JPS596306A - Blast furnace operation in iron and steel making process - Google Patents

Abstract

PURPOSE:To operate a blast furnace with minimum consumed energy by a method in which the minimum value of the function indicating the consumed energy in total process by applying the improved Nelder-Mead process, and the specially determined value of the variable satisfying said minimum value of the function is used as the index of the furnace operation. CONSTITUTION:The operative condition of a blast furnace to minimize the energy consumed in iron and steel making process operated by energy generated in its self and energy supplied from outside is obtained, and the blast furnace is operated by using said condition as an index. Then, blast temperature, blast humidity, fuel blown amount, nitrogen blown amount, oxygen blown amount etc. are used as variables x in the function f(x) indicating the amount of energy consumed in total iron and steel making process. Next, the upper and lower limit value of each of these variables, bosh gas amount, combustion temperature in front of tuyeres, estimated temperature of the furnace core and the limited value determined by others are as restriction conditions. Improved Nelder-Mead process is applied in the range which is capable of operation and is determined by these restriction condition. The minimum value of said function f(x) is obtained and the specially determined value of said variables (x) satisfying this minimum value is used as the index for the blast furnace operation.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to a method of operating a blast furnace in a steel process that aims to minimize energy consumption.

Technical Background The steel process forms a complex energy supply and demand system that not only uses externally supplied or purchased energy, but also generates energy as a side effect.Blast furnaces, in particular, consume and generate large amounts of energy. It has a major impact on the energy balance of the entire steel process. In other words, the blast furnace is loaded with ore and coke, blown with hot air and fuel, and generates blast furnace gas to drive and output the TRT (Top Furnace Pressure Recovery Turbine), but the amount of blast furnace gas (BFG) generated is due to co-firing with BFG. Changes in the amounts of coke oven gas COG, liquefied natural gas LNG, and converter gas LDG used in thermal equipment have a large impact on the gas balance of steel processes. In addition, the amount of hot air blown directly affects the gas balance and the power of the auxiliary equipment and hence the amount of purchased electricity when the blower is driven by steam, and directly affects the amount of purchased electricity when the blower is driven by an electric motor. Similarly, the TRT output affects the amount of purchased electricity, the blown fuel ratio affects the purchased fuel, and the coke ratio affects the amount of purchased coal and gas balance.

As described above, the blast furnace has a large influence on the energy balance of the entire steel process, so it is extremely important to comprehensively and systematically determine a blast furnace operating method that minimizes the energy consumption of the entire steel process.

Prior Art and Problems Conventional methods for determining blast furnace operating conditions that minimize the energy consumption of the entire steel process can be broadly classified into the following two methodologies. The first method is to determine the blast furnace operating conditions for several cases, such as the blast temperature, blast humidity, and blown fuel ratio, based on the given raw material conditions and production volume, and then use the blast furnace operation model to determine the blast furnace operating conditions for each case. After calculating the operational specifications and using these specifications to calculate the energy balance for the entire location using an energy optimal distribution model, the energy consumption of the entire process is calculated. Here, the blast furnace operating model is composed of operating formulas used in the blast furnace department for formulating needle plot values, analyzing operations, etc. In addition, the optimal energy allocation model determines the usage of self-generated energy and purchased energy in the most economical manner so as to satisfy the required amount of heat for each heat investment, as well as to satisfy the fuel usage quota and supply equipment constraints. It is something.

Once the energy consumption corresponding to each of several cases of blast furnace operating conditions is determined, these are compared and the specific case with the minimum energy consumption is determined as the optimal blast furnace operating condition. In addition, energy consumption in several cases will be compared and considered, and if necessary, based on the experience of blast furnace operators, new blast furnace operating conditions that are expected to further reduce energy consumption will be determined.

The second method adds a slightly more mathematical approach to the first method, and combines the blast furnace operation model and the energy optimal allocation model into a single model, and converts this model into a Ta
After linear approximation at a specific blast furnace operating point using ylar expansion,
Linear programming (Llnear Program+uing
) to find the optimal operating conditions.

The disadvantage of the first method is that it is only possible to compare the energy consumption corresponding to discrete and limited blast furnace operating conditions within the originally continuous and viable operating range, and it is difficult to compare the energy consumption corresponding to the true minimum energy. There is no guarantee that the operating conditions will be set to , then the corresponding energy consumption must be found, and the number of calculations increases exponentially with the number of dimensions of the conditions (variables), which is very inefficient. Therefore, in this method, in order to avoid searching for countless grid points, grid points are selected based on the experience of blast furnace operators, and new operating conditions that can be expected to further reduce energy consumption are determined.

However, in reality, it is difficult for blast furnace operators to comprehensively and quantitatively understand the energy balance of the iron mesh process and to systematically update operating conditions.

The main drawback of the second linearization method is that it linearly approximates the blast furnace operating model, which is essentially a nonlinear model, at a specific operating point. As a result, the search for an optimal solution is actually limited within the operating range near a specific operating point. Furthermore, when the operating rate, operating method, and energy supply and demand structure of the blast furnace change, linear approximation at a specific operating point must be performed again, resulting in a lack of flexibility in the model. In addition, if there are different operating methods for whether the air humidity should be humidified or dehumidified, it is impossible to formulate them simultaneously into one linear programming problem, and therefore all possible This means that it will not be possible to search for an optimal solution within the area.

Purpose of the Invention The present invention aims to improve these problems, namely, to find the blast furnace operating conditions that minimize the true energy within the entire practicable operating range, to minimize the number of calculations required, and to use the operator's experience judgment. The aim is to provide a method for searching blast furnace operating conditions that does not require analysis requirements, that the model for blast furnace operation and optimal energy allocation does not need to be nonlinear, and that can flexibly deal with changes in conditions.

Structure of the Invention In the present invention, an improved Nelder mead (Nelder-
By applying the (Mead) method to search for blast furnace operating conditions that minimize energy consumption in the steel process, the operating point update procedure can be performed without compromising the nonlinearity of the blast furnace operation optimization problem and based on the experience of blast furnace operators. This makes it possible to systematically search for optimal operating conditions within all feasible regions.

In the NM method, in an n-dimensional space called a simplex (
Using a geometric figure with n + 1) or more points as vertices, the reflection point of the point with the worst objective function value among the vertices of this simplex with respect to the center of the remaining vertices has the objective function value. We take the idea that we can expect improvement,
In addition to this mirroring process, correction operations such as expansion, contraction, and contraction are added depending on the shape of the objective function to serve as means for searching for an optimal solution. This NM method is a method proposed for optimization problems without constraints, but the improved NM method is a method that is extended to optimization problems with constraints.

The advantage of the NM method is that there is no need to clearly define the objective function f (x) to be minimized as a function of the decision variable It is possible to search for the optimal point by simply calculating and using the objective function value f(x') corresponding to any variable, without defining a single objective function with a complex energy supply and demand structure. Suitable for analysis in difficult steel processes.

Further, since there are various operational constraints in blast furnace operation, which restrict the practicable range, an improved NM method is suitable.

In the present invention, the problem of determining the blast furnace operating conditions that minimize the energy consumption of the entire steel process is mathematically solved by the independent variable
−X + , ..., objective function f(
x), lj≦xj≦uj, j=1. ...1m (n<
Consider the problem to be minimized under m constraints of the form m). Just L% ``n+j' ...+ X quiet independent variable X
h..., a function of Xn, upper and lower limits of the constraint conditions,
u−+ 1 is a constant. That is, J subj , to 1j≦Xj≦l1jj=1.・
..., n 12) lj≦g=J (
x l, ..., xn)≦ujj=n+1. ...
, m + 3+ In the blast furnace operation optimization problem, the objective function f (X) is the amount of energy consumed in the entire steel process. The decision variables x=(x+, . . . xn) include the temperature of the air, the humidity of the air, the amount of fuel blown, the amount of nitrogen blown, and the amount of oxygen blown.

