CN105867169B - A kind of Modelling of Dynamic System method based on state observer - Google Patents
A kind of Modelling of Dynamic System method based on state observer Download PDFInfo
- Publication number
- CN105867169B CN105867169B CN201610247941.4A CN201610247941A CN105867169B CN 105867169 B CN105867169 B CN 105867169B CN 201610247941 A CN201610247941 A CN 201610247941A CN 105867169 B CN105867169 B CN 105867169B
- Authority
- CN
- China
- Prior art keywords
- state
- controlled device
- data
- observer
- dynamic
- 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
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
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The Modelling of Dynamic System method based on state observer that the invention discloses a kind of, it realizes especially by following steps: (1) input and output data of live controlled device being acquired, the data of acquisition include the data segment that controlled device response curve is transitioned into steady-state process by certain dynamic process;(2) one section of dynamic transition of dynamic process is chosen to the continuous data of stable state, sets prediction model, parameter and the state observer pole of controlled device, and the original state for setting observer is zero, and application state observer observes the observation state of controlled device;(3) using the prediction model of controlled device using observation state as original state, remaining data segment emulation after being chosen to step (2), seek the error sum of squares of simulation curve Yu controlled device reality output, and using error sum of squares as objective function, the optimizing of hypothesized model parameter is carried out using intelligent optimization algorithm.The present invention is not necessarily to carry out controlled device special modeling, more practical, conveniently.
Description
Technical field
The Modelling of Dynamic System method based on state observer that the present invention relates to a kind of, more particularly to one kind is with operational process
In Dynamic Closed Loop data be modeling initial data, application state observer to system dynamic course carry out state observation, and with
The observation state is system initial state, and the method for carrying out the modeling of optimizing to prediction model parameter using intelligent optimization algorithm belongs to
Modeling method field.
Background technique
When the modeling process of traditional industry controlled device generally requires system and is in stable state, to system be added step disturbance into
Row modeling.This modeling method requires system to be in stable state, is difficult to realize in the actual industrial process.Moreover, when system adds
When entering step disturbance, the stability of system can be made to decline, or even cause accident, be unfavorable for the safe and economical operation of system.
Summary of the invention
To overcome above-mentioned the deficiencies in the prior art, technical problem to be solved by the invention is to provide it is a kind of it is easy to use,
Safely and effectively, and without adjusting system it is in stable state, the dynamic based on state observer of step disturbance is added without to system
State system modeling method.
To solve the above problems, the technical solution used in the present invention is:
A kind of Modelling of Dynamic System method based on state observer, uses following steps to realize:
Step 1, on-site data gathering:
The input and output data of live controlled device are acquired, the data segment of acquisition should include that controlled device responds
Curve by certain dynamic transition to steady-state process data segment;The data segment is divided into two parts, and first part data segment ab is
Data segment comprising partial dynamic process, second part data segment bc are comprising remainder dynamic process data and whole stable states
The data segment of data;The beginning phase of the ending of first part's data segment ab and the second part data segment bc
Even;
Step 2, controlled device state estimation:
The prediction model of controlled device and the parameter of the prediction model are set, and using the prediction model as state observation
The benchmark model of device design, one piece of data, that is, first part data segment ab of dynamic process starting in selecting step 1, using shape
State observer observation Prediction System state sets the pole of state observer in observation process, and thinks the initial of observer
State is zero;
Step 3, controlled device Dynamic Process Modeling:
Using set prediction model using observation state in step 2 as original state, to second part data segment in step 1
Bc is emulated, and seeks the error sum of squares of simulation curve Yu controlled device reality output, and using the error sum of squares as mesh
Scalar functions carry out the optimizing of hypothesized model parameter using intelligent optimization algorithm.
Further, the data segment of acquisition is divided into two parts about in step 1, because identification process needs dynamic mistake
The continuous data to stable state is crossed, and recognizes the system initial state at the data point started and state observer progress state is needed to estimate
Meter, so the ending of state observer observation data must be connected with the beginning of identification process, data amount check does not have
Specific requirement, as long as meeting segmentation described above.The state estimation of controlled device and the identification process of controlled device
Must be it is end to end, the state estimated in this way is just significant to identification process, and state estimation is first part's data segment ab,
It is recognized as second part data segment bc.
Further, the setting about the prediction model of the controlled device in the step 2, due to controlled pair of many industry
As all carrying out modeling with modelling by mechanism either Experiment Modeling, model structure is relatively more fixed, and only parameter can be due to reality
The characteristic of the controlled device on border is different, when only prediction model structure is improper, recognizes the curve of output and reality of model
The curve of output degree of fitting on border is relatively poor, can change prediction model structure again, model again.So even if prediction model does not have
Have and refers to and can also be determined by test.And the prediction model structure type of industrial object is limited, can be selected with test
It takes.
