JPH0461925A - Coal compounding plan control apparatus - Google Patents

Abstract

PURPOSE:To perform the quality planning control and coke manufacturing cost control of the whole of a coke process by calculating the optimum coal compounding ratio with respect to the request of the object alteration of coke quality, the alteration of operation and equipment restrictions and the alteration of an operation purpose from other system and an external apparatus. CONSTITUTION:In the coal compounding plan control system of equipment using coal and a coal yard, an objective value, a restricting condition, an objective function and operation result data are received from other system and an external apparatus and, further, the upper and lower limit values of a plurality of coal compounding ratios and the upper and lower limit values of the objective value are formed to be set to the restricting conditions. Further, the upper and lower limit values of a substitute coal compounding ratio are formed when no solution of the optimum calculation is generated and a plurality of the optimum calculations are executed when the solution is generated. A plurality of the optimum calculation results are transmitted to solution gathering and a preferential order is applied thereto, and the optimum solution of a coal compounding ratio is determined from the above-mentioned means. As a result, the quality planning control and coke manufacturing cost control of the whole of a coke process can be carried out.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] This invention relates to a coal blending planning control system for equipment using coal and a coal yard.

[Conventional technology]

The coke production process at a steelworks consists of two steps: receiving coal transported from a coal transport ship to a designated location in the yard, and discharging coal of a designated brand from the yard based on a production plan. The process consists of a blending process in which the discharged coal is pulverized and blended, a coke oven process in which the blended raw materials are turned into coke, and a transport process in which the coke product is sent to the next process such as a blast furnace. Among the above steps, the coal blending step estimates coke quality from the components of each coal brand,
A coal blend ratio planning control system that determines the blend ratio of each coal brand to achieve coke quality that meets the required standards for blast furnaces is essential.

For this purpose, for example, ``Operations Research '86-June issue 330-335'' has ``On the application of mixed integer programming (MIP) in coal blend planning'' (
A mathematical programming method based on mixed integer programming is described in the literature entitled 1).
CAMP-ISIJ VOL, 2 (1989)-1
012J is titled "Development of Coke Quality Control System" (Reference 2), and includes coke quality targets (DI Nidrum Index, ash content, etc.), quality constraints (hot strength, sulfur content),
Guides the formulation that satisfies the formulation constraints using goal programming,
A coke control system has been proposed that is highly responsive to changes in the target DI value simply by adjusting the raw material mixture.
Fujitsu Journal '90-Nα222-28 "Application of Expert System to Coking Coal Blending Planning Work at Hishima Steel Works, Kawasaki Steel Corporation" (Reference 3) A method of planning) has been proposed.

In the conventional operation method, a monthly plan was created using a method similar to the mixed integer programming method described in the above-mentioned document 1, and a blending plan within the framework of the monthly plan was implemented using a process computer.

For this reason, it takes several hours to obtain the desired coke quality from the start of the blending calculation, and the generation of defective coke during this time is unavoidable. Additionally, although there was a need for planned control of the coal blending ratio, it was not practicable.

[Problem to be solved by the invention]

However, the above-mentioned mixture planning method has the following problems.

For example, the mixed integer programming method of Reference 1 places emphasis on the supply and demand balance, and has not been able to satisfy the coke quality required by blast furnaces with high calculation accuracy. Further, in the mathematical programming method based on the goal programming method of Document 2, there is no guarantee that the required quality will always be output. Further, in Document 3, it is not possible to calculate the optimal solution to the objective function (for example, the production cost Min of the coke process).

In addition, a high-speed calculation processing method (processing time of 10 minutes to 2
0 minutes) is also an issue.

Therefore, the present invention aims to solve the above problems.

[Means to solve the problem]

In order to achieve the above objectives, the coal blending planning control system of the present invention transfers quality target values, constraints, objective functions, operational performance data, etc. from other systems and external devices during the coke manufacturing process. (ii) Means for generating coal blending ratio general constraints; (iii) Means for generating coke quality conditions; (iv) Means for generating alternative blending ratios; (V) Objective function, quality constraints, blending. means for generating a goal program from ratio constraints; (vi) means for performing optimal calculations according to the goal program and transferring the results to a solution set; (vj) means for selecting an optimal coal blending ratio solution from the solution set; (vi) coal (ix) means for selecting an alternative and recalculating when the optimum coal blending ratio solution does not exist; and (ix) means for transmitting the optimum coal blending ratio solution to other systems for setting control and monitoring.

[Effect]

The present invention uses other systems and external devices to calculate the optimum coal blending ratio in response to changes in coke quality targets, requests for changes in operational/equipment constraints, and changes in operational objectives, thereby controlling the quality design of the entire coke process and managing coke manufacturing costs. Let's do it.

〔Example〕

Hereinafter, the present invention will be specifically explained based on Examples.

FIG. 1 is a block diagram showing an example of a coke production system to which the present invention is applied.

As shown in Figure 1, this system uses each A of (A) to (E).
It consists of a process controller, an electric PLC (programmable logic controller), and an instrumentation DDC (direct digital controller) as a lower mechanism of the I (artificial intelligence or expert) system. As for the division of functions, the AI system performs planning and quality control between processes, the process computer performs setting control, performance collection, and upper level transmission, and the electrical PLC/instrumentation DDC performs target process control and performance management.

For blending planning, long-term planning is performed by the host central computer, and short-term planning and control is performed by the total system of this AI system, process combi coater, and instrumentation DDC. The frequency of control operations is, for example, 1 to 2 times/day.

The system according to the present invention differs from general mathematical programming or inference systems in that (1) macro qualitative inference is divided into inference mechanism and micro calculation processing is divided into 9 mathematical needle drawing methods, and both are integrated. The aim is to reduce processing time by adopting a system configuration that is (ii) The drawbacks of the conventional coal mix planning method using mathematical programming were;
The calculation under the given calculation conditions has no solution.Not Fou
ndS is abbreviated as NF. ), it is possible to perform optimal calculations within the given calculation conditions instead of stopping calculations or interrupting subsequent calculations due to abnormal processing, and as an inverse problem, how to determine the mixture ratio within predetermined quality constraints. In other words, it has a calculation mechanism that waits for new input conditions to the inference a structure or an external device (for example, a CRT).

The coal blend calculation items used in this example include quality items: (1) coal ash content (cOAL ASH), (ii) coal volatile matter (cOAL VM), and (iii) coal maximum fluidity (
cOAL MF), (iv) Coal extensibility (cOAL D
M), (V) Coal strong caking ratio, (■1) Coke ash content (OKεAS) l), (vii) Coke oxygen-containing ash content (
c[lKE ASII-O), (v, ii): l-x rotation strength (cOKE D+), (IX) strength after coke reaction (OKE C3R), (X) coke reaction rate (
cOKE CRI), (xl) coke total sulfur content (cOKE TS).

Coke manufacturing cost items include (i) coal material cost, (
(iii) Conversion of generated COG cost, (iii) Conversion of generated tar cost, (IV) Molded coal production cost, and (V) Total cost of coke production.

In addition to the above quality items and coke production cost items, the objective function is to (i) minimize the difference from the monthly plan, (ii) minimize each quality target value, and (ii) from the target blending ratio of each coal. is the minimization of

The last item is the brand of coal to be used.

Next, the structure of the calculation formula for each item above is (i) The quality item is Ql = (Σtai − Ri + (error term)) ÷
(Σ+bi Ri+ (error term)) or mouth 1=(Σ
+al −R++(error term)), (ii) Cost item is OF, =(Σ+ci・Ri+(error term))÷(Σ,
di −Ri+(error term)) or 0F1=ΣIC!・
R1 + (error term), (ii) The minimum difference between the objective function and the monthly plan is (minimum difference from the monthly plan) = Σ, wi ・IRi”
-Ri l or ΣIw1・(Rlo-Ri)". However, Q represents quality item I, and 1
F1 represents cost item i, ai, bi, ci,
di, wi are constant terms or terms that do not include (Ri), R1 is the blending ratio of coal brand 1, R10 is the blending ratio of the monthly plan, and (error term) is the difference between the current model formula and the analytical measurement value. be.

