Optimization device, optimization method and procedure

By introducing a condition and prediction information acquisition unit into the multi-objective function optimization device, the constraint information is obtained for optimization, which solves the problem of poor calculation accuracy caused by unsuitable constraint functions in the prior art and achieves higher precision optimization effect.

CN123076877APending Publication Date: 2026-11-10MITSUBISHI ELECTRIC CORP
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202480090662.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-04-16
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing technologies, devices used for multi-objective function optimization often suffer from poor computational accuracy due to unsuitable constraint functions.

Method used

The system employs a condition acquisition unit, a prediction information acquisition unit, and a constraint information acquisition unit. It acquires conditional and prediction information through a learning model, obtains constraint information based on the prediction information, and optimizes a multi-objective function.

Benefits of technology

It improves the computational accuracy of multi-objective function optimization, enabling more appropriate optimization, approaching the operation of a skilled operator, suitable for the current problem, and capable of handling unknown constraints.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123076877A_ABST
    Figure CN123076877A_ABST
Patent Text Reader

Abstract

The optimization device (2) is configured to include: a condition acquisition unit (11) that acquires the conditions used in the optimization calculation; a prediction information acquisition unit (12) that acquires prediction information as information based on condition prediction; a restriction information acquisition unit (13) that acquires restriction information for imposing restrictions on the optimization calculation based on the prediction information; and an optimization unit (14) that optimizes a multi-objective function based on the conditions and restriction information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This disclosure relates to optimization apparatus, optimization methods, and procedures. Background Technology

[0002] There exists an optimization apparatus for optimizing a multi-objective function that serves as multiple objective functions.

[0003] For example, Patent Document 1 discloses an apparatus for optimizing a multi-objective function based on a constraint function.

[0004] Patent Document 1: Japanese Patent Application Publication No. 2013-120521

[0005] The constraint function used in the device disclosed in Patent Document 1 is pre-designed and may not be suitable for the conditions used in optimized calculations. Therefore, there is a problem that the accuracy of optimized calculations may sometimes deteriorate. Summary of the Invention

[0006] This disclosure is made to solve the problems mentioned above, and aims to obtain an optimization apparatus that can perform multi-objective function optimization more appropriately.

[0007] The optimization apparatus disclosed herein is an apparatus for optimizing a multi-objective function that is multiple objective functions. The optimization apparatus includes: a condition acquisition unit that acquires conditions used in the optimization calculation; a prediction information acquisition unit that acquires prediction information as information based on condition prediction; a constraint information acquisition unit that, based on the prediction information, acquires constraint information for imposing constraints on the optimization calculation; and an optimization unit that optimizes the multi-objective function based on the conditions and constraint information.

[0008] According to this disclosure, it is possible to optimize multi-objective functions more appropriately. Attached Figure Description

[0009] Figure 1 This is a structural diagram showing the optimization device 2 involved in Implementation Method 1.

[0010] Figure 2 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Implementation Method 1.

[0011] Figure 3 This is a hardware structure diagram of a computer when the optimization device 21 is implemented by software or firmware.

[0012] Figure 4 This is an illustrative diagram illustrating an example of a wastewater treatment plant that performs inflow and water quality prediction.

[0013] Figure 5This is an illustrative diagram representing an example of the optimization of objective functions f1(x) and f2(x).

[0014] Figure 6 It means Figure 1 The processing sequence of the optimization device 2 shown is a flowchart of the optimization method.

[0015] Figure 7 This is an explanatory diagram representing an example of predicted information and an example of a solution representing the optimization result.

[0016] Figure 8A This is an explanatory diagram showing an example of a solution represented in mere notation of the optimization results of a multi-objective function based on the device disclosed in Patent Document 1. Figure 8B It means based on Figure 1 The diagram illustrates an example of a solution representing the optimization result of the multi-objective function of the optimization device 2 shown.

[0017] Figure 9 This is a structural diagram showing the other optimization device 2 involved in Implementation 1.

[0018] Figure 10 These are explanatory diagrams showing an example of constraint information when the objective function variable is the predicted water level, and another example of constraint information when the objective function value is the pumping volume.

[0019] Figure 11 This is an explanatory diagram showing the optimization results of a multi-objective function under the constraint of changing water level within a limited range.

[0020] Figure 12 This is a structural diagram showing the optimization device 2 involved in Embodiment 3.

[0021] Figure 13 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Implementation Method 3.

[0022] Figure 14 This is an explanatory diagram representing the objective function space configured with prediction information.

[0023] Figure 15 This is an explanatory diagram representing preference range information obtained based on a reference point.

[0024] Figure 16 This is a structural diagram showing the optimization device 2 involved in embodiment 4.

[0025] Figure 17 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Embodiment 4.

[0026] Figure 18 This is a flowchart showing the processing sequence of the optimization unit 16.

[0027] Figure 19 This is a structural diagram showing the optimization device 2 involved in embodiment 5.

[0028] Figure 20 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Implementation 5.

[0029] Figure 21 This is a structural diagram showing the optimization device 2 involved in embodiment 6.

[0030] Figure 22 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Implementation 6.

[0031] Figure 23 This is a structural diagram showing the optimization device 2 involved in Embodiment 7.

[0032] Figure 24 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Embodiment 7. Detailed Implementation

[0033] The following description, in order to illustrate the present disclosure in more detail, describes the form used to implement the present disclosure in accordance with the accompanying drawings.

[0034] Implementation Method 1

[0035] Figure 1 This is a structural diagram showing the optimization device 2 involved in Implementation Method 1.

[0036] Figure 2 This is a hardware structure diagram showing the hardware of the optimization device 2 involved in Implementation Method 1.

[0037] exist Figure 1 In this context, learning model 1 is a learned model that is given the conditions used in the optimized computation and the information based on the conditions and predictions, i.e., the prediction information, as learning data during the learning process, and learns the correspondence between the conditions and the prediction information.

[0038] When learning model 1 makes inferences, it outputs prediction information corresponding to the conditions used in the optimized computation.

[0039] The optimization device 2 is a device for optimizing a multi-objective function, and includes a condition acquisition unit 11, a prediction information acquisition unit 12, a constraint information acquisition unit 13, and an optimization unit 14.

[0040] Condition acquisition section 11, for example, by Figure 2 The conditions shown are obtained by circuit 31.

[0041] The condition acquisition unit 11 acquires the conditions used in the optimization calculation.

[0042] The condition acquisition unit 11 outputs the conditions to the prediction information acquisition unit 12 and the optimization unit 14 respectively.

[0043] For example, the predictive information acquisition unit 12 is composed of... Figure 2 The prediction information acquisition circuit 32 shown is implemented.

[0044] The prediction information acquisition unit 12 acquires the conditions used in the optimization calculation from the condition acquisition unit 11.

[0045] The prediction information acquisition unit 12 acquires information based on conditional prediction, i.e., prediction information.

[0046] Specifically, the prediction information acquisition unit 12 provides conditions to the learning model 1 and obtains prediction information from the learning model 1.

[0047] The prediction information acquisition unit 12 outputs the prediction information to the restriction information acquisition unit 13.

[0048] Restricted Information Access Department 13, for example, by Figure 2 The restriction information acquisition circuit 33 shown is implemented.

[0049] The restricted information acquisition unit 13 acquires forecast information from the forecast information acquisition unit 12.

[0050] The restriction information acquisition unit 13 acquires restriction information for imposing restrictions on the optimization calculation based on the prediction information.

[0051] The restriction information acquisition unit 13 outputs the restriction information to the optimization unit 14.

[0052] Optimization unit 14, for example, is composed of Figure 2 The optimized circuit 34 shown is implemented.

[0053] The optimization unit 14 obtains conditions from the condition acquisition unit 11 and obtains restriction information from the restriction information acquisition unit 13.

[0054] The optimization unit 14 optimizes the multi-objective function based on the conditions and constraints.

[0055] The optimization unit 14, for example, outputs the optimization results of the multi-objective function to the outside.

[0056] exist Figure 1 In the envisioned optimization device 2, the constituent elements namely the condition acquisition unit 11, the prediction information acquisition unit 12, the constraint information acquisition unit 13, and the optimization unit 14 are respectively composed of… Figure 2 The dedicated hardware implementation shown is as follows. That is, the optimization device 2 is envisioned to be implemented by a condition acquisition circuit 31, a prediction information acquisition circuit 32, a constraint information acquisition circuit 33, and an optimization circuit 34.

