CN110689160B - Parameter configuration optimization method and device for large-scale complex system - Google Patents
Parameter configuration optimization method and device for large-scale complex system Download PDFInfo
- Publication number
- CN110689160B CN110689160B CN201910610017.1A CN201910610017A CN110689160B CN 110689160 B CN110689160 B CN 110689160B CN 201910610017 A CN201910610017 A CN 201910610017A CN 110689160 B CN110689160 B CN 110689160B
- Authority
- CN
- China
- Prior art keywords
- fidelity
- low
- gaussian process
- process regression
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/17—Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method
- G06F17/175—Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method of multidimensional data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/445—Program loading or initiating
- G06F9/44505—Configuring for program initiating, e.g. using registry, configuration files
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Pure & Applied Mathematics (AREA)
- Strategic Management (AREA)
- Mathematical Optimization (AREA)
- Human Resources & Organizations (AREA)
- Computational Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Algebra (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Databases & Information Systems (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
A parameter configuration optimization method and device for a large-scale complex system comprise the following steps: step S1, training a low-fidelity Gaussian process regression model according to the low-fidelity experimental observation data; s2, acquiring a high fidelity experimental evaluation point according to a prediction curve output by the low fidelity Gaussian process regression model; step S3, training a high-fidelity Gaussian process regression model according to the high-fidelity experimental data; s4, fusing the prediction output of the low-fidelity Gaussian process regression model and the prediction output of the high-fidelity Gaussian process regression model, determining a next high-fidelity experiment evaluation point S5, and returning to the step S3 to iterate if the high-fidelity experiment times are less than a preset value; otherwise, selecting an optimal point from the high fidelity experimental evaluation points and outputting. The method adopts a Bayesian model fusion mechanism to fully utilize low-fidelity experimental data, and obtains optimal system configuration parameter values with less experimental cost.
Description
Technical Field
The invention belongs to the technical field of data optimization, and particularly relates to a parameter configuration optimization method and device for a large-scale complex system.
Background
In many fields such as industrial production, academic research, meteorological prediction and the like, large-scale complex systems exist, and the operation state of the system depends on the configuration condition of system parameters. The large-scale complex system parameter configuration optimization process has the characteristics of more parameters to be optimized, high evaluation experiment cost, no analytic form of an optimization objective function and no available gradient information in the optimization process, and great challenges are brought to the search of optimal system configuration parameter values. The common non-gradient optimization method comprises manual parameter adjustment and a heuristic optimization search method based on evolutionary computation, if enough good parameter configuration values need to be obtained by using the methods, multiple high-fidelity experiments need to be carried out, and because the cost of each experiment is high, the actually allowed times of the high-fidelity experiments are limited, the parameter configuration values output by the methods are far from the optimum.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a parameter configuration optimization method and device for a large-scale complex system.
The invention provides a parameter configuration optimization method for a large-scale complex system, which comprises the following steps:
step S1, training a low-fidelity Gaussian process regression model according to the low-fidelity experimental observation data;
s2, acquiring a high-fidelity experimental evaluation point according to a prediction curve output by the low-fidelity Gaussian process regression model;
step S3, training a high-fidelity Gaussian process regression model according to the high-fidelity experimental data;
step S4, fusing the prediction output of the low-fidelity Gaussian process regression model and the prediction output of the high-fidelity Gaussian process regression model, and determining the next high-fidelity experiment evaluation point;
s5, if the high fidelity experiment times are less than the preset value, returning to the S3 for iteration; otherwise, selecting an optimal point from the high fidelity experimental evaluation points and outputting.
As a further technical scheme of the invention, the low-fidelity Gaussian process regression model stores and encapsulates low-fidelity experimental data information.
Further, in step S2, the low-fidelity gaussian process regression model predicts the experimental response at any experimental evaluation point, and selects a high-fidelity experimental evaluation point according to the predicted value.
Furthermore, an optimal search model is adopted for selecting the high fidelity experimental evaluation points, and the objective function of the optimal search model is determined according to the predicted value of the low fidelity Gaussian process regression model.
Further, in step S4, the prediction output of the low-fidelity gaussian process regression model and the prediction output of the high-fidelity gaussian process regression model are weighted and fused through a dynamic weighting mechanism based on bayesian.
