Processing system, control device and processing method

Abstract

To provide a processing system, a controller, and a processing method capable of suppressing loads imposed on an operator in association with operation control.SOLUTION: A processing system includes a processing facility which executes processing according to a set value and a controller which calculates a set value, and the controller uses a first model for executing regression analysis to calculate a first estimated value in the processing facility upon execution of processing according to a candidate of the set value by the processing facility, and uses a second model for executing black box optimization to calculate a set value from the first estimated value.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to a processing system, a control device, and a processing method. [Background technology]

[0002] BACKGROUND ART Various systems have been proposed for automatically controlling various types of processing equipment (hereinafter also simply referred to as processing equipment) (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2009-048524 Summary of the Invention [Problem to be solved by the invention]

[0004] Here, in the processing facility, it is desirable to reduce the burden on workers involved in operation control, for example. [Means for solving the problem]

[0005] The processing system of the present invention is a processing system comprising processing equipment that performs processing according to a set value, and a control device that calculates the set value, wherein the control device uses a first model that performs regression analysis to calculate a first estimated value for the processing equipment when the processing equipment performs processing according to the candidate set value, and calculates the set value from the first estimated value by using a second model that performs black-box optimization. [Effects of the Invention]

[0006] According to the processing system of the present invention, it is possible to reduce the burden on workers involved in operation control. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a diagram illustrating a configuration of a processing system 1000 according to the first embodiment. [Figure 2] FIG. 2 is a diagram illustrating the hardware configuration of the control device 1. As shown in FIG. [Figure 3] FIG. 3 is a functional block diagram of the control device 1 in the first embodiment. [Figure 4] FIG. 4 is a diagram illustrating an outline of the estimation process according to the first embodiment. [Figure 5] FIG. 5 is a flowchart illustrating details of the estimation process according to the first embodiment. [Figure 6] FIG. 6 is a diagram illustrating details of the estimation process in the first embodiment. [Figure 7] FIG. 7 is a flowchart illustrating the estimation process according to the second embodiment. [Figure 8] FIG. 8 is a diagram illustrating the estimation process according to the second embodiment. [Figure 9] FIG. 9 is a flowchart illustrating the estimation process according to the third embodiment. [Figure 10] FIG. 10 is a functional block diagram of the control device 1 in the fourth embodiment. [Figure 11] FIG. 11 is a flowchart illustrating the learning process according to the fourth embodiment. [Figure 12] FIG. 12 is a flowchart illustrating the estimation process according to the fourth embodiment. [Figure 13] FIG. 13 is a diagram illustrating the estimation process according to the fourth embodiment. [Figure 14] FIG. 14 is a diagram illustrating the estimation process according to the fourth embodiment. [Figure 15] FIG. 15 is a diagram illustrating the configuration of the processing system 1000. As shown in FIG. [Figure 16] FIG. 16 is a diagram illustrating a specific example of a plurality of pieces of training data DP. [Figure 17]FIG. 17 is a diagram illustrating a specific example of a plurality of pieces of training data DE. [Figure 18] FIG. 18 is a diagram illustrating the configuration of a first modified example of the processing system 1000. In FIG. [Figure 19] FIG. 19 is a diagram illustrating the configuration of a second modified example of the processing system 1000. In FIG. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. However, such descriptions should not be interpreted in a limiting sense, and do not limit the subject matter described in the claims. Furthermore, various changes, substitutions, and modifications can be made without departing from the spirit and scope of the present disclosure. Furthermore, different embodiments can be combined as appropriate.

[0009] [Processing system 1000 according to the first embodiment] First, a description will be given of an example of the configuration of a processing system 1000 according to the first embodiment. Fig. 1 is a diagram illustrating the configuration of a processing system 1000 according to the first embodiment.

[0010] As shown in FIG. 1, the treatment system 1000 includes, for example, a treatment facility 100, a monitoring device 2 that monitors the treatment facility 100, and a control device 1 that controls the operation of the treatment facility 100 based on the monitoring results of the monitoring device 2. In the example shown in FIG. 1, the control device 1 and the monitoring device 2 are accessible to each other, for example, via a network NW. The following description will be given assuming that the treatment facility 100 is an incineration facility 100a that incinerates materials to be incinerated, and that the monitoring device 2 monitors the incineration status of the incineration facility 100a. Note that the following description will be given assuming that the materials to be incinerated are sludge such as sewage sludge (hereinafter simply referred to as sludge), but the incineration facility 100a may also incinerate materials to be incinerated other than sludge (for example, garbage, etc.).

[0011] The monitoring device 2 is, for example, an electronic device having an electronic circuit. Specifically, the monitoring device 2 is, for example, one or more physical machines or one or more virtual machines having a CPU (Central Processing Unit) and memory. The monitoring device 2 monitors the incineration state of sludge in the incineration facility 100a by acquiring, for example, actual values ​​of PVs (Process Variables) measured by a measuring device (not shown) installed in the incineration facility 100a, actual values ​​of SVs (Set Point Variables) set in the incineration facility 100a, and actual values ​​of MVs (Manipulated Variables) manipulated in the incineration facility 100a. Hereinafter, the actual values ​​of PVs measured by a measuring device installed in the incineration facility 100a will also be referred to as actual PVs or measured values. Also, hereinafter, the actual values ​​of SVs set in the incineration facility 100a will also be referred to as actual SVs. Furthermore, hereinafter, the actual values ​​of MVs manipulated in the incineration facility 100a will also be referred to as actual MVs.

[0012] The control device 1 is, for example, an electronic device having an electronic circuit. Specifically, the control device 1 is, for example, one or more physical machines or one or more virtual machines having a CPU and memory. The control device 1 then performs, for example, a process (hereinafter also referred to as an intervention target calculation process) to calculate an estimated value of the SV that is preferably set for the intervention target in the incineration facility 100a or an estimated value of the MV that is preferably operated for the intervention target in the incineration facility 100a. The intervention target is, for example, at least one of the multiple pieces of equipment that make up the incineration facility 100a. Hereinafter, the estimated value of the SV that is preferably set for the intervention target is also referred to as the intervention target SV or setting value. Also, the estimated value of the MV that is preferably operated for the intervention target is also referred to as the intervention target MV. Thereafter, the control device 1 controls the incineration state of sludge in the incineration facility 100a, for example, by instructing the monitoring device 2 to perform a setting corresponding to the calculated intervention target SV or an operation corresponding to the intervention target MV. The following description will be given assuming that the calculation of the intervention target SV is performed in the intervention target calculation process.

[0013] Specifically, the control device 1 in this embodiment uses, for example, a PV estimation model (hereinafter also referred to as the first model) that performs regression analysis to calculate an estimated value (hereinafter also referred to as the first estimated value) of the optimization target in the incineration equipment 100a when the incineration equipment 100a executes processing according to a candidate SV for intervention (hereinafter also simply referred to as a candidate SV for intervention).

[0014] The optimization target is, for example, the PV (actual PV) measured in the incineration facility 100a, the SV that is preferably set as the intervention target in the incineration facility 100a, the MV that is preferably operated in the intervention target in the incineration facility 100a, or a variable calculated by using at least one of these (hereinafter simply referred to as a composite variable). That is, the first estimated value is, for example, a predicted value of the PV measured in the incineration facility 100a (hereinafter also referred to as the first predicted PV), an estimated value of the SV that is preferably set as the intervention target in the incineration facility 100a (intervention target SV), an estimated value of the MV that is preferably operated in the intervention target in the incineration facility 100a (intervention target MV), or an estimated value of the composite variable. In this way, the optimization target can be, for example, any variable. The following description will be given assuming that the first estimated value is the first predicted PV.

[0015] The PV estimation model ME may use, for example, a gradient boosted tree (GDBT) as a regression analysis method. The PV estimation model ME may also use, for example, deep learning such as a long short term memory (LSTM) as a regression analysis method.

[0016] Then, the control device 1 in this embodiment calculates the intervention target SV from the first estimated value by using, for example, an inverse problem analysis model (hereinafter also referred to as a second model) that executes black-box optimization.

[0017] The inverse problem analysis model may use, for example, grid search, random search, Bayesian optimization, or the like as a black-box optimization method.

[0018] More specifically, the control device 1 in this embodiment calculates, as the intervention target SV, a new candidate for the intervention target SV whose first estimated value approaches a predetermined target value (hereinafter also simply referred to as the target value), for example.

[0019] That is, the control device 1 in this embodiment calculates an intervention target SV that can bring the first estimated value in the incineration facility 100a closer to a target value, for example, by using a PV estimation model that can calculate (estimate) a first estimated value from candidates for the intervention target SV, and an inverse problem analysis model that can calculate new candidates for the intervention target SV from the first estimated value. Then, the control device 1 controls the incineration state of the material to be incinerated in the incineration facility 100a, for example, by instructing the monitoring device 2 to set the calculated intervention target SV.

[0020] As a result, the control device 1 in this embodiment can, for example, maintain the incineration state of the materials to be incinerated in the incineration facility 100a in a normal state. Furthermore, even if the incineration state of the materials to be incinerated in the incineration facility 100a is not normal, the control device 1 can improve this. Therefore, the control device 1 can, for example, automatically control the operation of the incineration facility 100a, and reduce the burden on the workers compared to when the operation of the incineration facility 100a is manually controlled by the workers.

[0021] In addition, the monitoring device 2 may, for example, set the intervention target SV calculated by the control device 1 in the incineration equipment 100a, thereby performing pressure control (PIC control), flow rate control (FIC control), and temperature control (TIC control) of each piece of equipment included in the incineration equipment 100a, and control the incineration state of sludge in the incineration equipment 100a.

[0022] [Hardware configuration of control device 1] Next, a description will be given of the hardware configuration of the control device 1. Fig. 2 is a diagram illustrating the hardware configuration of the control device 1.

[0023] 2, the control device 1 includes, as electronic circuits, a CPU 101 which is a processor, a memory 102, a communication device 103, and a storage medium 104. Each unit is connected to one another via a bus 105.

[0024] The storage medium 104 has, for example, a program storage area (not shown) that stores a program 110 for performing the intervention target calculation process. The storage medium 104 also has, for example, a storage unit 130 (hereinafter also referred to as information storage area 130) that stores information used when performing the intervention target calculation process. The storage medium 104 may be, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

[0025] The CPU 101 performs the intervention target calculation process by executing a program 110 loaded from the storage medium 104 to the memory 102, for example.

[0026] The communication device 103 accesses the monitoring device 2, for example, via a network NW. The communication device 103 also accesses, for example, an operation terminal (not shown) operated by an operator, via the network NW. The network NW may be, for example, the Internet. That is, the control device 1 may be located on a cloud, for example. The network NW may also be, for example, a LAN (Local Area Network).

[0027] Hereinafter, the case where the control device 1 has a CPU 101 and a memory 102 will be described, but the present invention is not limited to this. Specifically, the control device 1 may have, for example, an electronic circuit (not shown) capable of executing the intervention target calculation process instead of or together with the CPU 101 and the memory 102, and the intervention target calculation process may be executed in this electronic circuit. The electronic circuit capable of executing the intervention target calculation process may be, for example, an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit).

[0028] [Functions of the control device 1 in the first embodiment] Next, the functions of the control device 1 in the first embodiment will be described. Figure 3 is a block diagram of the functions of the control device 1 in the first embodiment.

[0029] As shown in FIG. 3, the control device 1 realizes various functions including a setting estimation unit 116 by organic cooperation between hardware such as a CPU 101 and a memory 102 and a program.

[0030] 3, the control device 1 stores, for example, an actual PV 131, an actual SV 132, and an intervention target SV 133 in the information storage area 130. Furthermore, the control device 1 stores, for example, a PV estimation model ME and an inverse problem analysis model PG in the information storage area 130. Note that, for example, the actual MV may also be stored in the information storage area 130. Hereinafter, the PV estimation model ME and the inverse problem analysis model PG will be collectively referred to simply as the SV estimation model.

[0031] In the intervention target calculation process, the process of calculating the intervention target SV133 is also referred to as the estimation process. That is, the estimation process is a process of calculating the intervention target SV133 that makes it possible to improve the incineration state of the incineration facility 100a, for example, when the current incineration state of the incineration facility 100a is not in a normal state or when there is a possibility that the incineration state of the incineration facility 100a will not be in a normal state in the future.

[0032] The setting estimation unit 116 inputs, for example, the actual performance SV132 of the intervention target in the incineration facility 100a (the actual performance SV132 of the intervention target stored in the information storage area 130) into the PV estimation model ME. Furthermore, the control device 1 inputs, for example, a candidate for the intervention target SV133 (hereinafter also referred to as candidate for the intervention target SV 133a) output from the inverse problem analysis model PG into the PV estimation model ME. Then, the setting estimation unit 116 acquires, for example, a first predicted PV (hereinafter also referred to as first predicted PV131a) output from the PV estimation model ME.

