CN103048926A - Online neural network inverse controller in biological fermentation process and construction method of controller - Google Patents
Online neural network inverse controller in biological fermentation process and construction method of controller Download PDFInfo
- Publication number
- CN103048926A CN103048926A CN2012105565540A CN201210556554A CN103048926A CN 103048926 A CN103048926 A CN 103048926A CN 2012105565540 A CN2012105565540 A CN 2012105565540A CN 201210556554 A CN201210556554 A CN 201210556554A CN 103048926 A CN103048926 A CN 103048926A
- Authority
- CN
- China
- Prior art keywords
- controller
- neural network
- inverse
- input
- output
- 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
Images
Landscapes
- Feedback Control In General (AREA)
- Apparatus Associated With Microorganisms And Enzymes (AREA)
Abstract
The invention discloses an online neural network inverse controller in a biological fermentation process. The controller comprises a neural network inverse offline decoupling controller construction module and a neural network inverse online learning module. The control method of the controller comprises the following steps of: connecting an offline trained neural network inverse system and a fermentation system in series to realize the linear decoupling of the system; and realizing the online update by the online learning of the neural network inverse online learning module. A neural network inverse system decoupling method and an inverse system online identification method are combined to realize the decoupling control of a multivariable fermentation system. A neural network inverse system technology is adopted to realize the decoupling of the biological fermentation system without relying on the model and parameters of the process. The inverse controller leaning module is designed based on neural network inverse decoupling, so that the controller can be adapted to the time-varying characteristics of parameters of the biological fermentation system, and thus the real-time control of variables in the fermentation process can be realized, and the controller meets the needs of practical engineering application.
Description
Technical field
The present invention relates to the online decoupling controller of a kind of biological fermentation process and building method thereof, belong to the Optimized-control Technique field of biological fermentation process.
Background technology
The biofermentation system is the multiple-input and multiple-output nonlinear system that a class has time variation, uncertainty and strong coupling, and traditional nonlinear control method is difficult to the optimal control of implementation procedure variable.For the nonlinear system of multiple-input and multiple-output, generally adopt first certain rule that system linear is changed into a plurality of linear subsystems, then each subsystem design closed loop linear controller is realized the purpose of decoupling zero control.With the Neural network inverse control method that method of inverse combines with neural net method, do not rely on the accurate model of system, only need know that the relative rank of system can realize linearization of nonlinear system and decoupling zero, in Practical Project, obtain many successful Application.
The Neural network inverse control method is applied in the biological fermentation process control, though do not rely on the mathematical model of system, generalization ability, the external interference of its decoupling zero control performance and neural network have much relations.Systematic parameter changes in the insufficient or control procedure of off-line training neural network, all can cause the generalization ability of neural network to descend, thereby make the closed-loop control hydraulic performance decline.When if Fermentation Process of Parameter changes, neural network training can not satisfy again the real-time of process control again.
Summary of the invention
For nerve network reverse controller in the prior art and method above shortcomings thereof, the invention provides a kind of biological fermentation process line neural network inverse controller and building method thereof.
The technical scheme that biological fermentation process line neural network inverse controller of the present invention and building method thereof adopt is:
A kind of biological fermentation process line neural network inverse controller comprises nerve network reverse off-line decoupling controller constructing module and nerve network reverse on-line study module.
Described nerve network reverse off-line decoupling controller constructing module is analyzed the reversibility of fermentation system, the inverse system input variable of fermentation system is output variable and the derivative thereof of fermentation system, the inverse system output variable is the control inputs variable of fermentation system, realize the input variable of inverse system and the nonlinear relationship between the output variable by constructing neural network, Neural Network Inverse System is series at the input and output decoupling zero that fermentation system is realized in the fermentation system front.
The pseudo-linear compound system design closed loop linear controller of described nerve network reverse on-line study module after to decoupling zero realized the high-performance decoupling zero control of fermentation system, the output that is input as the real attenuation system of described nerve network reverse on-line study module (
,
), export target be the real attenuation system input (
,
), be used for the described nerve network reverse controller of regular update.
