CN110174841A - The parallel control method of industrial process large time delay inertia system based on parameter optimization - Google Patents
The parallel control method of industrial process large time delay inertia system based on parameter optimization Download PDFInfo
- Publication number
- CN110174841A CN110174841A CN201910363062.1A CN201910363062A CN110174841A CN 110174841 A CN110174841 A CN 110174841A CN 201910363062 A CN201910363062 A CN 201910363062A CN 110174841 A CN110174841 A CN 110174841A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- artificial
- loop
- closed
- control
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
- G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The parallel control method of industrial process large time delay inertia system based on parameter optimization that the invention discloses a kind of.Parallel control method of the invention is related to artificial system and actual control system, and the actual control system uses the anticipation function closed-loop control based on dull integral coefficient, builds corresponding close loop control circuit in artificial system and genetic optimization program is arranged；The method uses following steps: firstly, carrying out closed loop modeling to controlled device using the closed-loop identification algorithm based on deep learning；Secondly, plant model is placed in artificial system, control performance is assessed, using genetic algorithm optimizing optimum apjustment coefficient；Finally, artificial system optimizing parameter is placed in actual control system, parameter optimization interaction is realized.The present invention realizes the parameter optimization of actual control system by the reciprocation of two systems, improves actual control system Control platform.
Description
Technical field
The invention belongs to industrial process control system field, specifically a kind of industrial process based on parameter optimization is big
Lag the parallel control method of inertia system.
Background technique
A kind of large time delay inertia system is commonly present in industrial process, for the control performance for improving the large time delay inertia system,
It is often used Model Predictive Control, it is higher to the model needs of controlled device, control effect variation is often resulted in after model variation, is needed
Adjustment Model Predictive Control parameter in time.
Chinese Patent Application No. 201810184793.5 discloses a kind of dull integral coefficient prediction letter based on intelligent optimizing
Number control parameter tuning methods comprising following steps: the first order inertial loop in industrial process control system adds pure delay pair
As obtaining anticipation function optimal control law using a jump function as predictive control model, prediction step is enabled to obtain for 1
A kind of single step optimal control law；Using single regulation coefficient method, dull integral coefficient anticipation function optimal control law is obtained；Using
The setting method of dull integral coefficient uses genetic algorithm optimized tuning.According to said method is added to any one order inertia pure delay pair
As can get optimal regulation coefficient, guarantee that Predictive function control has stronger robustness.But after object model variation,
If intra-prediction model does not change, control performance can obviously be deteriorated, and need to change adjustment using the method for hereditary optimizing at this time
Coefficient improves control performance.
Summary of the invention
The technical problem to be solved by the present invention is to overcome the problems of the above-mentioned prior art, provide a kind of based on parameter
The parallel control method of industrial process large time delay inertia system of optimization, uses parallel control method, to effectively improve large time delay
The Control platform of inertia system.
For this purpose, the present invention adopts the following technical scheme that: the industrial process large time delay inertia system based on parameter optimization is flat
Row control method, the method are related to artificial system and actual control system, and the actual control system uses base
Corresponding close loop control circuit is built in the anticipation function closed-loop control of dull integral coefficient, artificial system and is arranged hereditary excellent
Change program；
The method uses following steps: firstly, using the closed-loop identification algorithm based on deep learning to controlled device
Carry out closed loop modeling；Secondly, plant model is placed in artificial system, control performance is assessed, using genetic algorithm
Optimizing optimum apjustment coefficient；Finally, artificial system optimizing parameter is placed in actual control system, realize that parameter optimization is handed over
Mutually.
Further, the closed-loop identification algorithm based on deep learning, its step are as follows: firstly, in anticipation function control
Forward and reverse step excitation signal is added in default value processed, obtains the closed loop input data set A and output data of controlled device
Collect B, data set A generates data set C by the internal model in Predictive function control system, using data set C and data set B as
Deep learning inactivates the training data of neural network at random, is trained using regression algorithm, obtains deep learning closed-loop identification
Model, to accurately reflect the feature of controlled device.
Further, after obtaining plant model, which is placed in the anticipation function control based on dull integral coefficient
In the artificial system of system, the internal model of artificial system is constant, and plant model is distinguished using deep learning closed loop
Know model, optimum apjustment coefficient is obtained to artificial system optimizing using genetic Optimization Algorithm；Artificial system optimizing ginseng
When number merging actual control system, regulation coefficient is gradual to optimum apjustment coefficient.
Further, the deep learning inactivate at random neural network (DNN network) forming process it is as follows:
Firstly, input data is normalized between given maximum value and minimum value；Secondly, being trained data set and survey
Data set classification is tried, ReLU activation primitive, setting nerve cell layer, learning rate and random inactivation rate are selected；It is calculated using returning
Method carries out deep learning neural metwork training and accuracy rate measuring and calculating, obtains deep learning and inactivates neural network at random.