[0057] The condition acquisition circuit 31, the prediction information acquisition circuit 32, the constraint information acquisition circuit 33, and the optimization circuit 34 are respectively equivalent to a single circuit, a composite circuit, a program-based instruction execution processor, a parallel program-based instruction execution processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0058] The components of the optimization device 2 are not limited to being implemented by dedicated hardware; the optimization device 2 can also be implemented by software, firmware, or a combination of software and firmware.

[0059] Software or firmware is stored as a program in the computer's memory. A computer refers to the hardware that executes programs, such as CPU (Central Processing Unit), GPU (Graphics Processing Unit), central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor).

[0060] Figure 3 This is a hardware structure diagram of a computer when the optimization device 2 is implemented by software or firmware.

[0061] When the optimization device 2 is implemented by software or firmware, a program for executing the respective processing order of the computer's condition acquisition unit 11, prediction information acquisition unit 12, restriction information acquisition unit 13, and optimization unit 14 is stored in the memory 51. Then, the computer's processor 52 executes the program stored in the memory 51.

[0062] In addition, Figure 2 The diagram shows an example where the components of the optimization device 2 are implemented by dedicated hardware. Figure 3 The example shown is an optimization device 2 implemented by software or firmware. However, this is just one example; it is also possible that some components of the optimization device 2 are implemented by dedicated hardware, while the remaining components are implemented by software or firmware.

[0063] As the application object of optimization performed by optimization device 2, for example, Figure 4 As shown, there are objective functions for predicting the inflow of water into the sewage treatment plant and for optimizing the objective function for predicting the water quality of water discharged from the sewage treatment plant into rivers, etc.

[0064] Figure 4 This is an illustrative diagram illustrating an example of a wastewater treatment plant that performs inflow and water quality prediction.

[0065] As the application object optimized by the optimization device 2, it is not limited to Figure 4 The objects shown, for example, also include the optimization of objective functions for predicting the electrical power consumed in an air conditioning unit and objective functions for predicting the room temperature adjusted by the air conditioning unit.

[0066] In addition, as an application object to be optimized by the optimization device 2, there are, for example, objective functions for predicting the integration degree of multiple components disposed on the substrate and objective functions for predicting the cost of wiring multiple components disposed on the substrate.

[0067] Figure 5 This is an illustrative diagram representing an example of the optimization of objective functions f1(x) and f2(x).

[0068] exist Figure 5 In the example, the objective function f1(x) is the function for predicting risk, and the objective function f2(x) is the function for predicting cost.

[0069] exist Figure 5 In the diagram, ○ represents the values ​​of the objective function f1(x) and the objective function f2(x) when x is a certain value.

[0070] exist Figure 5 In the example, the values ​​of objective function f1(x) and objective function f2(x) are trade-offs: when the value of objective function f1(x) increases, the value of objective function f2(x) decreases. Conversely, when the value of objective function f1(x) decreases, the value of objective function f2(x) increases.

[0071] exist Figure 5 In the diagram, the solid line represents the Pareto front, and the circle on the Pareto front represents the optimized value of the multi-objective function, i.e., the Pareto solution.

[0072] Next, the explanation Figure 1 The operation of the optimization device 2 shown.

[0073] Figure 6 It means Figure 1 The processing sequence of the optimization device 2 shown is a flowchart of the optimization method.

[0074] The condition acquisition unit 11 acquires conditions for use in the optimization calculation, for example, conditions at multiple times. Figure 6 Step ST1).

[0075] The application objects optimized by the optimization device 2 are, for example, Figure 4As shown, if the objective function is to predict the inflow water and the objective function is to predict the water quality, then the condition acquisition unit 11 obtains the observation values ​​at multiple times as a condition used in the optimization calculation, for example, as shown below.

[0076] • Observations from rainfall sensors installed in rivers or other water bodies upstream of wastewater treatment plants

[0077] • Observations from a pumping speed sensor that monitors the pumping speed of pumps drawing water from the sewage treatment plant's pumping wells into the water treatment tanks.

[0078] • Observed values ​​from the water level sensor that monitors the water level in the water treatment tank

[0079] The condition acquisition unit 11 outputs the acquired conditions to the prediction information acquisition unit 12 and the optimization unit 14 respectively.

[0080] The prediction information acquisition unit 12 acquires the conditions used in the optimization calculation from the condition acquisition unit 11.

[0081] The prediction information acquisition unit 12 acquires prediction information based on conditions. Figure 6 Step ST2).

[0082] Specifically, the prediction information acquisition unit 12 provides conditions to the learning model 1. When the learning model 1 is given conditions from the prediction information acquisition unit 12, it outputs prediction information corresponding to the conditions to the prediction information acquisition unit 12.

[0083] The prediction information acquisition unit 12 acquires the prediction information output from the learning model 1. As prediction information, such as... Figure 7 As shown, for example, to obtain timing information.

[0084] Figure 7 This is an explanatory diagram showing an example of predictive information and an example of a solution representing the optimization result. Figure 7 Examples of predicted pumping capacity and water level are provided.

[0085] The prediction information acquisition unit 12 outputs the prediction information to the restriction information acquisition unit 13.

[0086] The restricted information acquisition unit 13 acquires forecast information from the forecast information acquisition unit 12.

[0087] The constraint information acquisition unit 13 acquires constraint information for imposing constraints on the optimization calculation based on the prediction information. Figure 6 Step ST3).

[0088] As a method for obtaining constraint information, one could be to provide the constraint information to a regression model that has a formula representing the correspondence between the predicted information and the constraint information, thereby obtaining the constraint information from the regression model; or one could obtain the constraint information corresponding to the predicted information from a database that represents the correspondence between the predicted information and the constraint information.

[0089] As a constraint information corresponding to the prediction information, there is, for example, the magnitude of change of the predicted value at each time point shown in the prediction information. The magnitude of change of the predicted value represents the range between the upper and lower limits of each predicted value at each time point.

[0090] The restriction information acquisition unit 13 outputs the restriction information to the optimization unit 14.

[0091] The optimization unit 14 obtains conditions from the condition acquisition unit 11 and obtains restriction information from the restriction information acquisition unit 13.

[0092] The optimization unit 14 optimizes a multi-objective function based on the conditions and constraints. Figure 6 Step ST4).

[0093] Specifically, the optimization unit 14 extracts conditions that include the variation range of the predicted value from the conditions obtained by the self-condition acquisition unit 11 at multiple times. Conditions that include the variation range of the predicted value are more likely to be suitable for use in the optimization calculation. On the other hand, conditions that are not included in the variation range of the predicted value are less likely to be suitable for use in the optimization calculation.

[0094] Then, the optimization unit 14 uses the extracted conditions to optimize the multi-objective function.

[0095] The optimization unit 14, for example, outputs the optimization results of the multi-objective function to the outside.

[0096] Figure 8A This is an explanatory diagram representing an example of a solution that represents the optimization result of a multi-objective function based on the device disclosed in Patent Document 1.

[0097] Figure 8B It means based on Figure 1 The diagram illustrates an example of a solution representing the optimization result of the multi-objective function of the optimization device 2 shown.

[0098] Based on Figure 1 The solution represented by the optimization result of the multi-objective function of the optimization device 2 shown is a solution with higher computational accuracy compared to the solution represented by the optimization result of the multi-objective function based on the device disclosed in Patent Document 1.

[0099] In Embodiment 1 described above, the optimization apparatus 2 is configured to include a condition acquisition unit 11 for acquiring conditions used in the optimization calculation, a prediction information acquisition unit 12 for acquiring prediction information as information based on condition prediction, a restriction information acquisition unit 13 for acquiring restriction information for imposing restrictions on the optimization calculation based on the prediction information, and an optimization unit 14 for optimizing a multi-objective function based on the conditions and restriction information. Therefore, the optimization apparatus 2 is capable of performing multi-objective function optimization more appropriately.

[0100] Furthermore, by optimizing based on constraint information, it is possible to achieve an optimization suitable for this problem, similar to that of a skilled operator.