The beneficial effects of adopting the technical scheme are that
This method is not necessarily to carry out special modeling to controlled device, provides a kind of based on live day-to-day operation data
More practical, convenient modeling method, have well promote and practical value.
This method need not selecting system steady working condition, need not also step disturbance be added to controlled device, site of deployment adopt
The day-to-day operation data of collection carry out closed loop modeling to system mode, easy to operate, safe and efficient, applied widely.That chooses shows
There is field data system to export the feature from unstable state to stable state, and split data into two parts, and first part's data segment is packet
The data segment of the process containing partial dynamic;Second part data segment is comprising remainder dynamic process data and whole steady state datas
Data segment.Controlled device is observed using first part's data and prediction model in the state of the segment data end position, and
Using the state as the original state of prediction model;Intelligence is carried out using parameter of the second part data to controlled device prediction model
Optimizing.This method application state observer carries out status tracking to prediction model, and using final tracking mode as controlled device
Initial state compensates for controlled device modeling process due to being unable to estimate controlled device status, must be requested that controlled device is initial
The shortcomings that state is zero;Using intelligent algorithm prediction model parameter is adjusted during, when prediction model parameter with
When practical plant model parameter matches preferable, the controlled device original state that observer observes is also relatively accurate, thus
The output bias of the output and realistic model that make the prediction model at moment locating for second part data is also smaller, ensure that controlled pair
The consistency of parameter regulation in the state estimation procedure and model parameter searching process of elephant.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, can also be obtained according to these attached drawings other attached drawings.
Fig. 1 be acquisition controlled device response curve by certain dynamic transition to stable state process state diagram.
Fig. 2 is the input data curve of the controlled device of acquisition.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below with reference to FIG. 1 to FIG. 2 and specific embodiment
Clear, complete description is carried out to the present invention.
As depicted in figs. 1 and 2, the present embodiment need not selecting system steady working condition, also need not to controlled device be added rank
The day-to-day operation data of jump disturbance, site of deployment acquisition carry out closed loop modeling to system mode, and the field data of selection, which has, is
Feature of the system output from unstable state to stable state, and two sections are splitted data into, wherein one section includes part steady-state process and dynamic mistake
Journey, referred to as data segment bc, for carrying out testing mould to hypothesized model;Another section includes remaining data, referred to as data segment ab, for estimating
Controlled device is counted in the state of the binding site of two segment datas.1. state Observer Design
Assuming that the transmission function structure type of controlled device is following formula (1):
Wherein: G (s) is controlled device transmission function;
K is proportional gain;
T is inertia time constant;
N is controlled device order;
According to the knowledge of modern control theory it can be concluded that the state-space expression of controlled device is following formula (2)
With formula (3):
Y=xn (3)
Wherein: [x1 x2 … xn]TFor state vector;
U is input;
Y is output;
It is obvious that controlled device is entirely capable of seeing, and coefficient matrix is following formula (4):
Wherein: A is sytem matrix;
B is control matrix;
C is observing matrix;
To controlled device progress state observation is assumed, if the state observation matrix G of state observer is following formula (5):
By modern control theory knowledge it is found that the proper polynomial P of closed loop observation system is following formula (6):
P=| SI-A+GC | (6)
To make the state estimation of the state observer of controlled device level off to state true value, need state observer
POLE PLACEMENT USING might as well set the pole matrix F of observation system in the left side of s plane as following formula (7):
F=[- a1 -a2 … -an] (7)
Determine that desired character multinomial is following formula (8) by desired observation pole:
(s+a1)(s+a2)…(s+a3)=α0+α1s+…+αn-1sn-1+sn (8)
Since the coefficient of A and C is it is known that have:
| SI-A+GC |=β0+β1s+……+βn-1sn-1+sn (9)
Then have, according to the equal relationship of coefficient of correspondence, the following formula of n algebraic equation (10) can be obtained:
The element value of state observation matrix can be solved.