The invention has the following advantages: The present invention gives a kind of, the industrial process large time delay based on parameter optimization is used
The property parallel control method of system realizes the parameter optimization of actual control system by the reciprocation of two systems, improves
Actual control system Control platform.Wherein, need to only be swashed in setting value using better simply forward and reverse step in closed-loop identification
Encourage signal, can it is easy, accurately pick out system plant model, by identification model merging simulation control subsystem
In, after carrying out optimizing to control parameter, optimizing parameter is placed in actual control system, large time delay inertia system can be improved
Performance.The present invention has important practice significance for the application controlled in parallel.
Detailed description of the invention
Fig. 1 is the principle of the present invention figure；
Fig. 2, which is the present invention, learns closed-loop identification schematic diagram (G in figure for dull integral coefficient Predictive function control system depth
It (s) is transmission function object model, GmIt (s) is anticipation function internal model, kmFor internal model gain, m is dull integral coefficient, A
To set the forward and reverse pumping signal data set of input step, B is object closed loop input data set, and C is object closed loop output data
Collection, D are prediction internal model output data set, and E is closed-loop identification model output data set)；
Fig. 3 is the formation figure that deep learning of the present invention inactivates neural network at random.
Fig. 4 is that (in figure, sp is setting value to the artificial analogue system schematic diagram of the present invention, and GNN is deep learning closed-loop identification mould
Type, GmIt (s) is anticipation function internal model, kmFor internal model gain, m is dull integral coefficient, and pv is controlled device output)；
Fig. 5 be Optimal Parameters switching system schematic diagram of the present invention (sp is setting value, and G (s) is transmission function object model,
GmIt (s) is anticipation function internal model, kmFor internal model gain, m1 is original dull integral coefficient, and m2 is the dull whole system of optimization
Number, pv are controlled device output)；
Fig. 6 is DNN network output curve diagram after deep learning closed loop of the present invention modeling；
Fig. 7 is system output curve diagram after parameter optimization of the present invention.
Specific embodiment
For a clearer understanding of the technical characteristics, objects and effects of the present invention, this hair of Detailed description of the invention is now compareed
Bright specific embodiment.
Embodiment
The parallel control method of industrial process large time delay inertia system based on parameter optimization that the present embodiment provides a kind of, relates to
And two systems: artificial system and actual control system, actual control system use the prediction letter based on dull integral coefficient
Closed-loop control is counted, corresponding close loop control circuit is built in artificial system and genetic optimization program is set.
Parallel control method of the invention is as follows: firstly, using the closed-loop identification algorithm based on deep learning to controlled pair
As carrying out closed loop modeling；Secondly, plant model is placed in artificial system, control performance is assessed, is calculated using heredity
Method optimizing optimum apjustment coefficient；Finally, artificial system optimizing parameter is placed in actual control system, realize that parameter optimization is handed over
Mutually.
According to Fig. 1, actual control system uses the anticipation function closed-loop control based on dull integral coefficient, artificial system
In build corresponding close loop control circuit and genetic optimization program be set.The first step, by artificial system modelling, using based on deep
The closed-loop identification algorithm for spending study carries out closed loop modeling to controlled device；Plant model is placed in artificial by second step
System carries out optimizing to control parameter using genetic Optimization Algorithm；Third step, simultaneously with artificial system by actual control system
It lifts, control parameter is optimized in a parallel manner.
Closed-loop identification algorithm shown in Fig. 2 based on deep learning.Firstly, being added in control system setting value forward and reverse
Step excitation signal obtains closed loop input data set A, control data set B and output data set C, data set the B warp of controlled device
The internal model crossed in Predictive function control system generates data set D, and data set D and data set C is random as deep learning
The training data for inactivating neural network, is trained using regression algorithm, can obtain deep learning closed-loop identification model, the mould
Type can accurately reflect the feature of controlled device.
Deep learning shown in Fig. 3 inactivates neural network at random, and forming process is as follows:
Firstly, input data is normalized between given maximum value and minimum value；Secondly, being trained data set and survey
Data set classification is tried, ReLU activation primitive, setting nerve cell layer, learning rate and random inactivation rate are selected；It is calculated using returning
Method carries out deep learning neural metwork training and accuracy rate measuring and calculating, obtains deep learning and inactivates neural network at random.
According to Fig. 4, after obtaining plant model, which is placed in the Predictive function control based on dull integral coefficient
In analogue system, i.e. the survey function internal model of artificial system is constant, and plant model is distinguished using deep learning closed loop
Know model, optimum apjustment coefficient is obtained to system optimizing using genetic Optimization Algorithm.