[0101] In Implementation 1, the optimization device 2 is configured such that the prediction information is the information output from the learning model 1 given the input conditions. Therefore, the optimization device 2 is able to obtain prediction information corresponding to the conditions.

[0102] In addition, the optimization device 2 can cope with unknown limiting information by making predictions through machine learning.

[0103] In Implementation 1, the optimization device 2 is configured such that the predicted information is time-series information. Therefore, the optimization device 2 can perform optimization of a multi-objective function whose value changes over time. As a result, the optimization device 2 can be applied to wastewater treatment plants and other facilities where time-series prediction is highly necessary.

[0104] exist Figure 1 In the optimization device 2 shown, the prediction information acquisition unit 12 acquires the information output from the learning model 1, which has been given input conditions, as prediction information. However, this is just one example; the prediction information acquisition unit 12 can also be as follows: Figure 9 As shown, prediction information is obtained from storage unit 3 based on conditions.

[0105] Figure 9 This is a structural diagram showing the other optimized device 2 involved in Embodiment 1. Figure 9 In, with Figure 1 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0106] Storage unit 3 stores the prediction information corresponding to the conditions.

[0107] Even if the optimization device 2 is configured such that the prediction information acquisition unit 12 obtains prediction information from the storage unit 3 based on conditions, it is still possible to obtain prediction information corresponding to the conditions. In addition, the predictive information acquisition unit 12 improves explanatory power by obtaining prediction information corresponding to similar past operations from the storage unit 3.

[0108] exist Figure 1In the optimization apparatus 2 shown, the restriction information acquisition unit 13 acquires restriction information for imposing restrictions on the optimization calculation based on the prediction information. The restriction information acquisition unit 13 can also accept changes to the prediction information acquired by the prediction information acquisition unit 12 and acquire restriction information based on the changed prediction information. In this case, the restriction information acquisition unit 13 has a user interface for accepting change operations performed by the user, or an input / output interface for receiving change operations of prediction information from an external source, and modifies the prediction information acquired by the prediction information acquisition unit 12.

[0109] By configuring the optimization device 2 such that the restriction information acquisition unit 13 accepts changes to the prediction information acquired by the prediction information acquisition unit 12 and acquires restriction information based on the changed prediction information, an optimization result that reflects the user's intention can be obtained. As a result, operation that closely approximates the operator's preferences can be achieved.

[0110] Implementation method 2.

[0111] In Implementation 2, it is described that the prediction information includes at least one of the following: information for predicting the variables of the objective function, information for predicting the objective function, and information for predicting the constraint function for constraint optimization.

[0112] The structure of the optimization device 2 in Embodiment 2 is the same as that in Embodiment 1. The structural diagram of the optimization device 2 in Embodiment 2 is shown below. Figure 1 or Figure 9 .

[0113] When the prediction information acquisition unit 12 acquires prediction information containing information on the prediction of variables of the objective function, the restriction information acquisition unit 13 acquires restriction information based on the information on the prediction of variables of the objective function.

[0114] As an objective function, for example, there is an objective function whose value is water level or an objective function whose value is water volume. As variables of the objective function, for example, there are predicted values ​​of water level or predicted values ​​of water volume.

[0115] When the prediction information acquisition unit 12 acquires prediction information containing information on the prediction of variables of the objective function, the restriction information acquisition unit 13 acquires restriction information based on the information on the prediction of variables of the objective function.

[0116] If the variables in the objective function are predicted values ​​such as water level, then as follows: Figure 10 As shown, as limiting information, there is, for example, information used to limit the predicted values ​​such as water level to a certain range from the conditions given to the optimization unit 14, i.e., the observed values.

[0117] Figure 10These are explanatory diagrams showing an example of constraint information when the objective function variable is the predicted water level, and an example of constraint information when the objective function value is the pumping volume.

[0118] When the prediction information acquisition unit 12 acquires prediction information that includes information for predicting the objective function, the restriction information acquisition unit 13 acquires restriction information based on the information for predicting the objective function.

[0119] When the function value is an objective function such as the water lifting volume, for example... Figure 10 As shown, as limiting information, there is, for example, information for limiting the predicted value of water pumping volume, etc., to a certain range from the conditions given to the optimization unit 14, i.e., the observed values.

[0120] When the prediction information acquisition unit 12 acquires prediction information that includes information on predicting the constraint function for constraint optimization, the restriction information acquisition unit 13 acquires restriction information based on the information on predicting the constraint function.

[0121] When the constraint function is a function that outputs a range of values ​​that restricts the variables of the objective function, such as the predicted value of the water level, to a certain range, the constraint information may include, for example, information for extracting observations within the constraint range from the conditions given to the optimization unit 14, i.e., the observations.

[0122] The constraint information obtained by the constraint information acquisition unit 13 can be any constraint conditions that constrain the optimization calculation. The constraint conditions are conditions that constrain the objective function, or the variables of the objective function, i.e., the descriptive variables, in a way that makes them included within the range set based on the prediction information, i.e., the constraint range.

[0123] In this case, the optimization unit 14, as follows Figure 11 As shown, optimization calculations are performed within constraints, such as imposing constraints that cause changes in water level or pumping volume.

[0124] Figure 11 This is an explanatory diagram showing the optimization results of a multi-objective function under the constraint of changing water level within a limited range.

[0125] In Embodiment 2 described above, the optimization apparatus 2 is configured such that the prediction information includes at least one of the following: information for predicting the variables of the objective function, information for predicting the objective function itself, and information for predicting the constraint function for constraint optimization. Therefore, the optimization apparatus 2 can more appropriately perform multi-objective function optimization and can obtain constraint information based on various prediction information. Since any one of the information for predicting the variables of the objective function, the information for predicting the objective function itself, and the information for predicting the constraint function for constraint optimization is important, by including at least one of these pieces of information in the prediction information, constraint information useful for optimization can be obtained.

[0126] In Implementation 2, the optimization device 2 is configured such that the limiting information serves as a constraint on the optimized calculation. Therefore, the optimization device 2 is capable of imposing constraints on the optimized calculation.

[0127] In Implementation 2, the optimization device 2 is configured such that the limiting conditions are such that the objective function or the description variables of the objective function are included within a range set based on prediction information. Therefore, the optimization device 2 can perform optimization calculations within the constrained range.

[0128] Implementation method 3.

[0129] In Implementation 3, we will describe the case where the limiting information is information that represents the range that multiple objective functions can achieve in the objective function space, based on the prediction information.

[0130] Figure 12 This is a structural diagram showing the optimization device 2 involved in Embodiment 3. Figure 12 In the middle, to and Figure 1 or Figure 9 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0131] Figure 13 This is a hardware structure diagram illustrating the hardware of the optimization device 2 according to Embodiment 3. Figure 13 In the middle, to and Figure 2 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0132] The restricted information acquisition unit 15 acquires forecast information from the forecast information acquisition unit 12.

[0133] The restriction information acquisition unit 15, based on the prediction information, acquires information representing the range that multiple objective functions can achieve in the objective function space, using restriction information as the restriction information.

[0134] The restriction information acquisition unit 15 outputs the restriction information to the optimization unit 14.

[0135] Figure 12 The optimization device 2 shown applies the restriction information acquisition unit 15 to... Figure 1 The optimization device 2 shown is an example. However, this is just one example; the information acquisition restriction unit 15 can also be applied to... Figure 9 The optimization device 2 shown.

[0136] Restricted Information Access Department 15, for example, by Figure 12 The restriction information acquisition circuit 35 shown is implemented.

[0137] exist Figure 12 In the envisioned optimization device 2, the constituent elements namely the condition acquisition unit 11, the prediction information acquisition unit 12, the constraint information acquisition unit 15, and the optimization unit 14 are respectively composed of… Figure 13 The dedicated hardware implementation shown is as follows. That is, the optimization device 2 is envisioned to be implemented by a condition acquisition circuit 31, a prediction information acquisition circuit 32, a constraint information acquisition circuit 35, and an optimization circuit 34.

[0138] The condition acquisition circuit 31, the prediction information acquisition circuit 32, the constraint information acquisition circuit 35, and the optimization circuit 34 are respectively equivalent to a single circuit, a composite circuit, a processor for program-based instruction execution, a processor for parallel program-based instruction execution, an ASIC, an FPGA, or a circuit composed of these.