2. controlled device state estimation
Since the state estimation of state observer levels off to state true value, it might as well think that state observer original state is
Zero and assume system model it is accurate, output and input using the system in first part data segment ab, to the system shape of b point
State is observed, if above-mentioned hypothesized model is accurate, in the sufficiently long situation of first part data segment ab, the state of b point is estimated
Meter is inevitable accurate.By modern control theory it is found that the state-space expression of state observer is following formula (11)~(12):
Wherein:For observation state vector;
For observation output;
Controlled device is some system, so what system referred to herein is exactly controlled device, systematic discrete system side
Journey is following formula (13):
Wherein, TsFor sampling time interval;
The selection of state observation matrix G can determine the speed that evaluated error e (t) goes to zero, when the pole of observation system is matched
When setting place far from the imaginary axis on the left of s plane, band system band is very wide, declines to the rejection ability of noise, due to that can select
Enough time segment length are taken, so herein, by the POLE PLACEMENT USING of observer from the closer place of the imaginary axis, generally selection pole
Point position is less than 1/T, to can not only guarantee the accuracy of state observation, but also can guarantee the stability of observation data.People
The observation speed for being adjustment state observer to the state of controlled device is acted on for setting state observer pole.
When hypothesized model is accurate and first part's data segment ab long enough, State Viewpoint is carried out to first part data segment ab
It surveys, it is believed that the State Viewpoint measured value of b point is consistent with system virtual condition, to solve the initial shape in Dynamic Process Modeling
The puzzlement that state can not determine.
3. system dynamic course models
It is obtained in the dynamic process of system after the state of b point by state observer, is new system initial state using the state,
Assuming that estimation model is accurate, system is emulated, after seeking difference using the simulation data and reality output at each moment, seeks difference
Quadratic sum, objective function whether in this, as judgment models accurately.Using intelligent optimization algorithm to the parameter of hypothesized model
It modifies, optimizing is carried out to objective function, to establish the plant model of dynamic process.
State observer is following formula (14) in the State Viewpoint measured value of b point:
Then think that practical object by the virtual condition of original state of b point is following formula (15):
Then have, system emulated using b point as starting point, the discrete equation of system is following formula (16):
The input of simulation process is consistent with actually entering, and continues to systematic steady state, records the output of simulation process,
After seeking difference with the output of process and reality output, then seeking quadratic sum as objective function is following formula (17):
With formula (17) for objective function, optimizing, optimizing are carried out using model parameter of the intelligent optimization algorithm to hypothesized model
Process are as follows: provide the parameter area of certain hypothesized model, intelligent optimization algorithm gives certain group model parameter to subprogram, sub- journey
Sequence seeks corresponding state observation matrix G according to the group model parameter, and is estimated with the system mode that the matrix parameter carries out b point
Meter, using the estimated state of b point as the initial state of b point, emulates subsequent process, seeks the objective function of system, so past
It is multiple, to seek optimal model parameter.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment or equivalent replacement of some of the technical features;And
These are modified or replaceed, the spirit and model of technical solution of the embodiment of the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (2)
1. a kind of Modelling of Dynamic System method based on state observer, characterized by the following steps:
Step 1, on-site data gathering:
The input and output data of live controlled device are acquired, the data of acquisition include controlled device response curve by certain
Dynamic process is transitioned into the data segment of steady-state process;
Step 2, controlled device state estimation:
One section of dynamic transition of dynamic process is to the continuous data of stable state in selecting step 1, set controlled device prediction model and
The parameter of the prediction model, and state observer pole is set, and the original state for setting observer is zero, application state
The observation state of observer observation controlled device;
Step 3, controlled device Dynamic Process Modeling:
The prediction model of controlled device is selected using observation state described in step 2 as original state in step 2 in applying step 2
It takes rear remaining data segment to be emulated, seeks the error sum of squares of simulation curve Yu controlled device reality output, and with described
Error sum of squares is objective function, and the optimizing of hypothesized model parameter is carried out using intelligent optimization algorithm.