A kind of building method of biological fermentation process line neural network inverse controller, by the Neural Network Inverse System of off-line training is connected with fermentation system, the realization system linearity is dissolved coupling, by the on-line study of nerve network reverse on-line study module, realizes the online updating to inverse controller.Concrete steps are:
(1) model of sweat is carried out reversibility Analysis, determine the input and output amount of inverse system; System's current time output and previous moment thereof are output as the input variable of inverse system, and the control inputs amount of system is as the output quantity of inverse system;
(2) gather representative sweat inputoutput data;
(3) choice structure is single hidden layer BP neural network, utilizes the data neural network training that gathers, and makes its training error reach given precision;
(4) neural network that trains in the step (3) is series at before the controlled fermentation system, realizes linearization and the decoupling zero of fermentation system;
(5) take biofermentation system input parameter as input, the biofermentation system is input as output, by nerve network reverse on-line study module, realizes the contrary on-line study of biosystem;
(6) regularly upgrade nerve network reverse controller with nerve network reverse on-line study module.
The beneficial effect of biological fermentation process line neural network inverse controller of the present invention and building method thereof is:
1, Neural Network Inverse System decoupling method and inverse system on-line identification method are combined, realized the decoupling zero control of multivariate fermentation system.
2, adopt the decoupling zero of Neural Network Inverse System technology realization biofermentation system, do not rely on the model and parameter of process.Designed the inverse controller study module on the basis of neural network reversed decoupling, can adapt to the characteristics of the parameter time varying of biofermentation system, can realize the real-time control of procedure variable of fermenting, the demand that realistic engineering is used.
3, the present invention is applied to the decoupling zero control of biological fermentation process cell concentration and substrate concentration, is a kind of simple and practical procedure variable of fermenting control method, for the quality and yield that improves tunning provides technical support.
Description of drawings
Fig. 1 is the control block diagram of biological fermentation process line neural network inverse controller of the present invention and building method thereof.
Among the figure: 1, nerve network reverse off-line decoupling controller constructing module; 2, nerve network reverse on-line study module; 3, fermentation system.
Embodiment
Below in conjunction with accompanying drawing the present invention is further elaborated.
Biological fermentation process line neural network inverse controller of the present invention comprises two parts:
One, nerve network reverse off-line decoupling controller constructing module 1.
Export the biofermentation systems for having probabilistic two inputs two, according to the right invertibility of Inverse System Theory analytic system, determine the relative rank of system and the input/output variable of inverse system; Off-line acquisition system data neural network training approximate inverse-system is connected into pseudo-linear compound system with neural network and the controlled system that trains, and realizes linearization and the decoupling zero of controlled system.
Two, nerve network reverse on-line study module 2.
Can realize the decoupling zero control of system to the pseudo-linear compound system design closed loop linear controller after the decoupling zero, but its control performance is subject to the uncertain factor impacts such as Neural Network Inverse System modeling error, fermentation system parameter time varying and external interference, by designing a Neural Network Online study module, realize the high-performance decoupling zero control of fermentation system 3.
On the basis of existing Neural Network Inverse System decoupling method, for improving the Neural network inverse control performance, the present invention proposes line neural network inverse controller building method.By the Neural Network Inverse System of off-line training is connected with fermentation system, the realization system linearity is dissolved coupling.The characteristics that become when complicated for sweat dynamics have designed an inverse system study module, by on-line study, realize the online updating to inverse controller.The specific implementation step is as follows:
(1) model of sweat is carried out reversibility Analysis, determine the input and output amount of inverse system; System's current time output and previous moment thereof are output as the input variable of inverse system, and the control inputs amount of system is as the output quantity of inverse system;
(2) gather representative sweat inputoutput data;
(3) choice structure is single hidden layer BP neural network, utilizes the data neural network training that gathers, and makes its training error reach given precision;
(4) neural network that (3) is trained is series at before the controlled fermentation system 3, realizes linearization and the decoupling zero of fermentation system 3;
(5) take biofermentation system 3 input parameters as input, biofermentation system 3 is input as output, by nerve network reverse line study module, realizes the contrary on-line study of biosystem;
(6) regularly upgrade nerve network reverse controller with nerve network reverse on-line study module.
The decoupling control method that the present invention proposes is applicable to set up on Monod equation basis a class biological fermentation process of thalli growth and substrate consumption model.The present invention is take penicillin fermentation process as object, the decoupling controller of cell concentration and substrate concentration in the design sweat, and the implementation step is as follows:
1, the penicillin fermentation system is with feed rate
With the charging substrate concentration
Be control inputs, with cell concentration
And substrate concentration
Be output.The right-inverge input-output equation of penicillin fermentation system is:
2, the various control inputs amounts in the given concrete practical operation scope, the output data of collection penicillin fermentation system consist of
NGroup neural metwork training data set:
3, the data that gather are carried out the normalization pre-service.