[0139] The components of the optimization device 2 are not limited to being implemented by dedicated hardware; the optimization device 2 can also be implemented by software, firmware, or a combination of software and firmware.

[0140] When the optimization device 2 is implemented by software or firmware, the program for determining the processing order of the computer execution condition acquisition unit 11, prediction information acquisition unit 12, restriction information acquisition unit 15, and optimization unit 14 is stored in [the software]. Figure 3 In the memory 51 shown. Then, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0141] in addition, Figure 13 This illustrates an example where the components of optimization device 2 are implemented using dedicated hardware. Figure 3 This example illustrates that the optimization device 2 is implemented by software or firmware. However, this is just one example; it is also possible that some components of the optimization device 2 are implemented by dedicated hardware, while the remaining components are implemented by software or firmware.

[0142] Next, the explanation Figure 12 The operation of the optimization device 2 shown. Except for the restriction information acquisition unit 15, the rest are the same as... Figure 1 The optimization device 2 shown is the same.

[0143] The restricted information acquisition unit 15 acquires forecast information from the forecast information acquisition unit 12.

[0144] The restriction information acquisition unit 15 acquires information (hereinafter referred to as "preference range information") representing the range that multiple objective functions can achieve in the objective function space, based on the prediction information, as restriction information.

[0145] The preference range information is set based on the predicted values ​​among multiple predicted values ​​shown in the prediction information, such as the predicted value closest to the Pareto front. Specifically, it is set based on the function value of the objective function f1(x) corresponding to the predicted value closest to the Pareto front, and the function value of the objective function f2(x) corresponding to that predicted value.

[0146] Figure 14 This is an explanatory diagram representing the objective function space configured with prediction information.

[0147] exist Figure 14 In the diagram, the horizontal axis represents the function value of the objective function f1(x), and the vertical axis represents the function value of the objective function f2(x).

[0148] The solid line represents the Pareto front, ● represents the predicted value shown in the prediction information, and ○ represents the solution existing on the Pareto front.

[0149] exist Figure 14 In the left figure, among the function values ​​of the objective function f1(x) existing on the Pareto front, the function value of the objective function f1(x) closest to the predicted value is 0.4, and among the function values ​​of the objective function f2(x) existing on the Pareto front, the function value of the objective function f2(x) closest to the predicted value is 0.5. Therefore, in Figure 14 In the example in the left figure, the preference region of the objective function space, which is shown as the preference range information, is set as the range where the function value of the objective function f1(x) is approximately 0.1 to 0.4 and the function value of the objective function f2(x) is approximately 0.1 to 0.5.

[0150] The restriction information acquisition unit 15 outputs the preference range information to the optimization unit 14 as restriction information.

[0151] The optimization unit 14 optimizes a multi-objective function based on conditions and preference range information, which serves as constraint information. The optimization process performed by the optimization unit 14 is the same as in Embodiment 1, so detailed descriptions are omitted.

[0152] The restriction information acquisition unit 15 has a human-machine interface for accepting changes to preference range information made by the user, or an input / output interface for receiving change information of preference range information from the outside.

[0153] When a user wishes to change the preference range information, the restriction information acquisition unit 15 accepts the change of preference range information and outputs the changed preference range information to the optimization unit 14 as restriction information.

[0154] The following is a detailed explanation of the change processing of preference range information performed by the restricted information acquisition unit 15.

[0155] First, when the information acquisition unit 15 acquires the predicted value shown in the prediction information, it configures the predicted value in the objective function space, for example, as follows: Figure 14 As shown, the objective function space configured with predicted values ​​is displayed on the touch panel (omitted in the diagram).

[0156] If a user's preferred region is not the region they prefer displayed in the objective function space of a touch panel or similar device, they can set other regions as preferred conditions.

[0157] For example, if the region where the function value of the objective function f1(x) is about 0.0 or higher and the function value of the objective function f1(x) is about 0.2 or lower is used as the preference condition, if the user touches the region on the touch panel corresponding to the preference condition, the restriction information acquisition unit 15 acquires preference range information representing the region corresponding to the preference condition.

[0158] Figure 14 The right figure shows the region corresponding to the preference conditions, as indicated by the preference range information. That is, the preference region is changed to the range where the function value of the objective function f1(x) is approximately 0.0 to 0.2 and the function value of the objective function f2(x) is approximately 0.3 to 0.7.

[0159] Here, the restriction information acquisition unit 15 accepts changes to the preference range information and outputs the changed preference range information to the optimization unit 14 as restriction information. However, this is just one example; the restriction information acquisition unit 15 could of course also... Figure 14 The preference range information representing the preference region shown in the left figure is output to the optimization unit 14 as is.

[0160] Alternatively, the restricted information acquisition unit 15 may refuse to accept changes to the preference range information and instead explore the preference area through optimization processing.

[0161] exist Figure 14 The example shown is an instance where the constraint information represents the range that multiple objective functions can achieve in the objective function space based on the prediction information. However, this is just one example; the constraint information acquisition unit 15 can also use the constraint information as a reference point in the objective function space corresponding to the prediction information, and acquire preference range information based on the reference point.

[0162] At this time, the information acquisition unit 15 can also determine the range that multiple objective functions can obtain based on a reference point, or it can determine the range that one objective function can obtain, or the range that each objective function can obtain, based on multiple reference points.

[0163] The following describes in detail the process of obtaining preference range information by the restriction information acquisition unit 15.

[0164] The restriction information acquisition unit 15 randomly selects one or more predicted values ​​from the multiple predicted values ​​shown in the prediction information, and uses the selected one or more predicted values ​​as reference points respectively.

[0165] Figure 15 This is an explanatory diagram representing preference range information obtained based on a reference point.

[0166] exist Figure 15 In the left graph, ○ represents the predicted value shown in the prediction information. Figure 15 In the left figure, four predicted values ​​are shown. Among the four predicted values, the second and fourth predicted values ​​from the left are selected and are referred to as reference points. Hereinafter, the second predicted value from the left will be referred to as reference point (1), and the fourth predicted value from the left will be referred to as reference point (2).

[0167] Here, for convenience, two reference points (1) and (2) are set to illustrate that the range that the objective function f1(x) can achieve is determined by reference point (1). Figure 5 Examples of the range that the objective function f2(x) can achieve are expressed as "preference interval" and determined by the reference point (2).

[0168] The information acquisition unit 15 restricts the use of reference point (1) as a reference point to determine the range that the objective function f1(x) can achieve, and reference point (2) as a reference point to determine the range that the objective function f2(x) can achieve.

[0169] Then, the information acquisition unit 15 determines the region to which the reference point (1) belongs from among the multiple regions that can be used as preferences in the function values ​​of the objective function f1(x) as the range that the objective function f1(x) can obtain.

[0170] There are multiple regions that can be used as preferences, such as regions with a value of 0.0 or higher and less than 0.3, regions with a value of 0.1 or higher and less than 0.4, regions with a value of 0.2 or higher and less than 0.5, regions with a value of 0.3 or higher and less than 0.6, regions with a value of 0.4 or higher and less than 0.7, regions with a value of 0.5 or higher and less than 0.8, regions with a value of 0.6 or higher and less than 0.9, or regions with a value of 0.7 or higher and less than 1.0.

[0171] In this case, the information acquisition unit 15 restricts the range that the objective function f1(x) can obtain from multiple regions that can be used as preferences, and determines the region to which the reference point (1) belongs, that is, the region above 0.1 and less than 0.4.

[0172] In addition, the information acquisition unit 15 determines the region to which the reference point (2) belongs from multiple regions that can be used as preferences among the function values ​​of the objective function f2(x) to be the range that the objective function f2(x) can obtain.

[0173] There are multiple regions that can be used as preferences, such as regions with a value of 0.0 or higher and less than 0.3, regions with a value of 0.1 or higher and less than 0.4, regions with a value of 0.2 or higher and less than 0.5, regions with a value of 0.3 or higher and less than 0.6, regions with a value of 0.4 or higher and less than 0.7, regions with a value of 0.5 or higher and less than 0.8, regions with a value of 0.6 or higher and less than 0.9, or regions with a value of 0.7 or higher and less than 1.0.

[0174] In this case, the information acquisition unit 15 restricts the range that the objective function f2(x) can obtain from multiple regions that can be used as preferences to a region that is greater than 0.1 and less than 0.4.