[0033] As a result, for example, if the acquired first predicted PV 131a does not satisfy a predetermined condition (hereinafter also simply referred to as the predetermined condition), the setting estimation unit 116 selects (searches) a candidate SV for intervention 133a by using the inverse problem analysis model PG. The predetermined condition is, for example, a first condition or a second condition, which will be described later.

[0034] Then, the setting estimation unit 116 inputs, for example, the selected intervention target SV candidate 133a into the PV estimation model ME. After that, the setting estimation unit 116 again acquires, for example, the first predicted PV 131a output from the PV estimation model ME.

[0035] As a result, for example, if the reacquired first predicted PV131a satisfies a predetermined condition, the setting estimation unit 116 outputs the candidate intervention target SV 133a corresponding to the reacquired first predicted PV131a as the intervention target SV133 (new intervention target SV133). Specifically, the setting estimation unit 116, for example, transmits the intervention target SV133 to the monitoring device 2. Then, in this case, the monitoring device 2 controls the incineration facility 100a by using the received intervention target SV133, for example.

[0036] On the other hand, for example, if the reacquired first predicted PV 131a does not satisfy the predetermined condition, the setting estimation unit 116 selects (searches) again for the intervention target SV candidate 133a by using the reacquired first predicted PV 131a. Thereafter, the setting estimation unit 116 repeatedly performs the process of acquiring the first predicted PV 131a and the process of selecting the intervention target SV candidate 133a until, for example, the first predicted PV 131a output from the PV estimation model ME satisfies the predetermined condition.

[0037] Note that, although the following description will be given of a case where one control device 1 executes the intervention target calculation process, the present invention is not limited to this. Specifically, the functions required to execute the intervention target calculation process may be distributed among a plurality of control devices 1 (hereinafter simply referred to as a plurality of control devices 1) that can access each other via a network NW. In this case, the intervention target calculation process may be executed by, for example, the plurality of control devices 1 cooperating with each other.

[0038] [Outline of Estimation Processing in the First Embodiment] Next, an outline of the estimation process in the first embodiment will be described. Fig. 4 is a diagram for explaining an outline of the estimation process in the first embodiment. Note that the intervention target SVs 133 (candidates 133a for intervention target SVs) shown in Fig. 4 are, for example, SVs of different types.

[0039] As shown in Figure 4, the control device 1, for example, uses the PV estimation model ME to calculate the first predicted PV131a measured in the incineration equipment 100a when the incineration equipment 100a performs processing according to the intervention target SV133 (candidate SV for intervention target SV 133a).

[0040] Specifically, as shown in Fig. 4, the control device 1 inputs, for example, a candidate for intervention target SV 133a output from the inverse problem analysis model PG to the PV estimation model ME. Then, the control device 1 acquires, for example, a first predicted PV 131a output from the PV estimation model ME in conjunction with the input of the intervention target SV 133 (candidate for intervention target SV 133a).

[0041] When inputting the actual performance SV132 of the intervention target in the incineration facility 100a into the PV estimation model ME, the control device 1 may also input, for example, actual performance PV131 and actual performance MV that are highly relevant to the intervention target in the incineration facility 100a. The actual performance PV131 that is highly relevant to the intervention target is, for example, the actual performance PV131 measured in the intervention target or the actual performance PV131 measured in other equipment that is highly relevant to the intervention target. Furthermore, the actual performance MV that is highly relevant to the intervention target is, for example, the actual performance MV that is required when setting the actual performance SV132 in the intervention target or the actual performance MV that is required when setting the actual performance SV132 in other equipment that is highly relevant to the intervention target.

[0042] Similarly, when inputting the intervention target SV133 in the incineration facility 100a into the PV estimation model ME, the control device 1 may also input, for example, the actual PV131 and intervention target MV that are highly relevant to the intervention target in the incineration facility 100a. The intervention target MV that is highly relevant to the intervention target is, for example, the intervention target MV that is required when setting the intervention target SV133 in the intervention target, or the intervention target MV that is required when setting the intervention target SV133 in other equipment that is highly relevant to the intervention target.

[0043] Next, the control device 1 calculates a new candidate 133a for the intervention target SV that can bring the first predicted PV 131a closer to the target value by using, for example, the inverse problem analysis model PG, as shown in Fig. 4. That is, the control device 1 calculates, for example, a new candidate 133a for the intervention target SV that can be input to the PV estimation model ME to cause the PV estimation model ME to output the first predicted PV 131a that is closer to the target value.

[0044] Specifically, the control device 1 inputs, for example, a first predicted PV 131a output from the PV estimation model ME to the inverse problem analysis model PG. Then, the control device 1 acquires, for example, a new candidate 133a for the intervention target SV output from the inverse problem analysis model PG in response to the input of the first predicted PV 131a.

[0045] That is, the control device 1 in this embodiment calculates the intervention target SV 133 that can bring the first predicted PV 131a closer to the target value, for example, by using a PV estimation model ME that can calculate the first predicted PV 131a from the intervention target SV candidate 133a and an inverse problem analysis model PG that can calculate a new intervention target SV candidate 133a from the first predicted PV 131a. Then, the control device 1 controls the incineration state of the material to be incinerated in the incineration facility 100a, for example, by instructing the monitoring device 2 to set the calculated intervention target SV 133.

[0046] As a result, the control device 1 in this embodiment can, for example, maintain the incineration state of the materials to be incinerated in the incineration facility 100a in a normal state. Furthermore, even if the incineration state of the materials to be incinerated in the incineration facility 100a is not normal, the control device 1 can improve this. Therefore, the control device 1 can, for example, automatically control the operation of the incineration facility 100a, and reduce the burden on the workers compared to when the operation of the incineration facility 100a is manually controlled by the workers.

[0047] [Details of Estimation Processing in the First Embodiment] Next, the details of the estimation process in the first embodiment will be described. Fig. 5 is a flowchart illustrating the details of the estimation process in the first embodiment. Fig. 6 is a diagram illustrating the details of the estimation process in the first embodiment.

[0048] 5, the setting estimation unit 116 inputs, for example, the actual performance SV132 of the intervention target in the incineration facility 100a into the PV estimation model ME. Then, the setting estimation unit 116 acquires, for example, the value output from the PV estimation model ME as the first predicted PV 131a (step S101).

[0049] Specifically, the setting estimation unit 116 obtains the first predicted PV131a by inputting, for example, the actual SV132 of the intervention target in the incineration facility 100a (for example, the actual SV132 at the most recent timing) into the PV estimation model ME, as shown in "first time" in Figure 6(B).

[0050] Next, the setting estimation unit 116 selects (searches for) candidates 133a for intervention target SV in the incineration facility 100a, for example, by using the inverse problem analysis model PG (step S102).

[0051] Specifically, when performing step S102 for the first time, the setting estimation unit 116 searches for candidates 133a for the intervention target SV by, for example, inputting the first predicted PV 131a acquired in step S101 to the inverse problem analysis model PG, as shown in Fig. 6(A). Also, when performing step S102 for the second or subsequent times, the setting estimation unit 116 searches for candidates 133a for the intervention target SV by, for example, inputting the first predicted PV 131a acquired in step S103 (step S103 performed immediately before) described below to the inverse problem analysis model PG.

[0052] More specifically, for example, when the PV estimation model ME is generated by learning multiple pieces of training data DE shown in Figure 17 described later, the setting estimation unit 116 searches for candidate intervention target SVs 133a (combinations of candidate intervention target SVs 133a) for each of the "dryer sludge input amount," "dryer inlet temperature," "dryer rotation speed," and "dryer outlet temperature."

[0053] Next, the setting estimation unit 116 inputs, for example, the intervention target SV candidate 133a searched for in step S102 into the PV estimation model ME. Then, the setting estimation unit 116 acquires, for example, the first predicted PV 131a output from the PV estimation model ME (step S103).

[0054] Specifically, the setting estimation unit 116 obtains the first predicted PV 131a by, for example, inputting the candidate intervention target SV 133a searched for in step S102 into the PV estimation model ME, as shown in "second time onwards" in Figure 6(B).

[0055] More specifically, for example, when the PV estimation model ME is generated by learning multiple pieces of training data DE shown in Figure 17 described later, the setting estimation unit 116 acquires a value corresponding to the "temperature inside the incinerator" as the first predicted PV 131a.

[0056] Thereafter, the setting estimation unit 116 determines whether or not the number of times steps S102 and S103 have been executed satisfies a first condition (step S104). The first condition is, for example, that the number of times steps S102 and S103 have been executed has reached a predetermined number of times (hereinafter also referred to as the first number of times).

[0057] As a result, when it is determined that the number of times steps S102 and S103 have been executed does not satisfy the first condition (NO in step S104), the setting estimation unit 116 performs the processes from step S102 onwards, for example, again.

[0058] That is, for example, when it is determined that the number of times steps S102 and S103 are executed does not satisfy the first condition, the control device 1 determines that the search for the intervention target SV candidate 133a has not been performed sufficiently, and continues the estimation process.

[0059] On the other hand, if it is determined that the number of times steps S102 and S103 have been executed satisfies the first condition (YES in step S104), the setting estimation unit 116 ends the estimation process, for example.

[0060] That is, for example, when it is determined that the number of times steps S102 and S103 have been executed satisfies the first condition, the control device 1 determines that the search for the intervention target SV candidates 133a has been sufficiently performed, and ends the estimation process.

[0061] Thereafter, the setting estimation unit 116 selects, for example, one of the first predicted PVs 131a acquired in step S103 and identifies the candidate intervention target SV 133a used in calculating the selected first predicted PV 131a as the intervention target SV 133. Specifically, the setting estimation unit 116 selects, for example, a first predicted PV 131a that satisfies a second condition from the first predicted PVs 131a acquired in step S103. The second condition may be, for example, that the first predicted PV 131a is within a predetermined allowable range or that the difference between the first predicted PV 131a and the target value is equal to or less than a threshold. In this case, the setting estimation unit 116 then transmits, for example, the identified intervention target SV 133 to the monitoring device 2.

[0062] In step S104, the setting estimation unit 116 may determine whether the execution times of steps S102 and S103 satisfy a first condition. In this case, the first condition may be, for example, that a predetermined time (e.g., 10 minutes) has elapsed since the initial execution start timing of step S102.

[0063] [Specific example (1) of estimation process in the first embodiment] Next, a specific example of the estimation process in the first embodiment will be described.

[0064] For example, when executing step S102 for the first time, the setting estimation unit 116 starts a counter (not shown) whose initial value is 0. Then, as shown in FIG. 6(B), for example, the setting estimation unit 116 inputs the intervention target SV 133 currently set in the incineration facility 100a into the PV estimation model ME to obtain a first predicted PV 131a, and uses the obtained first predicted PV 131a as an input to the inverse problem analysis model PG (step S101 and the first step S102). Furthermore, as shown in FIG. 6(B), for example, the setting estimation unit 116 inputs the intervention target SV candidate 133a searched for in the first step S102 into the PV estimation model ME to obtain the first predicted PV 131a (first step S103).

[0065] Thereafter, the setting estimation unit 116 repeatedly executes steps S102 and S103 until the number of times indicated by the counter reaches the first number of times (NO in step S104), for example.

[0066] Here, the setting estimation unit 116 stores, for example, the candidates 133a for the intervention target SV searched for in step S102 and the first predicted PV 131a obtained in step S103 in the memory 102. That is, the setting estimation unit 116 stores, for example, a large number of candidates 133a for the intervention target SV and a large number of first predicted PVs 131a in the memory 102 by repeatedly performing steps S102 and S103.

[0067] Then, for example, when it is determined that the number indicated by the counter has reached the first number (YES in step S104), the setting estimation unit 116 selects the first predicted PV 131a that is closest to the target value from among the many first predicted PVs 131a stored in the memory 102. The target value may be, for example, the median value in the allowable range of the first predicted PVs 131a.

[0068] Thereafter, the setting estimation unit 116 transmits, for example, the candidate intervention target SV 133a used in the calculation of the selected first predicted PV 131a as the intervention target SV 133 to the monitoring device 2. Furthermore, in this case, the setting estimation unit 116 resets, for example, a counter.

[0069] [Specific example (2) of estimation process in the first embodiment] Next, another specific example of the estimation process in the first embodiment will be described.

[0070] For example, when executing step S102 for the first time, the setting estimation unit 116 starts an elapsed time timer (not shown) whose initial value is 0. Then, as shown in FIG. 6(B), for example, the setting estimation unit 116 inputs the intervention target SV 133 currently set in the incineration facility 100a into the PV estimation model ME to obtain a first predicted PV 131a, and uses the obtained first predicted PV 131a as an input to the inverse problem analysis model PG (step S101 and the first step S102). Then, as shown in FIG. 6(B), for example, the setting estimation unit 116 inputs the intervention target SV candidate 133a searched for in the first step S102 into the PV estimation model ME to obtain the first predicted PV 131a (first step S103).

[0071] Thereafter, the setting estimation unit 116 repeatedly executes steps S102 and S103 until the time indicated by the elapsed timer reaches a predetermined time (for example, 10 minutes) (NO in step S104).

[0072] Here, the setting estimation unit 116 stores in the memory 102, for example, the intervention target SV candidate 133a searched for in step S102 and the first predicted PV 131a obtained in step S103.

[0073] Then, for example, when it is determined that the time indicated by the elapsed time timer has reached a predetermined time (YES in step S104), the setting estimation unit 116 selects the first predicted PV 131a that is closest to the target value from among the many first predicted PVs 131a stored in the memory 102.

[0074] Thereafter, the setting estimation unit 116 transmits, for example, the candidate intervention target SV 133a used in calculating the selected first predicted PV 131a as the intervention target SV 133 to the monitoring device 2. Furthermore, in this case, the setting estimation unit 116 resets, for example, the elapsed time timer.

[0075] In this way, the control device 1 in this embodiment uses, for example, the PV estimation model ME to calculate the first predicted PV131a that will be measured in the incineration facility 100a when the incineration facility 100a executes processing according to the candidate intervention target SV 133a. Then, for example, the control device 1 uses, for example, the inverse problem analysis model PG to calculate, as the intervention target SV 133, a new candidate intervention target SV 133a where the first predicted PV131a approaches the target value.

[0076] Specifically, the control device 1 repeatedly performs the process of calculating the first predicted PV 131a and the process of calculating the candidate intervention target SV 133a until, for example, the number of times or the execution time of the process of calculating the first predicted PV 131a and the process of calculating the candidate intervention target SV 133a satisfy a first condition. Then, the control device 1 calculates, as the intervention target SV 133, the candidate intervention target SV 133a corresponding to the first predicted PV 131a that is closest to the target value among the repeatedly calculated first predicted PVs 131a.

[0077] More specifically, the control device 1 calculates, as the intervention target SV 133, a candidate intervention target SV 133a corresponding to the first predicted PV 131a that is closest to the target value among the repeatedly calculated first predicted PVs 131a.

[0078] That is, the control device 1 calculates the intervention target SV 133 that can bring the actual PV 131 in the incineration facility 100a closer to the target value, for example, by using a PV estimation model ME that can calculate the first predicted PV 131a from the intervention target SV candidate 133a, and an inverse problem analysis model PG that can calculate a new intervention target SV candidate 133a from the first predicted PV 131a. Then, the control device 1 controls the incineration state of the material to be incinerated in the incineration facility 100a, for example, by instructing the monitoring device 2 to set the calculated intervention target SV 133.

[0079] Specifically, the control device 1 adopts, for example, from among the candidate intervention target SVs 133a selected in the repeatedly performed step S102, the candidate intervention target SV 133a used to calculate the first predicted PV 131a that can be determined to be close to the target value as the candidate intervention target SV 133a that can maintain the incineration state in the incineration equipment 100a in a normal state (the candidate intervention target SV 133a that can improve the incineration state in the incineration equipment 100a).

[0080] This allows the control device 1 to, for example, more accurately maintain and improve the incineration state of materials to be incinerated in the incineration facility 100a.

[0081] The setting estimation unit 116 may, for example, determine the tendency of a series of first predicted PVs 131a sequentially stored in the memory 102. When the setting estimation unit 116 determines that a series of first predicted PVs 131a sequentially stored in the memory 102 are approaching a target value, the setting estimation unit 116 may select a first predicted PV 131a that can be determined to be close to the target value, or transmit (to the monitoring device 2) a candidate intervention target SV 133a corresponding to the selected first predicted PV 131a. On the other hand, when the setting estimation unit 116 determines that a series of first predicted PVs 131a sequentially stored in the memory 102 are not approaching the target value, or when a first predicted PV 131a that is not within the allowable range is stored, the setting estimation unit 116 may not select a first predicted PV 131a that can be determined to be close to the target value, or may not transmit a candidate intervention target SV 133a corresponding to the selected first predicted PV 131a.

[0082] [Estimation process in the second embodiment] Next, the estimation process in the second embodiment will be described. Fig. 7 is a flowchart illustrating the estimation process in the second embodiment. Fig. 8 is a diagram illustrating the estimation process in the second embodiment.

[0083] The estimation process in the second embodiment differs from the estimation process in the first embodiment in that, for example, the types of intervention target SVs whose values ​​are changed are narrowed down.

[0084] 7, the setting estimation unit 116 inputs the actual performance SV132 of the intervention target in the incineration facility 100a into the PV estimation model ME, for example, in the same manner as in step S101 of Fig. 5. Then, the setting estimation unit 116 acquires, for example, the value output from the PV estimation model ME as the first predicted PV131a (step S111).

[0085] Next, the setting estimation unit 116 selects (searches) candidate intervention target SVs 133a (combinations of candidate intervention target SVs 133a) in the incineration facility 100a, for example, by using the inverse problem analysis model PG, similar to step S102 in Figure 5 (step S112).

[0086] Next, the setting estimation unit 116 inputs the intervention target SV candidate 133a searched for in step S112 into the PV estimation model ME, for example, in the same manner as in step S103 of Fig. 5. Then, the setting estimation unit 116 acquires, for example, the first predicted PV 131a output from the PV estimation model ME (step S113).

[0087] Thereafter, the setting estimation unit 116 determines whether or not the number of times steps S112 and S113 are executed satisfies a third condition (step S114). The third condition is, for example, that the number of times steps S112 and S113 are executed reaches a predetermined number of times (hereinafter also referred to as a second number of times).

[0088] As a result, when it is determined that the number of times steps S112 and S113 have been executed does not satisfy the third condition (NO in step S114), the setting estimation unit 116 performs, for example, the processes from step S112 onwards again.

[0089] On the other hand, if it is determined that the number of times steps S112 and S113 are executed satisfies the third condition (YES in step S114), the setting estimation unit 116 identifies one or more types of the intervention target SV 133, for example, based on each of the first predicted PVs 131a repeatedly obtained in step S113 and each of the candidate intervention target SVs 133a repeatedly selected in step S112 (step S115).

[0090] Specifically, for example, the setting estimation unit 116 identifies one or more types that can be determined to have a large influence on the first predicted PV131a output from the PV estimation model ME from among the types of intervention target SV133 input to the PV estimation model ME in step S113. In other words, for example, the setting estimation unit 116 identifies one or more types that can be determined to have a large contribution rate to bringing the first predicted PV131a output from the PV estimation model ME closer to the target value from among the types of intervention target SV133 input to the PV estimation model ME in step S113.

[0091] That is, for example, when there are N types of intervention target SV133 (N is an integer of 1 or more), the setting estimation unit 116 identifies one or more types of intervention target SV133 that can be determined to have a large influence on the first predicted PV 131a from among the N types. In other words, in this case, the setting estimation unit 116 narrows down the types of intervention target SV133 that can be determined to have a large influence on the first predicted PV 131a, for example.

[0092] More specifically, the setting estimation unit 116, for example, by referring to each of the first predicted PVs 131a repeatedly acquired in step S113 and each of the candidate intervention target SVs 133a repeatedly selected in step S112, calculates, for each type (e.g., N types) of intervention target SVs 133, the likelihood (e.g., probability) that the first predicted PVs 131a satisfying the second condition will be output from the PV estimation model ME by changing the candidate intervention target SVs 133a corresponding to each type. That is, for example, when a combination of candidate intervention target SVs 133a in which only the candidate intervention target SVs 133a corresponding to each type is changed is input to the PV estimation model ME, the setting estimation unit 116 calculates, as the likelihood corresponding to each type, the likelihood that the first predicted PVs 131a satisfying the second condition will be output from the PV estimation model ME. Then, the setting estimation unit 116 identifies, for example, among the types of intervention target SVs 133, a type whose calculated likelihood satisfies a fifth condition. The fifth condition is, for example, that the likelihood is equal to or greater than a predetermined threshold.

[0093] For example, when the PV estimation model ME is generated by learning multiple pieces of training data DE shown in Figure 17 described below, if the accuracy corresponding to the "dryer sludge input amount" is greater than the accuracy corresponding to each of the "dryer inlet temperature," "dryer rotation speed," and "dryer outlet temperature," the setting estimation unit 116 identifies the "dryer sludge input amount" in step S115.

[0094] Returning to FIG. 7, the setting estimation unit 116 selects (searches) candidates 133a for intervention target SVs (combinations of candidates 133a for intervention target SVs) in the incineration facility 100a, for example, by using the inverse problem analysis model PG (step S116).

[0095] Specifically, the setting estimation unit 116 selects new candidates 133a for intervention target SVs by, for example, changing at least the candidates 133a for intervention target SVs corresponding to each of one or more types identified in step S115 from among the candidates 133a for intervention target SVs selected in the immediately preceding step S112 or the immediately preceding step S116.

[0096] More specifically, the setting estimation unit 116 selects a new candidate 133a for the intervention target SV by, for example, changing only the candidate 133a for the intervention target SV that corresponds to any of the one or more types identified in step S115, from among the candidate 133a for the intervention target SV selected in the immediately preceding step S112 or the immediately preceding step S116.

[0097] Next, the setting estimation unit 116 inputs the intervention target SV candidate 133a searched for in step S116 to the PV estimation model ME, for example, in the same manner as in step S113. Then, the setting estimation unit 116 acquires, for example, the first predicted PV 131a output from the PV estimation model ME (step S117).

[0098] Thereafter, the setting estimation unit 116 determines whether the number of times steps S116 and S117 have been executed satisfies a fourth condition (step S118), for example, similarly to step S114. The fourth condition is, for example, that the number of times steps S116 and S117 have been executed reaches a predetermined number of times (hereinafter also referred to as a third number of times).

[0099] As a result, when it is determined that the number of times steps S116 and S117 have been executed does not satisfy the fourth condition (NO in step S118), the setting estimation unit 116 performs the processes from step S116 onwards again, for example.

[0100] On the other hand, if it is determined that the number of times steps S116 and S117 have been executed satisfies the fourth condition (YES in step S118), the setting estimation unit 116 ends the estimation process, for example.

[0101] Thereafter, the setting estimation unit 116 selects, for example, one of the first predicted PVs 131a acquired in step S117, and identifies the candidate intervention target SV 133a used in calculating the selected first predicted PV 131a as the intervention target SV 133. Then, in this case, the setting estimation unit 116 transmits, for example, the identified intervention target SV 133 to the monitoring device 2.

[0102] In step S114, the setting estimation unit 116 may determine whether the execution time of steps S112 and S113 satisfies a third condition, for example. In this case, the third condition may be, for example, that a predetermined time (for example, five minutes) has elapsed since the initial execution start timing of step S112.

[0103] Furthermore, in step S118, the setting estimation unit 116 may determine whether the execution time of steps S116 and S117 satisfies a fourth condition, for example. In this case, the fourth condition may be, for example, that a predetermined time (for example, five minutes) has elapsed since the initial execution start timing of step S116.

[0104] Thus, the intervention target SV 133 in this embodiment is, for example, a combination of multiple types of intervention target SV 133 set in the incineration facility 100a. Then, the control device 1 calculates, for example, a first predicted PV 131a when the incineration facility 100a executes processing according to the new candidate SV 133a for the intervention target SV, by using the PV estimation model ME. Furthermore, the control device 1 calculates, for example, a new candidate SV for the intervention target SV 133a for which the first predicted PV 131a approaches a target value, by using the inverse problem analysis model PG. Thereafter, the control device 1 repeatedly performs the process of calculating the first predicted PV 131a and the process of calculating the new candidate SV for the intervention target SV 133a, until, for example, the number of executions or execution time of the process of calculating the first predicted PV 131a and the process of calculating the new candidate SV for the intervention target SV satisfy a third condition.

[0105] Next, the control device 1 identifies one or more types of the intervention target SV 133 based on, for example, each of the repeatedly calculated first predicted PVs 131a and each of the repeatedly calculated new candidates 133a for the intervention target SV. Subsequently, the control device 1 repeatedly performs the process of calculating the first predicted PV 131a and the process of calculating the new candidate 133a for the intervention target SV while calculating new candidates 133a for the intervention target SV in which the intervention target SV 133 corresponding to one or more types has been changed, for example, until the number of times that the process of calculating the first predicted PV 131a and the process of calculating the new candidate 133a for the intervention target SV are executed satisfies a fourth condition. Then, the control device 1 calculates, as the intervention target SV 133, the new candidate 133a for the intervention target SV corresponding to the first predicted PV 131a that satisfies the second condition from among the first predicted PVs 131a repeatedly calculated.

[0106] Specifically, for example, the control device 1 calculates, as the intervention target SV 133, a new candidate 133a for the intervention target SV corresponding to the first predicted PV 131a that is closest to the target value among the first predicted PVs 131a that have been repeatedly calculated.

[0107] Furthermore, the control device 1 calculates, for example, from each of the repeatedly calculated first predicted PVs 131a and each of the repeatedly calculated new candidates 133a for the intervention target SV, the probability that the first predicted PV 131a that satisfies the second condition will be calculated by changing the new candidates 133a for the intervention target SV corresponding to each type of intervention target SV 133. Then, the control device 1 identifies, for example, from the types of intervention target SV 133, types whose calculated probability satisfies the fifth condition as one or more types.

[0108] That is, the control device 1 in this embodiment, for example, repeatedly performs step S112 and step S113 to identify one or more types of intervention target SV133 that can be determined to have a large influence on the first predicted PV131a output from the PV estimation model ME. Then, the control device 1, for example, repeatedly performs step S116 and step S117 while changing the identified one or more types of intervention target SV133 to calculate a new intervention target SV133.

[0109] Specifically, in the estimation process, the control device 1 first performs a process (hereinafter also referred to as a screening process) of narrowing down the types of intervention target SV133 that can be determined to have a large influence on the first predicted PV131a output from the PV estimation model ME. Then, after the screening process, the control device 1 performs a process (hereinafter also referred to as a search process) of searching for candidates 133a for intervention target SV while changing the types of intervention target SV133 narrowed down by the screening process.

[0110] More specifically, as shown in Fig. 8, the control device 1 performs the screening process by repeatedly executing each process from step S112 to step S114 K times and then executing step S115. Thereafter, the control device 1 performs the search process by repeatedly executing each process from step S116 to step S118 M times.

[0111] This allows the control device 1 in this embodiment to efficiently search for an intervention target SV133 that can improve the incineration state of the incineration facility 100a, for example. Therefore, the control device 1 can acquire an intervention target SV133 that can improve the incineration state of the incineration facility 100a with greater accuracy. Therefore, the control device 1 can improve the incineration state of the incineration facility 100a with greater accuracy by performing control using the acquired intervention target SV133, for example.

[0112] For example, when the estimation process in this embodiment is repeatedly performed, the control device 1 may perform the search process by using the processing result of the screening process performed in the past (the processing result of step S115). That is, in this case, the control device 1 may perform the search process without performing the screening process.

[0113] Furthermore, the setting estimation unit 116 may select, for example, a search method from among increasing a value, decreasing a value, and maintaining a value (hereinafter also simply referred to as a search method) in step S116. That is, for example, when there are N types of intervention target SVs 133 and a search is performed to change only the candidate intervention target SV 133a corresponding to any one of the N types, the setting estimation unit 116 may perform processing to select one coordinate point from an N-dimensional space having N × 3 coordinate points.

[0114] Specifically, for example, the setting estimation unit 116 may calculate, for each coordinate point selected in step S116 up to the previous time, an evaluation value indicating the proximity of the first predicted PV 131a to the target value when each coordinate point is selected. Then, in step S116, for example, the setting estimation unit 116 may preferentially select, from among N×3 coordinate points in the N-dimensional space, coordinate points located in an area near coordinate points with high evaluation values.

[0115] Furthermore, the setting estimation unit 116 may, for example, refer to each of the first predicted PVs 131a repeatedly acquired in step S113 and each of the intervention target SV candidates 133a repeatedly selected in step S112, and change the intervention target SV candidate 133a corresponding to each combination for each combination of the type of intervention target SV 133 and the search method (hereinafter simply referred to as a combination) to calculate the probability that the first predicted PV 131a satisfying the first condition will be output from the PV estimation model ME. Then, the control device 1 may, for example, identify, for each type of intervention target SV 133, the search method included in the combination corresponding to the highest probability among the probabilities corresponding to combinations including each type. Thereafter, in step S116, the setting estimation unit 116 may, for example, select a new intervention target SV candidate 133a by changing the intervention target SV candidate 133a corresponding to each type using the search method identified in step S115.

[0116] [Estimation process in the third embodiment] Next, the estimation process in the third embodiment will be described with reference to a flowchart shown in FIG.

[0117] In the estimation process in the third embodiment, for example, unlike the estimation process in the first embodiment, the estimation process is continued until a first predicted PV 131a within the allowable range (a first predicted PV 131a that satisfies the second condition) is found.

[0118] 9, the setting estimation unit 116 inputs the actual performance SV132 of the intervention target in the incineration facility 100a into the PV estimation model ME, for example, in the same manner as in step S101 of Fig. 5. Then, the setting estimation unit 116 acquires, for example, the value output from the PV estimation model ME as the first predicted PV131a (step S121).

[0119] Next, the setting estimation unit 116 selects (searches for) candidates 133a for intervention target SV in the incineration facility 100a by using the inverse problem analysis model PG, for example, in the same way as in step S102 of FIG. 5 (step S122).

[0120] 5, the setting estimation unit 116 inputs the intervention target SV candidates 133a (combination of intervention target SV candidates 133a) searched for in step S122 to the PV estimation model ME. Then, the setting estimation unit 116 acquires, for example, the first predicted PV 131a output from the PV estimation model ME (step S123).

[0121] After that, the setting estimation unit 116 determines, for example, whether the first predicted PV 131a calculated in step S123 satisfies a second condition (step S124).

[0122] As a result, when it is determined that the first predicted PV 131a acquired in step S123 does not satisfy the second condition (NO in step S124), the setting estimation unit 116 performs, for example, the processes from step S122 onwards again.

[0123] That is, for example, if it is determined that the first predicted PV 131a calculated in step S123 does not satisfy the second condition, the control device 1 determines that a candidate SV 133a for intervention that can return the operating status of the incineration equipment 100a to a normal state has not yet been identified, and continues the estimation process.

[0124] On the other hand, if it is determined that the first predicted PV 131a acquired in step S123 satisfies the second condition (YES in step S124), the setting estimation unit 116 ends the estimation process, for example.

[0125] That is, for example, if it is determined that the first predicted PV131a calculated in step S123 satisfies the second condition, the control device 1 determines that a candidate SV 133a for intervention that can return the operating status of the incineration equipment 100a to a normal state has been identified, and terminates the estimation process.

[0126] Thereafter, the setting estimation unit 116 selects, for example, one of the first predicted PVs 131a acquired in step S123, and identifies the candidate intervention target SV 133a used in calculating the selected first predicted PV 131a as the intervention target SV 133. Then, in this case, the setting estimation unit 116 transmits, for example, the identified intervention target SV 133 to the monitoring device 2.

[0127] In this way, the control device 1 in this embodiment repeatedly performs the process of calculating the first predicted PV 131a and the process of calculating the candidate intervention target SV 133a until the first predicted PV 131a satisfies the second condition, for example. Then, the control device 1 calculates the candidate intervention target SV 133a corresponding to the first predicted PV 131a that satisfies the second condition as the intervention target SV 133.

[0128] This allows the control device 1 in this embodiment to acquire, for example, an intervention target SV133 that can accurately improve the incineration state of the incineration facility 100a. Therefore, the control device 1 can more accurately improve the incineration state of the incineration facility 100a by performing control using the acquired intervention target SV133, for example.

[0129] [Intervention target calculation process in the fourth embodiment] Next, the intervention target calculation process in the fourth embodiment will be described.

[0130] The intervention target calculation process in the fourth embodiment differs from the estimation process in the first embodiment in that, for example, in addition to the PV estimation model ME and the inverse problem analysis model PG, a time series prediction model MP (hereinafter also referred to as the third model MP) is also used. The time series prediction model MP is a model that outputs, for example, a predicted value (hereinafter also referred to as the first predicted value) a predetermined time after the time corresponding to the input actual PV 131. The first predicted value is, for example, a predicted value (hereinafter also referred to as the second predicted PV 131b) of the PV (actual PV 131) measured in the incineration system 100a, a predicted value of the SV that is preferably set as the intervention target in the incineration system 100a, a predicted value of the MV that is preferably operated in the intervention target in the incineration system 100a, or a predicted value of a composite variable. In other words, the time series prediction model MP is a model that predicts, for example, the future value (future value) of the actual PV 131 in the incineration system 100a. Note that, for example, the actual SV and actual MV may be further input to the time series prediction model MP in addition to the actual PV. The following description will be given assuming that the first predicted value is the second predicted PV 131b.

[0131] [Functions of the control device 1 in the fourth embodiment] First, the functions of the control device 1 in the fourth embodiment will be described. Fig. 10 is a block diagram of the functions of the control device 1 in the fourth embodiment.

[0132] As shown in FIG. 10, the control device 1 realizes various functions including a data acquisition unit 111, a data extraction unit 112, a model generation unit 113, an operation monitoring unit 114, a measurement prediction unit 115, and a setting estimation unit 116 by organically cooperating with hardware such as a CPU 101 and a memory 102 and a program.

[0133] 10 , the control device 1 stores, for example, an actual PV 131, an actual SV 132, and an intervention target SV 133 in the information storage area 130. Furthermore, the control device 1 stores, for example, a plurality of pieces of teacher data DP, a plurality of pieces of teacher data DE, a time series prediction model MP, a PV estimation model ME, and an inverse problem analysis model PG in the information storage area 130.

[0134] [Learning processing functions] First, among the functions in the intervention target calculation process, the functions in the process of generating the time series prediction model MP and the PV estimation model ME (hereinafter also referred to as the learning process) will be described.

[0135] The data acquisition unit 111 acquires the actual PV 131 (actual PV 131 measured by the monitoring device 2) measured by a measuring device (not shown) provided in the incineration facility 100a at regular intervals, such as every 10 minutes. Specifically, the data acquisition unit 111 acquires, as the actual PV 131, the temperature inside the incinerator 16 (described below) measured by a thermometer provided in the incinerator 16, for example. Then, the data acquisition unit 111 stores the acquired actual PV 131 in the information storage area 130, for example.

[0136] Furthermore, the data acquiring unit 111 acquires the actual SV132 set by an operator in the incineration facility 100a (the actual SV132 set by the monitoring device 2) at regular intervals, such as every 10 minutes. Specifically, the data acquiring unit 111 acquires, for example, the number of rotations of a rotating drum provided in a dryer 12 (described below) as the actual SV132. Then, the data acquiring unit 111 stores the acquired actual SV132 in the information storage area 130, for example.

[0137] The data acquisition unit 111 may acquire, for example, the actual MV of the operation performed by the worker in the incineration facility 100a. The data acquisition unit 111 may store the acquired actual MV in the information storage area 130.

[0138] The data extraction unit 112 extracts multiple pieces of teacher data DP from, for example, the actual performance PV 131 acquired by the data acquisition unit 111 (the actual performance PV 131 stored in the information storage area 130). Specifically, the data extraction unit 112 performs, for example, preprocessing (hereinafter also simply referred to as first preprocessing) on ​​the actual performance PV 131 acquired by the data acquisition unit 111, and then generates each piece of teacher data DP by including the actual performance PV 131 that has been subjected to the first preprocessing. The first preprocessing is, for example, processing to convert the format of the actual performance PV 131, which is so-called raw data, into an appropriate format (for example, a predetermined format) as the format of data to be included in the teacher data DP.

[0139] Each of the multiple pieces of teacher data DP is teacher data including, for example, multiple actual PV131 acquired by the data acquisition unit 111 at multiple times (multiple times included in a predetermined period). In other words, each of the multiple pieces of teacher data DP includes, for example, an actual PV131 acquired by the data acquisition unit 111 and an actual PV131 acquired by the data acquisition unit 111 a predetermined time after the actual PV131 (for example, several hours later) (hereinafter also referred to as another actual PV131). Specifically, each of the multiple pieces of teacher data DP may include, for example, the temperature inside the incinerator 16 at a first time (one of the times included in the predetermined period) as the actual PV131, and the temperature inside the incinerator 16 at a second time (one of the times included in the predetermined period) a predetermined time after the first time as the other actual PV131.

[0140] Then, the data extraction unit 112 stores, for example, the extracted plurality of pieces of teacher data DP in the information storage area 130. Note that the plurality of pieces of teacher data DP may further include, for example, an achievement SV 132 and an achievement MV.

[0141] Furthermore, the data extraction unit 112 extracts multiple pieces of teacher data DE from, for example, the performance PV131 acquired by the data acquisition unit 111 (the performance PV131 stored in the information storage area 130) and the performance SV132 of the intervention target acquired by the data acquisition unit 111 (the performance SV132 of the intervention target stored in the information storage area 130). Specifically, the data extraction unit 112 performs preprocessing (hereinafter also referred to as second preprocessing) on ​​each of the performance PV131 acquired by the data acquisition unit 111 and the performance SV132 of the intervention target acquired by the data acquisition unit 111, and then generates each piece of teacher data DE by including the performance PV131 subjected to the second preprocessing and the performance SV132 of the intervention target subjected to the second preprocessing. The second preprocessing is, for example, processing to convert the format of the performance PV131 and the performance SV132 of the intervention target, which are so-called raw data, into an appropriate format (for example, a predetermined format) as the format of data to be included in the teacher data DE.

[0142] Each of the plurality of teacher data DE is teacher data including, for example, actual PV131 acquired by the data acquisition unit 111 and actual SV132 of the intervention target when the actual PV131 was measured. Specifically, each of the plurality of teacher data DE may include, for example, the temperature inside the incinerator 16 as the actual PV131 and the rotation speed of the rotary drum in the dryer 12 as the actual SV132.

[0143] Then, the data extracting unit 112 stores, for example, the extracted plurality of pieces of teacher data DE in the information storage area 130. Note that the plurality of pieces of teacher data DE may further include, for example, an achievement MV.

[0144] The model generation unit 113 generates the time series prediction model MP by, for example, learning the plurality of pieces of teacher data DP (the plurality of pieces of teacher data DP stored in the information storage area 130) extracted by the data extraction unit 112. Note that the time series prediction model MP may be, for example, a model that does not learn the plurality of pieces of teacher data DP (for example, a model made up of mathematical expressions, etc.).

[0145] Furthermore, the model generation unit 113 generates the PV estimation model ME by, for example, learning the plurality of pieces of teacher data DE extracted by the data extraction unit 112 (the plurality of pieces of teacher data DE stored in the information storage area 130).

[0146] [Functions in estimation processing] Next, the functions in the estimation process among the functions in the intervention target calculation process will be described. Below, the functions in the estimation process of the fourth embodiment will be described.

[0147] The data acquisition unit 111 acquires the actual PV 131 measured by a measuring device installed in the incineration facility 100a at regular intervals, for example, every 10 minutes. The data acquisition unit 111 also acquires the actual SV 132 set in the incineration facility 100a at regular intervals, for example, every 10 minutes. The data acquisition unit 111 may also acquire the actual MV of the incineration facility 100a at regular intervals, for example, every 10 minutes.

[0148] The operation monitoring unit 114 determines whether the incineration facility 100a is in operation, for example, based on at least one of the actual PV 131 and actual SV 132 acquired by the data acquisition unit 111. Specifically, for example, each time the data acquisition unit 111 acquires the actual PV 131, the operation monitoring unit 114 determines whether the incineration facility 100a is in operation by referencing the acquired actual PV 131. Also, for example, each time the data acquisition unit 111 acquires the actual SV 132, the operation monitoring unit 114 determines whether the incineration facility 100a is in operation by referencing the acquired actual SV 132. Note that the operation monitoring unit 114 may determine whether the incineration facility 100a is in operation by, for example, referencing the actual MV acquired by the data acquisition unit 111, a composite variable calculated from the actual PV 131 acquired by the data acquisition unit 111, etc.

[0149] For example, when the operation monitoring unit 114 determines that the incineration facility 100a is in operation, the measurement prediction unit 115 inputs the actual PV 131 acquired by the data acquisition unit 111 to the time series prediction model MP. Specifically, for example, each time the data acquisition unit 111 acquires the actual PV 131, the measurement prediction unit 115 inputs the acquired actual PV 131 to the time series prediction model MP. Then, the measurement prediction unit 115 acquires, for example, a second predicted PV 131b output from the time series prediction model MP.

[0150] The setting estimation unit 116 determines, for example, whether the second predicted PV 131b acquired by the measurement prediction unit 115 satisfies the second condition. Specifically, for example, each time the measurement prediction unit 115 acquires the second predicted PV 131b, the setting estimation unit 116 determines whether the acquired second predicted PV 131b satisfies the second condition. In other words, the setting estimation unit 116 determines, for example, whether the actual PV 131 acquired a predetermined time after the data acquisition unit 111 acquired the actual PV 131 satisfies the second condition.

[0151] The setting estimation unit 116 may determine whether the actual PV 131, actual SV 132, actual MV, or composite variable satisfies a second condition. In this case, the second condition may be, for example, that the actual PV 131, actual SV 132, actual MV, or composite variable is within a predetermined allowable range, or that the difference between the actual PV 131, actual SV 132, actual MV, or composite variable and the target value is equal to or less than a threshold.

[0152] Then, when it is determined that the second predicted PV 131b does not satisfy the second condition, the setting estimation unit 116 selects (searches for), for example, a candidate intervention target SV 133a. That is, in this case, the setting estimation unit 116 determines that, for example, the incineration state of the incineration facility 100a may become abnormal in the future, and starts the process of estimating an intervention target SV 133 that can improve the incineration state of the incineration facility 100a.

[0153] Next, the setting estimation unit 116 inputs, for example, the selected intervention target SV candidate 133a into the PV estimation model ME. Then, the setting estimation unit 116 acquires, for example, the first predicted PV 131a output from the PV estimation model ME.

[0154] As a result, for example, if the acquired first predicted PV 131a does not satisfy the second condition, the setting estimation unit 116 selects (searches for) a candidate 133a for the intervention target SV by using the first predicted PV 131a. Specifically, in this case, the setting estimation unit 116 selects (searches for) a candidate 133a for the intervention target SV by using, for example, the inverse problem analysis model PG. Then, the setting estimation unit 116 inputs, for example, the selected candidate 133a for the intervention target SV into the PV estimation model ME. Thereafter, the setting estimation unit 116 re-acquires, for example, the first predicted PV 131a output from the PV estimation model ME.

[0155] As a result, for example, if the reacquired first predicted PV131a satisfies the second condition, the setting estimation unit 116 outputs the candidate intervention target SV 133a corresponding to the reacquired first predicted PV131a as the intervention target SV133 (new intervention target SV133). Specifically, in this case, the setting estimation unit 116 transmits, for example, the intervention target SV133 to the monitoring device 2. Then, the monitoring device 2 controls the incineration facility 100a by using, for example, the received intervention target SV133.

[0156] On the other hand, for example, if the reacquired first predicted PV 131a does not satisfy the second condition, the setting estimation unit 116 selects (searches for) the intervention target SV candidate 133a again by using the reacquired first predicted PV 131a. Thereafter, the setting estimation unit 116 repeatedly performs the process of acquiring the first predicted PV 131a and the process of selecting the intervention target SV candidate 133a until the first predicted PV 131a output from the PV estimation model ME satisfies the second condition.

[0157] [Learning Process in the Fourth Embodiment] Next, the learning process in the fourth embodiment will be described. Fig. 11 is a flowchart illustrating the learning process in the fourth embodiment. Specifically, Fig. 11(A) is a flowchart illustrating the learning process when a time series prediction model MP is generated, and Fig. 11(B) is a flowchart illustrating the learning process when a PV estimation model ME is generated.

[0158] 11(A), the data extraction unit 112 extracts, for example, multiple pieces of teacher data DP (step S1). Specifically, the data extraction unit 112 performs step S1 in response to, for example, an operator inputting information indicating that multiple pieces of teacher data DP will be generated. Then, the data extraction unit 112 stores, for example, the extracted multiple pieces of teacher data DP in the information storage area 130. Note that the data extraction unit 112 may also store, for example, multiple pieces of teacher data DP manually extracted by an operator in the information storage area 130.

[0159] Thereafter, the model generation unit 113 generates a time series prediction model MP by, for example, learning the plurality of pieces of training data DP extracted in step S1 (the plurality of pieces of training data DP stored in the information storage area 130) (step S2).

[0160] 11(B), the data extraction unit 112 extracts, for example, multiple pieces of teacher data DE (step S3). Specifically, the data extraction unit 112 performs step S3 in response to, for example, an operator inputting information indicating that multiple pieces of teacher data DE will be extracted. Then, the data extraction unit 112 stores, for example, the extracted multiple pieces of teacher data DE in the information storage area 130. Note that the data extraction unit 112 may store, for example, multiple pieces of teacher data DE manually extracted by an operator in the information storage area 130.

[0161] Thereafter, the model generation unit 113 generates a PV estimation model ME by, for example, learning the plurality of pieces of teacher data DE extracted in step S3 (the plurality of pieces of teacher data DE stored in the information storage area 130) (step S4).

[0162] [Estimation process in the fourth embodiment] Next, the estimation process in the fourth embodiment will be described. Fig. 12 is a flowchart illustrating the estimation process in the fourth embodiment. Figs. 13 and 14 are diagrams illustrating the estimation process in the fourth embodiment.

[0163] As shown in FIG. 12, the data acquisition unit 111 acquires the actual PV131 measured by a measuring device (not shown) installed in the incineration equipment 100a and the actual SV132 set in the incineration equipment 100a at regular intervals, such as every 10 minutes (step S11).

[0164] Specifically, for example, when a time series prediction model MP is generated by learning multiple training data DP shown in Figure 16 described below, the data acquisition unit 111 acquires a value corresponding to the "temperature inside the incinerator" as the actual PV131, and acquires values ​​corresponding to each of the "dryer sludge input amount," "dryer inlet temperature," "dryer rotation speed," and "dryer outlet temperature" as the actual SV132 (combination of actual SV132).

[0165] 13, the measurement prediction unit 115 inputs, for example, the actual PV 131 and actual SV 132 acquired in step S11 to the time series prediction model MP. Then, the measurement prediction unit 115 acquires, for example, a second predicted PV 131b output from the time series prediction model MP (step S12).

[0166] Specifically, for example, when the time series prediction model MP is generated by learning multiple pieces of training data DP shown in Figure 16 described below, the measurement prediction unit 115 acquires, for example, a value corresponding to "temperature inside the incinerator" as the second predicted PV131b.

[0167] Next, the setting estimation unit 116 determines whether the second predicted PV 131b acquired in step S12 satisfies a second condition (step S13). For example, if the second predicted PV 131b is a value corresponding to the "temperature inside the incinerator," the second condition may be 750°C or higher and 800°C or lower.

[0168] As a result, when it is determined that the second predicted PV 131b acquired in step S12 satisfies the second condition (YES in step S13), the control device 1, for example, ends the estimation process. That is, in this case, the control device 1 determines that the operating state of the incineration facility 100a is normal and ends the estimation process.

[0169] On the other hand, if it is determined that the second predicted PV131b obtained in step S12 does not satisfy the second condition (NO in step S13), the setting estimation unit 116 selects (searches for), for example, an intervention target SV133 (a new intervention target SV133) that is preferably set in the incineration facility 100a (step S14).

[0170] Specifically, in step S14, the setting estimation unit 116 executes, for example, the estimation process in the first embodiment, the estimation process in the second embodiment, or the estimation process in the third embodiment.

[0171] Thereafter, the setting estimation unit 116 transmits, for example, one of the candidates 133a for the intervention target SV searched for in step S14 to the monitoring device 2 as the intervention target SV 133.

[0172] In this way, the control device 1 in this embodiment repeatedly performs the process of calculating the first predicted PV 131a and the process of calculating the candidate intervention target SV 133a until the first predicted PV 131a satisfies the second condition, for example. Then, the control device 1 calculates the candidate intervention target SV 133a corresponding to the first predicted PV 131a that satisfies the second condition as the intervention target SV 133.

[0173] Specifically, for example, when the actual PV 131 measured by the monitoring device 2 does not satisfy the second condition, the control device 1 calculates the first predicted PV 131a by using the PV estimation model ME.

[0174] More specifically, the control device 1 calculates a second predicted PV 131b at a second timing after the first timing from the actual PV 131 at the first timing by using, for example, the time-series prediction model MP. Then, for example, when the calculated second predicted PV 131b does not satisfy a second condition, the control device 1 calculates a first predicted PV 131a by using the PV estimation model ME.

[0175] In other words, in this embodiment, if the second predicted PV131b calculated by the time series prediction model MP does not satisfy the second condition, the control device 1 determines that the sludge incineration state in the incineration equipment 100a may become abnormal in the future, and begins controlling the sludge incineration state in the incineration equipment 100a.

[0176] Specifically, as shown in Figure 14, even if the actual PV131 at the current time, t1 (first timing), is V1, which is smaller than V2, which is the upper threshold within the normal range (hereinafter simply referred to as the upper threshold), if the second predicted PV131b of the actual PV131 at time t3 (second timing), a predetermined time after time t1, is V3 and predicted to exceed V2, which is the upper threshold (NO in step S13 in Figure 12), the control device 1 determines, for example, that there is a possibility that the incineration state of the sludge in the incineration equipment 100a will no longer be normal at time t3, and starts control of the incineration state of the sludge in the incineration equipment 100a (step S14 in Figure 12).

