claims (1)

[Claims] 1. A first prediction result for predicting a short-term change in furnace heat is fetched by taking in at least one blast furnace data showing a situation inside the blast furnace and comparing the blast furnace data with a reference value. 1, the reference value is changing every moment according to the blast furnace data in a predetermined period in the past, at least one blast furnace data indicating the in-furnace condition of the blast furnace and the hot metal temperature are taken in, and these data are obtained. , Which predicted long-term changes in furnace heat, taking into account the effects of operational actions
From the second step of outputting the prediction result of, and the furnace heat level of each of the first and second prediction results by a membership function, and from these furnace heat levels and the first rule determined in advance, The third step of obtaining the total furnace heat level by fuzzy inference and the operation data indicating the total furnace heat level and the current operation status are inferred by collating them with the second rule set in advance, thereby A fourth step of selecting an execution operation action which is an operation action to be performed for stabilizing the operation, and an operation action amount which is a control amount of the execution operation action, the execution operation using the integrated reactor heat level as an argument. A fifth step of obtaining by a function peculiar to the action, and a sixth step of executing the execution operation action according to the operation action amount are further included. Blast furnace heat control method provided.