According to Fig. 5, when artificial system optimizing parameter is placed in actual control system, regulation coefficient was by 0.0001/ second
Rate is gradual to optimal parameter.
Verification experimental verification
Object model usesUsing the Predictive function control based on dull integral coefficient, internal model is adopted
WithRegulation coefficient m adjusting is 1.5, when object model property variation isThis
When internal model still use Gm, (i.e. originally using the parallel control method of industrial process large time delay inertia system based on parameter optimization
Invention) carry out parameter optimization.
Firstly, recognizing mould by forward and reverse step excitation signal data and object model output data training deep learning
Type, then deep learning identification model is placed in artificial system, using genetic Optimization Algorithm optimizing optimal parameter, finally,
Optimal Parameters are placed in former control system by the rate of setting.
Input signal data integrates as behind reversed after positive step 1,800 second step -1,1600 after positive step 1,2400 second
Reversed step -1, by network training, is measured after data set C, D are input to deep learning identification model using mean square deviation (MSE)
Method (i.e. the average value of square-error between object output and identifier output), the MSE value of acquisition is 0.0054, accurate to express
The feature of output data.
Deep learning closed-loop identification model is placed in artificial system, is most preferably joined using genetic Optimization Algorithm optimizing
Number, the response curve of acquisition such as Fig. 7 have larger improvement compared to original identification response curve such as Fig. 6.
The foregoing is merely the schematical specific embodiment of the present invention, the range being not intended to limit the invention.It is any
Those skilled in the art, made equivalent changes and modifications under the premise of not departing from design and the principle of the present invention,
It should belong to the scope of protection of the invention.
Claims (4)
1. the parallel control method of industrial process large time delay inertia system based on parameter optimization, which is characterized in that the method
It is related to artificial system and actual control system, the actual control system uses the anticipation function based on dull integral coefficient
Closed-loop control builds corresponding close loop control circuit in artificial system and genetic optimization program is arranged；
The method uses following steps: firstly, being carried out using the closed-loop identification algorithm based on deep learning to controlled device
Closed loop modeling；Secondly, plant model is placed in artificial system, control performance is assessed, using genetic algorithm optimizing
Optimum apjustment coefficient；Finally, artificial system optimizing parameter is placed in actual control system, parameter optimization interaction is realized.
2. the industrial process large time delay inertia system parallel control method according to claim 1 based on parameter optimization,
It is characterized in that, the closed-loop identification algorithm based on deep learning, its step are as follows: firstly, being set in Predictive function control system
Forward and reverse step excitation signal is added in definite value, obtains the closed loop input data set A and output data set B of controlled device, data
Collect A and generate data set C by the internal model in Predictive function control system, using data set C and data set B as deep learning
The training data of random inactivation neural network, is trained using regression algorithm, deep learning closed-loop identification model is obtained, with standard
The really feature of reflection controlled device.
3. the industrial process large time delay inertia system parallel control method according to claim 2 based on parameter optimization,
It is characterized in that, after obtaining plant model, by the artificial of Predictive function control of the model merging based on dull integral coefficient
In analogue system, the internal model of artificial system is constant, and plant model uses deep learning closed-loop identification model, adopts
Optimum apjustment coefficient is obtained to artificial system optimizing with genetic Optimization Algorithm；The merging of artificial system optimizing parameter is practical
When control system, regulation coefficient is gradual to optimum apjustment coefficient.