Furthermore, the fused weighting coefficient is determined by the model prediction value and the actual response value.
The invention also provides a parameter configuration optimization device for the large-scale complex system, which is characterized by comprising a low-fidelity experimental data storage device, a low-fidelity Gaussian process regression prediction device, a high-fidelity data acquisition device, a high-fidelity Gaussian process regression prediction device, a prediction fusion device and an optimal decision device;
the low-fidelity experimental data storage device is used for storing terminal equipment of low-fidelity experimental observation data;
the low-fidelity Gaussian process regression prediction device is used for analyzing and processing low-fidelity experimental observation data and determining the parameter values of a low-fidelity Gaussian process regression model;
high fidelity data acquisition device. The system is used for collecting and storing high-fidelity experimental evaluation data;
the high-fidelity Gaussian process regression prediction device is used for analyzing and processing high-fidelity experimental observation data and determining high-fidelity Gaussian process regression model parameter values;
and the prediction fusion device is used for performing weighted fusion on the low-fidelity Gaussian process regression prediction and the high-fidelity Gaussian process regression prediction and giving a fusion prediction response value to any experimental evaluation point.
And the optimal decision device is used for selecting the optimal next iteration experiment evaluation point according to the fusion prediction response value given by the prediction fusion device.
Further, the low fidelity gaussian process regression prediction device outputs the initial high fidelity experimental evaluation points.
Furthermore, the prediction fusion device and the optimal decision device adopt a dynamic weighting mechanism based on Bayes to carry out weighted fusion on the prediction output of the low-fidelity Gaussian process regression model and the prediction output of the high-fidelity Gaussian process regression model.
Furthermore, the fused weighting coefficient is determined by the model prediction value and the actual response value.
The method of the invention adopts a Bayesian model fusion mechanism to fully utilize low-fidelity experimental data and obtain optimal system configuration parameter values with less experimental cost.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a block diagram of the apparatus of the present invention.
Detailed Description
Referring to fig. 1, the present embodiment provides a method for optimizing parameter configuration for a large-scale complex system, including the following steps:
step S1, training a low-fidelity Gaussian process regression model according to the low-fidelity experimental observation data;
s2, acquiring a high fidelity experimental evaluation point according to a prediction curve output by the low fidelity Gaussian process regression model;
step S3, training a high fidelity Gaussian process regression model according to the high fidelity experimental data;
s4, fusing the prediction output of the low-fidelity Gaussian process regression model and the prediction output of the high-fidelity Gaussian process regression model, and determining the next high-fidelity experiment evaluation point;
step S5, if the high fidelity experiment times are less than the preset value, returning to the step S3 for iteration; otherwise, selecting an optimal point from the high fidelity experimental evaluation points and outputting.
And storing and packaging the low-fidelity experimental data information by the low-fidelity Gaussian process regression model.
In step S2, the low-fidelity gaussian process regression model predicts the experimental response at any experimental evaluation point, and selects a high-fidelity experimental evaluation point according to the predicted value.
And selecting the high fidelity experimental evaluation points by adopting an optimal search model, wherein the target function of the optimal search model is determined according to the predicted value of the low fidelity Gaussian process regression model.
In step S4, the prediction output of the low-fidelity gaussian process regression model and the prediction output of the high-fidelity gaussian process regression model are weighted and fused by a bayesian-based dynamic weighting mechanism.
The fused weighting coefficient is determined by the model predicted value and the actual response value.
As shown in fig. 2, a parameter configuration optimization device for a large-scale complex system is characterized by comprising a low-fidelity experimental data storage device, a low-fidelity gaussian process regression prediction device, a high-fidelity data acquisition device, a high-fidelity gaussian process regression prediction device, a prediction fusion device and an optimal decision device;
the low-fidelity experimental data storage device is used for storing terminal equipment of low-fidelity experimental observation data;
the low-fidelity Gaussian process regression prediction device is used for analyzing and processing low-fidelity experimental observation data and determining the parameter values of a low-fidelity Gaussian process regression model;
high fidelity data acquisition device. The device is used for collecting and storing high fidelity experimental evaluation data;
the high-fidelity Gaussian process regression prediction device is used for analyzing and processing high-fidelity experimental observation data and determining high-fidelity Gaussian process regression model parameter values;
and the prediction fusion device is used for performing weighted fusion on the low-fidelity Gaussian process regression prediction and the high-fidelity Gaussian process regression prediction and giving a fusion prediction response value to any experimental evaluation point.