[0177] More specifically, for example, if the time from when a new intervention target SV133 is set in the incineration equipment 100a until the sludge incineration state in the incineration equipment 100a changes (improves) is time T, the control device 1 begins to control the sludge incineration state in the incineration equipment 100a, for example, by time t2, which is the time T before time t3.

[0178] As a result, the control device 1 in this embodiment can perform automatic control so that the incineration state of the incineration facility 100a is maintained in a normal state not only when the current incineration state of the incineration facility 100a is not in a normal state, but also when there is a possibility that the incineration state of the incineration facility 100a will not be in a normal state in the future. Therefore, the control device 1 can further reduce the burden on the worker compared to when the operation of the incineration facility 100a is manually controlled by the worker, for example.

[0179] [Configuration of processing system 1000] Next, the configuration of the treatment system 1000 in the first to fourth embodiments will be described. Figure 15 is a diagram illustrating the configuration of the treatment system 1000. Hereinafter, of the intervention target SV133 in the incineration facility 100a, the intervention target SV133 that is determined without being influenced by other intervention target SV133 will also be referred to as the primary intervention target SV133, and the intervention target SV133 that may be influenced by the primary intervention target SV133 will also be referred to as the secondary intervention target SV133. Note that the positions and numbers of lines (pipes) and valves shown in Figure 15 are not limited to those shown here.

[0180] As shown in Figure 15, the incineration facility 100a includes, for example, a feeder 11, a dryer 12, a feeder 13, a carbonization furnace 14, a dust collector 15, an incinerator 16, a heat exchanger 17, a heat exchanger 18, a heat exchanger 19, and an exhaust gas treatment system 20. The incineration facility 100a also includes, for example, a sludge transfer pump P1, an air circulation pump P2, an induced draft machine F1, a blower B1, and a blower B2. The induced draft machine F1 is a device that has the function of drawing air, such as a fan or a blower. The blower B1 and the blower B2 are each a device that has the function of blowing air, such as a fan or a blower.

[0181] The following description will be given assuming that the incineration facility 100a has a carbonization furnace 14, but the present invention is not limited to this. Specifically, the incineration facility 100a may not have a carbonization furnace 14. The incineration facility 100a may also have a gasification furnace instead of the carbonization furnace 14.

[0182] The cutting device 11 is, for example, a device that cuts out sludge (dehydrated cake) that has been supplied via a previous stage facility (for example, a sludge dehydration device) not shown.

[0183] The sludge transfer pump P1 is a pump that supplies (transfers) the sludge cut out by the cutting device 11 to the dryer 12, for example.

[0184] Specifically, the sludge transfer pump P1 supplies sludge to the dryer 12 in accordance with an intervention target SV133 specified by, for example, the monitoring device 2. More specifically, the control device 1 calculates the intervention target SV133 for the amount of sludge to be input to the dryer 12, for example, by executing an intervention target calculation process. Then, the monitoring device 2 transmits the intervention target SV133 calculated by the control device 1 to the sludge transfer pump P1, for example.

[0185] The dryer 12 reduces the moisture content of the sludge by, for example, drying the sludge supplied from the feeding device 11 by the sludge transfer pump P1. Then, the dryer 12 supplies the dried sludge to the feeding device 13, for example.

[0186] Specifically, the dryer 12 dries the sludge supplied from the excretion device 11 in accordance with, for example, an intervention target SV133 specified by the monitoring device 2. More specifically, the control device 1 calculates, for example, an intervention target SV133 for the rotation speed of a rotating drum (not shown) provided in the dryer 12 by executing an intervention target calculation process. Then, the monitoring device 2 transmits, for example, the intervention target SV133 calculated by the control device 1 to the dryer 12.

[0187] 15, the air used to dry the sludge in the dryer 12 is supplied to a heat exchanger 19 by, for example, an air circulation pump P2. The air heated by the heat exchanger 19 is then further heated by, for example, a burner 12a that burns a gaseous fuel such as city gas (hereinafter also simply referred to as gaseous fuel), and is then supplied into the dryer 12 to be used again to dry the sludge. Furthermore, cooling water is supplied to the dryer 12 as needed, for example, to lower the temperature inside the dryer 12.

[0188] 15, the valve V1 adjusts the supply amount of gaseous fuel to the burner 12a by adjusting the opening degree in accordance with the intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (first intervention target SV133) for the air temperature on the inlet side of the dryer 12, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (second intervention target SV133) for the flow rate of gaseous fuel at the valve V1, for example, from the intervention target SV133 calculated by the control device 1 and an actual PV for the air temperature (for example, the current temperature) on the inlet side of the dryer 12. Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V1, for example, from the intervention target SV133 calculated by the monitoring device 2 and an actual PV for the flow rate of gaseous fuel at the valve V1 (for example, the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V1, for example.

[0189] Furthermore, the valve V2 adjusts the amount of cooling water supplied to the dryer 12 by adjusting its opening in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (primary intervention target SV133) for the air temperature on the inlet side of the dryer 12, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (secondary intervention target SV133) for the flow rate of cooling water at the valve V2, for example, from the intervention target SV133 calculated by the control device 1 and an actual PV for the air temperature (e.g., the current temperature) on the inlet side of the dryer 12. Furthermore, the monitoring device 2 calculates the intervention target MV for the opening of the valve V2, for example, from the intervention target SV133 calculated by the monitoring device 2 and an actual PV for the flow rate of cooling water at the valve V2 (e.g., the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V2, for example.

[0190] The feeding device 13 is, for example, a device that temporarily accumulates the sludge supplied from the dryer 12 and sequentially feeds the accumulated sludge into the carbonization furnace 14.

[0191] Specifically, the feeding device 13 feeds the sludge supplied from the dryer 12 into the carbonization furnace 14 in accordance with the intervention target SV133 specified by, for example, the monitoring device 2. More specifically, the control device 1 calculates the intervention target SV133 for the rotation speed of a feeder (not shown) provided in the feeding device 13, for example, by executing an intervention target calculation process. Then, the monitoring device 2 transmits the intervention target SV133 calculated by the control device 1 to the feeding device 13, for example.

[0192] The carbonization furnace 14 is a furnace that, for example, decomposes the sludge supplied from the feeding device 13 into charcoal (solid matter) and gas (combustible gas) by thermal decomposition. The carbonization furnace 14 then discharges the gas generated by the thermal decomposition to the incinerator 16, and discharges the charcoal generated by the thermal decomposition to the outside of the carbonization furnace 14.

[0193] 15, the air used for pyrolysis in the carbonization furnace 14 is, for example, air (atmospheric air) supplied by a blower B1, and is heated by a heat exchanger 17 before being supplied to the carbonization furnace 14. Furthermore, the carbonization furnace 14 is supplied with cooling water, for example, to lower the temperature inside the carbonization furnace 14 as needed. Furthermore, the carbonization furnace 14 is supplied with gaseous fuel, for example, to raise the temperature inside the carbonization furnace 14 by combustion as needed.

[0194] 15, the valve V3 adjusts the amount of air supplied to the carbonization furnace 14 by adjusting its opening according to an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (primary intervention target SV133) for the temperature inside the carbonization furnace 14, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (secondary intervention target SV133) for the air flow rate at the valve V3, for example, from the intervention target SV133 calculated by the control device 1 and the actual PV for the temperature inside the carbonization furnace 14 (for example, the current temperature). The intervention target SV133 for the air flow rate at the valve is, for example, the opening of the valve. Furthermore, the monitoring device 2 calculates the intervention target MV for the opening of the valve V3, for example, from the intervention target SV133 calculated by the monitoring device 2 and the intervention target SV133 for the air flow rate (for example, the current flow rate) at the valve V3. Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V3, for example.

[0195] Furthermore, the valve V4 adjusts the supply amount of cooling water to the carbonization furnace 14 by adjusting the opening degree in accordance with the intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (first intervention target SV133) for the temperature inside the carbonization furnace 14, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (second intervention target SV133) for the flow rate of cooling water at the valve V4, for example, from the intervention target SV133 calculated by the control device 1 and the actual PV for the temperature inside the carbonization furnace 14 (for example, the current temperature). Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V4, for example, from the intervention target SV133 calculated by the monitoring device 2 and the actual PV for the flow rate of cooling water at the valve V4 (for example, the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V4, for example.

[0196] Furthermore, the valve V5 adjusts the supply amount of gaseous fuel to a burner (not shown) provided in the carbonization furnace 14 by adjusting the opening degree in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (primary intervention target SV133) for the temperature inside the incinerator 16, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (secondary intervention target SV133) for the flow rate of gaseous fuel at the valve V5, for example, from the intervention target SV133 calculated by the control device 1 and the actual PV for the temperature inside the incinerator 16 (for example, the current temperature). Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V5, for example, from the intervention target SV133 calculated by the monitoring device 2 and the actual PV for the flow rate of gaseous fuel at the valve V5 (for example, the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V5, for example.

[0197] Dust collector 15 is installed, for example, downstream of carbonization furnace 14, and removes carbonized materials contained in the gas (gas generated by thermal decomposition in carbonization furnace 14) supplied from carbonization furnace 14 to incinerator 16.

[0198] The incinerator 16 is a furnace that burns (completely combusts) the gas supplied from, for example, the carbonization furnace 14. The exhaust gas generated by, for example, the combustion of the gas in the incinerator 16 is drawn by an induction machine F1 to the exhaust gas treatment system 20 through a pipe that connects the outlet of the incinerator 16 to the inlet of the exhaust gas treatment system 20. This pipe is provided with, for example, heat exchangers 17, 18, and 19.

[0199] The air used for burning gas in the incinerator 16 includes, for example, a portion of the air used for drying sludge in the dryer 12 (air supplied by the air circulation pump P2). The air used for burning gas in the incinerator 16 is, for example, a portion of the air (atmosphere) supplied by the blower B2, and includes air that is heated by the heat exchanger 18 and then supplied to the incinerator 16. Cooling water is supplied to the incinerator 16 as needed, for example, to lower the temperature inside the incinerator 16. Furthermore, gaseous fuel is supplied to the incinerator 16 as needed, for example, to raise the temperature inside the incinerator 16 by combustion. In the incinerator 16, the burner 16a burns the gaseous fuel using a portion of the air supplied from the blower B2, thereby raising the temperature inside the incinerator 16.

[0200] 15, the valve V6 adjusts the amount of air supplied from the blower B2 to the incinerator 16 by adjusting the opening degree in accordance with the intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (primary intervention target SV133) for the temperature inside the incinerator 16, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (secondary intervention target SV133) for the air flow rate at the valve V6, for example, from the intervention target SV133 calculated by the control device 1 and the actual PV for the temperature inside the incinerator 16 (for example, the current temperature). Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V6, for example, from the intervention target SV133 calculated by the monitoring device 2 and the intervention target SV133 for the air flow rate (for example, the current flow rate) at the valve V6. Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V6, for example.

[0201] Furthermore, the valve V7 adjusts the amount of air supplied from the blower B2 to the burner 16a by adjusting the opening degree in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (first intervention target SV133) for the temperature inside the incinerator 16, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (second intervention target SV133) for the air flow rate at the valve V7, for example, from the intervention target SV133 calculated by the control device 1 and the actual PV for the temperature inside the incinerator 16 (for example, the current temperature). Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V7, for example, from the intervention target SV133 calculated by the monitoring device 2 and the intervention target SV133 for the air flow rate at the valve V7 (for example, the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V7, for example.

[0202] Furthermore, the valve V8 adjusts the supply amount of cooling water to the incinerator 16 by adjusting the opening degree in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (first intervention target SV133) for the temperature of the flue gas discharged from the incinerator 16, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (second intervention target SV133) for the flow rate of cooling water at the valve V8, for example, from the intervention target SV133 calculated by the control device 1 and an actual PV for the temperature of the flue gas discharged from the incinerator 16 (for example, the current temperature). Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V8, for example, from the intervention target SV133 calculated by the monitoring device 2 and an actual PV for the flow rate of cooling water at the valve V8 (for example, the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V8, for example.

[0203] Furthermore, the valve V9 adjusts the supply amount of gaseous fuel to the burner 16a by adjusting the opening degree in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (first intervention target SV133) for the temperature inside the incinerator 16, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (second intervention target SV133) for the flow rate of the gaseous fuel at the valve V9, for example, from the intervention target SV133 calculated by the control device 1 and an actual PV for the temperature inside the incinerator 16 (for example, the current temperature). Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V9, for example, from the intervention target SV133 calculated by the monitoring device 2 and an actual PV for the flow rate of the gaseous fuel at the valve V9 (for example, the current flow rate). Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V9, for example.