[0175] The restriction information acquisition unit 15 outputs the preference range information, which represents the range that the objective function f1(x) can achieve, and the preference range information, which represents the range that the objective function f2(x) can achieve, to the optimization unit 14.

[0176] The optimization unit 14 optimizes the multi-objective function based on the conditions and the preference range information as limiting information.

[0177] Figure 15 The right figure shows the optimization results of the multi-objective function performed by the optimization unit 14, indicating that the solution of the multi-objective function optimization results is restricted to the preference range.

[0178] Here, reference point (1) is used as the reference point for determining the range that the objective function f1(x) can achieve, and reference point (2) is used as the reference point for determining the range that the objective function f2(x) can achieve. However, this is just one example; reference point (1) can also be used as the reference point for determining the range that the objective function f2(x) can achieve, and reference point (2) can be used as the reference point for determining the range that the objective function f1(x) can achieve.

[0179] Alternatively, reference point (1) can be used as a reference point to determine the range that the objective function f1(x) can achieve, and also as a reference point to determine the range that the objective function f2(x) can achieve.

[0180] Alternatively, reference point (2) can be used as a reference point to determine the range that the objective function f1(x) can achieve, and also as a reference point to determine the range that the objective function f2(x) can achieve.

[0181] The restricted information acquisition unit 15 has a human-machine interface for accepting changes to reference points, or an input / output interface for receiving change information of reference points from external sources.

[0182] If the user wishes to change the reference point (hereinafter referred to as "the first reference point"), the restriction information acquisition unit 15 may also accept the change of the first reference point, and based on the changed reference point, i.e. the second reference point, acquire preference range information representing the range that multiple objective functions can achieve as restriction information.

[0183] Here, it is shown that when a user wishes to change the first reference point, the restriction information acquisition unit 15 accepts the change of the first reference point and obtains preference range information as restriction information based on the changed reference point, i.e., the second reference point. However, this is just one example. Even without the aforementioned human-machine interface or input / output interface, the restriction information acquisition unit 15 can automatically accept the change of the first reference point and obtain preference range information as restriction information based on the changed reference point, i.e., the second reference point.

[0184] The following is a detailed explanation of the reference point change processing performed by the restricted information acquisition unit 15.

[0185] First, the information access restriction department 15, such as Figure 15 As shown in the left figure, the objective function space containing multiple predicted values ​​with the predicted information is displayed on a touch panel (omitted in the figure).

[0186] Users can change a prediction value other than the one set as an arbitrary reference point from among multiple prediction values ​​displayed on a touch panel, etc.

[0187] In the four predicted values, when the second predicted value from the left in the figure is set as the first reference point (1) and the fourth predicted value from the left is set as the first reference point (2), for example, if you want to change the first predicted value from the left to the second reference point (1) and the third predicted value from the left to the second reference point (3), then the user, for example, performs an operation by touching the first predicted value from the left and then touching the third predicted value from the left.

[0188] If the user performs the operation described above, the restriction information acquisition unit 15 will change the first predicted value from the left to the second reference point (1), and the third predicted value from the left to the second reference point (3). However, the above operation is just one example, and the reference point can also be changed through other operations.

[0189] The information acquisition unit 15 restricts the range that the objective function f1(x) can obtain from multiple regions that can be used as preferences, for example, the region to which the second reference point (1) belongs.

[0190] In addition, the information acquisition unit 15 restricts the range that the objective function f2(x) can obtain from multiple regions that can be used as preferences, such as the region to which the second reference point (2) belongs.

[0191] The restriction information acquisition unit 15 outputs the preference range information, which represents the range that the objective function f1(x) can achieve, and the preference range information, which represents the range that the objective function f2(x) can achieve, to the optimization unit 14.

[0192] Furthermore, the optimization of the multi-objective function performed by the optimization unit 14 in Implementation 1 is based on constraint information. Specifically, for example, the pumping volume is explored based on constraint information, and as a result, a solution that satisfies the water level is explored. As a method for such exploration, for example, there is the constraint-based NSGA (Non-dominated Sorting Genetic Algorithm) II.

[0193] The optimization of the multi-objective function performed by the optimization unit 14 in Implementation Method 3 further considers optimization of reference points, etc. Specifically, it is an optimization close to the reference point, without constraints on the pumping volume. As a method for such exploration, R-NSGAII is an example. By utilizing predictions with amplitude, it is also possible to set the predicted values ​​passed to R-NSGAII to multiple points and obtain solutions close to multiple points.

[0194] In the above-described embodiment 3, the optimization device 2 is configured such that the constraint information represents the range that multiple objective functions can achieve in the objective function space based on the prediction information. Therefore, the optimization device 2 can perform multi-objective function optimization more appropriately and can obtain constraint information based on the prediction information.

[0195] In Implementation 3, the optimization device 2 is configured such that the constraint information is defined as information representing the range that multiple objective functions can achieve, with the points in the objective function space corresponding to the prediction information serving as reference points. Therefore, the optimization device 2 can more appropriately optimize multiple objective functions and obtain constraint information based on the reference points corresponding to the prediction information.

[0196] In embodiment 3, the optimization device 2 is configured by setting multiple reference points in the objective function space based on prediction information. Therefore, the optimization device 2 can perform multi-objective function optimization more appropriately and can obtain constraint information based on multiple reference points.

[0197] In embodiment 3, the optimization device 2 is configured such that the constraint information is information representing the range that multiple objective functions can achieve based on a second reference point changed by the user from the first reference point. Therefore, the optimization device 2 can perform multi-objective function optimization more appropriately and can obtain constraint information based on the reference point desired by the user.

[0198] Implementation method 4.

[0199] In Embodiment 4, an optimization device 2 is described, in which the optimization unit 16 optimizes a multi-objective function through evolutionary calculation.

[0200] Figure 16 This is a structural diagram showing the optimization device 2 involved in Embodiment 4. Figure 16 In the middle, to and Figure 1 , Figure 9 or Figure 12 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0201] Figure 17 This is a hardware structure diagram illustrating the hardware of the optimization device 2 according to embodiment 4. Figure 17 In the middle, to and Figure 2 or Figure 13 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0202] The optimization unit 16 obtains conditions from the condition acquisition unit 11 and obtains constraint information representing constraint conditions and constraint information representing preference range information from the constraint information acquisition unit 15.

[0203] The optimization unit 16 optimizes the multi-objective function through evolutionary computation. For example, a genetic algorithm can be used for evolutionary computation.

[0204] Specifically, the optimization unit 16 performs optimization processing on the first group of more than one solution candidate, which is based on the constraint conditions that constrain the optimization calculation contained in the constraint information, and performs optimization processing on the second group of solution candidate that is different from the first solution candidate, which is based on the information contained in the constraint information and the information that the prediction information represents the range that multiple objective functions can achieve in the objective function space, i.e., preference conditions.

[0205] During the optimization process based on evolutionary computation, the optimization unit 16 performs an exchange process between a portion of the first candidate solution group that has undergone constraint-based optimization and a portion of the second candidate solution group that has undergone preference-based optimization.

[0206] Then, the optimization unit 16 performs preference-based optimization on a portion of the first solution candidate group and constraint-based optimization on a portion of the second solution candidate group.

[0207] The optimization unit 16, for example, outputs the optimization results of a multi-objective function to the outside.

[0208] Optimization Unit 16, for example, is composed of Figure 17 The optimized circuit 36 ​​shown is implemented.

[0209] exist Figure 17 In the envisioned optimization device 2, the constituent elements namely the condition acquisition unit 11, the prediction information acquisition unit 12, the constraint information acquisition unit 15, and the optimization unit 16 are respectively composed of… Figure 17 The dedicated hardware implementation shown is as follows. That is, the optimization device 2 is envisioned to be implemented by a condition acquisition circuit 31, a prediction information acquisition circuit 32, a constraint information acquisition circuit 35, and an optimization circuit 36.

[0210] The condition acquisition circuit 31, the prediction information acquisition circuit 32, the constraint information acquisition circuit 35, and the optimization circuit 36 ​​are respectively equivalent to a single circuit, a composite circuit, a processor for program-based instruction execution, a processor for parallel program-based instruction execution, an ASIC, an FPGA, or a circuit composed of these.

[0211] The components of the optimization device 2 are not limited to being implemented by dedicated hardware; the optimization device 2 can also be implemented by software, firmware, or a combination of software and firmware.

[0212] When the optimization device 2 is implemented by software or firmware, the program for determining the processing order of the computer execution condition acquisition unit 11, prediction information acquisition unit 12, restriction information acquisition unit 15, and optimization unit 16 is stored in the software. Figure 3 In the memory 51 shown. Then, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0213] in addition, Figure 17 This illustrates an example where the components of optimization device 2 are each implemented using dedicated hardware. Figure 3 This example illustrates that the optimization device 2 is implemented by software or firmware. However, this is just one example; it is also possible that some components of the optimization device 2 are implemented by dedicated hardware, while the remaining components are implemented by software or firmware.

[0214] Next, the explanation Figure 16 The operation of the optimization device 2 shown. Except for the optimization unit 16, the rest are related to... Figure 12 The optimization device 2 shown is the same. Therefore, only the operation of the optimization unit 16 will be described here.

[0215] Figure 18 This is a flowchart showing the processing sequence of the optimization unit 16.

[0216] The optimization department 16 obtains conditions from the condition acquisition department 11.

[0217] In addition, the optimization unit 16 obtains constraint information from the constraint information acquisition unit 15, which represents the constraint conditions used to impose restrictions on the optimization calculation, and obtains constraint information representing the preference range information.

[0218] The optimization unit 16 performs optimization processing on one or more candidate solutions, namely the first candidate solution group, based on the conditions and constraints obtained from the condition acquisition unit 11. Figure 18 Step ST11). This optimization process is the same as the optimization process based on optimization unit 14 involved in Embodiment 1.

[0219] The optimization unit 16 performs optimization processing on the candidate solutions, i.e., the second candidate solution group, based on the conditions obtained from the condition acquisition unit 11 and the preference range information shown therein. Figure 18 Step ST12). This optimization process is the same as the optimization process based on optimization unit 14 involved in Embodiment 3.

[0220] Before each optimization process in steps ST11 and ST12 is completed, the optimization unit 16 performs an exchange process on a portion of the first candidate solution group that has undergone constraint-based optimization and a portion of the second candidate solution group that has undergone preference-based optimization. Figure 18 Step ST13). The swapping process is equivalent to the so-called crossover process. The swapping process itself is a well-known technique, so a detailed description is omitted.

[0221] After performing the exchange process, the optimization unit 16 performs optimization processing based on preference conditions on a portion of the first candidate solution group. Figure 18 Step ST14), performs constraint-based optimization on a portion of the second candidate solution group ( Figure 18 Step ST15).

[0222] The optimization unit 16, for example, outputs the optimization results of each processed multi-objective function to the outside.

[0223] In Embodiment 4 above, the optimization device 2 is configured as follows: the optimization unit 16 performs multi-objective function optimization through evolutionary computation; the first group of candidate solutions is optimized based on constraint conditions constraining the optimization computation using constraint information; the second group of candidate solutions, which is different from the first group of candidate solutions, is optimized based on preference conditions, which represent the range achievable by multiple objective functions in the objective function space using prediction information using constraint information; and during the evolutionary computation-based optimization process, a portion of the first group of candidate solutions optimized based on constraint conditions and a portion of the second group of candidate solutions optimized based on preference conditions are swapped; the first group of candidate solutions is optimized based on preference conditions, and the second group of candidate solutions is optimized based on constraint conditions. Therefore, the optimization device 2 can perform multi-objective function optimization more appropriately and can improve the accuracy of the multi-objective function optimization results. In Embodiment 4, both the advantages of constraint-based optimization and preference-based optimization can be combined for optimization.

[0224] Implementation method 5.

[0225] In Embodiment 5, an optimization device 2 is described, which includes an output unit 17 that outputs multiple optimization results including the results optimized by the optimization unit 14.

[0226] Figure 19 This is a structural diagram illustrating the optimization device 2 according to Embodiment 5. Figure 19 In the middle, to and Figure 1 , Figure 9 , Figure 12 or Figure 16 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0227] Figure 20 This is a hardware structure diagram illustrating the hardware of the optimization device 2 according to Embodiment 5. Figure 20 In the middle, to and Figure 2 , Figure 13 or Figure 17 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0228] Output unit 17 outputs multiple optimization results, including the results optimized by optimization unit 14, to processing unit 18.

[0229] The processing unit 18 processes the optimization result selected by the user from among the multiple optimization results output from the output unit 17.

[0230] Figure 19 The optimization device 2 shown applies the output unit 17 and the processing unit 18 to respectively Figure 1 The optimization device 2 shown is an example. However, this is only one example; the output unit 17 and the processing unit 18 can also be applied to... Figure 9 The optimization device 2 shown Figure 12 The optimization device 2 shown, or Figure 16 The optimization device 2 shown.

[0231] Output section 17, for example, is composed of Figure 20 The output circuit 37 shown is implemented, and the processing unit 18 is, for example, implemented by... Figure 20 The processing circuit 38 shown is implemented.

[0232] exist Figure 19 In the envisioned optimization device 2, the constituent elements namely the condition acquisition unit 11, the prediction information acquisition unit 12, the constraint information acquisition unit 13, the optimization unit 14, the output unit 17, and the processing unit 18 are respectively composed of… Figure 20 The dedicated hardware implementation shown is as follows. That is, the optimization device 2 is envisioned to be implemented by a condition acquisition circuit 31, a prediction information acquisition circuit 32, a constraint information acquisition circuit 33, an optimization circuit 34, an output circuit 37, and a processing circuit 38.

[0233] The condition acquisition circuit 31, the prediction information acquisition circuit 32, the constraint information acquisition circuit 33, the optimization circuit 34, the output circuit 37, and the processing circuit 38 are respectively equivalent to, for example, a single circuit, a composite circuit, a processor for program-based instruction execution, a processor for parallel program-based instruction execution, an ASIC, an FPGA, or a circuit composed of these.

[0234] The components of the optimization device 2 are not limited to being implemented by dedicated hardware; the optimization device 2 can also be implemented by software, firmware, or a combination of software and firmware.

[0235] When the optimization device 2 is implemented by software or firmware, the program for determining the processing order of each of the computer execution condition acquisition unit 11, prediction information acquisition unit 12, restriction information acquisition unit 13, optimization unit 14, output unit 17, and processing unit 18 is stored. Figure 3 In the memory 51 shown. And, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0236] in addition, Figure 20 Examples are shown where the components of the optimization device 2 are implemented by dedicated hardware. Figure 3 The example shown is an optimization device 2 implemented by software or firmware. However, this is just one example; it is also possible that some components of the optimization device 2 are implemented by dedicated hardware, while the remaining components are implemented by software or firmware.

[0237] Next, the explanation Figure 19The operation of the optimization device 2 shown. Except for the output unit 17 and the processing unit 18, the rest are related to... Figure 1 The optimization device 2 shown is the same. Therefore, only the operation of the output unit 17 and the processing unit 18 will be described here.

[0238] In addition to obtaining the optimization results of the multi-objective function from the optimization unit 14, the output unit 17 also obtains one or more Pareto solutions that exist on the Pareto front and are different from the solutions shown in the optimization results as optimization results.

[0239] Output unit 17 outputs multiple optimization results to processing unit 18.

[0240] The processing unit 18 has a human-machine interface for receiving the selection of optimization results, or an input / output interface for receiving selection information of optimization results from the outside.

[0241] The processing unit 18 obtains multiple optimization results from the output unit 17.

[0242] The processing unit 18 accepts the selection of any one of multiple optimization results based on the selection operation of the optimization results performed by the user.

[0243] The processing unit 18 performs processing based on the optimization results selected by the user. As part of the processing performed by the processing unit 18, if the application of the optimization is a sewage treatment plant, there is, for example, control of the pumping volume, etc., and if the application of the optimization is an air conditioning unit, there is, for example, control of the room temperature of the air conditioning unit, etc.

[0244] In the above-described embodiment 5, the optimization device 2 is configured such that it includes an output unit 17 that outputs multiple optimization results including the results optimized by the optimization unit 14, and a processing unit 18 that processes the optimization results selected by the user from the multiple optimization results output from the output unit 17. Therefore, the optimization device 2 can perform multi-objective function optimization more appropriately and can perform processing based on optimization results that reflect the user's intention.

[0245] Users can select the desired optimization result from multiple optimization results output by output unit 17, and thus choose a solution from multiple optimization results.

[0246] Implementation method 6.

[0247] In Embodiment 6, an optimization device 2 is described, which includes a receiving unit 19 that allows for the selection of whether to optimize a multi-objective function using constraint information or not.

[0248] Figure 21 This is a structural diagram showing the optimization device 2 according to embodiment 6. Figure 21 In the middle, to and Figure 1 , Figure 9 , Figure 12 , Figure 16 or Figure 19 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0249] Figure 22 This is a hardware structure diagram illustrating the hardware of the optimization device 2 according to Embodiment 6. Figure 22 In the middle, to and Figure 2 , Figure 13 , Figure 17 or Figure 20 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0250] The receiving unit 19 accepts the choice between optimizing a multi-objective function using restrictive information or optimizing a multi-objective function without using restrictive information.

[0251] The receiving unit 19 outputs the selected receiving result to the optimization unit 20.

[0252] The optimization unit 20 obtains conditions from the condition acquisition unit 11 and obtains restriction information from the restriction information acquisition unit 13.

[0253] In addition, the optimization unit 20 obtains the selected acceptance results from the acceptance unit 19.

[0254] The optimization unit 20 optimizes the multi-objective function based on the selection received by the receiving unit 19.

[0255] Specifically, if the acceptance result received by the receiving unit 19 represents the selection of the main objective of optimizing a multi-objective function using constraint information, then the optimization unit 20 and... Figure 1 Similarly, the optimization unit 14 shown performs multi-objective function optimization based on condition and constraint information.

[0256] If the acceptance result received by the receiving unit 19 indicates that the selection is based on the principle of not using restrictive information to optimize the multi-objective function, then the optimization unit 20 optimizes the multi-objective function based on the conditions.

[0257] The optimization unit 20, for example, outputs the optimization results of a multi-objective function to the outside.

[0258] Figure 21 The optimization device 2 shown applies the receiving unit 19 and the optimization unit 20 to respectively Figure 1 The optimization device 2 shown is an example. However, this is only one example; the receiving unit 19 and the optimization unit 20 can also be applied to... Figure 9 The optimization device 2 shown Figure 12 The optimization device 2 shown Figure 16 The optimization device 2 shown, or Figure 19The optimization device 2 shown.

[0259] For example, by the receiving department 19 Figure 22 The receiving circuit 39 shown is implemented, and the optimization unit 20 is, for example, made by... Figure 22 The optimized circuit 40 shown is implemented.

[0260] exist Figure 21 In the envisioned optimization device 2, the constituent elements namely the condition acquisition unit 11, the prediction information acquisition unit 12, the restriction information acquisition unit 13, the receiving unit 19, and the optimization unit 20 are respectively composed of… Figure 22 The dedicated hardware implementation shown is as follows. That is, the optimization device 2 is envisioned to be implemented by a condition acquisition circuit 31, a prediction information acquisition circuit 32, a constraint information acquisition circuit 33, a receiving circuit 39, and an optimization circuit 40.

[0261] The condition acquisition circuit 31, the prediction information acquisition circuit 32, the restriction information acquisition circuit 33, the receiving circuit 39, and the optimization circuit 40 are respectively equivalent to, for example, a single circuit, a composite circuit, a processor for program-based instruction execution, a processor for parallel program-based instruction execution, an ASIC, an FPGA, or a circuit composed of these.

[0262] The components of the optimization device 2 are not limited to being implemented by dedicated hardware; the optimization device 2 can also be implemented by software, firmware, or a combination of software and firmware.

[0263] When the optimization device 2 is implemented by software or firmware, the program for determining the processing order of the computer execution condition acquisition unit 11, prediction information acquisition unit 12, restriction information acquisition unit 13, receiving unit 19, and optimization unit 20 is stored in the software. Figure 3 In the memory 51 shown. Then, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0264] in addition, Figure 22 This illustrates an example where the components of optimization device 2 are each implemented using dedicated hardware. Figure 3 This example illustrates that the optimization device 2 is implemented by software or firmware. However, this is just one example; it is also possible that some components of the optimization device 2 are implemented by dedicated hardware, while the remaining components are implemented by software or firmware.

[0265] Next, the explanation Figure 21 The operation of the optimization device 2 shown. Except for the receiving unit 19 and the optimization unit 20, it is related to... Figure 1 The optimization device 2 shown is the same. Therefore, only the operation of the receiving unit 19 and the optimization unit 20 will be described here.

[0266] The receiving unit 19 has a human-machine interface for accepting the choice of whether to optimize a multi-objective function using constraint information or not, or an input / output interface for receiving the aforementioned selection information from the outside.

[0267] The receiving unit 19 accepts the choice of whether to optimize the multi-objective function using constraint information or not, and outputs the selection result to the optimization unit 20.

[0268] The optimization unit 20 obtains conditions from the condition acquisition unit 11 and obtains restriction information from the restriction information acquisition unit 13.

[0269] In addition, the optimization unit 20 obtains the selected acceptance results from the acceptance unit 19.

[0270] If the acceptance result received by the receiving unit 19 represents the selection of the main objective of optimizing a multi-objective function using constraint information, then the optimization unit 20 and Figure 1 Similarly, the optimization unit 14 shown performs multi-objective function optimization based on condition and constraint information.

[0271] If the acceptance result received by the receiving unit 19 indicates that the selection is based on the principle of not using restrictive information to optimize the multi-objective function, then the optimization unit 20 optimizes the multi-objective function based on the conditions.

[0272] The optimization unit 20, for example, outputs the optimization results of a multi-objective function to the outside.

[0273] In Embodiment 6 described above, the optimization device 2 is configured such that it includes a receiving unit 19 that accepts a choice between optimizing a multi-objective function using constraint information or optimizing a multi-objective function without using constraint information, and an optimization unit 20 optimizes a multi-objective function based on the choice made by the receiving unit 19. Therefore, the optimization device 2 can perform multi-objective function optimization more appropriately and can perform optimization using a calculation method that reflects the user's intent. In Embodiment 6, when it is not suitable to optimize a multi-objective function using constraint information, normal multi-objective function optimization can be performed.

[0274] Implementation method 7.

[0275] In Embodiment 7, an optimization device 2 is described, which includes a learning unit 21 that uses the result optimized by the optimization unit 14 as learning data to enable the learning model 1 to learn.

[0276] Figure 23 This is a structural diagram showing the optimization device 2 according to Embodiment 7. Figure 23 In the middle, to and Figure 1 , Figure 9 , Figure 12 , Figure 16 , Figure 19 or Figure 21 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0277] Figure 24 This is a hardware structure diagram illustrating the hardware of the optimization device 2 according to Embodiment 7. Figure 24 In the middle, to and Figure 2 , Figure 13 , Figure 17 , Figure 20 or Figure 22 The same symbols represent the same or equivalent parts, so detailed explanations are omitted.

[0278] The learning unit 21 obtains the optimization results of the multi-objective function from the optimization unit 14.

[0279] The learning unit 21 uses the optimization results of the multi-objective function as learning data to enable the learning model 1 to learn.

[0280] Figure 23 The optimization device 2 shown applies the learning unit 21 to... Figure 1 The optimization device 2 shown is an example. However, this is just one example; the learning unit 21 can also be applied to... Figure 9 The optimization device 2 shown Figure 12 The optimization device 2 shown Figure 16 The optimization device 2 shown Figure 19 The optimization device 2 shown, or Figure 21 The optimization device 2 shown.

[0281] Learning Department 21, for example, by Figure 24 The learning circuit 41 shown is implemented.

[0282] exist Figure 23 In the envisioned optimization device 2, the constituent elements namely the condition acquisition unit 11, the prediction information acquisition unit 12, the constraint information acquisition unit 13, the optimization unit 14, and the learning unit 21 are respectively composed of… Figure 24 The dedicated hardware implementation shown is as follows. That is, the optimization device 2 is envisioned to be implemented by a condition acquisition circuit 31, a prediction information acquisition circuit 32, a constraint information acquisition circuit 33, an optimization circuit 34, and a learning circuit 41.