Next, symbols used in this embodiment will be explained.

RO= (Rri; Lri≦Rri≦Uri, ri=
1~rn) Mixing ratio general constraint C= (cci; ci=1xcn) Mixing calculation item Rj = (Rrj; Lrj≦Rri≦tlrj,
rj = 1~rn) CJ blending ratio constraint Q = (Qqi; qi = 1~qn) Ranking of quality constraints Qj = (Qqi; Lj, qi ≦Q
qi ≦Roj, Qi, Qi=1~qn) Quality constraint formula % formula % () Due to the quality constraint formula 〇a, alternative mixture ratio AR (Qabc...) when ``NF without mixture ratio solution'' J Quality constraint formula Qa, Q
b, Qc... is the cause Alternative mixture ratio during MF = (OFfl; fi = 1 ~ qn) Objective function family OFj = (OFj; fJ = 1 ~ qm) Subset of objective function family (qm≦qn) D(i, j, k) = (Qi, Rj, 0Fk
) Quality constraint Qi, blending ratio constraint RJ.

Optimal calculation conditions D(i, j, k) (AR(Qa)) = (Qi, R
j, OFk, AR(Qa)) Quality constraint Q1. Polymerization ratio constraint RJ.

Under the condition of the objective function OFk, it becomes NF due to the quality constraint Qa, and recalculate with the mixture ratio of AR (Qa) D(i, j, k) (AR(Qabc...))
= (Qi, Rj, OFk, AR (Qabc-)) Quality constraint Qi, Mixing ratio constraint RJ.

Quality under the condition of objective function OFk Due to the constraint Qa, it becomes NF and AR When recalculating with the blending ratio of (Qa) is also NF, and hereafter Qb, Qc.

. . . Next, the present embodiment will be explained according to the above symbols and the flowchart of FIG. 2.

Step 1: Coal blending plan External specification input processing From other systems (e.g. coke quality stabilization guidance and process computer in Figure 1), coke/coal quality action items and amounts, action date/time or timing, molten coal ratio, softening pitch usage ratio, furnace wall temperature, furnace temperature, holding time, quality constraint priority, correction term, coke/coal analysis value,
The current operational performance values, the next candidate brand after use of coal brand i, the upper and lower limits of the blending ratio that coal brand i must strictly adhere to, etc. are transferred or transmitted. Outer tribe! (For example, CR, T) indicates the future conditions of operation or equipment, such as unusable blending tanks,
Transfer the objective function items, operational policy (for example, minimizing fluctuations in the current blending ratio or allowing significant changes to the current blending ratio with strict adherence to quality), etc.

Using these data, the sensitivity analysis method uses the objective function,
Applies to quality and cost items. The above-mentioned transfer or transmission data and sensitivity analysis data are transferred to the inference section.

Step 2: Coal blending ratio general constraint processing This process is performed by inference to generate general upper and lower limits of the coal blending ratio to be used in the following steps. The structure consists of blending tank, brand, coal yard and elapsed usage time, coal yard and shipboard plan information, other systems or external devices, and coal special information constraints.

(i) Blending tank constraints are constraints on the blending tank.

As an example, thermal coal per tank is ｜Xa%, MINb%
The special charcoal is designated as a' and b'.

(ii) Brand constraints are constraints for each brand.

As an example, steam coal is rated at Xc%, special coal is set at a', and brands that have no experience in use are set at MAX d%.

(ii) Coal yard and usage elapsed time constraints are constraints resulting from coal yard information and its usage history in the coke oven. As an example, when the yard inventory is el) or more, the current ratio is 02%, and the number of inventory maintenance days is 03 or more, MAX is set to f% and MIN is set to g%.

(iv) Constraints from the coal yard and shipping plan are constraints that are limited by the current ratio, shipping mortar, current storage capacity, and monthly plan. As an example, if the yard inventory is over h1 ton,
If the current ratio is h2%, the inventory maintenance period is h3 days, and the large-sized mortar of this brand is within h4 days, it will be Xf'% and MINg'%.

(V) If there are constraints from other systems or external devices, use that value more strongly than this determined value.

(vi) Coal special information constraints search internal pattern information to see if the coal blending ratio obtained in (i) to (V) above deviates significantly from past usage experience, and makes corrections or issues a warning. Output both results to an external device. The reason is that because coal is an organic substance, it does not easily fit the quality estimation formula described above, and there is a risk of abnormal quality.

This inference result is defined as a coal blending ratio general constraint RO and output to an external device. The output value of this result is the following formula RO= (Rr
i; Lri≦Rri≦Uri, ri=1~rn)
(1) Equation A-h above are constants set by an external device.

Each compounding ratio defined by formula (1) does not exceed each compounding ratio used in the following steps. The following formula holds true.

Regarding the blending ratio Ri of coal brand l, Li, [IiεROLi', tli' ERj (
j=1) L1≦Li′≦R1≦U1′≦Ll i
(2) Equation step 3: Coke quality condition generation process This process uses inference to order or weight the quality items used in the goal program described below.
Next, a tolerance range for the target value of the quality item is generated. The case of ordering will be explained below (weighting can also be done by a similar operation).

This process consists of quality priorities and quality constraints.

(i) The quality priority order is determined according to the designation if specified by an external device or other system, and is uniquely determined by the internal configuration if there is no particular designation. For example, if the action items are quality a, a', a', the quality priority is C5R,
D1. Coke ASH... priority order.

The output value of this result is as follows.

0= (Qqi; qi =1~qn)
(3) Equation (u) Quality constraint conditions provide upper and lower limits that are permissible for input targets from an external device or other system, and are stipulated for each quality item. As an example, if the quality item is coke ASH, the target ASH is equal to or higher than the analysis value, and when there is no action amount, the lower limit value is the target value, and the upper limit value is the target value + b. The output value of this result is as follows.

Qj= (QQi; Lj, qi ≦Loq1≦Uj, q
i, Qi=1~qn) (4) Formula, however, ``=1 and j
=2.3.4 are patterns Nα of loosening the blending ratio, deleting brands, and adding brands, respectively.

Step 4: Alternative blending ratio generation process In this process, brands are grouped based on the coal organization chart, calculation items are classified, and blending ratios are generated according to the classification items.

(1) The classification of calculation items is that calculation condition C1 is the combination calculation plan specified by formula (1), calculation condition C2 is the combination calculation plan generated from the combination calculation result specified by formula (1), and calculation condition C3 is the combination calculation plan specified by formula (1). Mixture calculation plans generated within grouped groups,
Calculation condition C4 is a combination calculation plan generated between groups, and calculation condition C5 is a combination calculation plan generated without group constraints. When expressed as a logical formula, it becomes equation (5).

Calculation item classification C= (cci; ci=l~5)
Equation (5) To explain in more detail, calculation condition C2 is calculated by using equations (1) and (4) as constraint conditions, and calculating the difference between the current mixing ratio or the monthly planned mixing ratio (this is set as Ri), that is, the function Σ1IRi” −Ri l as Min, W
Mixing ratio range for ax (Rmin(i))≦R1≦Rm
There is a large variation in the mixture calculation results within ax (i) ), and Rmax (i) - R + n1n (i)>
0.5, Li=Rmin(i), [Ji=R
Let min(i)+0.5.

Calculation condition C3 uses sensitivity analysis data and has a sensitivity of quality Q1 within each group! Search for Jax and Min brands and find the difference DQi (i) (where j is group Nα)
Calculate. DQI(j)...D for each quality Q1
on('') satisfies a certain condition (constant written in the frame).
i(jl), DQi°12), ..., DQi(
jk) (i=1 to n). When the action is higher than the current one, the Max stock of DQi (jl) (
1) When there is no brand that satisfies this condition, the current blending ratio is applied to generate a coal blending ratio constraint. In this way, one set of coal blending ratio constraints is generated.

Calculation condition C4 generates one set of coal blending ratio constraints by performing the same operations as in calculation condition C3 between groups.

Calculation condition C5 performs the same operation as calculation condition C3 except for removing the group constraint to generate 1m of coal blending ratio constraint.

When these are collectively expressed as a logical formula, the formula (6) is obtained.

Coal blending ratio constraint R corresponding to Cj = (Rrj; Lrj≦Rri≦l1rj, rj
=1-rn) (6) Formula (ii) Calculation item C3
, C4, and C5 are calculated using the mixture plan of formula (6), and an alternative mixture plan is generated as a countermeasure against NF (Not Found) without a mixture ratio solution. If the method has 3 conditions, when NF occurs due to the quality constraint formula ○a, the above D
Qa (j 1), DQa (j2), ...+,
From DQa (jk), the mixture ratio of formula (1) is newly applied to the Max brand (used Min brand) of DQa (Jl). This is called a mixing ratio renormalization operation.

The same applies to C3, C4, and C5 below. When these are collectively expressed as a logical expression, the formula (7) is obtained.

^R(Qa) = (^Rqj(Qa) ;Lqj (
Qa)≦ARqj (Qa)≦IJqj (Qa),
qj=1~qn) (7) Below 2, similarly, quality constraint expressions Qa, Qb, Qc...
generates an alternative blending ratio ^R (Qabc...) at the time of NF.

Step 5: Generation of goal program objective function family O
From F= (OFfi; fi-1~qn), objective function OFj to be calculated = (OFj; fj = Q1~q++
Select 1). Under the conditions of quality constraint Q1, mixture ratio constraint Rj1 and objective numerical number OFk, NF becomes AR due to quality constraint Ωa.
The blending ratio of (Qa) results in NF even when recalculating, and below Qb
, Qc, ・.

Generate a goal program with a tree structure that transitions to... When expressed as a logical formula, it becomes equation (8).

D(i, j, k)(AR(Qabc”))=(mouth i,
Rj, OFk, ^R(QabC-)) (8) Formula Step 6; Mixing ratio calculation and analysis process D(i, j, k) for each Ci (AR(Qabc...))
=(Qi, Rj.

OFk AR(Qabc...)), and if an optimal solution exists during the calculation, transfer the calculation result of C1 to the solution set and move on to calculation of C1+1. When CI does not have an optimal solution (
i) Qa is the quality constraint Qa, b... that is NF.

b... Semi-optimal calculation as EOF and transfer to solution set (il)
Transfer the inverse optimal calculation analysis result of the coal blending ratio to the solution set (the inverse optimal calculation problem is the objective function Σ+lRt”R1
Step 7: Evaluation/verification process of alternatives (i) Compare each solution from the solution set and prioritize.

When there is a solution and it is a one-time process, the corresponding objective function is prioritized (for example, the total cost).If it is Min, the highest priority is assigned in the order of the lowest total cost. ).

When using multiple objectives, calculate the norm between each objective function and prioritize in descending order of norm.

(u) Select the solution with the highest priority (defined as the optimal coal blending ratio solution).

Step 8; Alternative plan modification process (i) Output to an external device, and when the optimum coal blending ratio solution is OK through operator intervention, it is transmitted to other systems for setting control.

(The inference and calculation are stopped.) (il) If the coal blending ratio optimum solution is not OK, the upper and lower blending limit values are corrected and recalculation is performed from step 2.

Step 9: Alternative plan selection process If the optimum coal blending ratio solution is not OK due to operator intervention in step 8, press the item selection button, ie, blending ratio relaxation, brand deletion, brand addition, from the external device and proceed to step 2. Here, relaxing the mix ratio means expanding the upper and lower limits of the mix plan defined in the above steps, deleting stocks means deleting stocks that are subject to calculation, and adding stocks means that stocks that are not subject to calculation this time. Brands that exist in the yard at the current or control start date.