4, selecting structure is single hidden layer BP neural network of 4 * 10 * 2, will
As nerve network reverse controller training input quantity,
For nerve network reverse controller training expectation output quantity, nerve network reverse controller (1) is trained up, make training error reach given expectation value.
5, the neural network (1) that trains is series at before the penicillin fermentation system, as shown in Figure 1, first output that is about to nerve network reverse controller is connected with first input of fermentation system, second output of nerve network reverse controller is connected with second input of fermentation system, is decoupled into two sub-systems.
6, design simultaneously a Neural Network Inverse System on-line identification module (2) identical with the inverse controller structure, the output that is input as real attenuation system (3) of this module, export target is the input of real attenuation system (3).
7, every designated period, upgrade nerve network reverse controller with inverse system on-line study module.
Claims (2)
1. biological fermentation process line neural network inverse controller, it is characterized in that: described inverse controller comprises nerve network reverse off-line decoupling controller constructing module (1) and nerve network reverse on-line study module (2);
Described nerve network reverse off-line decoupling controller constructing module (1) is analyzed the reversibility of fermentation system, the inverse system input variable of fermentation system is output variable and the derivative thereof of fermentation system, the inverse system output variable is the control inputs variable of fermentation system, realize the input variable of inverse system and the nonlinear relationship between the output variable by constructing neural network, Neural Network Inverse System is series at the input and output decoupling zero that fermentation system is realized in fermentation system (3) front;
Described nerve network reverse on-line study module (2) is to the pseudo-linear compound system design closed loop linear controller after the decoupling zero, realize the high-performance decoupling zero control of fermentation system (3), the output that is input as the real attenuation system of described nerve network reverse on-line study module (2) (
,
), export target be real attenuation system (3) input (
,
), be used for the described nerve network reverse controller of regular update.
2. the building method of a biological fermentation process line neural network inverse controller, it is characterized in that: Neural Network Inverse System and the fermentation system (3) of off-line training are connected, the realization system linearity is dissolved coupling, by the on-line study of nerve network reverse on-line study module (2), realize the online updating to inverse controller; Concrete steps are:
(1) model of sweat is carried out reversibility Analysis, determine the input and output amount of inverse system; System's current time output and previous moment thereof are output as the input variable of inverse system, and the control inputs amount of system is as the output quantity of inverse system;
(2) gather representative sweat inputoutput data;
(3) choice structure is single hidden layer BP neural network, utilizes the data neural network training that gathers, and makes its training error reach given precision;
(4) neural network that trains in the step (3) is series at controlled fermentation system (3) before, realizes linearization and the decoupling zero of fermentation system (3);
(5) take biofermentation system (3) input parameter as input, biofermentation system (3) is input as output, by nerve network reverse on-line study module (2), realizes the contrary on-line study of biosystem;
(6) regularly use nerve network reverse on-line study module (2) to upgrade nerve network reverse controller.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012105565540A CN103048926A (en) | 2012-12-20 | 2012-12-20 | Online neural network inverse controller in biological fermentation process and construction method of controller |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012105565540A CN103048926A (en) | 2012-12-20 | 2012-12-20 | Online neural network inverse controller in biological fermentation process and construction method of controller |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103048926A true CN103048926A (en) | 2013-04-17 |
Family
ID=48061609
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2012105565540A Pending CN103048926A (en) | 2012-12-20 | 2012-12-20 | Online neural network inverse controller in biological fermentation process and construction method of controller |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103048926A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103345159A (en) * | 2013-07-03 | 2013-10-09 | 江苏大学 | Hybrid electric vehicle BSG system control method based on neural network self-adaptation inversion |