[0283] The condition acquisition circuit 31, the prediction information acquisition circuit 32, the constraint information acquisition circuit 33, the optimization circuit 34, and the learning circuit 41 are respectively equivalent to a single circuit, a composite circuit, a processor for program-based instruction execution, a processor for parallel program-based instruction execution, an ASIC, an FPGA, or a circuit composed of these.

[0284] The components of the optimization device 2 are not limited to being implemented by dedicated hardware; the optimization device 2 can also be implemented by software, firmware, or a combination of software and firmware.

[0285] When the optimization device 2 is implemented by software or firmware, the program for determining the processing order of each of the computer execution condition acquisition unit 11, prediction information acquisition unit 12, restriction information acquisition unit 13, optimization unit 14, and learning unit 21 is stored in... Figure 3 In the memory 51 shown. Then, Figure 3 The processor 52 shown executes the program stored in the memory 51.

[0286] in addition, Figure 24 This illustrates an example where the components of optimization device 2 are each implemented using dedicated hardware. Figure 3 This example illustrates that the optimization device 2 is implemented by software or firmware. However, this is just one example; it is also possible that some components of the optimization device 2 are implemented by dedicated hardware, while the remaining components are implemented by software or firmware.

[0287] Next, the explanation Figure 23 The operation of the optimization device 2 shown. Except for the learning unit 21, the rest are related to... Figure 1 The optimization device 2 shown is the same. Therefore, only the operation of the learning unit 21 will be described here.

[0288] The learning unit 21 obtains the optimization results of the multi-objective function from the optimization unit 14.

[0289] The learning unit 21 uses the optimization results of the multi-objective function as learning data to enable the learning model 1 to learn.

[0290] Specifically, when the learning unit 21 obtains the optimization result of the multi-objective function, it enables the learning model 1 to learn the correspondence between the conditions output by the condition acquisition unit 11 and the prediction information output by the prediction information acquisition unit 12.

[0291] By performing such learning, when conditions similar to those used to obtain the optimization result of the multi-objective function are obtained by the condition acquisition unit 11, the learning model 1 given such conditions outputs prediction information similar to the prediction information used to obtain the optimization result of the multi-objective function.

[0292] In the above-described embodiment 7, the optimization device 2 is configured such that it includes a learning unit 21 that uses the optimization result from the optimization unit 14 as learning data to enable the learning model 1 to learn. Therefore, the optimization accuracy of the optimization device 2 based on the optimization unit 14 is improved. In other words, by using the optimized result as learning data to enable the learning model 1 to learn, situations where correct inferences cannot be made due to the lack of past data can be avoided.

[0293] Furthermore, this disclosure allows for free combination of various embodiments, modification of any constituent element of each embodiment, or omission of any constituent element in each embodiment.

[0294] [Industry Applicability]

[0295] This disclosure is applicable to the optimization of apparatus, optimization methods and procedures.

[0296] Explanation of reference numerals in the attached figures

[0297] 1: Learning model; 2: Optimization device; 3: Storage unit; 11: Condition acquisition unit; 12: Prediction information acquisition unit; 13: Restriction information acquisition unit; 14: Optimization unit; 15: Restriction information acquisition unit; 16: Optimization unit; 17: Output unit; 18: Processing unit; 19: Receiving unit; 20: Optimization unit; 21: Learning unit; 31: Condition acquisition circuit; 32: Prediction information acquisition circuit; 33: Restriction information acquisition circuit; 34: Optimization circuit; 35: Restriction information acquisition circuit; 36: Optimization circuit; 37: Output circuit; 38: Processing circuit; 39: Receiving circuit; 40: Optimization circuit; 41: Learning circuit; 51: Memory; 52: Processor.

Claims

1. An optimization apparatus for optimizing multiple objective functions, i.e., multi-objective functions, characterized in that, have: The condition acquisition unit acquires the conditions used in the optimization calculation; The prediction information acquisition unit acquires prediction information as information predicted based on the conditions. The restriction information acquisition unit acquires restriction information for imposing restrictions on the optimization calculation based on the prediction information; as well as The optimization unit optimizes the multi-objective function based on the conditions and the constraints.

2. The optimization device according to claim 1, characterized in that, The prediction information is output by the learned model after being fed the conditions.

3. The optimization device according to claim 1, characterized in that, The prediction information acquisition unit obtains the prediction information from the storage unit based on the conditions.

4. The optimization device according to any one of claims 1 to 3, characterized in that, The prediction information is time-series information.

5. The optimization apparatus according to any one of claims 1 to 4, characterized in that, The restriction information acquisition unit accepts changes to the prediction information obtained by the prediction information acquisition unit, and acquires the restriction information based on the changed prediction information.

6. The optimization apparatus according to any one of claims 1 to 5, characterized in that, The prediction information includes at least one of the following: information predicting the variables of the objective function, information predicting the objective function, and information predicting the constraint function constraining the optimization.

7. The optimization apparatus according to any one of claims 1 to 6, characterized in that, The constraint information refers to the constraints that constrain the optimized calculation.

8. The optimization device according to claim 7, characterized in that, The constraint is a condition that restricts the objective function or the descriptive variables of the objective function in a manner that includes them within a range set based on the prediction information.

9. The optimization apparatus according to any one of claims 1 to 5, characterized in that, The constraint information is information representing the range that the multiple objective functions can achieve in the objective function space based on the prediction information.

10. The optimization device according to claim 9, characterized in that, The constraint information is information that uses the points in the objective function space corresponding to the prediction information as reference points, and represents the range that the multiple objective functions can achieve based on the reference points.

11. The optimization apparatus according to claim 10, characterized in that, The reference points are set in multiple ways on the objective function space based on the prediction information.

12. The optimization apparatus according to claim 10 or 11, characterized in that, The reference point is the first reference point. The limitation information is based on the range that the multiple objective functions can achieve, represented by the second reference point after the user changes the reference point from the first reference point.

13. The optimization apparatus according to any one of claims 1 to 12, characterized in that, The optimization unit optimizes the multi-objective function through evolutionary computation. The first candidate solution group, which is a candidate for a solution, is optimized based on the constraints contained in the constraint information that constrain the optimization calculation. The second candidate solution group, which is different from the first candidate solution group, is optimized based on the information contained in the constraint information that represents the range that the multiple objective functions can achieve in the objective function space based on the prediction information, i.e., the preference conditions. Midway through the optimization process based on the evolutionary calculation, a part of the first candidate solution group that has undergone optimization based on the constraint conditions and a part of the second candidate solution group that has undergone optimization based on the preference conditions are swapped. The part of the first candidate solution group is then optimized based on the preference conditions, and the part of the second candidate solution group is optimized based on the constraint conditions.

14. The optimization apparatus according to any one of claims 1 to 13, characterized in that, have: The output unit outputs multiple optimized results from the results optimized by the optimization unit. as well as The processing unit processes the optimization results selected by the user from among the plurality of optimization results output from the output unit.

15. The optimization apparatus according to any one of claims 1 to 14, characterized in that, It includes a receiving unit that accepts the choice of whether to optimize the multi-objective function using the constraint information or not using the constraint information. The optimization unit optimizes the multi-objective function based on the selections accepted by the acceptance unit.

16. The optimization device according to claim 2, characterized in that, It has a learning unit that uses the results optimized by the optimization unit as learning data to enable the learned model to learn.

17. An optimization method for optimizing multiple objective functions, i.e., multi-objective functions, characterized in that, The condition acquisition unit acquires the conditions used in the optimization calculation. The prediction information acquisition unit acquires prediction information as information based on the conditions. Based on the prediction information, the constraint information acquisition unit acquires constraint information for imposing constraints on the optimization calculation, and The optimization unit optimizes the multi-objective function based on the conditions and the constraints.

18. A program for executing on a computer an optimization apparatus that optimizes multiple objective functions, i.e., multi-objective functions, characterized in that, The program is used to cause the computer to perform the following processes: The processing of obtaining the conditions used in the optimized calculation; Processing to obtain prediction information as information based on the conditions; Based on the prediction information, a process is performed to obtain constraint information for imposing restrictions on the optimization calculation; as well as Based on the conditions and constraints, the optimization of the multi-objective function is performed.

Citation Information

Patent Citations

  • Pump operation planning device

    JP2013120521A