Here, when you press the item selection button, that is, relaxation of mixture ratio, deletion of brand, and addition of brand, to proceed to step 2, the information added after step 2 is the calculation result of the inverse optimal calculation problem for the given mixture plan and the achievement. Information on deviations from quality targets that cannot be achieved, and in the case of additional brands, information on the additional brands is added.

FIG. 3 shows an example of the output of the mixture plan calculation results. This searched for a combination of brands that would increase coke ASH by 0.02 from the current level and have a total cost that was 50 yen/ton of coke lower than the current level, and although the quality was satisfied in all respects, the overall cost was unsuccessful (20 yen/ton). There is no cheaper combination)
Setting control is being executed.

The above is the present embodiment, and it is easy to completely automate the present embodiment. That is, in step 8, if the coal blending ratio optimum solution completely satisfies the given planning conditions, it is transmitted to another system, and if not, the process moves to step 9.

In step 9, the terms of easing the blending ratio, deleting brands, and adding brands are moved to step 2 in the order of easing blending ratios, deleting brands, and adding brands. In this way, complete automation is possible.

The calculation time of this embodiment is about 10 minutes from step 1 to step 8 on a computer with an MIPS value (number of standard instruction executions per second, unit: 1 million) of 0.9.

In order to increase the processing speed, in computers with existing RETE-based inference mechanisms, the bottleneck is the transfer time from the file to the frame of the inference mechanism. ing. Of course, if this bottleneck is eliminated and data is immediately transferred to the inference mechanism after the optimal calculation, the calculation time can be further reduced. Furthermore, since the optimal calculation has an inherent problem of the amount of calculation, a mixing ratio renormalization operation is performed to avoid unnecessarily increasing the calculation variables.

The mathematical programming method of this example uses multi-objective bounded variable linear programming to shorten calculation time, and in this system, 1
The number of calculations for linear programming is at least 48 times, and usually 6 times.
0 to 120 The same level of calculation is performed, and the processing time required for mathematical programming is approximately 5 minutes.

Although the present embodiment concerns coal blending planning control for a coke oven, it is not particularly limited to coke ovens, and can be easily applied to, for example, a thermal power plant that uses coal and has a coal yard.

〔Effect of the invention〕

As explained above, the present invention has the following effects.

This system is not a system with a small degree of freedom based on strict causal laws based on mathematical programming, but a structure with a certain degree of freedom to automatically derive the optimal solution for the coal blend ratio with good calculation accuracy that can meet the demands of other systems. Contribute to solving real control problems.

Setting and controlling operations that conventionally took several hours can be done in about 10 minutes, which not only saves labor but also shortens the time during which defective coke is generated.

Variations in quality can be reduced by improving calculation accuracy and setting control frequency.

If the overall cost Min is selected as the objective function, the operation will always be at the cost Min, and even if the operating environment changes to another objective, this system can be followed flexibly.

If the system allows manual intervention, it will be easier to plan as intended by the operator, and even if planning fails, there will be a change in the amount of Min deviation from the quality target value, or conversely, if the coal blend planning range is out of range, there will be a replacement. A plan is output, and the optimal solution intended by the operator can usually be obtained in at least two cycles.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing an example of a coke production system to which the present invention is applied, Fig. 2 is a flowchart showing the procedure of coal blending process planning according to the present invention, and Fig. 3 is a blending plan CR.
It is a T output output water diagram. Patent applicant Nippon Steel Corporation

Claims (1)

[Claims] 1. In a coal blend planning control system for equipment that uses coal and a coal yard, (a) means for receiving target values, constraints, objective functions, operational performance data, etc. from other systems and external devices; , (b) means for generating a plurality of upper and lower limits of coal blending ratios, (c
) Means for generating upper and lower limits of the target value and using them as constraint conditions; (d) Means for generating upper and lower limits of alternative coal blending when no solution occurs in the optimal calculation; (e) Means for multiple optimal calculations without no solution occurring. (f) means for transferring a plurality of optimal calculation results to a solution set and attaching a priority to each calculation result; and (g) above (f).
A coal blending planning control system comprising means for determining an optimal coal blending ratio solution.