CN110157613A (en) * | 2019-05-23 | 2019-08-23 | 湖南民康生物技术研究所 | A kind of layer frame endless track formula solid state fermentation equipment and process for solid state fermentation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0710901A1 (en) * | 1994-11-01 | 1996-05-08 | The Foxboro Company | Multivariable nonlinear process controller |
US20050149234A1 (en) * | 2003-12-30 | 2005-07-07 | Vian John L. | Methods and systems for analyzing engine unbalance conditions |
CN1661004A (en) * | 2004-12-13 | 2005-08-31 | 东南大学 | Contrary flexible method in neural network for measuring procedure variable of fermenting erythromycin and structure of flexible instrument |
CN1790197A (en) * | 2005-12-22 | 2006-06-21 | 安徽工业大学 | Simple method for neural network decoupling of multi-variable system based on model reference adaptive control |
-
2012
- 2012-12-20 CN CN2012105565540A patent/CN103048926A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0710901A1 (en) * | 1994-11-01 | 1996-05-08 | The Foxboro Company | Multivariable nonlinear process controller |
US20050149234A1 (en) * | 2003-12-30 | 2005-07-07 | Vian John L. | Methods and systems for analyzing engine unbalance conditions |
CN1661004A (en) * | 2004-12-13 | 2005-08-31 | 东南大学 | Contrary flexible method in neural network for measuring procedure variable of fermenting erythromycin and structure of flexible instrument |
CN1790197A (en) * | 2005-12-22 | 2006-06-21 | 安徽工业大学 | Simple method for neural network decoupling of multi-variable system based on model reference adaptive control |
Non-Patent Citations (1)
Title |
---|
于霜等: "多变量生物发酵过程的在线学习神经网络逆控制", 《第三十一届中国控制会议论文集B卷》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103345159A (en) * | 2013-07-03 | 2013-10-09 | 江苏大学 | Hybrid electric vehicle BSG system control method based on neural network self-adaptation inversion |
CN110157613A (en) * | 2019-05-23 | 2019-08-23 | 湖南民康生物技术研究所 | A kind of layer frame endless track formula solid state fermentation equipment and process for solid state fermentation |
CN110157613B (en) * | 2019-05-23 | 2022-08-26 | 湖南民康生物技术研究所 | Layer-frame circulating crawler-type solid-state fermentation equipment and solid-state fermentation method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106022521B (en) | Short-term load prediction method of distributed BP neural network based on Hadoop architecture | |
CN103064292A (en) | Biological fermentation adaptive control system and control method based on neural network inverse | |
CN101286044B (en) | Coal-burning boiler system steam-temperature mixing modeling method | |
CN102411305B (en) | Design method of comprehensive disturbance rejection control system for single-rotor wing helicopter/turboshaft engine | |
CN103218669A (en) | Intelligent live fish cultivation water quality comprehensive forecasting method | |
CN107527116B (en) | Short-term load prediction method based on support vector regression | |
CN104049536B (en) | The building method of chassis nonlinear system fuzzy-neural network generalized inverse controller | |
CN101763036A (en) | Lysine fermentation process feeding prediction control system and method based on fuzzy neural network | |
CN109634108A (en) | The different factor full format non-model control method of the MIMO of parameter self-tuning | |
CN103792845B (en) | A kind of Ferment of DM process mends the method and system of sugared rate optimized control | |
CN105974801A (en) | Greenhouse environment modeling and control technology based on data drive | |
Wang et al. | A robust decoupling control method based on artificial bee colony-multiple least squares support vector machine inversion for marine alkaline protease MP fermentation process | |
CN102594245B (en) | Sliding mode control method of under-actuated double-coupling motor | |
CN106200379A (en) | A kind of distributed dynamic matrix majorization method of Nonself-regulating plant | |
CN106950824A (en) | Stalk fermentation alcohol fuel process feeding prediction control system and method based on fuzzy neural network | |
CN102645894A (en) | Fuzzy adaptive dynamic programming method | |
CN102393645A (en) | Control method of high-speed electro-hydraulic proportional governing system | |
CN103048926A (en) | Online neural network inverse controller in biological fermentation process and construction method of controller | |
CN108614431A (en) | A kind of Hammerstein-Wiener systems multi model decomposition and control method based on angle | |
CN104182854A (en) | Mixed energy consumption measuring method for enterprise energy management system | |
CN1234056C (en) | Self-adaptation nonlinear time varying controller and controlling method thereof | |
CN109814389A (en) | The tight format non-model control method of the different factor of the MIMO of parameter self-tuning | |
CN105069323A (en) | Memetic algorithm based microbial fermentation control and optimization method | |
CN107423853A (en) | A kind of determination method and system of per unit area yield meteorology variation coefficient | |
CN114384931A (en) | Unmanned aerial vehicle multi-target optimal control method and device based on strategy gradient |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130417 |