4. the industrial process large time delay inertia system parallel control method according to claim 2 based on parameter optimization,
It is characterized in that, the forming process that the deep learning inactivates neural network at random is as follows:
Firstly, input data is normalized between given maximum value and minimum value；Secondly, being trained data set and test number
Classify according to collection, selects ReLU activation primitive, setting nerve cell layer, learning rate and random inactivation rate；Using regression algorithm into
Row deep learning neural metwork training and accuracy rate measuring and calculating, obtain deep learning and inactivate neural network at random.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910363062.1A CN110174841A (en) | 2019-04-30 | 2019-04-30 | The parallel control method of industrial process large time delay inertia system based on parameter optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910363062.1A CN110174841A (en) | 2019-04-30 | 2019-04-30 | The parallel control method of industrial process large time delay inertia system based on parameter optimization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110174841A true CN110174841A (en) | 2019-08-27 |
Family
ID=67690397
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910363062.1A Pending CN110174841A (en) | 2019-04-30 | 2019-04-30 | The parallel control method of industrial process large time delay inertia system based on parameter optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110174841A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582562A (en) * | 2020-04-20 | 2020-08-25 | 杭州意能电力技术有限公司 | Neural network prediction control method based on optimization control platform |
| CN114237043A (en) * | 2021-11-29 | 2022-03-25 | 东南大学溧阳研究院 | Gas turbine equipment transfer function closed-loop identification method based on deep learning |
| WO2024016556A1 (en) * | 2022-07-22 | 2024-01-25 | 中控技术股份有限公司 | Model autonomous learning method suitable for predictive control in process industry |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5740324A (en) * | 1990-10-10 | 1998-04-14 | Honeywell | Method for process system identification using neural network |
| CN108333947A (en) * | 2018-03-07 | 2018-07-27 | 国网浙江省电力有限公司电力科学研究院 | Dull integral coefficient Predictive function control parameter tuning method based on intelligent optimizing |
-
2019
- 2019-04-30 CN CN201910363062.1A patent/CN110174841A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5740324A (en) * | 1990-10-10 | 1998-04-14 | Honeywell | Method for process system identification using neural network |
| CN108333947A (en) * | 2018-03-07 | 2018-07-27 | 国网浙江省电力有限公司电力科学研究院 | Dull integral coefficient Predictive function control parameter tuning method based on intelligent optimizing |
Non-Patent Citations (6)
| Title |
|---|
| 单志敏: "钕铁硼氢粉碎过程平行控制方法研究", 《中国优秀博硕士学位论文全文数据库(硕士) 工程科技Ⅱ辑》 * |
| 段艳杰 等: "深度学习在控制领域的研究现状与展望", 《自动化学报》 * |
| 王飞跃 等: "复杂系统的平行控制理论及应用", 《复杂系统与复杂性科学》 * |
| 王飞跃 等: "智能控制：从学习控制到平行控制", 《控制理论与应用》 * |
| 王飞跃: "平行控制：数据驱动的计算控制方法", 《自动化学报》 * |
| 陈凯: "深度学习模型的高效训练算法研究", 《中国优秀博硕士学位论文全文数据库(博士) 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582562A (en) * | 2020-04-20 | 2020-08-25 | 杭州意能电力技术有限公司 | Neural network prediction control method based on optimization control platform |
| CN114237043A (en) * | 2021-11-29 | 2022-03-25 | 东南大学溧阳研究院 | Gas turbine equipment transfer function closed-loop identification method based on deep learning |
| CN114237043B (en) * | 2021-11-29 | 2024-04-16 | 东南大学溧阳研究院 | Gas turbine equipment transfer function closed-loop identification method based on deep learning |
| WO2024016556A1 (en) * | 2022-07-22 | 2024-01-25 | 中控技术股份有限公司 | Model autonomous learning method suitable for predictive control in process industry |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110174841A (en) | The parallel control method of industrial process large time delay inertia system based on parameter optimization | |
| CN106453293B (en) | A kind of network security situation prediction method based on improved BPNN | |
| CN105488528B (en) | Neural network image classification method based on improving expert inquiry method | |
| CN110084367A (en) | A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model | |
| CN105608512A (en) | Short-term load forecasting method | |
| CN103166830B (en) | A kind of Spam Filtering System of intelligent selection training sample and method | |
| CN109492287A (en) | A kind of solid electrolyte ionic conductivity prediction technique based on BP neural network | |
| CN110059852A (en) | A kind of stock yield prediction technique based on improvement random forests algorithm | |
| CN109146121A (en) | The power predicating method stopped in the case of limited production based on PSO-BP model | |
| CN109242236A (en) | Based on the PMU primary frequency modulation load forecasting method for being associated with full Connection Neural Network and LSTM | |
| CN107909145A (en) | A kind of training method of convolutional neural networks model | |
| CN109143408B (en) | Dynamic region combined short-time rainfall forecasting method based on MLP | |
| CN104865827B (en) | A kind of pumping production optimization method based on multi-state model | |
| CN113762603B (en) | Photovoltaic base station short-term photovoltaic power prediction method based on improved sparrow algorithm optimization | |
| CN107944495A (en) | A kind of household electricity load classification recognition methods based on deep layer forest algorithm | |
| CN105701562B (en) | Training method, applicable method for predicting generated power and respective system | |
| CN106997495A (en) | A kind of Methods of electric load forecasting | |
| CN101520856B (en) | Adaptive modeling method for support vector regression based on KKT condition and nearest neighbor method | |
| CN104820877A (en) | Photovoltaic system generation power prediction method based on cloud adaptive PSO-SNN | |
| CN106296434A (en) | A kind of Grain Crop Yield Prediction method based on PSO LSSVM algorithm | |
| CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
| CN106681133B (en) | A kind of Hydropower Unit model refinement type subspace closed-loop identification method | |
| CN108985323A (en) | A kind of short term prediction method of photovoltaic power | |
| CN110298506A (en) | A kind of urban construction horizontal forecast system | |
| CN113033898A (en) | Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190827 |