In a steel process that has multiple blast furnaces, the specifications of multiple blast furnaces are compounded and affect the energy balance of all locations, so instead of formulating a minimization problem for each blast furnace, a single It must be formulated as a problem.

Therefore, the dimension of the decision variable X is the number of operating conditions times the number of blast furnaces. The blast furnace operation constraints include the upper and lower limits of the above-mentioned decision variables. 1 (King) includes the amount of Bo Soche gas, the combustion temperature at the tuyere tip, and the estimated furnace core temperature for each blast furnace. The method of determining the blast furnace operation index of the present invention will be explained below.

Examples of the invention First, define the symbols as follows.

xK + Coordinates of the vertex of the simplex in n-dimensional space (xlK , . . . The value of f (X) at the maximum point, and
The minimum of the objective function in the 8i simplex is G, fG
i Centers of all vertices except the worst point xL, and xG
The value of f (x) at (1) Initial arrangement: In a problem with constraints, in order to prevent simplex degeneracy, the number of vertices of the simplex is taken as large as K = 2n. Constraint expression +
21. (Initial point (xl'+...) that satisfies 31.

xnO), and the remaining (K-1) points are determined using uniform random numbers as xj==lj+rj(uj-1j), j=1. ...,
Determined by n(4). Here, 0≦r and ≦1 are uniform random numbers. That is, the first point of the initial configuration is determined based on the feasible blast furnace operating conditions (for example, the current operating conditions), but the remaining 11 initial configurations are generated by the system. (4)
Sincerity x=(x+) generated by the formula.

..., ×n) violates constraint (3), the set of points whose sincerity has already been selected (including the initial point)
to the center of the image point. However, if repeating this regression process, there is a possibility that a sincerity that is the same as the already selected sincerity will be selected, it is better to select a new sincerity using equation (4) using a different random number.

After the sincerity xi, l”l, ..., K of K blast furnace operating conditions are selected, the blast furnace operation model and the optimal energy allocation model are used to correspond to the sincerity X1 as shown in Fig. 1. The amount of consumed energy fj is determined.That is, first, the blast furnace operation specifications corresponding to the nose bottle are calculated from the blast furnace operation model using raw material conditions, production volume, energy consumption unit, etc. as given conditions.

At this time, in addition to the constraint values of Bosch gas amount, tuyere tip combustion temperature, and estimated core temperature, BFG generation amount, BFG raw gas calories, etc., which are directly related to energy balance,
The required heat amount of the hot air stove, fuel ratio, TRT output, etc. are calculated.

Next, an optimal energy distribution model for the entire process is performed based on the blast furnace operating specifications and the given conditions. The optimal energy allocation model determines the amount of purchased energy and self-generated energy to be used in the most economical manner while satisfying the required amount of heat from the thermal equipment of each process, as well as satisfying equipment constraints, purchased fuel allowances, etc. It is. After the energy balance of the entire process is calculated in this way, the amount of consumed energy is calculated as an energy balance of (purchased energy - externally sold energy). Energy types include coke (coal), electricity, BFG, LDOSLNG, COG, heavy oil,
Industrial water, etc. Energy such as oxygen, nitrogen, steam, etc. generated and consumed within the process system is
Calculated based on the amount of energy such as electricity.

As described above, the energy consumption of the entire blast furnace can be uniquely calculated once one blast furnace operating condition is determined, so there is no need to clearly define the functional form f(x). For the operating conditions of the initial arrangement, '+ 1=1゜..., the amount of energy consumption f't1'''L as the objective function corresponding to K.
. . . Once K is determined, a simplex update procedure is next performed, and the outline of the procedure will be explained with reference to FIG. 3 for a simple example where n=2.

In Figure 3, the horizontal axis is the operating factor! +, the vertical axis shows the operating factor x2, 11 and u + + It 2 and u2 are the factors XI.

These are the upper and lower limits of x2. Curves C1, C2... are x
+.