And the optimal decision device is used for selecting the optimal next iteration experiment evaluation point according to the fusion prediction response value given by the prediction fusion device.
The low fidelity gaussian process regression prediction device outputs the initial high fidelity experimental evaluation points.
And the prediction fusion device and the optimal decision device adopt a dynamic weighting mechanism based on Bayes to perform weighted fusion on the prediction output of the low-fidelity Gaussian process regression model and the prediction output of the high-fidelity Gaussian process regression model.
The fused weighting coefficient is determined by the model predicted value and the actual response value.
The method and the device have the characteristics of more parameters to be optimized, high evaluation experiment cost, no analytic form of an optimization objective function and no available gradient information in the optimization process, fully utilize low-fidelity experiment data by adopting a Bayesian model fusion mechanism, and obtain the optimal system configuration parameter value with less experiment cost.
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are intended to further illustrate the principles of the invention, and that various changes and modifications may be made without departing from the spirit and scope of the invention, which is intended to be protected by the appended claims. The scope of the invention is defined by the claims and their equivalents.
Claims (10)
1. A parameter configuration optimization method for a large-scale complex system is characterized by comprising the following steps of S1, training a low-fidelity Gaussian process regression model according to low-fidelity experimental observation data;
s2, acquiring a high fidelity experimental evaluation point according to a prediction curve output by the low fidelity Gaussian process regression model;
step S3, training a high-fidelity Gaussian process regression model according to the high-fidelity experimental data;
s4, fusing the prediction output of the low-fidelity Gaussian process regression model and the prediction output of the high-fidelity Gaussian process regression model, and determining the next high-fidelity experiment evaluation point;
s5, if the high fidelity experiment times are less than the preset value, returning to the S3 for iteration; otherwise, selecting an optimal point from the high fidelity experimental evaluation points and outputting.
2. The method of claim 1, wherein the low-fidelity Gaussian process regression model stores and encapsulates low-fidelity experimental data information.
3. The method for optimizing parameter configuration for large-scale complex systems according to claim 1, wherein in step S2, the low-fidelity gaussian process regression model predicts experimental responses at any experimental evaluation point, and selects a high-fidelity experimental evaluation point according to the predicted value.
4. The method according to claim 3, wherein the selection of the high fidelity experimental evaluation point adopts an optimal search model, and an objective function of the optimal search model is determined according to a predicted value of a low fidelity Gaussian process regression model.
5. The method for optimizing parameter configuration for large-scale complex systems according to claim 1, wherein in step S4, the predicted output of the low-fidelity gaussian process regression model and the predicted output of the high-fidelity gaussian process regression model are weighted and fused by a dynamic weighting mechanism based on bayesian.
6. The method according to claim 5, wherein the fused weighting coefficients are determined by model prediction values and actual response values.
7. A parameter configuration optimization device for a large-scale complex system is characterized by comprising a low-fidelity experimental data storage device, a low-fidelity Gaussian process regression prediction device, a high-fidelity data acquisition device, a high-fidelity Gaussian process regression prediction device, a prediction fusion device and an optimal decision device;
the low-fidelity experimental data storage device is used for storing terminal equipment of low-fidelity experimental observation data;
the low-fidelity Gaussian process regression prediction device is used for analyzing and processing low-fidelity experimental observation data and determining a low-fidelity Gaussian process regression model parameter value;
the high-fidelity data acquisition device is used for acquiring and storing high-fidelity experimental evaluation data;
the high-fidelity Gaussian process regression prediction device is used for analyzing and processing high-fidelity experimental observation data and determining high-fidelity Gaussian process regression model parameter values;
the prediction fusion device is used for performing weighted fusion on the low-fidelity Gaussian process regression prediction and the high-fidelity Gaussian process regression prediction and giving a fusion prediction response value to any experimental evaluation point;
and the optimal decision device is used for selecting an optimal next iteration experiment evaluation point according to the fusion prediction response value given by the prediction fusion device.