[0204] Furthermore, the valve V10 adjusts the supply amount of air used to dry sludge in the dryer 12 to the incinerator 16 by adjusting the opening degree in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (primary intervention target SV133) for the pressure at the outlet of the dryer 12, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (secondary intervention target SV133) for the air flow rate at the valve V10, for example, from the intervention target SV133 calculated by the control device 1 and the actual PV for the outlet pressure (e.g., current pressure) of the dryer 12. Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V10, for example, from the intervention target SV133 calculated by the monitoring device 2 and the intervention target SV133 for the air flow rate (e.g., current flow rate) at the valve V10. Thereafter, the monitoring device 2 transmits, for example, the calculated intervention target MV to the valve V10.

[0205] Furthermore, the valve V11 adjusts the amount of air supplied from the dryer 12 to the heat exchanger 19 by adjusting the opening degree in accordance with an intervention target SV133 specified by the monitoring device 2, for example. Specifically, the control device 1 calculates the intervention target SV133 (primary intervention target SV133) for the air temperature at the outlet or inlet side of the dryer 12, for example, by executing an intervention target calculation process. Then, the monitoring device 2 calculates the intervention target SV133 (secondary intervention target SV133) for the air flow rate at the valve V11, for example, from the intervention target SV133 calculated by the monitoring device 2 and the actual PV for the air temperature (e.g., the current temperature) at the outlet or inlet side of the dryer 12. Furthermore, the monitoring device 2 calculates the intervention target MV for the opening degree of the valve V11, for example, from the intervention target SV133 calculated by the monitoring device 2 and the intervention target SV133 for the air flow rate (e.g., the current flow rate) at the valve V11. Thereafter, the monitoring device 2 transmits the calculated intervention target MV to the valve V11, for example.

[0206] The heat exchanger 17 exchanges heat between, for example, the exhaust gas discharged from the incinerator 16 and the air (atmosphere) supplied by the blower B1. Specifically, the heat exchanger 17 uses, for example, the waste heat of the exhaust gas supplied from the incinerator 16 to heat the air supplied from the blower B1. Then, the heat exchanger 17 supplies, for example, the heated air to the carbonization furnace 14.

[0207] Heat exchanger 18 performs heat exchange between, for example, the exhaust gas supplied from heat exchanger 17 and a portion of the air (atmosphere) supplied by blower B2. Specifically, heat exchanger 18 uses, for example, waste heat of the exhaust gas supplied from heat exchanger 17 to heat a portion of the air supplied from blower B2. Then, heat exchanger 18 supplies, for example, the heated air to incinerator 16.

[0208] The heat exchanger 19 exchanges heat, for example, between the exhaust gas supplied from the heat exchanger 18 and a portion of the air supplied from the dryer 12 (air circulation pump P2). Specifically, the heat exchanger 19 uses the waste heat of the exhaust gas supplied from the heat exchanger 18 to heat a portion of the air supplied from the dryer 12. Then, the heat exchanger 19 supplies the heated air to the dryer 12 (burner 12a), for example.

[0209] The exhaust gas treatment system 20 includes, for example, a white smoke prevention air preheater (not shown) that generates heated air (white smoke prevention air) that prevents water vapor in the exhaust gas supplied from the heat exchanger 19 from appearing as white smoke, a dust collector (not shown) that collects impurities in the exhaust gas, and a filter (not shown) that removes SO from the exhaust gas by contacting the SO in the exhaust gas with an agent. X and a smoke washing tower (not shown) for removing components such as:

[0210] The inducer F1 is provided, for example, at the rear of the exhaust gas treatment system 20, and induces the gas generated in the carbonization furnace 14 to the incinerator 16, and also induces the exhaust gas generated in the incinerator 16 to the exhaust gas treatment system 20. That is, the inducer F1 induces, for example, the gas generated in the carbonization furnace 14, and controls the pressure inside the carbonization furnace 14 by a valve installed on the primary side of the inducer F1. The exhaust gas induced by the inducer F1 is released to the outside via a chimney (not shown).

[0211] As described above, in the treatment system 1000 according to the first to fourth embodiments, when the treatment facility 100 is an incineration facility 100a, the incineration facility 100a may include, for example, an incinerator 16 that incinerates sludge, and a treatment device that treats at least one of the materials to be incinerated in the incinerator 16 (e.g., sludge), substances discharged from the incinerator 16 (e.g., incineration ash), and fluids discharged from the incinerator 16 (e.g., exhaust gas) by using at least a portion of the energy recovered from the waste heat of the incinerator 16. Specifically, the treatment device may be, for example, the dryer 12, the inducer F1, or the exhaust gas treatment system 20. In this case, the actual PV131 may be, for example, the actual PV131 of the incinerator 16. In this case, the intervention target SV133 may be, for example, the intervention target SV133 in the treatment device.

[0212] [Specific example of multiple training data DP] Next, a specific example of the plurality of pieces of training data DP will be described when the treatment facility 100 is the incineration facility 100a. Fig. 16 is a diagram for explaining a specific example of the plurality of pieces of training data DP.

[0213] Each of the plurality of teacher data DP shown in Fig. 16 has, for example, an item "Inside Incinerator Temperature," which sets the temperature inside incinerator 16 (e.g., each of the temperatures at a plurality of timings) measured by a thermometer (not shown) provided in incinerator 16. Each of the plurality of teacher data DP shown in Fig. 16 also has, for example, an item "Dryer Sludge Input Amount," which sets the amount of sludge measured by a flow meter (not shown) provided in the piping between extrusion device 11 and dryer 12 (the amount of sludge input from extrusion device 11 to dryer 12), and a item "Dryer Inlet Temperature," which sets the temperature measured by a thermometer (not shown) provided in the piping between heat exchanger 19 and dryer 12 on the inlet side of dryer 12 (the temperature of air supplied from heat exchanger 19 to dryer 12). Furthermore, each of the multiple teaching data DP shown in FIG. 16 has, for example, items such as "dryer rotation speed," which sets the rotation speed of the rotating drum installed in the dryer 12 (a ratio (e.g., percentage, etc.) to the maximum rotation speed), and "dryer outlet temperature," which sets the temperature (the temperature of the air supplied from the dryer 12 to the heat exchanger 19) measured by a thermometer (not shown) installed on the outlet side of the dryer 12 in the piping between the dryer 12 and the heat exchanger 19.

[0214] Specifically, in the training data DP in the first row shown in Figure 16, for example, the "temperature inside the incinerator" is set to "950 (°C)", the "dryer sludge input amount" is set to "3.0 (t / h)", the "dryer inlet temperature" is set to "450 (°C)", the "dryer rotation speed" is set to "22 (%)", and the "dryer outlet temperature" is set to "180 (°C)".

[0215] 16, for example, "950 (°C)" is set as the "temperature inside the incinerator," "3.0 (t / h)" is set as the "sludge input amount into the dryer," "450 (°C)" is set as the "temperature in the dryer," "20 (%)" is set as the "speed of rotation of the dryer," and "180 (°C)" is set as the "temperature out of the dryer." Explanation of the other teaching data DP shown in FIG. 16 will be omitted.

[0216] [Specific example of multiple training data DE] Next, a specific example of the plurality of pieces of training data DE will be described when the treatment facility 100 is the incineration facility 100a. Fig. 17 is a diagram for explaining a specific example of the plurality of pieces of training data DE.

[0217] Each of the multiple teaching data DE shown in Figure 17 has the following items, for example, as described in Figure 16: "Temperature inside incinerator," "Amount of sludge input into dryer," "Temperature at inlet of dryer," "Rotation speed of dryer," and "Temperature at outlet of dryer."

[0218] Specifically, in the training data DE in the first row shown in Figure 17, for example, the "temperature inside the incinerator" is set to "950 (℃)", the "dryer sludge input amount" is set to "2.7 (t / h)", the "dryer inlet temperature" is set to "450 (℃)", the "dryer rotation speed" is set to "18 (%)", and the "dryer outlet temperature" is set to "180 (℃)".

[0219] 17, for example, "950 (°C)" is set as the "temperature inside the incinerator," "2.7 (t / h)" is set as the "sludge input amount into the dryer," "460 (°C)" is set as the "temperature in the dryer," "18 (%)" is set as the "speed of rotation of the dryer," and "180 (°C)" is set as the "temperature out of the dryer." Explanation of the other training data DE shown in FIG. 17 will be omitted.

[0220] In addition, the multiple teaching data DP and the multiple teaching data DE may include, for example, in addition to the ``dryer sludge input amount,'' ``dryer inlet temperature,'' ``dryer rotation speed,'' and ``dryer outlet temperature,'' items such as the rotation speed of a feeder (not shown) installed in the input device 13, the temperature measured by a thermometer (not shown) installed in the carbonization furnace 14 (the temperature inside the carbonization furnace 14), the pressure measured by a pressure gauge (not shown) installed in the carbonization furnace 14 (the pressure inside the carbonization furnace 14), and the temperature measured by a thermometer (not shown) installed on the outlet side of the incinerator 16 in the piping between the incinerator 16 and the heat exchanger 17 (the temperature of the exhaust gas discharged from the incinerator 16 to the heat exchanger 17).

[0221] That is, the control device 1 may estimate the primary intervention target SV133 in the incineration facility 100a (for example, the primary intervention target SV133 described in FIG. 15) by using the PV estimation model ME, for example. The monitoring device 2 may then calculate the secondary intervention target SV133 in the incineration facility 100a (for example, the secondary intervention target SV133 set in FIG. 15) by using the primary intervention target SV133 estimated by the control device 1, for example.

[0222] This enables the control device 1 to, for example, control the secondary intervention target SV133 that is affected by the primary intervention target SV133 with high accuracy. Also, in this case, the control device 1 does not, for example, estimate the secondary intervention target SV133 in the intervention target calculation process, and therefore it is possible to reduce the processing load associated with the execution of the intervention target calculation process. Note that the control device 1 may, for example, estimate the primary intervention target SV133 in the incineration facility 100a by using a time series prediction model MP and a PV estimation model ME.

[0223] Furthermore, for example, even in the estimation process, the data extraction unit 112 may continue to extract (generate) new teacher data DP and new teacher data DE by using the actual PV 131 acquired by the data acquisition unit 111 and the intervention target SV 133 set in the incineration facility 100a. And, for example, even in the estimation process described below, the model generation unit 113 may continue to train the time series prediction model MP using the new teacher data DP and train the PV estimation model ME using the new teacher data DE.

[0224] This enables the control device 1 to further improve the determination accuracy of the time series prediction model MP and the PV estimation model ME, for example.

[0225] [Configuration of Modified Example of Processing System 1000 (1)] Next, a first modified example of the processing system 1000 according to the first to fourth embodiments (hereinafter simply referred to as the first modified example) will be described. Fig. 18 is a diagram illustrating the configuration of the first modified example of the processing system 1000.

[0226] The treatment facility 100 in the first modification is, for example, a water purification facility 100b that performs water purification treatment on water to be treated (hereinafter simply referred to as water to be treated), such as raw water. The monitoring device 2 in the first modification monitors, for example, the treatment status in the water purification facility 100b.

[0227] As shown in FIG. 18, the water purification system 100b includes, for example, a gritification basin 210, a receiving well 220, a mixing basin 230, a flocculation basin 240 (hereinafter simply referred to as the flocculation basin 240), a sedimentation basin 250, a filtration basin 260, a purified water basin 270, a distribution basin 280, a storage tank T, and a pump P11.

[0228] The settling basin 210 is, for example, a tank into which the water to be treated first flows, and is a tank in which sedimentation and removal of sediment and the like contained in the water to be treated is performed.

[0229] The receiving well 220 is, for example, a tank that adjusts the amount of water to be treated supplied from the settling basin 210 and supplies it to the mixing basin 230.

[0230] The mixing basin 230 is, for example, a tank in which a flocculant is injected into the water to be treated supplied from the receiving well 220 .

[0231] The flocculation basin 240 is a tank that, for example, agitates the water to be treated supplied from the mixing basin 230, thereby flocculating suspended solids contained in the water to be treated supplied from the mixing basin 230 with a coagulant to form flocs.

[0232] The settling basin 250 is, for example, a tank in which flocs contained in the water to be treated supplied from the flocculation basin 240 are allowed to settle and are separated from the water to be treated.

[0233] The filtration basin 260 is a tank that filters the water to be treated supplied from the settling basin 250 by using a filter body (not shown) made of, for example, sand, gravel, or the like.

[0234] The purified water reservoir 270 is, for example, a tank that temporarily stores the water to be treated supplied from the filtration reservoir 260 (for example, the water to be treated after being disinfected by chlorine downstream of the filtration reservoir 260) and supplies it to the distributing reservoir 280.

[0235] The distributing reservoir 280 temporarily stores the water to be treated (treated water) supplied from the purified water reservoir 270, for example, and supplies it to homes and the like (not shown).

[0236] The storage tank T is, for example, a tank for storing a flocculant to be injected into the water to be treated, and supplies the flocculant to the pump P.

[0237] The pump P11 is, for example, a pump provided in a pipe (not shown) that connects the storage tank T and the mixing basin 230. Specifically, the pump P11 supplies the coagulant to the mixing basin 230 at an injection rate that corresponds to a predetermined injection rate determined by, for example, the manager of the water purification facility 100b.

[0238] Then, the control device 1 acquires the actual PV131 (the actual PV131 measured by the monitoring device 2) measured by a measuring device (not shown) provided in the water purification facility 100b at regular intervals, for example, every 10 minutes. Specifically, the control device 1 acquires, for example, the settling velocity of flocs formed in the flocculation basin 240 (for example, the average value of the settling velocity) as the actual PV131.

[0239] The control device 1 also acquires the actual SV132 (the actual SV132 set by the monitoring device 2) set by an operator in the water purification facility 100b at regular intervals, such as every 10 minutes. Specifically, the control device 1 acquires, for example, the injection rate of the coagulant by the pump P11 as the actual SV132. The control device 1 also acquires, for example, the stirring intensity and stirring time in the flocculation basin 240 (hereinafter, these will also be collectively referred to as stirring intensity, etc.) as the actual SV132.

[0240] 3 etc., the control device 1 extracts, for example, a plurality of pieces of training data DP from the acquired actual result PV 131 and actual result SV 132, and generates a time series prediction model MP by using the extracted plurality of pieces of training data DP. The control device 1 also extracts, for example, a plurality of pieces of training data DE from the acquired actual result PV 131 and actual result SV 132, and generates a PV estimation model ME by using the extracted plurality of pieces of training data DE.

[0241] Specifically, the control device 1 generates a PV estimation model ME that estimates the settling velocity of flocs in the flocculation basin 240 using, for example, the injection rate of the coagulant by the pump P11, the stirring intensity in the flocculation basin 240, etc. as inputs.

[0242] [Configuration of Modified Example of Processing System 1000 (2)] Next, a second modified example of the processing system 1000 according to the first to fourth embodiments (hereinafter simply referred to as the second modified example) will be described. Fig. 19 is a diagram illustrating the configuration of the second modified example of the processing system 1000.

[0243] The treatment facility 100 in the second modified example is, for example, a sewage treatment facility 100c that treats water to be treated, such as sewage (hereinafter also simply referred to as water to be treated). The monitoring device 2 in the second modified example monitors the treatment status in the sewage treatment facility 100c, for example.

[0244] As shown in FIG. 19, the sewage treatment facility 100c includes, for example, a primary sedimentation tank 310, a wastewater treatment device 320, a final sedimentation tank 330, a thickening tank 340, a thickening device 350, a digestion tank 360, and a heater 370.

[0245] The primary sedimentation tank 310 separates, for example, organic matter and suspended matter contained in the water to be treated by settling. The primary sedimentation tank 310 then discharges the separated organic matter and suspended matter as primary sedimentation sludge to a concentration tank 340, and discharges the water to be treated from which the organic matter and suspended matter have been separated to a wastewater treatment device 320.

[0246] The wastewater treatment device 320 treats the water to be treated by biological treatment such as a standard activated sludge process or a circulating nitrification-denitrification process. Specifically, the wastewater treatment device 320 has, for example, a denitrification tank (not shown) in which anaerobic denitrifying bacteria produce nitrogen from nitrite nitrogen and nitrate nitrogen (denitrification), and a nitrification tank (not shown) located downstream of the denitrification tank in which aerobic nitrifying bacteria nitrify ammonia nitrogen to produce nitrite nitrogen and nitrate nitrogen. The wastewater treatment device 320 then discharges the water to be treated into, for example, a final sedimentation tank 330.

[0247] The final settling tank 330, for example, separates and settles sludge contained in the water to be treated discharged from the sewage treatment device 320, and discharges the separated sludge as activated sludge. The final settling tank 330 then supplies, for example, a portion of the activated sludge to the thickener 350 as excess sludge, and returns the activated sludge other than the excess sludge to the sewage treatment device 320 as returned sludge. The final settling tank 330 also discharges the water to be treated (supernatant) from which the sludge has been separated, for example, to a downstream sterilization treatment device (not shown). Thereafter, the sterilization treatment device (not shown) sterilizes the water to be treated discharged from the final settling tank 330, for example, and discharges the sterilized treated water.

[0248] The thickening tank 340 thickens the primary sludge discharged from the primary sedimentation tank 310 and supplies the thickened sludge to the digestion tank 360 .

[0249] The thickener 350 thickens excess sludge discharged from the final settling tank 330 and supplies the thickened sludge to the digestion tank 360, for example.

[0250] The digestion tank 360 is a tank that stores, for example, anaerobic bacteria and sludge. The anaerobic bacteria in the digestion tank 360 anaerobically digest (decompose) organic matter in the sludge, including the primary sludge supplied from the thickening tank 340 and the excess sludge supplied from the thickener 350, through a biological reaction to produce digested sludge. The anaerobic bacteria in the digestion tank 360 also produce digestion gases, such as methane gas, during the digestion process.

[0251] The heater 370 heats the excess sludge before it is supplied to the digestion tank 360. Specifically, the heater 370 is, for example, a heat exchanger that heats (raises the temperature of) the excess sludge before it is supplied to the digestion tank 360 using the heat retained in a heat medium (a fluid such as water or thermal oil).

[0252] The control device 1 then acquires the actual PV131 (actual PV131 measured by the monitoring device 2) measured by a measuring device (not shown) provided in the sewage treatment facility 100c at regular intervals, such as every 10 minutes. Specifically, the control device 1 acquires, for example, the concentrations of ammonia nitrogen, nitrite nitrogen, and nitrate nitrogen in the sewage treatment device 320 as the actual PV131. The control device 1 also acquires, for example, the temperature of the sludge in the digestion tank 360 as the actual PV131.

[0253] Furthermore, the control device 1 acquires the actual SV132 (actual SV132 set by the monitoring device 2) set by an operator in the sewage treatment facility 100c at regular intervals, such as every 10 minutes. Specifically, the control device 1 acquires, for example, the aeration amount (aeration amount per unit time) in the sewage treatment device 320 as the actual SV132. Furthermore, the control device 1 acquires, for example, the temperature and amount of the heat medium supplied to the heater 370 (heat medium used to heat the excess sludge supplied to the digester 360) as the actual SV132.

[0254] 3 etc., the control device 1 extracts, for example, a plurality of pieces of training data DP from the acquired actual result PV 131 and actual result SV 132, and generates a time series prediction model MP by using the extracted plurality of pieces of training data DP. The control device 1 also extracts, for example, a plurality of pieces of training data DE from the acquired actual result PV 131 and actual result SV 132, and generates a PV estimation model ME by using the extracted plurality of pieces of training data DE.

[0255] Specifically, the control device 1 generates a PV estimation model ME that estimates one of ammonia nitrogen, nitrite nitrogen, and nitrate nitrogen in the wastewater treatment device 320, for example, using the aeration amount (aeration amount per unit time) in the wastewater treatment device 320 as an input. The control device 1 also generates a PV estimation model ME that estimates the temperature of the sludge in the digestion tank 360, for example, using the temperature and amount of the heat medium supplied to the heater 370 as an input.

[0256] The treatment facility 100 may be, for example, a treatment facility other than the incineration facility 100a, the water purification facility 100b, and the sewage treatment facility 100c (for example, a treatment facility in a chemical plant). [Explanation of symbols]

[0257] 1: Control device 2: Monitoring device 10: Incineration equipment 11: Cutting equipment 12: Dryer 12a: Burner 13: Feeding device 14: Carbonization furnace 15: Dust collector 16: Incinerator 16a: Burner 17: Heat exchanger 18: Heat exchanger 19: Heat exchanger 20: Exhaust gas treatment system 100: Treatment equipment 100a: Incineration facility 100b: Water purification facility 100c: Sewage treatment facility 101: CPU 102: Memory 103: Communication device 104: Storage medium 105: Bus 110: Program 111: Data acquisition section 112: Data extraction unit 113: Model generation unit 114: Operation monitoring unit 115: Measurement prediction unit 116: Setting estimation section 130: Storage section 131: Actual PV 131a: First predicted PV 131b: Second predicted PV 132: Intervention target SV 132a: Candidate for intervention SV 210: Settling pond 220: Landing well 230: Mixing pond 240: Flocculation tank 250: Sedimentation tank 260: Filtration pond 270: Purified water pond 280: Distribution reservoir 310: Primary sedimentation tank 320: Sewage treatment equipment 330: Final sedimentation tank 340: Concentrator tank 350: Concentrator 360: Digestion tank 1000: Treatment system DP: Teacher data DE: Teacher data PG: Inverse problem analysis model MP: Learning model ME: Learning model NW: Network

Claims

1. A processing system comprising a processing facility that executes processing according to a set value and a control device that calculates the set value, the set value is a combination of a plurality of types of set values ​​set in the processing equipment, The control device calculating a first estimated value in the processing equipment when the processing equipment executes a process according to the candidate combination by using a first model that performs regression analysis; calculating new candidates for the combination that bring the first estimate closer to a target value by using a second model that performs black-box optimization; Identifying one or more types from the plurality of types based on the calculated first estimate value and the calculated new candidate combination; a processing system that calculates, as the setting value, a new candidate for the combination in which each candidate for the setting value corresponding to the one or more identified types has been changed.

2. The control device repeating the process of calculating the first estimated value and the process of calculating the new candidate combination until the number of times or the execution time of the process of calculating the first estimated value and the process of calculating the new candidate combination satisfy a third condition; identifying one or more types from the plurality of types based on each of the first estimate values ​​that are repeatedly calculated and each of the new candidates for the combination that are repeatedly calculated; repeating the process of calculating the first estimated value and the process of calculating the new candidate for the combination while calculating the new candidate for the combination in which the candidate for each setting value corresponding to the one or more types has been changed, until the number of times that the process of calculating the first estimated value and the process of calculating the new candidate for the combination have been executed satisfies a fourth condition; The processing system according to claim 1 , wherein a new candidate for the combination corresponding to the first estimated value that satisfies a second condition is calculated as the setting value from among the first estimated values ​​that are repeatedly calculated again.

3. The control device calculating, for each of the plurality of types, a likelihood that the first estimated value that satisfies the second condition will be calculated by changing the candidate setting value corresponding to each type, from each of the first estimated values ​​that have been repeatedly calculated and each of the new candidates for the combination that have been repeatedly calculated; The processing system according to claim 2 , wherein a type of the plurality of types for which the calculated certainty satisfies a fifth condition is identified as the one or more types.

4. The treatment system according to claim 1 , wherein the treatment facility is an incineration facility that incinerates materials to be incinerated.

5. The incineration facility comprises: an incinerator for incinerating the material to be incinerated; and a treatment device that treats at least one of the material to be incinerated, the substance discharged from the incinerator, and the fluid discharged from the incinerator by using at least a portion of the energy recovered from the waste heat of the incinerator; The processing system according to claim 4 , wherein the setting value is a setting value set in the processing device.

6. The treatment system according to claim 1 , wherein the treatment facility is a water treatment facility that treats water to be treated.

7. A processing system comprising a processing facility that executes processing according to a set value and a control device that calculates the set value, The control device Acquire measurements measured in the processing equipment; calculating a first predicted value at a second timing later than the first timing from the measured value at the first timing by using another model different from the first model for performing regression analysis and the second model for performing black-box optimization; When the calculated first predicted value does not satisfy a condition, by using the first model, a first estimated value in the treatment equipment when the treatment equipment executes a process according to the candidate setting value is calculated; A processing system that calculates the set point from the first estimate by using the second model.

8. A control device that calculates a setting value to be set in a processing facility, the set value is a combination of a plurality of types of set values ​​set in the processing equipment, The control device calculating a first estimated value in the processing equipment when the processing equipment executes a process according to the candidate combination by using a first model that performs regression analysis; calculating new candidates for the combination that bring the first estimate closer to a target value by using a second model that performs black-box optimization; Identifying one or more types from the plurality of types based on the calculated first estimate value and the calculated new candidate combination; The control device calculates, as the setting values, new candidates for the combination obtained by changing the candidates for the setting values ​​corresponding to the one or more identified types.

9. A processing method for causing a computer to execute a process for calculating a setting value to be set in a processing facility, the set value is a combination of a plurality of types of set values ​​set in the processing equipment, calculating a first estimated value in the processing equipment when the processing equipment executes a process according to the candidate combination by using a first model that performs regression analysis; calculating new candidates for the combination that bring the first estimate closer to a target value by using a second model that performs black-box optimization; Identifying one or more types from the plurality of types based on the calculated first estimate value and the calculated new candidate combination; a processing method for causing a computer to execute a process of changing candidates for each setting value corresponding to the one or more identified types, and calculating new candidates for the combination as the setting value;

Images