The contour lines of the open number f (x) when x2 is changed variously,
Min is the eel minimum value. Straight line L1. L2 is the constraint expression hj(
The upper and lower limits of x IX 2 ) are the upper and lower limits, and the allowable range for the value of f(x) is the third
This is within the area 10 surrounded by diagonal lines in the figure.

The initial point X° is selected within the area 10 like point 1, and since it is n-2, the number of vertices of the simplex is 2n = 4, so the remaining three points are selected using uniform random numbers using equation (4), If these were at points 2.3.4 in Figure 3, squares 1, 2, .
3.4 is a simplex. Of the four vertices of the simplex, the one with the largest f (x), that is, the worst point, is vertex 1, so take a triangle consisting of 2, 3.4 excluding this, and calculate the point l about its center. Mirror point 5
make.

Next, square 2. 3. 4. 5, the worst point among them is 4, so remove it and find the reflection point 6 of point 4 with respect to the center of triangle 2,3°5. The same goes for the following, and is there a point with this operation? , 8.9, . . . , and reach point 12, which reaches the limit of area 10, this is taken as the minimum point to be found. Although it does not reach the actual swelling point Min,
This is due to the constraints, and thus the method can approach the minimum point as much as possible within the constraints. Moreover, the amount of calculation required is considerably small. Note that reflection refers to finding points that are equidistant in opposite directions with respect to the axis of symmetry.
In nature, they do not necessarily need to be at the same distance, and may be expanded or contracted in the same direction as appropriate. Next, a method for updating the above-mentioned operating conditions will be described.

(2) Simplex update: (:) Compare the objective function values at each vertex of the simplex, reflect the worst point xL by with respect to the center of the remaining vertices, and find the sincerity xK. α is the blast furnace to be solved It is determined appropriately depending on the operation optimization problem.The function value fK of sincerity xK obtained by reflection is fs, (H.

(Determine which correction action to take, elongation, contraction, or reduction of the simplex, by considering where it is classified compared to 1-1 and the constraint conditions.

(ii) When fK≦fH: This is a case where xK is a new simplex and does not become the worst point, so it is necessary to determine whether 1 is an acceptable point. In order to determine whether xK is an acceptable point, first determine whether xK violates the upper and lower limit constraints of the operating conditions itself, and if it does not violate the Bosch gas amount and impeller Tip combustion temperature,
It is determined whether or not the constraint h(x) of the estimated core temperature is violated, and at the same time, if the constraint is satisfied, it is determined to be an acceptable point.

When ×1 is an acceptable point; ■If rs<rK srl+ The search continues using xK instead of xL.

■When rK<tR, xK is the best point in the new simplex, so further improvement can be expected in that direction. Therefore x”= (1−
mγ) χc−αxt, (expansion coefficient r > 1) ・
(Extend by 61. The expansion coefficient is determined appropriately depending on the problem. The objective function value f0 at the expansion point X0 is incremented by 1. When r E < r K and xo is an allowable point,
Take 50 instead of XL, otherwise take zu instead of x1 and continue.

When xK is not a permissible point: Use the following permissive algorithm to obtain xKf, which is a permissible value of xK, and use it instead of XLO.

Toleration Algorithm (8) If sincerity If it is not satisfied, set the edge inside the boundary of the constraint by lp O-00001).

(bl If the constraint condition (3) of Bosch gas amount, tuyere tip combustion temperature, and estimated core temperature is not satisfied, return the sincerity Find xKf. If the acceptable point is still not obtained, determine whether it is acceptable with XG, and if it is still not an acceptable point, dx
= (xL-x') /10 is calculated by 1, dxzu□ゞi-
□□ Search in the G direction twice. If that point is also unacceptable, shrink the simplex toward xs.

The reduction operation will be explained in section (iii) (2) below.

In order to ensure more stable convergence, /2 vertices are expanded by p=ts times starting from the one with the smaller objective function value x
Take B1 (!=1. . . , /2) and compare it with the function value of xKf with 1 and K allowed, and the point that gives the minimum function value is adopted instead of XL. However, if the sea is not an acceptable point, it contracts by 1/3 toward Xo.

(iii) When fH<fK; This is a situation that occurs in a narrow valley or near the optimum point, and the sincerity xK becomes the worst point even in the new simplex. Let's consider two cases of thorns.

■(If L≦fK: Return xK that has gone too far,
Since fL≦1, the internal division point on the xL side can be expected to give a smaller function value, so xC=βXK+
(l-β)XG (shrinkage coefficient 0 x β<1)...
...(7) Calculate the objective function value fC by 1. If (C<11
If -, the contraction was successful and XC is used instead of xL.

If fC≧fL, the contraction fails and the simplex is reduced. Here, reduction is a method used when the size of the simplex is not appropriate for the problem.
Create a reduced simplex by moving all vertices by 1/2 toward xKt-x''.

When XC is not a permissible point: When the non-permissible region is surrounded by a simplex, the simplex is reduced to an appropriate size that matches the shape of the constrained region.

■When 1K<1L: Since fL>fK, xK(7
Since the function value can be expected to improve by returning to the internal division point on the ) side, when −r is an allowable point: perform the same process as in (iii) (■).

When XC is not a permissible point: Use the permissible algorithm described in the (i+) section to find a permissible )(cf) and use it instead of the sum.

(3) Convergence Judgment Condition When the standard deviation of the objective function value f(xK) of each vertex of the herring plex becomes ε or less, it is considered that the training point has converged.

ε is determined appropriately depending on the problem.

The variable X that minimizes the function f (x) obtained in this way serves as an index for blast furnace operation. That is, since a blast furnace is not operated only under the condition of minimum energy consumption, in reality, optimal blast furnace operation is carried out while taking into account other requirements.

A schematic flow of the above operation is not shown in FIG. These are programmed so that the required data can be input according to the guidance displayed on the CRT display. Figure 2 shows the equipment configuration. In this figure, 12 is an electronic computer, and 14 is a card reader, which is used to input a large amount of created programs and data. 16 is auxiliary memory,
The above data etc. are stored. 18 is a CRT display connected to the computer 12 via a transmission device and a transmission line;
It is also connected to the keyboard 20 and hard copy 24, and displays input guidance messages, input/output data, etc. 22 is a printer.

For example, the guidance message is displayed as follows.

*0BJBCTIIJt! FUNCTION FOR
OPTIM [ZITION
0 osset is not correct.

1.tlNI! RGY C05T (1)2. TOTA
LCALORIB(0)3, BF CALORIB(0
) In this example, the objective function to be fitted is energy cost, and 1 is set in parentheses in this column via the keyboard 20. The energy cost calculation program is then called up, and the operator inputs the necessary information according to the guidance message, and then instructs execution to start calculation.

When updating the simplex, it is necessary to determine the initial value
x) The minimum value may not be easily found. In extreme cases, the values normally employed in blast furnace operation may already be outside the allowable range 10 in Figure 3. In order for the initial value to be accurately within the range and to quickly and reliably converge to the minimum value, it is preferable to calculate the initial value as close to the minimum value as possible by calculating with coarse grid points. Therefore, in the present invention, after the upper and lower limits are set by displaying the table as shown in the table above, coarse grid points are created by displaying as shown below.

＊＊＊ Grass carp
MB...(31TB...(31N2...(510)
2-・-(1101L, R・-+1) CO^L, R-
(1) COG, R-・-+11 where MB is blast furnace air humidity, TB is blast furnace air temperature, and these range from 9 to 4.
Indicates that 0.1100 to 1280 is divided into three. N2
is the amount of blown N2, 02 is the amount of 02 blown, OIL is the blown fuel ratio, and these are 5, 1.1...
...Indicates division. Here, a variable whose division number is set to 1 means a variable that is not involved in the creation of the grid points. In this way, coarse grid points are created, and the computer calculates f (x) in this example, the consumed energy for each grid point, and displays it on the display for the operator to select, or automatically selects the smallest value as the initial value X°. . By performing such an operation, the problem of not converging to the minimum value (within the allowable range) will disappear. Next, examples of substrates will be given.

Prerequisites: All coke operation, 220kt-pig/
Moon, 2 blast furnaces (BF). The air blowing temperature is IBF 1100°C.
(1210), 28 F is 1100 (1200)
, the air humidity g/Nnf is IBF 9 (20), 2
BF is 10 (20), blown N2 amount Nffr/H is IBF
is 0 (21000), 28F is 3000 (21000
”), the energy consumption MCAL/l-pig is 3040
(3110). Note that the values in parentheses are initial values. At this initial stage, the simplex is updated, and as shown in Figure 5, the optimal value on the left side of the parentheses is obtained after around 20 searches, and 70MC
AL/t-pig could be reduced. This is 2.25
% savings.

In addition, the +8+ solid line in this figure is calories, the dotted line is cost,
(bl~(dl) The dotted line indicates IBF, and the solid line indicates 2BF.

Effects of the Invention According to the present invention, the optimum values of the blast furnace operating factors that minimize the energy consumption, namely, the blast temperature, the blast humidity, the amount of blown fuel, the amount of nitrogen blown, and the amount of blown oxygen, are determined by a nonlinear model and the amount of calculation required is small. It is easy to obtain, and if the current operating point is given as an initial point, it is possible to provide an operating index for how to aim for the optimum value from the current situation in the process of exploring penetration and cracking, and it is possible to take into account the current state of the blast furnace. The best method for operating a blast furnace can be easily determined. Furthermore, the optimal energy balance for the entire steel process can be easily determined, and if the objective function value is the energy cost for the entire process, a blast furnace operating method that minimizes the energy cost can be found in the same way. When finding a blast furnace operating method that simultaneously minimizes energy consumption and energy cost, the objective function value f is f=wf+ + (1w) f 2 (0≦W≦1) where the consumed energy is fl and the energy cost is f2.
) (11) Defined by the linear sum of the weighting coefficients W, the blast furnace operating method that minimizes f can be found in the same way. Weight W
If it approaches 1, an optimal solution that emphasizes minimizing energy consumption will be found, and if it approaches 0, an optimal solution that emphasizes minimizing energy cost will be found.

[Brief explanation of the drawing]

The 1st BM is an explanatory diagram showing the procedure for determining the consumed energy from the blast furnace operating conditions. Figure 2 is a block diagram showing the outline of the equipment configuration for determining the blast furnace operating index. Figure 3 is an explanatory diagram of the improved NM method. Figure 4 is a flowchart showing how to update the operating point, etc., and Figure 5 is a graph showing an example of test results. In the drawing! + + 4 is an example of a simplex used in the NM method. Applicant Minoru Aoyagi, patent attorney representing Nippon Steel Corporation ) Gun size n isushi

Claims (1)

[Claims] (11) A method of determining blast furnace operating conditions that minimize the energy consumption of a steel mesh process operated using self-generated energy and externally supplied energy, and operating the blast furnace using the conditions as an index, comprising: Humidity, the amount of blown fuel, the amount of blown nitrogen, the amount of blown oxygen, and others are set as variables 1 of the function f(x) that indicates the energy consumption of the entire steel process, and the independent upper and lower limit values of the variables and the Bosch gas amount, Using a limit value determined from the tuyere tip combustion temperature, estimated core temperature, and others as a constraint condition, find the minimum value of the function f (x) using the improved Nergoo-Mead method within the operable range determined by the constraint condition, A method of operating a blast furnace in a steel process, characterized in that a specific value of variable 1 that satisfies the minimum value is used as a blast furnace operation index. (2) The initial value to be set when applying the improved Nergoo-Mead method is Roughly divide the upper and lower limit ranges,
2. The method of operating a blast furnace in a steel process according to claim 1, wherein the value of the function f(x) is calculated for each lattice point, and the variable 5° is set to minimize the result.