8. The apparatus of claim 7, wherein the low fidelity gaussian process regression prediction means outputs initial high fidelity experimental evaluation points.
9. The device for optimizing parameter configuration for large-scale complex systems according to claim 7, wherein the prediction fusion device and the optimal decision device perform weighted fusion on the prediction output of the low-fidelity gaussian process regression model and the prediction output of the high-fidelity gaussian process regression model by using a bayesian-based dynamic weighting mechanism.
10. The apparatus according to claim 7, wherein the fused weighting coefficients are determined by the model predicted values and the actual response values.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910610017.1A CN110689160B (en) | 2019-07-08 | 2019-07-08 | Parameter configuration optimization method and device for large-scale complex system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910610017.1A CN110689160B (en) | 2019-07-08 | 2019-07-08 | Parameter configuration optimization method and device for large-scale complex system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110689160A CN110689160A (en) | 2020-01-14 |
| CN110689160B true CN110689160B (en) | 2022-08-23 |
Family
ID=69108166
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910610017.1A Active CN110689160B (en) | 2019-07-08 | 2019-07-08 | Parameter configuration optimization method and device for large-scale complex system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110689160B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220011760A1 (en) * | 2020-07-08 | 2022-01-13 | International Business Machines Corporation | Model fidelity monitoring and regeneration for manufacturing process decision support |
| CN113254472B (en) * | 2021-06-17 | 2021-11-16 | 浙江大华技术股份有限公司 | Parameter configuration method, device, equipment and readable storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
| CN109214605A (en) * | 2018-11-12 | 2019-01-15 | 国网山东省电力公司电力科学研究院 | Power-system short-term Load Probability prediction technique, apparatus and system |
-
2019
- 2019-07-08 CN CN201910610017.1A patent/CN110689160B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110689160A (en) | 2020-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Elmaz et al. | CNN-LSTM architecture for predictive indoor temperature modeling | |
| CN111444878B (en) | Video classification method, device and computer readable storage medium | |
| CN110832509B (en) | Black box optimization using neural networks | |
| KR101961421B1 (en) | Method, controller, and computer program product for controlling a target system by separately training a first and a second recurrent neural network models, which are initially trained using oparational data of source systems | |
| US9483739B2 (en) | Transductive feature selection with maximum-relevancy and minimum-redundancy criteria | |
| CN110689160B (en) | Parameter configuration optimization method and device for large-scale complex system | |
| CN105160444A (en) | Electrical equipment failure rate determining method and system | |
| CN110781595B (en) | Method, device, terminal and medium for predicting energy use efficiency (PUE) | |
| US12020166B2 (en) | Meta-learned, evolution strategy black box optimization classifiers | |
| US20230326191A1 (en) | Method and Apparatus for Enhancing Performance of Machine Learning Classification Task | |
| WO2021169478A1 (en) | Fusion training method and apparatus for neural network model | |
| JP2023542654A (en) | Experience Replay Algorithm with Task Priority for Reinforcement Learning | |
| CN110826810B (en) | Regional rainfall prediction method combining spatial reasoning and machine learning | |
| WO2009142011A1 (en) | Decision device with minimum power transmission loss system, method, and program | |
| CN112766402A (en) | Algorithm selection method and device and electronic equipment | |
| CN113783715B (en) | Opportunistic network topology prediction method adopting causal convolutional neural network | |
| CN107437112B (en) | A kind of mixing RVM model prediction methods based on the multiple dimensioned kernel function of improvement | |
| CN110705695B (en) | Method, device, equipment and storage medium for searching model structure | |
| CN111859625A (en) | Energy-saving control method and device based on big data and storage medium | |
| CN109726849A (en) | A kind of building microgrid load forecasting method of colored pollination algorithm optimization neural network | |
| US11156969B1 (en) | Environment factor control device and training method thereof | |
| Karkaria et al. | A Digital Twin Framework Utilizing Machine Learning for Robust Predictive Maintenance: Enhancing Tire Health Monitoring | |
| CN112016695A (en) | Method, apparatus and computer program for predicting a learning curve | |
| CN112348175A (en) | Method for performing feature engineering based on reinforcement learning | |
| CN116189176B (en) | Intelligent tea matching method, equipment and medium based on generation measurement TD3 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |