JP2005092827A - Scheduling system and program for making computer perform scheduling - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a scheduling system capable of forming a schedule minimizing energy consumption while satisfying manufacturing conditions such as a due date. <P>SOLUTION: This scheduling system is provided with an energy information analysis means 31 finding information matching energy consumption rates for respective processes including a procedure between processes as to a plurality of processes for each item, a process information storage means 32 storing process information for each item, and a scheduler forming a schedule 33 including energy demand forecast information from the information acquired by the energy information analysis means and the information stored in the process information storage means according to processing matching the item and the schedule condition. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a scheduling system that can be preferably used for, for example, a production facility or an inspection facility that performs a plurality of processes using a plurality of facilities, and a program for causing a computer to execute scheduling, and particularly relates to reduction of energy consumption. It is.

  As a conventional scheduling system, for example, in a production plan creation method (for example, see Patent Document 1) in which energy efficiency is reduced and energy efficiency is improved, or in a sheet winding processing facility, yield (yield) and operation rate are set. Based on the data on the raw rolls stored in large quantities in the mill roll storage and the order data collected over a certain period for the purpose of improving the productivity of raw materials and energy by improving the amount of loss of the belt-like sheet A method for producing a production plan with a computer (see, for example, Patent Document 2) has been proposed.

JP 2001-184114 A (5th page, FIG. 1) JP-A-5-318388 (5th page, FIG. 1)

  In the invention described in Patent Document 1, although the energy cost is lowered, there is a problem that the amount of energy used is not always lowered and the delivery date is not always satisfied. In addition, the invention of Patent Document 2 has a problem that extra stock costs occur because a large amount of raw material is stored.

  The present invention has been made to solve the above-described problems of the prior art, and a scheduling system and a scheduling capable of creating a schedule that minimizes the amount of energy used while satisfying various production conditions such as delivery time. An object of the present invention is to provide a program for causing a computer to execute.

  The scheduling system according to the present invention includes an energy information analyzing means for obtaining information according to an energy intensity for each process including a setup between processes for a plurality of processes by product type, and process information for storing process information by product type A schedule including energy demand prediction information is generated from the information obtained by the energy information analysis means and the information stored in the process information storage means according to the storage means and processing and schedule conditions according to the product type. And a scheduler.

  According to the present invention, there is an effect that it is possible to provide a scheduling system capable of creating a schedule that minimizes the amount of energy used while satisfying various production conditions such as delivery date. In addition, when energy demand prediction information is considered at the time of scheduling, it is possible to assign to a desired energy demand time zone, and it is possible to prevent an excess of the power contract amount during operation.

Embodiment 1 FIG.
FIG. 1 to FIG. 5 explain the outline of the scheduling system according to Embodiment 1 of the present invention. FIG. 1 is an overall configuration diagram schematically showing the relationship between general processes and the scheduling system, and FIG. FIG. 3 is a diagram showing an example of an energy load (demand) pattern obtained by the scheduler, and FIG. 4 is a diagram of the energy generation plan obtained by the energy generation plan planning function of the scheduler. FIG. 5 is a diagram showing an example of an in-house (in-house power generation) operation plan, and FIG. In the drawings, the same reference numerals indicate the same or corresponding parts.

  As shown in the drawing, the production / processing / inspection line 1 is used to process a workpiece A such as one or a plurality of raw materials or semi-finished products into a processed product B, or to perform measurement or inspection. Equipment 11, 12, 13,..., 1 n are provided, and processing of process 1 to process n is performed by each of these equipment 11 to 1 n. Before and after each step, there are setups 1, setups 2,..., Setups n for performing some operation such as moving, taking out, placing, placing, setting, waiting, or heat dissipation, respectively. Yes. The line 1 is provided with general process control means 2 for controlling each process.

  The scheduling system 3 includes an energy information analysis means 31 that obtains information according to an energy intensity for each process including a setup between processes for a plurality of processes by product type, and a process information storage that stores process information by product type The information obtained by the energy information analyzing means 31 and the information stored in the process information storage means 32 according to the means 32 and the processing and schedule conditions according to the product type, for example, the process basic unit, the process step Scheduler 33 for generating a plurality of production schedules using energy demand prediction information as an index from other information, production conditions (input plan, cost information), and other constraint conditions (delivery date, factory model, etc.), and a display that is an interpersonal interface Such as the information output means 34, the process conditions, and the keyboard used for inputting the delivery date. And a information input means 35 and energy information analyzer 31 information storing means 36 for storing information necessary for preparation and scheduling other than the information is analyzed and the process information storage means 32, the. In addition, this scheduling system 3 can be comprised using a normal general computer including a personal computer etc., for example.

  The energy information analyzing means 31 detects process information such as the type, process, equipment used, processing time, temperature, set temperature, processing amount, etc. Production progress information collecting means 31a that collects setup information such as time, dead time, and loss amount, and equipment energy consumption measuring means 31b that collects data obtained by measuring energy consumed by the equipment (electric power and heat) per unit time And an analysis means 31c for obtaining the energy intensity by type / process / equipment from the data measured by the production progress information collecting means 31a and the equipment energy consumption measuring means 31b, and for each kind / process / equipment. Energy intensity and setup energy consumption can be obtained.

  The above setup information is obtained from the measured equipment consumption energy and process information. However, the setup information refers to the energy consumption, setup time, and dead time (until the process is stabilized). Time), and the amount of lost material lost until the process is stabilized. In addition, here, “setup” including the setup change is called. These pieces of information are composed of measurement data (including probabilities).

  Furthermore, “energy consumption” and “energy usage” are basically synonymous, but for convenience of explanation in this document, the energy measured by the facility may be particularly referred to as energy consumption. The energy consumption includes the measured energy consumption and the energy amount totaled based on the measured energy consumption. In addition to the primary energy such as oil, coal, gas, hydrogen, sunlight, black liquor, wind, tide, wave power, etc., the above energy can be used, such as compressed air, heat such as hot / cold water, electric power, etc. Includes everything.

  Further, 41 is energy demand prediction information obtained by the scheduler 33, and 42 is energy for generating an energy supply plan based on the energy demand prediction information 41 and information such as a primary energy unit price, a heat unit price, a compressed air unit price, and a power unit price. The supply planning means 43 includes energy recovered from the line 1 by the energy supply planning means 42, such as recovered energy such as compressed air, heat and electric power, various primary energy such as oil and gas, and commercial power. Energy supply means for supplying at least one to the line 1.

  Next, an outline of the operation of the first embodiment configured as described above will be described with reference to the flowchart of FIG. 2 constituting the computer program. Production progress information is collected by the production progress information collecting means 31a of the energy information analyzing means 31 (step S1), and the equipment (process) and the energy consumption for each setup are measured by the equipment energy consumption measuring means 31b (step S2). ) Based on the obtained data, the analysis unit 31c obtains the energy intensity by type, process, and facility, and the amount of setup energy consumption (step S3). The various information obtained is stored in the process information storage means 32, and the database is updated sequentially.

  Based on the energy intensity by type / process / equipment determined in step S3 and the amount of setup energy consumption, the scheduler 33 sets scheduling conditions (step S3b), scheduling (step S4), schedule plan, and energy demand forecast information. By repeating the process of storing (step S5), determining whether to end the scheduling (step S6), and changing the scheduling condition (step S7), the minimum delivery delay, the minimum operation time, the minimum energy use (consumption) amount, Create multiple schedules that optimize scheduling goals such as minimum energy cost, minimum loss, and minimum power excess.

  Next, a plurality of schedules (drafts) obtained as described above are displayed on the information output means 34 such as a display, and the information can be replaced by the information input means 35 such as a keyboard and a mouse. Let the plan be selected. For the selected schedule alternative, for example, the energy demand prediction information shown in FIG. 3 is displayed on the information output means 34 such as a display by the scheduler 33. Further, in step S8, the scheduler 33 creates an in-house operation plan shown in FIG. 4, for example, as an energy supply plan by the energy supply plan planning function, and an energy trading plan shown in FIG. Is calculated. In addition, A, B, C, and D shown in FIG. 4 show the generator 1, the generator 2, the generator 3, and the generator 4, respectively.

  Next, in step S9, by the multi-purpose evaluation function of the schedule alternative of the scheduler 33, for example, the relationship between the energy usage amount and the delivery delay time, the relationship between the operation time and the delivery delay time, and the relationship between the operation time and the energy usage amount. A plurality of solutions are displayed on the information output means 34 such as a display, and multipurpose evaluation is performed by the operator.

  Based on the evaluation in step S9, it is determined in step S10 whether or not the scheduling including the energy supply plan is to be terminated. If it is not determined that the process is terminated, the scheduling condition is changed in step S6, and the process is returned to the scheduling means in step S4. If it is determined that the process is terminated, the series of processes is terminated.

In addition, as a response | compatibility in the said step S6, for example,
(1) If there is a lot that cannot meet the delivery date, the change is made to advance the time for inserting the lot.
(2) When the delivery date can be observed and there is a margin in the delivery date, a change is made to delay the lot insertion time.
(3) If there is a time zone that exceeds the contracted power consumption, a change is newly made to insert a schedule of the pseudo maintenance work for the facility in that time zone.
(4) Change the schedule conditions by combining the conditions of the highly evaluated schedule plan.
It is possible to make appropriate alternatives as appropriate.

  In the above description, in order to facilitate understanding of the invention, the process has been described with an example in which the process is simplified as shown in FIG. 1, but the present invention is not limited to this example. , When joining again after branching, when supplying other materials on the way, when things that were flowing from the beginning in several steps are joined in the middle, and even when there are multiple equipment of the same or different processes, etc. The scheduling system of the present invention can be used for any process. The private power generation device may be, for example, a hot water generator using waste heat, and the surplus energy utilization device is an optional function and may be a device that does not have this function. Further, for example, equipment that does not include a processing step such as an air compressor and an air conditioner, and an office building can be added to the target of energy consumption, or one or more factories as a whole can be targeted.

In addition to the above,
Calculate energy demand forecast information from the production schedule and input the energy demand forecast information load pattern to the energy supply planning means to formulate an optimum operation plan.
・ Calculate energy costs from the planning results.
・ Calculate product costs.
・ Present the evaluation results of multiple schedules with the evaluation index as the opposite axis.
Such functions may be added. In this case, the degree of relative evaluation of the schedule can be grasped. Furthermore, the importance of each evaluation index and the priority order of alternatives can be determined by paired comparative evaluation. Then, the result may be used for the next evaluation to automate the evaluation.

As described above, according to the first embodiment,
(1) By obtaining a schedule that minimizes the amount of energy used for setup, the overall amount of energy used can be minimized, and the amount of energy used can be reduced as compared with the prior art.
(2) By obtaining a schedule that minimizes the amount of material loss during setup, the material cost can be reduced as compared with the prior art.
(3) By considering the energy demand forecast information, which is the forecast information of how much energy is required in which time zone, at the time of scheduling, it becomes possible to allocate to the desired energy demand time zone. It is possible to prevent the power contract amount from being exceeded.
(4) The energy cost can be minimized by formulating an optimum operation plan of the energy supply system that minimizes the energy cost using the energy demand prediction information as input information to the energy supply planning means.
(5) In addition to energy trading plans that consider selling energy to the outside, it is possible to obtain not only energy but also factory operation schedules, energy supply system supply plans, and energy trading plans that maximize overall profits. it can.
(6) In addition to minimizing energy consumption, multiple schedule proposals can minimize delays in delivery, minimize operating time, minimize setup time, minimize material loss, prevent excess contract power, and profit A schedule that takes into account multiple scheduling objectives is obtained by presenting on the display an evaluation window with multiple scheduling evaluation indices such as maximization as the opposite axis and allowing humans to select the schedule plan that best suits the scheduling objective. be able to.
Effects such as can be obtained.

Embodiment 2. FIG.
6 to 23 are explanatory diagrams when the scheduling system according to the second embodiment is used in a test process for testing a semiconductor in, for example, a semiconductor test apparatus. FIG. 6 shows the relationship between the process time and the setup time. FIG. 7 is an explanatory diagram showing the influence of jig constraints, FIG. 8 is an explanatory diagram showing the relationship between setup time and energy usage, and FIG. 9 is a configuration diagram showing a specific example of a scheduler in the scheduling system 10 is a flowchart showing an outline of the flow of the scheduling system, FIG. 11 is a flowchart showing details of steps G6 and G7 in the flowchart of FIG. 10, and FIG. 12 is a multipurpose evaluation function of a schedule alternative of the scheduler. Fig. 12 (a) shows the relationship between delivery delay time and energy consumption, and Fig. 12 (b) shows operating time and delivery date. Relationship between the time that, FIG. 12 (c) is a diagram, Figure 26 illustrates the operation according to the scheduling system both the invention from FIG. 13 shows the relationship between operating time and energy consumption.

  In the test process of the semiconductor test apparatus, for example, a test is performed as to whether the product functions normally at a temperature (for example, high temperature, normal temperature, or low temperature) set in each test process. Each process is performed by connecting a change kit corresponding to a jig and a test board to a tester corresponding to an apparatus not shown. Hereinafter, elements to be considered when the present invention is applied will be briefly described.

a. Relationship between process time and setup time If the process is different, setup change time may occur. This time varies greatly depending on the set temperature. For example, in FIG. 6, “(a) when flowing from a low process set temperature” is heated by a heater, so it can be changed in a relatively short time, but conversely, “(b) high process set temperature. When it is flowed from, it takes a long time due to natural cooling.

b. On the other hand, as described above, the process process is performed by installing a change kit and a test board (jig) in a tester (device). However, “(a) No jig restriction in FIG. For example, if there are restrictions on the jig, such as a change kit and test board corresponding to the process being used elsewhere, even if the tester is vacant, "(b) Jig" in FIG. As shown in “When there is a restriction”, a jig waiting time may occur and the process may not be processed.

c. Relationship between energy consumption and operating time Also, from the viewpoint of energy consumption, changing the temperature to a higher one requires more energy than changing to a lower one. Therefore, there is a trade-off between minimizing operating hours and minimizing energy usage. For example, as shown in the measurement result of FIG. 8, on the basis of the heating pattern (energy consumption 64) of the apparatus 1, in the apparatus 2, “(a) When there is much heating setup” (no jig waiting time) , "(B) When heating setup is low" (with jig waiting time), when (a) Heating setup is large in apparatus 2, the operation time is short because there is no jig restriction, but the amount of energy used It can be seen that (b) the amount of energy used 28 when the heating setup is small is smaller.

d. Relationship between minimizing setup changes and observing delivery schedules Furthermore, setup changes require setup time and extra energy, so it is necessary to reduce setup changes as much as possible. However, since there is a product that cannot meet the delivery date unless the setup change is made, there is a trade-off relationship between minimizing the setup change (minimizing the operation time) and observing the delivery date (minimizing the delivery date delay).

  The scheduler 33 in the scheduling system of the present invention sets a new scheduling condition by combining scheduling conditions of a plurality of highly evaluated schedules, and repeatedly generates a schedule corresponding to the new scheduling condition. Search for the optimal schedule. As an example, a method for generating a schedule by simulation will be described, but the method for generating a schedule is not limited to simulation. The simulation conditions used in the following means the same as the scheduling conditions.

Next, the outline of the algorithm in the scheduler 33 will be described more specifically with reference to FIGS. 9, 10 and 11. FIG.
Based on the energy usage information 361 stored in the process information storage means 32 and the information storage means 36, the simulator production line information setting means 110 sets the simulator production line information 362 necessary for the production line simulation and reads it into the simulator 331. (Step G0). Next, the simulation condition setting means 334 sets simulation conditions such as the number of initial candidate groups for the schedule, evaluation indices, device priority rules, ranking rules, lot insertion date and time, and setup information (step G1). The schedule solution candidate indicates a schedule plan that is a simulation result.

  Next, a simulation execution (step G2) process is performed by the simulator 331, and after the completion of step G2, a schedule plan evaluation process (step G3) is performed by the schedule evaluation unit 336, and the schedule 364 and simulation are performed by the simulation condition and result storage unit 333. The storage process (step G4) of the condition 365 and the schedule evaluation value 366 is performed.

  Next, the number of initial solution candidate groups in the schedule set in step G1 is compared with the number of solution candidates in the schedule stored in step G4 (step G5). If the number of solution candidates in the stored schedule is small, simulation is performed. For example, the condition change means 335 changes the simulation conditions related to the input date and time so as to satisfy the delivery date by the input date and time setting means 335c (step G6), returns to the simulation execution step G2, and the number of stored schedule proposals is the initial solution candidate group of the schedule Steps G6, G2, G3, and G4 are repeatedly executed until the number becomes equal.

  Next, a solution candidate of a new schedule is generated by combining the solution candidates of the initial solution candidate group of the schedule. Specifically, the evaluation index selection setting unit 335a and the schedule solution candidate selection setting unit 335b select solution candidates for a plurality of schedules that are likely to improve the schedule from the solution group of schedules, and set the input date and time. The simulation condition is changed by combining the solution candidate conditions of the schedule selected by the means 335c, the priority rule setting means 335d, and the ranking rule setting means 335e (step G7), and the simulation is executed (step G8). Generate solution candidates.

Next, the schedule evaluation means 336 evaluates the solution candidates for the schedule (step G9). If a solution candidate for a schedule that has the same evaluation as the solution candidate for the saved schedule is generated, the existing solution is saved to save the storage capacity. The solution is replaced with a solution candidate for a schedule having the same evaluation value. In other cases, a schedule plan as a simulation result is stored in the simulation result storage means 410 (step G10), and the number of solution candidates for the schedule is increased. The type of schedule solution candidates means the number of different schedules existing in a stored schedule solution candidate group.

  Next, the stored solution candidates for the schedule are evaluated, and it is determined whether or not the scheduling optimization is finished by checking whether there exists a solution solution for the schedule that satisfies the scheduling purpose (step G11). If a solution candidate for a schedule that satisfies the above is obtained, the process is terminated. If a schedule solution candidate that satisfies the scheduling objective is not obtained, a scheduling time is set in advance until a schedule solution candidate that satisfies the scheduling objective is obtained or until a person decides to end the scheduling. The process returns to step G7 and steps G7, G8, G9, G10, and G11 are repeated until the scheduled end time is reached.

Next, step G6 for changing the simulation conditions will be described with reference to FIG.
First, the loading date / time of the lot is set by the loading date / time setting means 335c (step G6-1). Specifically, the delivery time margin D _i of each lot _i is calculated from the previous simulation result by the equation (1).
D _i = DD _i -CT _i (1)
Here, DD _i and CT _i represent the delivery date of lot i and the completion date and time of the final process, respectively.

When D _i > = 0, that is, when there is a margin in the delivery time, the insertion date / time TT _i (t) of the lot i is updated based on the formula (2) (the insertion date / time is delayed by 1 / γ of the margin).
TT _i (t + 1) = TT _i (t) + D _i / γ (2)
Here, t represents the generation of simulation (t-th simulation). γ is a number of 1 or more (for example, 2).

When D _i <0, that is, when the delivery date is delayed, the input date / time TT _i (t) of the lot i is updated based on the expression (3) (the input date / time is advanced by the delay).
TT _i (t + 1) = TT _i (t) + D _i (3)
The determination of whether to set these input dates / times for all lots or for the largest one is made before the simulation is executed.

  Next, the priority rule setting means 335d randomly selects and determines priority rules from those prepared in advance (step G6-2). Then, the ranking rule setting means 335e randomly selects and determines a ranking rule for the device from those required in advance (step G6-3).

Next, step G7 for changing the simulation condition by combining a plurality of schedule solution candidate conditions will be described with reference to the flowchart of FIG.
Arbitrary one is selected from the apparatus priority rules, ranking rules, lot insertion date and time, etc., which are the setting items of the simulation conditions (step G7-1).
In addition, as the priority rule of the device, the first-in first-out priority rule, the minimum setup change time rule, the same setup priority rule, the setup energy usage minimum priority rule, and the ranking rule are ranked in ascending order of input date and time, current time Ranking from the shortest time to delivery date / time, ranking from the earliest delivery date / time, ranking from the fewest number of remaining processes, ranking from the smallest remaining process processing time, ranking from the fewest processes in the subsequent process, ranking in priority order, process There are rankings in the order of shorter processing time, rankings in the order of longer processing time, and the like, but are not limited thereto.
One evaluation index j is randomly selected for each setting item by the evaluation index selection setting means 335a (step G7-2).

A schedule solution candidate i is randomly selected from the schedule solutions stored by the schedule solution candidate selection setting means 335b, and an expected value E _{i, j} regarding the evaluation index j of the schedule solution candidate i is calculated based on the equation (4). Calculate (step G7-3).
E _{i, j} = 1 / (1 + V _j −V _{i, j} ) (4)
Here, V _{i, j} represents an evaluation value related to the evaluation index j of the solution candidate for the schedule, and V _j represents the maximum value of the evaluation index j.

Next, a value r from 0 to 1 as a selection criterion is randomly selected (step G7-4) and compared with E _{i, j} (G7-5).
If E _{i, j} > r, the setting value of the solution candidate for the schedule is set as the setting value of the solution candidate for the new schedule, and the process proceeds to the next step. In other cases, Steps G7-2, G7-3, G7-4, and G7-5 are repeated.
It is checked whether all the setting items are set (G7-6), and if all the setting items are set, the process is terminated. Otherwise, the process returns to Step G7-1.

  In addition, as an evaluation index of this example, (1) delivery delay number, (2) delivery delay time, (3) operation period, (4) energy consumption, (5) contract power excess number, (6) contract power Although the excess amount, (7) the amount of loss of setup replacement, and (8) the manufacturing cost were used, it is not limited to these. For example, a part of the evaluation index may be changed, another index may be added, or the index may be reduced.

<Evaluation of solution candidates>
The scheduling problem handled in the present invention is a multi-objective optimization problem that minimizes the evaluation index, and there are a plurality of Pareto optimal solutions (a Pareto optimal solution cannot improve a plurality of evaluation indexes simultaneously. Is a feasible solution). In the evaluation of the scheduling system according to this embodiment, the user freely selects two evaluation indexes and maps the solution candidates to these two-dimensional coordinates.

  For example, FIG. 12 shows a result of mapping solution candidates by selecting two evaluation axes from three evaluation indexes of energy usage (FIG. 12a), delivery delay time (FIG. 12b), and operation period (FIG. 12c). From these results, the solution A is the optimal solution for energy consumption, the solution B is for the delivery delay time, and the solution C is for the operation period, and there is no solution that simultaneously improves these evaluation indices. I understand. Also, it can be seen from these figures that the initial solution has been improved by searching. In this example, only three evaluation indexes are compared for easy understanding, but the number of evaluation indexes is not limited.

  Next, which of these is selected as an execution solution depends on the actual schedule as shown in FIG. 13 by the evaluator (in this example, the three lots m304, m305, and m202 are delayed in delivery). It must be judged together.

<Scheduling example>
Hereinafter, the energy saving production scheduling used in the second embodiment will be described more specifically.
Production system model The scale of the production system used in the second embodiment is as follows.
-Number of varieties: 3 varieties.
-Input amount: 20 lots.
-Number of process flows: One process flow for each product type. A total of three process flows.
-Number of processes: 2 varieties have 2 processes. One product has 3 processes.
-Number of devices: 4 units.
-Number of jigs: 3 units.
-Process processing time: The processing time of each process is 2 hours.
-Setup change time: 0 to 190 minutes.

Incidentally, the energy consumption CE _p step p, stage energy SE _p required for the replacement, for example, equipment such as shown in FIG. 14 (a) of (device 1, 2, 3) for each energy consumption measurement results at that time From the process information, the energy intensity is measured for each apparatus and process as shown in FIG.

  In addition, as shown in FIGS. 15 to 17, the production system model used in the second embodiment includes an apparatus model (FIG. 15A), a jig model (FIG. 15B), and a process flow model. (FIG. 15 (c)), which is composed of six models: a product model (FIG. 15 (d)) showing the relationship between product types and process flow, an input plan model (FIG. 16), and a setup matrix model (FIG. 17) The

(I) Device Model As shown in FIG. 15A, the device model describes the relationship among device groups, devices, priority rules, and ranking rules. Note that the ranking rule designates what index is used to order the priority order when a lot is selected.
(Ii) Jig Model As shown in FIG. 15B, the jig model describes the relationship between the jig group and the jig.
(Iii) Process Flow Model As shown in FIG. 15C, in the process flow model, the process flow, process, device group, processing time, processing time unit, processing unit, setup, set temperature, jig group, alternative Describe the relationship between processes, energy consumption, and energy consumption units.
(Iv) Product Type Model As shown in FIG. 15 (d), the product model describes the relationship between the product type and the process flow.
(V) Input Plan Model As shown in FIG. 16, the input plan model describes the relationship between the order name, lot name, product type, input date / time, delivery date / time, and processing amount of one lot.
(Vi) Setup Matrix Model As shown in FIG. 17, the setup matrix model is a unit of the time required for the setup to change from the current setup to the new next setup, its energy usage, and energy usage. Is described.

Solution Result Distribution of Scheduling Result (A) Search Result FIG. 18 shows the distribution of solution candidates when energy usage (X axis) and operation time (Y axis) are selected as evaluation scales. In this evaluation scale, it can be seen that the solution candidates 18, 101, 141, and 199 do not have any other solution candidates that simultaneously improve the operation time and the energy usage. The numbers in the figure represent numbers assigned in the order of generation of solution candidates. FIG. 19 shows the distribution of solutions when the delivery delay number (X axis) and the contract power consumption excess number (Y axis) are selected as the solution evaluation scale. It can be seen that the solution candidate 199 is optimal for these two evaluation measures.

  When an arbitrary number of solution candidates is selected, the scheduler according to the present invention, as shown in FIG. 20, energy consumption, delivery delay (total time), delivery delay (number), operation time, contract power overage The number of times (times) and the total number of contracted electric power excess (kWh) are displayed, and the Gantt chart display function, energy demand forecast information display function, and solution candidate deletion function buttons displayed on the display screen are clicked with a pointing device, for example. Thus, specific charts can be displayed.

  Thus, it may be possible to select an optimal solution by evaluating solution candidates from various evaluation scales. However, when the delivery date delay is not important for the evaluator and the evaluation index of the energy consumption is regarded as important, the solution candidate 141 is selected as the optimal solution.

(B) Schedule Diagram The schedule diagrams (Gantt charts) generated for each solution candidate 18, 101, 141, 199 are shown in FIGS. 21, 22, 23, and 24, respectively. In the figure, th and PH represent a device and a jig, respectively.
These figures also show that the operation time of the solution candidate 199 in FIG. 24 is the shortest.
(C) Power load pattern In FIG. 25 and FIG. 26, the first day (September 14, 2002) and the second day (September 15, 2002) of the power demand prediction information (30-minute unit) of the solution candidate 199 are shown. Each is shown. In these drawings, the horizontal axis represents date and time, and the vertical axis represents power demand (load) (kWh).
It can be seen from the Gantt chart of FIG. 24 and the power demand prediction information of FIGS. 25 and 26 that the power demand rises at the time of setup change (setup change that raises the set temperature) with a short setup time.
(D) Energy Supply Plan From the power demand prediction information of FIGS. 25 and 26, or the required power amount and the required heat amount that are the energy demands of the factory shown in FIG. 3 of the first embodiment, FIG. A self-running operation plan as shown is made in the same manner as in the first embodiment, and an energy trading plan result as shown in FIG. 5 is obtained. Thereby, the optimal energy cost can be predicted.

  FIG. 14 is a plot of the relationship between the production amount measured for each facility and the power consumption at that time based on the measurement data. From this, it is possible to calculate the amount of power used per unit (basic unit). FIG. 14B is a diagram in which the basic unit for each process of each product is predicted for each facility from the measurement result of FIG. 14A and the production progress information indicating which product is processed at that time.

As described above, according to the second embodiment,
(1) A new scheduling method combining a scheduling method based on simulation and a scheduling method based on evolutionary computation can efficiently obtain an optimal schedule for a production line having complicated constraints that has been difficult in the past.
(2) A schedule for minimizing the energy usage can be obtained by the priority rule for minimizing the energy used for the setup change.
(3) Comparing the completion date and time of delivery of each lot after scheduling and changing the input date and time of each lot at the next scheduling based on the difference, thereby making a schedule that minimizes the delivery time delay (deviation) Can be obtained.
(4) A schedule for optimizing a plurality of purposes (evaluation indices) by storing a scheduling plan (solution candidate) generated in the scheduling process and generating a new schedule by combining scheduling conditions of schedules with good evaluation indices Can be obtained efficiently.
Effects such as can be obtained.

Embodiment 3 FIG.
27 to 32 illustrate the scheduling system according to the third embodiment. FIG. 27 is an explanatory diagram showing an energy demand prediction method based on a probability density function of energy usage, and FIG. 28 is the presence or absence of a priori information. FIG. 29 is a diagram illustrating the relationship between the energy density prediction function and the probability density function, FIG. 29 is a diagram illustrating the energy demand prediction information when there is no prior information, and the probability that the actual result will be less than that, and FIG. FIG. 31 is an explanatory diagram showing overall energy demand prediction information from individual energy demand prediction information, and FIG. 32 is an explanatory diagram showing another example of the demand prediction method. It is.

  In the third embodiment, the uncertainty (energy supply plan blur) of the energy generation plan (energy supply plan) is quantified based on the probability density function obtained from the measurement result of the prediction accuracy of the energy demand. The plan error can be considered in advance, and the error of the energy consumption predicted by the scheduler is reduced using a priori information. The a priori information means information such as a product name, a product name, a process name, an equipment name, a process condition of a process, a process condition of a process processed immediately before with the same equipment, a season, an air temperature, and a process start time. Moreover, the process conditions of a process mean conditions, such as temperature set to the equipment which processes a process. In addition, the configuration of the scheduling system is the same as in the first and second embodiments.

  Note that FIG. 27 shows the energy consumption of each process, setup, idle, stop, etc. in step S3 of FIG. 2 from the energy information analysis means 31 and process information storage means 32 in FIG. 1 described in the first embodiment. This corresponds to a description of another example of means for obtaining the quantity and outputting the energy demand prediction information (41 in FIG. 1) at the time of scheduling (step S5 in FIG. 2). 28 to 31 are diagrams for supplementary explanation of FIG.

Next, a case where energy demand information is predicted based on a probability density function will be described with reference to FIGS.
Each equipment that is data on when (from what time to what time), what (what kind of process), where (with which equipment), and how (under what conditions) the equipment (P1) Operation history information data (P2a) and time series information data (P2b) of energy consumption such as heat, power, temperature, pressure, compressed air, etc. at each facility are stored in the energy consumption database (P3).

  By analyzing the energy consumption event and the corresponding energy consumption from the operation result information data and the time series information data of the energy consumption, for example, the energy for each type / process / equipment as shown in FIG. A probability density function using the consumption amount as a random variable, a probability density function using the energy consumption amount for each setup as a random variable, and a probability density function using the idle energy consumption amount of each facility as a random variable are created (P5).

  For example, the graph (P10) on the right side of FIG. 27 is an example of P5, and represents a probability density function using the basic unit (energy consumption per unit) of the type A process 1 as a random variable. The solid line is the probability density, and the broken line is the cumulative probability. For example, in FIG. 26, the probability that the basic unit is a is 10%, the probability that b is 35%, and the probability that c is 4%. The probability that the basic unit is a or less is 10% (r1), the probability that it is b or less is 70% (r2), and the probability that it is c or less is 96% (r3).

  These probability density functions specify conditions relating to seasons such as summer, spring / autumn, winter, etc., which are examples of a priori information as illustrated in FIG. Each probability density function (conditional probability) in accordance with the specified conditions can be created by specifying conditions such as the temperature-related conditions and the immediately preceding manufacturing conditions. The more the conditions, the smaller the variance of the probability density function and the higher the energy demand prediction probability.

  From the probability density function created by the above procedure and the planned production schedule (P4), calculate the energy usage by specifying the cumulative probability, and multiple proposals for the predicted energy demand when the production schedule is implemented The energy demand forecast information (demand scenario) of each plan is created (P6).

  For example, the energy usage (time) when processing the product A as shown in FIG. 28 using each energy usage (a in the figure) when the cumulative probabilities are all “optimistic” r1 (for example, 10%). F1 representing 8-12), F2 representing the amount of energy used when processing product B (time 12-18), and F3 representing the amount of energy used when processing product C (time 18-23) The energy demand forecast information F5 at the time 8-23 is obtained by adding each setup energy F4 on the time axis, and this is the demand forecast information when the most optimistic (when the amount of energy used is small) (see FIG. 29 (4)).

  Note that b shown in FIG. 29 indicates the energy usage in the case of “normal” r2 (for example, 70%), and c indicates the energy usage in the case of “worst” r3 (for example, 96%). Corresponds to the sign.

  Further, when obtaining the energy demand prediction information, by specifying the a priori information (FIG. 30), the variation (error) in the energy demand prediction information is reduced as compared with the case where there is no a priori information (FIG. 29). be able to. For example, the difference between (1) the probability r3 * r3 * r3 that is less than or equal to the prediction in FIG. 30 and (4) the probability r1 * r1 * r1 that is less than or equal to the prediction, and (1) and (4) in FIG. When the difference is compared, it can be seen that the difference between (1) and (4) in FIG. 30 is smaller. * Represents the product (x) (the same applies hereinafter).

  On the other hand, the probability that the energy demand falls within these energy demand prediction information can be obtained by the product of the cumulative probabilities r. For example, the probability that the energy demand falls within the demand prediction information of (1) in FIGS. 30 and 29 is r3 * r3 * r3.

  As shown in FIG. 31, by adding the demand prediction information of each facility on the time axis, it is possible to obtain demand prediction information for the entire facility (“(4) Overall energy demand” in FIG. 31). In addition, when the energy consumption amount for every equipment cannot be measured, the equipment may be calculated by replacing the production line. In the above example, the energy usage is described as a continuous variable. However, a discrete variable obtained by segmenting the energy usage may be used.

Furthermore, in the above example, the case of predicting energy demand has been explained. However, the amount of material loss at the time of setup and the time to stable operation (loss of operation time) are estimated by the probability density function (error). May be quantified. This demand forecast information is input to the energy supply plan P11 shown in FIG. 27 (this corresponds to the energy supply plan means 42 in FIG. 1 and step S8 in FIG. 2), and the optimum energy use amount and energy Costs, CO ₂ (carbon dioxide) emissions, etc. are output.

  Next, the case where the demand prediction information is predicted based on the learning model will be described with reference to FIG. First, with respect to the facility (J1), the operating state of the facility (J2a), temperature, heat (pressure, temperature, flow rate), electric energy, compressed air consumption, and the future operating state of the facility (J2b) Based on the data stored in the database (J3) and based on the data in this database, the future from the operating state of the equipment (J2a), temperature, heat (pressure, temperature, flow rate), electric energy, compressed air consumption and the future operating state of the equipment A facility model (J5) for predicting the heat amount (pressure, temperature, flow rate), power amount, and compressed air consumption of the facility is learned (J4). For example, the operating state of the equipment includes, for example, the process 1 of the product A, the setup from the process 1 of the product A to the process 1 of the product B, the idling, the stop, and the temperature condition α of the equipment.

  Next, input the operating state (J2a), temperature, heat (pressure, temperature, flow rate), electric energy, compressed air consumption and the operating state (J2b) of the future equipment to the learned equipment model (J5). Then, predicted values of the heat amount (pressure, temperature, flow rate), power amount, and compressed air consumption of the future equipment are output (J6). In J6, the solid line represents the actual value of energy use up to time t, and the broken line represents the predicted value after t (t + 1, t + 2).

  The amount of steam, power, and equipment at time t are s (t), p (t), a (t), and the predicted values of steam and power are so (t) and po (t), respectively. . As shown in FIG. 32, s (t), s (t-1), s (t-2), s (tk), p (t), p (t-1), p (t-2) ), P (t−k), a (t + j), a (t + 2), a (t + 1), a (t), a (t−1), a (t−2), a (t−k) , By inputting to the equipment learning model J7 (same as J5) at time t, the predicted values of energy usage at times t + 1 and t + 2 so (t + j), so (t + 2), so (t + 1), po (t + j), po (t + 2) and po (t + 1) are output. However, 2 <k <K, 2 <j <J. Note that K indicates a time-series range to be input, and J indicates a time-series range to be predicted.

  As a learning method, a learning model of a nonlinear system such as a neural network, a GMDH (general method of data handling), or an immune network which is a known technique can be used.

Also, the absolute value of prediction error for feedback learning,
| S (t + 1) −so (t + 1) |,
| S (t + 2) -so (t + 2) |,
| P (t + 1) -po (t + 1) | and | p (t + 2) -po (t + 2) |
May be added to each input.

  As described above, the facility learning model is configured, and the demand for energy is expressed without explicitly expressing the probability density model described above by sequentially updating t as t = t + 1 based on the manufacturing schedule (P4 in FIG. 27). Predictive information can also be obtained.

As described above, according to the third embodiment,
(1) When the probability density model is used, there is an effect that the variation in the energy demand prediction information and the probability at that time can be grasped in advance. Therefore, there is an effect that scheduling can be performed in consideration of the risk of deviating prediction and cost merit.
(2) Since the probability density model is sequentially updated with the accumulated data, there is an effect that the prediction accuracy is gradually increased.
(3) By classifying the probability density function in detail using a priori information, there is an effect that the prediction accuracy can be improved.
(4) By using the learning model, energy demand prediction information can be obtained without explicitly expressing the probability density model. In addition, by using this learning model for model predictive control of the energy supply system, it is possible to adjust a deviation generated between the plan (schedule) and the actual operation.
Effects such as can be obtained.

  By the way, in the description of the above embodiment, the energy information analyzing means 31 detects the type of product flowing in each process in order to calculate the energy (basic unit) by product type and process (setup), and processing The case of having a function of measuring (detecting) time, processing amount, energy consumption (demand / load) / energy requirement information has been described. For example, equipment used, temperature, set temperature, and generated energy You may add the function to measure. If there is generated energy, it can be used as recovered energy, which can be further used to save energy.

  In addition, the scheduler 33 not only creates a production schedule and alternatives as energy data obtained by the energy information analysis means 31, but also calculates, for example, energy demand prediction information, energy management, energy cost calculation, product It is desirable to add functions such as cost calculation and presenting the evaluation results of a plurality of schedules with the evaluation index as an opposite axis as options.

  In addition, energy supply devices (means) to supply factory processing equipment, production lines, offices, etc. are equipment that procure energy from outside, including power companies, gas companies, oil companies, etc., private power generators, heat generators , Including a driving force / compressed air generator, a heat converter, and the like. Furthermore, processing is not limited to mechanical processing, for example, assembly of parts, measurement / inspection, painting, heating / cooling, chemical reactions such as decomposition / synthesis / condensation / plating / dissolution, printing, bookbinding, etc. Those that consume energy can be included.

FIG. 3 is an overall configuration diagram schematically showing a relationship between a general process and a scheduling system according to the first embodiment. FIG. 3 is a flowchart schematically showing a flow of the scheduling system according to the first embodiment. The figure which shows the example of the energy load (demand) pattern obtained by the scheduler of Embodiment 1. FIG. The figure which shows the self-operation plan as an example of the energy generation plan obtained by the energy generation plan planning function of the scheduler of Embodiment 1. The figure which shows the energy trading plan of Embodiment 1. FIG. Explanatory drawing which shows the relationship between the process time in the scheduling system by Embodiment 2, and a setup time. Explanatory drawing which shows the influence of the jig | tool restrictions in Embodiment 2. FIG. Explanatory drawing which shows the relationship between the setup time and energy consumption in Embodiment 2. FIG. FIG. 6 is a configuration diagram showing a specific example of a scheduler in a scheduling system according to a second embodiment. The flowchart which shows the outline | summary of the flow of the scheduling system shown in FIG. The flowchart which shows the detail of step G6 in the flowchart of FIG. 10, and step G7. Obtained by the multi-purpose evaluation function of the scheduler schedule alternative in the second embodiment, (a) relationship between delivery delay time and energy consumption, (b) relationship between operation time and delivery delay time, and (c) The figure which shows the relationship between an operation period and energy consumption. FIG. 10 is a diagram showing a part of a schedule diagram of a solution 2 in the second embodiment. FIG. 10 is a diagram illustrating an example of a basic unit prediction result obtained from a measurement result in the second embodiment. The figure which shows the example of (a) apparatus model, (b) jig model, (c) process flow model, (d) kind model as a model used for Embodiment 2. FIG. The figure which shows the example of an input plan model as a model used for Embodiment 2. FIG. FIG. 6 is a diagram illustrating a setup matrix model as a model used in the second embodiment. The figure which illustrates distribution 1 of the solution candidate in Embodiment 2. The figure which illustrates distribution 2 of the solution candidate in Embodiment 2. The figure which shows the example of the evaluation screen of the solution in Embodiment 2. FIG. The Gantt chart figure calculated | required about the solution candidate 18 in Embodiment 2. FIG. The Gantt chart figure calculated | required about the solution candidate 101 in Embodiment 2. FIG. The Gantt chart figure calculated | required about the solution candidate 141 in Embodiment 2. FIG. The Gantt chart figure calculated | required about the solution candidate 199 in Embodiment 2. FIG. The figure which shows the energy demand prediction information (1st day) calculated | required about the solution candidate 199 in Embodiment 2. FIG. The figure which shows the energy demand prediction information (2nd day) calculated | required about the solution candidate 199 in Embodiment 2. FIG. Explanatory drawing which shows the energy demand prediction method based on the probability density function of the energy usage in the scheduling system by Embodiment 3. FIG. Explanatory drawing which shows the relationship between the presence or absence of a priori information and probability density function in Embodiment 3. FIG. Explanatory drawing which shows the probability that energy demand prediction information and results when there is no prior information in Embodiment 3 will be less than that. Explanatory drawing which shows the probability that energy demand prediction information in the case where there is a priori information in Embodiment 3, and a performance will become less than it. Explanatory drawing which shows the whole energy demand prediction information from the individual energy demand prediction information in Embodiment 3. FIG. Explanatory drawing which shows the other example of the demand prediction system in Embodiment 3. FIG.

Explanation of symbols

1 line, 11, 12, 13, ..., 1n equipment, 2 process control means, 3 scheduling system, 31 energy information analysis means, 31a production progress information collection means, 31b equipment energy consumption measuring means, 31c analysis means, 32 process information storage means, 33 scheduler, 34 information output means, 35 information input means, 36 information storage means, 41 energy demand prediction information, 42 energy supply planning means, 43 energy supply means, 331 simulator, 332 simulator production line information Setting means, 333 simulation condition and result storage means, 334 simulation condition setting means, 335 simulation condition changing means, 335a evaluation index selection setting means, 335b schedule solution candidate selection setting means, 335c input Setting means, 335d priority rule setting means, 335e ranking rules setting means, 336 schedule evaluation means, 361 energy use information, 362 simulator production line information of the, 363 simulation conditions, 364 schedule, 365 simulation conditions, 366 schedule evaluation value.

Claims (13)

  Energy information analysis means for obtaining information corresponding to the energy intensity of each process including the setup between processes for multiple processes by product type, process information storage means for storing process information by product type, and corresponding to the product type And a scheduler for generating a schedule including energy demand prediction information from the information obtained by the energy information analysis unit and the information stored in the process information storage unit according to the processing and schedule conditions. A featured scheduling system.   The energy information analysis means obtains process setup information including energy consumption generated during setup, setup time, dead time until process stabilization, and loss amount from the measured equipment energy consumption and process information. The scheduling system according to claim 1, wherein the scheduling system has a function.   The scheduling system according to claim 1, wherein the process includes a heating process and / or a cooling process.   The energy information analysis means includes an energy consumption database that stores energy consumption information obtained from the processing results of processes for equipment, and a function that generates an energy usage model based on the information of the energy consumption database. 4. The scheduling system according to claim 1, wherein energy demand prediction information is output from the energy use model and the production schedule.   The scheduling system according to any one of claims 1 to 4, wherein the scheduler includes a function of generating an energy supply plan.   The scheduling system according to claim 5, wherein the scheduler includes a function of generating an energy trading plan.   7. The scheduling system according to claim 1, wherein the scheduler includes a function for evaluating a schedule generated for the purpose of scheduling.   The scheduler sets a new scheduling condition by combining scheduling conditions of a plurality of highly evaluated schedules with respect to the plurality of schedules evaluated by the function for evaluating the schedule generated for the scheduling purpose. The scheduling system according to claim 7, wherein an optimum schedule is searched by repeating a procedure for newly generating a schedule corresponding to a scheduling condition.   The scheduling system according to any one of claims 1 to 8, wherein the scheduler includes a function for obtaining a relationship of a predicted delivery date with respect to an energy consumption amount as a plurality of alternatives.   The scheduling system according to claim 1, wherein the scheduler has a function of outputting a comparison result between predicted power consumption and contract power.   2. The scheduler according to claim 1, further comprising a function of quantifying the uncertainty of the energy generation plan based on the probability density obtained from the measurement result based on the a priori information, and increasing the prediction accuracy of the energy demand. Item 11. The scheduling system according to any one of Items 10.   Energy information analysis step for obtaining information corresponding to energy intensity for each process, including setup between processes for multiple processes by product type, process information storage step for storing process information by product type, and corresponding to product type Generating a schedule including energy demand prediction information from the information obtained by the energy information analysis step and the information stored by the process information storage step according to the processing and scheduling conditions performed. A program for causing a computer to execute scheduling. An evaluation step for evaluating the schedule generated for the above scheduling condition, and a plurality of schedules evaluated by this evaluation step are combined with the scheduling conditions of a plurality of highly evaluated schedules to set a new scheduling condition. 13. A condition resetting step to be set and a step of newly generating a schedule based on the condition resetting step, and searching for an optimal schedule by repeating these steps as necessary. A program that causes a computer to execute scheduling.

Images