2. a kind of Modelling of Dynamic System method based on state observer according to claim 1, it is characterised in that: described
The data of acquisition are divided into two sections, one section for by the initial position of dynamic process data to controlled device still in dynamic process certain
The position at moment, it further includes the data segment that controlled device enters stable state that another section, which had both included the remaining data segment of dynamic process,.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610247941.4A CN105867169B (en) | 2016-04-20 | 2016-04-20 | A kind of Modelling of Dynamic System method based on state observer |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610247941.4A CN105867169B (en) | 2016-04-20 | 2016-04-20 | A kind of Modelling of Dynamic System method based on state observer |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105867169A CN105867169A (en) | 2016-08-17 |
CN105867169B true CN105867169B (en) | 2019-03-08 |
Family
ID=56633231
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610247941.4A Active CN105867169B (en) | 2016-04-20 | 2016-04-20 | A kind of Modelling of Dynamic System method based on state observer |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105867169B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110806693B (en) * | 2019-10-31 | 2021-04-27 | 南京航空航天大学 | Gray wolf prediction control method for time lag of plate heat exchanger |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101639664A (en) * | 2009-08-14 | 2010-02-03 | 东莞理工学院 | Predictive control device and method |
CN102354104A (en) * | 2005-09-19 | 2012-02-15 | 克利夫兰州立大学 | Controllers, observers, and applications thereof |
CN102375442A (en) * | 2010-08-23 | 2012-03-14 | 同济大学 | Real-time on-line control system and method for miscellaneous nonlinear system |
CN103676651A (en) * | 2013-12-02 | 2014-03-26 | 国家电网公司 | Method for predicting and controlling steam temperatures of boilers on basis of state observation model |
CN104714564A (en) * | 2014-12-17 | 2015-06-17 | 浙江工业大学 | Distributed liquid level system control method based on extendstate observer |
CN105182756A (en) * | 2015-10-22 | 2015-12-23 | 东华大学 | Novel inner module compensation control system based on model reference adaptive control |
CN105371872A (en) * | 2015-12-24 | 2016-03-02 | 哈尔滨工业大学 | Extended high-gain observer based disturbance estimation method for gyrowheel system |
-
2016
- 2016-04-20 CN CN201610247941.4A patent/CN105867169B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102354104A (en) * | 2005-09-19 | 2012-02-15 | 克利夫兰州立大学 | Controllers, observers, and applications thereof |
CN101639664A (en) * | 2009-08-14 | 2010-02-03 | 东莞理工学院 | Predictive control device and method |
CN102375442A (en) * | 2010-08-23 | 2012-03-14 | 同济大学 | Real-time on-line control system and method for miscellaneous nonlinear system |
CN103676651A (en) * | 2013-12-02 | 2014-03-26 | 国家电网公司 | Method for predicting and controlling steam temperatures of boilers on basis of state observation model |
CN104714564A (en) * | 2014-12-17 | 2015-06-17 | 浙江工业大学 | Distributed liquid level system control method based on extendstate observer |
CN105182756A (en) * | 2015-10-22 | 2015-12-23 | 东华大学 | Novel inner module compensation control system based on model reference adaptive control |
CN105371872A (en) * | 2015-12-24 | 2016-03-02 | 哈尔滨工业大学 | Extended high-gain observer based disturbance estimation method for gyrowheel system |
Also Published As
Publication number | Publication date |
---|---|
CN105867169A (en) | 2016-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
DE112020003136T5 (en) | Method for generating a lane change decision model, method and device for lane change decision of an unmanned vehicle | |
CN105069826B (en) | The modeling method of elastomeric objects amoeboid movement | |
CN103984234A (en) | Electro hydraulic servo system self-correction fuzzy PID control method | |
CN108121215B (en) | Process control loops method of evaluating performance and device based on full loop reconstruct emulation | |
CN107273693A (en) | A kind of hydrocarbon fuel mechanism simplification method | |
CN110398029B (en) | Control method and computer-readable storage medium | |
CN104881512A (en) | Particle swarm optimization-based automatic design method of ripple-free deadbeat controller | |
CN105186959A (en) | Parameter setting method of sliding mode controller of servo system | |
CN105867169B (en) | A kind of Modelling of Dynamic System method based on state observer | |
CN110456799A (en) | A kind of online incremental learning method of automatic driving vehicle Controlling model | |
CN109344968A (en) | A kind of method and device of the hyper parameter processing of neural network | |
CN103411628B (en) | A kind of disposal route of MEMS gyro instrument Random Drift Error | |
CN110490319A (en) | Distributed deeply study based on fused neural network parameter | |
CN111513723A (en) | Motion attitude monitoring method, motion attitude adjusting device and terminal | |
CN106372366A (en) | Intelligent power generation control method based on hill-climbing algorithm | |
Yucheng et al. | Incremental learning method of least squares support vector machine | |
CN104035328B (en) | A kind of multiple movement bodies tracking and controlling method using interference estimator | |
CN111694272B (en) | Self-adaptive control method and device of nonlinear multi-agent based on fuzzy logic system | |
CN107168066A (en) | A kind of greenhouse self-adaptation control method | |
CN116995682A (en) | Adjustable load participation active power flow continuous adjustment method and system | |
Huang et al. | Introduction to TensorFlow 2 | |
CN110450164A (en) | Robot control method, device, robot and storage medium | |
CN111753423A (en) | XGboost-based inter-well dynamic connectivity quantification method | |
CN106094524A (en) | The rapid model prediction control method compensated based on input trend | |
CN103595741A (en) | A node in P2P stream media and a method for optimizing a neighbor node table of the node |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |