CN103499920A - Control parameter optimization method and system through vector time series prediction and expert fuzzy transformation ratio - Google Patents

Abstract

The invention discloses a control parameter optimization method and system through vector time series prediction and the expert fuzzy transformation ratio. The control parameter optimization method and system through vector time series prediction and the expert fuzzy transformation ratio can be applied to the fields such as the field of control, the field of decision making and the field of artificial intelligence. According to the scheme, input which serves as control parameter on-line optimization setting is predicted and is output through a vector time series prediction method according to the states of a control system and output time series, a transformation ratio value of the control parameter on-line optimization setting is predicted through an expert rule table or a fuzzy controller to carry out on-line optimization setting on control parameters, and the control parameter optimization method through vector time series prediction and the expert fuzzy transformation ratio is formed by combining the two points and a classic control method, further, a specific device and a specific connection relation are designed, and the control parameter optimization system through vector time series prediction and the expert fuzzy transformation ratio is formed. According to the control parameter optimization method and system through vector time series prediction and the expert fuzzy transformation ratio, on-line optimization setting can be carried out on the control parameters according to difference between the working environments and uncertainty of the working environments, a control system can deal with various sudden environmental changes, dynamic performance, anti-interference performance and static performance are improved, and the defect of overshoot is overcome.

Description

Vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method and system

Technical field

The present invention proposes a kind of vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method and system, may be used on the fields such as control, decision-making, artificial intelligence.

Background technology

How to make control system can simulate people's thinking, according to the applied environment automatic on-line adjust (optimization) control parameter, and obtain the target that better control performance is automation field scientist effort for a long time, be also the problem that the present invention solves.The experts and scholars of this area combine the method for artificial intelligence mode with classical control method has obtained various advanced persons' intelligent control method, can be divided into two classes: the first kind is based on the intelligent control method (unified referred to as the R-PID control method) of rule-based reasoning mechanism, expert PID (Expert PID for example, E-PID) on-line tuning method, fuzzy (Fuzzy PID, F-PID) on-line tuning method; Equations of The Second Kind is based on the intelligent control method (this research is unified referred to as the XNN-PID control system) of Neurocomputing Science, BP Neural network PID (Back propagation neural networks PID for example, BPNN-PID) on-line tuning method, RBF Neural network PID (Radial basis function neural networks PID, RBFNN-PID) on-line tuning method.

The problem that the above-mentioned first kind (Based Intelligent Control of rule-based inference mechanism) method (technology) exists is: (1) is controlled parameter, controlled quentity controlled variable sudden change and is caused that system output is unstable, process is not steady; (2) dynamic property of system (the quick property of system responses, anti-interference recover rapidity), static properties (static error) are poor, have much room for improvement; (3) control system degree of intelligence (predictive ability, learning ability, adaptive ability) is poor, has much room for improvement.

The problems referred to above be reflected in produce to give birth to or application in be mainly manifested in: (1) controller parameter can't carry out the on-line optimization control parameter of adjusting according to the difference of working environment, controlled device etc. and uncertainty, before control system is applied each time, the slip-stick artist is debugging control parameter repeatedly, and system is after operation a period of time, and the slip-stick artist has to such an extent that repeatedly check in the debugging control parameter; (2) control system lacks coping mechanism for the various sudden variation of environment, can only make control decision according to current situation, even control decision is harmful to for the output in system future, makes response after also can only by the time be harmful to the effect generation again; (3) dynamic property of control system and Immunity Performance have much room for improvement; (4) static properties of control system has much room for improvement; (5) can produce overshoot in some applications, and some engineering application do not allow to occur the situation of overshoot.

Summary of the invention

The present invention proposes a kind of vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method and system, can turn to following two cover embodiment by instantiation: (1) vector time series prediction no-load voltage ratio expert controls (VARMA-EA-PID) method and system; (2) vector time series prediction no-load voltage ratio fuzzy control (VARMA-EA-PID) method and system.

One, purpose of the present invention:

The objective of the invention is to solve 5 aspect problems in above-mentioned productive life application.Be that target 1 is to solve controller parameter can't carry out the on-line optimization control parameter of adjusting according to the difference of working environment, controlled device etc. and uncertainty, before control system is applied each time, the slip-stick artist is debugging control parameter repeatedly, and, after system operation a period of time, the slip-stick artist has to such an extent that repeatedly check the problem in the debugging control parameter; Target 2 is to solve the various sudden variation shortage coping mechanism of control system for environment, can only make control decision according to current situation, even control decision is harmful to for the output in system future, also can only by the time be harmful to after effect produces the problem of making again response; Target 3 is to solve the dynamic property of control system and the problem that Immunity Performance has much room for improvement; Target 4 is problems that the static properties of solution control system has much room for improvement; Target 5 is to solve in some applications to produce overshoot, and the application of some engineerings does not allow to occur the problem of the situation of overshoot.

Two, the technical problem to be solved in the present invention:

1. solve and how to make the output of control system be predicted according to the time series of each state in system and the time series of output, control decision (controlling optimizing and revising of parameter) can be calculated according to following system output;

2. how the result of prediction combines with controlling the online card of the parameter method of determining, and gives full play of the advantage of classical PID control system, artificial intelligence approach (expert system, Fuzzy control system) and Predictive Control System;

3. solution fuzzy and expert PID method are controlled parameter, the controlled quentity controlled variable sudden change causes unstable, the jiggly problem of process of system output; The dynamic property of fuzzy and expert PID method (the quick property of system responses, anti-interference recover rapidity), the poor problem of static properties (static error); The problem that fuzzy and expert PID method degree of intelligence (predictive ability, learning ability, adaptive ability) are poor.

Three, the technical solution used in the present invention

To achieve these goals, the present invention has taked following technical scheme:

A kind of vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method is characterized in that:

The method by step (001) to step (014) totally 14 steps form:

Step (001): initialization:

Initialization max calculation cycle Kmax: wherein Kmax=0 or Kmax ∈ N, N means natural number, if Kmax=0 means that control algolithm stops never;

The current computation period k:k=1 of initialization;

Target Rin (k): Rin (k)=f (k) is controlled in initialization, and wherein f (n) means Controlling object function, according to moment n, provides the control target;

The initialization prediction step is L _predict, L _predictfor positive integer, according to engineering experience, its span is referenced as: 1≤L _predict≤ 50;

Initialization prediction Startup time T _predict, according to engineering experience, its span is referenced as: 5%Kmax≤T _predict≤ 10%Kmax, if Kmax=0,10L _predict≤ T _predict≤ 100; Initialization system output yout (k)=0;

Determine expert fuzzy no-load voltage ratio inference rule and initialization expert fuzzy no-load voltage ratio inference rule parameter;

Step (002): system is truly exported collection: when the method for the present invention's proposition is used for real system; yout_real (k) is that the controlled device of necessary being produces; by harvester, gather controlled device and produce true output yout_real (k); The method proposed as the present invention is during for computer simulation experiment, and yout_real (k) is that the calculated with mathematical model by controlled device draws: during current calculatings moment k=1, and indirect assignment yout_real (k)=1; Current calculating is k>1 o'clock constantly, and yout_real (k) was calculated k-l performance period;

Step (003): judge whether to start prediction: if current time not yet arrives prediction Startup time, i.e. k<T _prsdict, do not start prediction, jump to step (004); If current time arrives prediction Startup time, i.e. k>=T _prsdict, start prediction, jump to step (005);

Step (004): indirect assignment: not yet start prediction, participate in the yout (k) that error calculates=yout_real (k-1);

Step (005): prediction output: started prediction, participated in the yout (k) that error calculates=y _predict(k); Computing method adopt the vector time series Forecasting Methodology: the time series that is input as following five variablees of vector time series prediction algorithm: proportional control parameter time series K _p(1), K _p(2) ..., K _p(k), integration control parameter time series K _i(1), K _i(2) ..., K _i(k), differential is controlled parameter time series K _d(1), K _d, (2) ..., K _d(k), controlled quentity controlled variable time series u (1), u (2) ..., u (k), control system output yout_real (1), yout_real (2) ..., yout_real (k); The vector time series prediction algorithm is output as the constantly rear L of k _predictsystem output y constantly _predict(k);

Step (006): error is calculated: calculate the error e (k) of current time according to sampled value yout (k) and the desired value rin (k) of controlled device of controlled device current time, wherein e (k)=yout (k)-rin (k);

Step (007): error rate calculates: the error rate ec (k) between the error e (k) of calculating current time and the error e (k-1) of previous moment, wherein ec (k)=e (k)-e (k-1);

Step (008): control the parameter no-load voltage ratio and calculate: according to Fuzzy Controller Parameters T _fAperhaps Expert Rules table T _eAadopt the method for Fuzzy Calculation or inquiry Expert Rules table, being input as of Fuzzy Calculation method or inquiry Expert Rules table method: current time error e (k) and error rate ec (k), Fuzzy Calculation method or inquiry Expert Rules table method are output as controls parameter no-load voltage ratio R _p(k), R _i(k), R _d(k);

Step (009): control calculation of parameter: according to formula 1, formula 2, formula 3, calculate the control parameter K _p(k), K _i(k), K _d(k);

K _p(k)=R _p(e (k), ec (k)) K _p(k-1) formula 1

K _i(k)=R _i(e (k), ec (k)) K _i(k-1) formula 2

K _d(k)=R _d(e (k), et (k)) K _d(k-1) formula 3

Step (010): calculate controlled quentity controlled variable: adopt Digital PID Algorithm, the input of Digital PID Algorithm is to control parameter K _p(k), K _i(k), K _d(k), output is controlled quentity controlled variable u (k);

Step (011): controlled device produces true output: when the method for the present invention's proposition is used for real system, controlled quentity controlled variable acts on controlled device, and controlled device produces true output yout_real (k); When the method for the present invention's proposition is used for computer simulation experiment; calculated with mathematical model by controlled device draws yout_real (k); the input of the mathematical model of above-mentioned controlled device is controlled quentity controlled variable u (k), and output is that controlled device produces true output yout_real (k);

Step (012): computation period is from increasing: i.e. current time k=k+1;

Step (013): judge whether to finish: if maximum performance period Kmax>0 and current time k>Kmax jump to step (014): finish; If maximum performance period Kmax=0, jump to step (002), repeat process;

Step (014): finish.

Wherein determine expert fuzzy no-load voltage ratio inference rule and initialization expert fuzzy no-load voltage ratio inference rule parameter, be specially:

If expert fuzzy no-load voltage ratio inference rule adopts the method for expert's no-load voltage ratio reasoning, adopt method for designing design specialist's no-load voltage ratio rule list of Expert Rules table in expert PID, in expert PID, the input of Expert Rules table and expert's no-load voltage ratio rule list is all moment error e (k) and error rate ec (k), and in expert PID, the output of Expert Rules table is to control parameter K _p(k), K _i(k), K _d, and the output of expert's no-load voltage ratio rule list is to control parameter no-load voltage ratio R (k) _p(k), R _i(k), R _d(k);

If expert fuzzy no-load voltage ratio inference rule adopts the method for fuzzy calculation inference, adopt the method for designing of indistinct logic computer in fuzzy to design fuzzy no-load voltage ratio inference machine, in fuzzy, the input of indistinct logic computer and fuzzy no-load voltage ratio inference machine is all moment error e (k) and error rate ec (k), and in fuzzy, the output of indistinct logic computer is to control parameter K _p(k), K _i(k), K _d, and the output of fuzzy no-load voltage ratio inference machine is to control parameter no-load voltage ratio R (k) _p(k), R _i(k), R _d(k).

Apply the system of described a kind of vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method, it is characterized in that: system is by 5 installation compositions: decision-making device (100), controller (101), controlled device (102), controlled quentity controlled variable sensing transducer (103), controlled device output sensing transducer (104);

Decision-making device can be that computing machine or Programmable Logic Controller or be input as controlled target (system input digital quantity), controlled quentity controlled variable (digital quantity), system and truly exported (digital quantity), is output as the device of controlling parameter; Wherein decision-making device is by its input equipment input control target (system input digital quantity); Decision-making device passes through RS232 interface or Ethernet interface or capture card interface and is connected with the controlled quentity controlled variable sensing transducer, receives up-to-date controlled quentity controlled variable (digital quantity); Decision-making device passes through RS232 interface or Ethernet interface or capture card interface and is connected with controlled device output sensing transducer, receives up-to-date system and truly exports (digital quantity);

Wherein, decision-making device can be a device, also can be by 3 sub-installation compositions: control parameter no-load voltage ratio on-line tuning device, control parameter on-line tuning device, fallout predictor; These 3 devices can be all computing machine or Programmable Logic Controller or the device that meets subordinate's annexation and input/output relation: control parameter no-load voltage ratio on-line tuning device by its input equipment input control target (system input digital quantity), by Ethernet or communication bus, with fallout predictor, be connected, receiving control system prediction output, be connected output controlled quentity controlled variable no-load voltage ratio value with control parameter on-line tuning device by Ethernet or communication bus; Control parameter on-line tuning device and be connected with controller with fallout predictor by Ethernet or communication bus, export up-to-date control parameter; Fallout predictor is connected with controlled quentity controlled variable sensing transducer and controlled device output sensing transducer by RS232 interface or Ethernet interface, capture card transmission interface, and receiving controlled quentity controlled variable (digital quantity) and system, truly to export (digital quantity) be to input;

Controller can be with the PLC of programing function and driver or with the programing function frequency converter or can produce the device that controlled quentity controlled variable drives controlled device according to pid control parameter; Controller is connected with control parameter on-line tuning device by industrial bus, data line, accepts to control parameter; Controller output controlled quentity controlled variable, is connected with controlled device by physical construction or conduction, transmission medium, to controlled device generation driving;

Controlled device can be the object of electric motor and driving object or pressure control equipment and object or the controlled quentity controlled variable impact that can controlled device produces;

The controlled quentity controlled variable sensing transducer can be voltage collecting device, pressure acquisition device, rotating speed harvester or analog acquisition device; The input of controlled quentity controlled variable sensing transducer and the output of controller are connected by wire or gearing or conduction device, obtain the analog quantity of controller output; The controlled quentity controlled variable sensing transducer is translated into controlled quentity controlled variable (digital quantity), then is connected pipage control amount (digital quantity) with decision-making device by RS232 interface or Ethernet interface, capture card transmission interface;

Controlled device output sensing transducer can be voltage collecting device, pressure acquisition device, rotating speed harvester or analog acquisition device; The input of controlled device output sensing transducer and controlled device are connected by wire or gearing or conduction device, obtain the analog quantity that controlled device is exported; Controlled device output sensing transducer is translated into system and truly exports (digital quantity), then is connected with decision-making device by RS232 interface or Ethernet interface, capture card transmission interface, and induction system is truly exported (digital quantity);

Marriage relation and the implementation procedure of system and method are as follows:

Step (001) initialization is realized by decision-making device; Specifically by control parameter no-load voltage ratio on-line tuning device reception input, set, initialization result is distributed to and controls parameter on-line tuning device and fallout predictor;

Step (002) system is truly exported to gather by controlled quentity controlled variable sensing transducer and controlled device output sensing transducer and is coordinated decision-making device to realize;

Step (003) judges whether to start to be predicted to step (005) prediction output and is realized by fallout predictor;

Step (006) error is calculated to step (008) and controls the parameter transformation-ratio meter parameter on-line tuning device realizes by controlling at last;

Step (009) control parameter transformation-ratio meter is realized by control parameter on-line tuning device at last;

Step (010) controlled quentity controlled variable is calculated and is realized by controller;

Step (011) controlled device produces true output and is realized by controlled device;

Step (012) computation period finishes from increasing to step (014), by decision-making device, is realized.

In sum, the technological means that the present invention adopts is: (1) adopts the prediction of vector time series Forecasting Methodology to export the input of adjusting as controlling the parameter on-line optimization according to the time series of each state of control system and output, (2) adopt no-load voltage ratio value that Expert Rules table or fuzzy controller PREDICTIVE CONTROL parameter on-line optimization adjust and then the on-line optimization control parameter of adjusting, (3) above-mentioned 2 classical control methods of combination have been formed to vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method, further design concrete device, linking relationship has also formed the system that has adopted said method.

Four, the present invention compares with existing technical scheme advantage and beneficial effect:

The present invention compares and has following advantage with the prior art scheme: (1) makes controller parameter to carry out the on-line optimization control parameter of adjusting according to the difference of working environment and uncertainty, before control system is applied each time, the slip-stick artist only needs roughly setup control parameter, automatically can be controlled the optimization of parameter after the online implementing operation and adjust; (2) control system can be tackled for the various sudden variation of environment, can do control decision according to the result of prediction, responds in advance and tackle bad control result; (3) dynamic property of control system and Immunity Performance improve; (4) static properties of control system improves; (5) overcome the situation that overshoot occurs in some applications.

2. the adaptive parameter on-line tuning of the expert fuzzy increment type optimization system that the present invention designs, the process that has realized whole controller parameter adjustment optimization is all realized by algorithm self, process nobody's that on-line parameter optimization is adjusted participation, control system can be according to self hardware characteristics and working environment automatic adjusting adaptation parameters.

3. the adaptive parameter on-line tuning of the expert fuzzy increment type optimization method that the present invention proposes, a plurality of control loops of material resources structural similarities have been solved because working environment changes or there are differences, the problem that controller parameter can't be optimized according to difference and the uncertain online self-tuning of working environment; A plurality of control loops of material resources structural similarities have been solved because there is difference in controlled device, the problem that controller parameter can't be optimized according to difference and the uncertain online self-tuning of controlled device; A plurality of control loops that solved the material resources structural similarities need to be controlled parameter tuning repeatedly, and controller parameter can't complete the problem of the work online self-tuning optimization of this repetition automatically.

The accompanying drawing explanation

The system and device annexation figure of Fig. 1 based on vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method;

The process flow diagram of Fig. 2 vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method;

Embodiment

Below in conjunction with accompanying drawing 1 and Fig. 2, the embodiment of the vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method and system that the present invention is proposed is described in detail and illustrates.

1. vector time series is predicted expert fuzzy no-load voltage ratio Optimization about control parameter method and system, is that scheme is as follows in fact:

1.1 the method by step (001) to step (014) totally 14 steps form:

1.1.1 step (001): initialization:

Initialization max calculation cycle Kmax:Kmax=0, mean that control algolithm stops never;

The current computation period k:k=1 of initialization;

Target (system input) Rin (k): Rin (k)=1 is controlled in initialization, and each moment n control target (system input) is a constant;

The initialization prediction step is L _predict: L _predict=20;

Initialization prediction Startup time T _predict: T _predict=100;

In the present embodiment, expert fuzzy no-load voltage ratio inference rule adopts the method for expert's no-load voltage ratio reasoning, adopts method for designing design specialist's no-load voltage ratio rule list of Expert Rules table in expert PID, as shown in formula 4, formula 5 and formula 6:

$R_{\mathrm{Kp}}(e\left(k\right),\mathrm{ec}\left(k\right))=$

$\left\{\begin{array}{ccccc}1& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1.002& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& {\mathrm{EC}}_{\mathrm{low}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 1.004& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\\ 0.998& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& {\mathrm{EC}}_{\mathrm{low}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 1.002& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\\ 0.998& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|<E_{\mathrm{low}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|E_{\mathrm{low}}& \mathrm{and}& {\mathrm{EC}}_{l\mathrm{ow}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|<E_{\mathrm{low}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\end{array}\right.$ Formula 4

$R_{\mathrm{Ki}}(e\left(k\right),\mathrm{ec}\left(k\right))=$

$\left\{\begin{array}{ccccc}1.002& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1.002& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& {\mathrm{EC}}_{\mathrm{low}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 1.002& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\\ 1& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& {\mathrm{EC}}_{\mathrm{low}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\\ 0.998& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|<E_{\mathrm{low}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 0.998& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|E_{\mathrm{low}}& \mathrm{and}& {\mathrm{EC}}_{l\mathrm{ow}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 0.998& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|<E_{\mathrm{low}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\end{array}\right.$ Formula 5

$R_{\mathrm{Kd}}(e\left(k\right),\mathrm{ec}\left(k\right))=$

$\left\{\begin{array}{ccccc}1.002& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& {\mathrm{EC}}_{\mathrm{low}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 0.998& ,\mathrm{when}& \left|e\left(k\right)\right|\&GreaterEqual;E_{\mathrm{high}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\\ 1.002& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& {\mathrm{EC}}_{\mathrm{low}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 0.998& ,\mathrm{when}& E_{\mathrm{low}}\&le;\left|e\left(k\right)\right|<E_{\mathrm{high}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\\ 1.002& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|<E_{\mathrm{low}}& \mathrm{and}& \left|\mathrm{ec}\left(k\right)\right|\&GreaterEqual;{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|E_{\mathrm{low}}& \mathrm{and}& {\mathrm{EC}}_{l\mathrm{ow}}\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{high}}\\ 1& ,\mathrm{when}& 0\&le;\left|e\left(k\right)\right|<E_{\mathrm{low}}& \mathrm{and}& 0\&le;\left|\mathrm{ec}\left(k\right)\right|<{\mathrm{EC}}_{\mathrm{low}}\end{array}\right.$ Formula 6

Wherein order:

E _high=0.5，

E _low=0.2，

EC _high=0.002，

EC _low=0.001。

R wherein _x(e (k), ec (k)) correction that the value that means the X parameter determines according to error current e (k) and error rate ec (k), X means numbering or the implication of parameter, e (k) means the error of current time k collection value y (k) and desired value r (k), E _high, E _lowmean to divide respectively the boundary value of combination in that state of error current e (k) and error rate ec (k); H, m, l means respectively e (k), ec (k) and E _high, E _lowthe state at the value place determined more afterwards;

Annotate: if expert fuzzy no-load voltage ratio inference rule adopts the method for fuzzy no-load voltage ratio reasoning, fuzzy controller can design (providing the Matlab Program) in the following manner:

a=newfis('fuzzpid');

a=addvar(a,'input','e',[-3,3]); %Parameter e

a=addmf(a,'input',1,'NB','zmf',[-3,1]);

a=addmf(a,'input',1,'Z','trimf',[-2,0,2]);

a=addmf(a,'input',1,'PB','smf',[-1,3]);

a=addvar(a,'input','ec',[-3,3]); %Parameter ec

a=addmf(a,'input',2,'NB','zmf',[-3,1]);

a=addmf(a,'input',2,'Z','trimf',[-2,0,2]);

a=addmf(a,'input',2,'PB','smf',[-1,3]);

a=addvar(a,'output','d_kp',[-0.004,0.004]); %Parameter kp

a=addmf(a,'output',1,'NB','zmf',[-0.004,0]);

a=addmf(a,'output',1,'Z','trimf',[-0.002,0,0.002]);

a=addmf(a,'output',1,'PB','smf',[0,0.004]);

a=addvar(a,'output','d_ki',[-0.004,0.004]); %Parameter ki

a=addmf(a,'output',2,'NB','zmf',[-0.004,0]);

a=addmf(a,'output',2,'Z','trimf',[-0.002,0,0.002]);

a=addmf(a,'output',2,'PB','smf',[0,0.004]);

a=addvar(a,'output','d_kd',[-0.004,0.004]); %Parameter kp

a=addmf(a,'output',3,'NB','zmf',[-0.004,0]);

a=addmf(a,'output',3,'Z','trimf',[-0.002,0,0.002]);

a=addmf(a,'output',3,'PB','smf',[0,0.004]);

a=addrule(a,rulelist);

a=setfis(a,'DefuzzMethod','centroid');

1.1.2 step (002): system is truly exported collection: when the method for the present invention's proposition is used for real system; yout_real (k) is that the controlled device of necessary being produces; by harvester, gather controlled device and produce true output yout_real (k); The method proposed as the present invention is during for computer simulation experiment, and yout_real (k) is that the calculated with mathematical model by controlled device draws: during current calculatings moment k=1, and indirect assignment yout_real (k)=1; Current calculating is k>1 o'clock constantly, and yout_real (k) was calculated k-1 performance period;

1.1.3 step (003): judge whether to start prediction: if current time not yet arrives prediction Startup time, i.e. k<T _prsdict, do not start prediction, jump to step (004); If current time arrives prediction Startup time, i.e. k>=T _prsdict, start prediction, jump to step (005);

1.1.4 step (004): indirect assignment: not yet start prediction, participate in the yout (k) that error calculates=yout_real (k-1);

1.1.5 step (005): prediction output: started prediction, participated in the yout (k) that error calculates=y _predict(k); Computing method adopt the vector time series Forecasting Methodology: the time series that is input as following five variablees of vector time series prediction algorithm: proportional control parameter time series K _p(1), K _p(2) ..., K _p(k), integration control parameter time series K _i(1), K _i(2) ..., K _i(k), differential is controlled parameter time series K _d(1), K _d(2) ..., K _d(k), controlled quentity controlled variable time series u (1), u (2) ..., u (k), control system output yout_real (1), yout_real (2) ..., yout_real (k); The vector time series prediction algorithm is output as the constantly rear L of k _predictsystem output y constantly _predict(k);

Being described below of vector time series Forecasting Methodology:

Control parameter K _p(t), K _i(t), K _d(t), controlled quentity controlled variable u (t), control system output y _system(t), there is interactional relation each other in l=5 (l representation element number or dimension) state or output variable altogether, forms 5 elementary time series models, referring to formula 7:

{ X _t}={ χ _1t, χ _2t..., χ _5t}={ K _pt, K _it, K _dt, u _t, y _{system t}formula 7

When fixed time t, X _t=[x _1t, x _2t..., x _5t] ^t=[K _pt, K _it, K _dt, u _t, y _{system t}] ^tthe time series vector of one 5 dimension, wherein each component x _rt(r=1,2 ..., 5) be to control parameter K _p(t), K _i(t), K _d(t), controlled quentity controlled variable u (t), control system output y _system(t) in the t value in the moment; When change moment t, { X _t5 yuan of (five components) vector time series have been formed.Due to increasing of time series unit number, polynary sequential { X _tthan monobasic sequential { x _tcontain abundanter information, it can not only reflect single sequential { x _rt(r=1,2 ..., 5) statistical relationship of internal data between successively, and can also reflect different sequential { x _rtand { x _st(r, s=1,2 ..., 5) between the mutual statistical relation.

For l=5 unit in this research steadily, normal state, zero-mean sequential { X _t}={ x _1t, x _2t..., x _5t(t=1,2 ..., N), if time series vector X _tvalue not only with each time series vector X of front n step _t-1, X _t-2..., X _t-nrelevant, but also with each noise vector A of front m step _t-1, A _t-2..., A _t-nrelevant, can be to { X _tset up l dimension model, referring to formula 8:

$X_t=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}_i\&CenterDot;X_{t-i}+\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{\mathrm{\&Theta;}}_j\&CenterDot;A_{t-j}+A_t$ Formula 8

Φ in formula 8 _i, Θ _jbe respectively auto-regressive parameter matrix and moving average parameter matrix, be l rank square formation; N, m are respectively the order of autoregression part and running mean part; X _t, A _tbe respectively l dimension random vector.Each matrix and vectorial form are referring to formula 9.

formula 9

In this research, the order of autoregression part and running mean part is 2 rank (n=2, m=2), and the seasonal effect in time series dimension is 5 dimensions (l=5), i.e. adopt VARMA (2,2,5) model in this research.

Adopt vector model Scalarizing Method estimation VARMA model parameter, the VARMA model is launched to l and become a Scalar Model, then estimate respectively the parameter of l Scalar Model by long autoregression calculating residual error method, then these parameter combinations are got up and obtained the VARMA model.

While adopting the VARMA algorithm, the span of nAR and nMA parameter is the positive integer in 5 to 10 closed intervals; In Matlab, the VARMA prediction algorithm is realized by vgxset (), vgxvarx (), three functions of vgxpred (), below provides the example that a method is implemented:

VARMA algorithmic function source program is as follows:

functionypredict=myarma5in2ar2ma(Y,PredictLength)

Ylog=diff(log(Y));

S=ceil(0.1*size(Ylog,1));

Ypre1=Ylog(1:S,:);

Yest1=Ylog((S+1):end,:);

VAR2MA2=vgxset('n',5,'nAR',2,'nMA',2,'Series',{'yout','u','kp','ki','kd'});

[EstSpec1,EstStdErrors1]=vgxvarx(VAR2MA2,Yest1,[],Ypre1,'CovarType','Diagonal','IgnoreMA','yes');

[ypredict,FYCov1]=vgxpred(EstSpec1,PredictLength,[],Yest1);

ypredict=[log(Y(end,:));ypredict];

ypredict=exp(cumsum(ypredict));

1.1.6 step (006): error is calculated: calculate the error e (k) of current time according to sampled value yout (k) and the desired value rin (k) of controlled device of controlled device current time, wherein e (k)=yout (k)-rin (k);

1.1.7 step (007): error rate calculates: the error rate ec (k) between the error e (k) of calculating current time and the error e (k-1) of previous moment, wherein ec (k)=e (k)-e (k-1);

1.1.8 step (008): control the parameter no-load voltage ratio and calculate: according to Fuzzy Controller Parameters T _fAperhaps Expert Rules table T _eAadopt the method for Fuzzy Calculation or inquiry Expert Rules table, being input as of Fuzzy Calculation method or inquiry Expert Rules table method: current time error e (k) and error rate ec (k), Fuzzy Calculation method or inquiry Expert Rules table method are output as controls parameter no-load voltage ratio R _p(k), R _i(k), R _d(k);

1.1.9 step (009): control calculation of parameter: according to the following control parameter K that calculates _p(k), K _i(k), K _d(k);

K _p(k)=R _p(e (k), ec (k)) K _p(k-1) formula 10

K _i(k)=R _i(e (k), ec (k)) K _i(k-1) formula 11

K _d(k)=R _d(e (k), ec (k)) K _d(k-1) formula 12

1.1.10 step (010): calculate controlled quentity controlled variable: adopt Digital PID Algorithm, the input of Digital PID Algorithm is to control parameter K _p(k), K _i(k), K _d(k), output is controlled quentity controlled variable u (k);

1.1.11 step (011): controlled device produces true output: when the method for the present invention's proposition is used for real system, controlled quentity controlled variable acts on controlled device, and controlled device produces true output yout_real (k); When the method for the present invention's proposition is used for computer simulation experiment; calculated with mathematical model by controlled device draws yout_real (k); the input of the mathematical model of above-mentioned controlled device is controlled quentity controlled variable u (k), and output is that controlled device produces true output yout_real (k);

1.1.12 step (012): computation period is from increasing: i.e. current time k=k+1;

1.1.13 step (013): judge whether to finish: if maximum performance period Kmax>0 and current time k>Kmax jump to step (014): finish; If maximum performance period Kmax=0, jump to step (002), repeat process;

1.1.14 step (014): finish;

System is by 5 installation compositions: decision-making device (100), controller (101), controlled device (102), controlled quentity controlled variable sensing transducer (103), controlled device output sensing transducer (104);

Decision-making device adopts DELL R620 server computer.In the present embodiment, decision-making device adopts a DELLR620 server computer.But also can be by 3 sub-installation compositions: control parameter no-load voltage ratio on-line tuning device, control parameter on-line tuning device, fallout predictor; These 3 devices can be all DELL R620 server computers: control parameter no-load voltage ratio on-line tuning device by its keyboard input control target (system input digital quantity), by Ethernet interface, with the Ethernet interface of fallout predictor, be connected, receiving control system prediction output, be connected output controlled quentity controlled variable no-load voltage ratio value with control parameter on-line tuning device by Ethernet; Control parameter on-line tuning device and be connected with controller with fallout predictor by Ethernet, export up-to-date control parameter; Fallout predictor is connected with controlled device output sensing transducer with the controlled quentity controlled variable sensing transducer by RS232 interface switching RS485 bus, receives controlled quentity controlled variable (digital quantity) and system and truly exports (digital quantity) for input;

Decision-making device is by keyboard input control target (system input digital quantity); Decision-making device is connected with the controlled quentity controlled variable sensing transducer by the RS232 interface, receives up-to-date controlled quentity controlled variable (digital quantity); Decision-making device is connected by transfer mode and the controlled device output sensing transducer of RS485 data bus of RS232, receives up-to-date system and truly exports (digital quantity);

Controller is by PID controller, major loop protective device, solid-state relay: it is YUDIAN-AI-7048 that the PID controller is selected model, and four standard transducer signal input parts of PID controller are connected respectively with after the 50 Ω resistance in parallel of four standard transducer signal output parts from two Pt100 temperature transmitters.Pt100 temperature transmitter output 4-20mA standard transducer current signal, by 50 Ω resistance in parallel, be converted to signal the PID controller normal voltage input signal of 200mV-1V.Control output and receive respectively the controlled quentity controlled variable input end of controlling actuating unit in performance element for four of the PID controller.

Performance element is by the major loop protective device, and the controlled device abnormal protection device, control actuating unit, and major loop supervising device and controlled device (load) access device five parts form, and this five part is followed in series to form major loop.

The major loop protective device is selected air switch; the controlled device abnormal protection device is selected overheat protector module and relay; control actuating unit and select solid-state relay, the major loop supervising device is selected current transformer, and controlled device (load) access device is selected the forceful electric power connection terminal.

The input end 1 of air switch, input end 2 are connected on respectively live wire and the zero line of major loop, and output terminal 3, output terminal 4 are respectively live wire and the zero line of late-class circuit, and effect is to cut off major loop when electric current is excessive in loop, protects other devices.Overheat protector module and the common formation of relay series connection thermal-shutdown circuit, the input end 1 of overheat protector module, input end 2 difference live wires are connected with zero line and are used for to the overheat protector module for power supply, the output terminal 3 of overheat protector module and the input end 2 of relay are connected with zero line, form reference zero, the output terminal 4 of bothering about protection module is connected with the input end 1 of relay, realization is bothered about protection module the switch of relay is controlled, the input end 9 of overheat protector module, input end 11 connects the temperature sensor that comes from controlled device, when temperature surpasses setting value, its output terminal 3 and output terminal 4 output cut-off signals, the output terminal 3 of relay, output terminal 4 series windings are connected on live wire, according to the output of overheat protector module, major loop is carried out to break-make control.The controlled quentity controlled variable input end "+" of solid-state relay, input end "-" are connected with the controlled quentity controlled variable output of controller, realize that controller passes through the control of the form of PWM break-make to major loop.Major loop imports the former limit of current transformer into, and the secondary of current transformer is connected with light emitting diode, realizes the monitoring to the major loop break-make.The series connection of forceful electric power connection terminal, in major loop, realizes controlled device is accessed to performance element.

Controller is connected with device (100) decision-making device by the RS485 data bus, accepts to control parameter; Controller output controlled quentity controlled variable drives heating furnace to produce the mode of heating with forceful electric power, with controlled device, is connected, and controlled device is produced and drives;

1.2.3 controlled device is heating furnace, the inside of heating furnace is a resistance wire spiraled, and when the input of heating furnace adds the driving voltage of 0～220V, resistance wire becomes heat energy by electric energy conversion, realizes the heating to stove inside.Gather temperature in body of heater by the Pt100 temperature sensor, produce corresponding resistance value;

1.2.4 the controlled quentity controlled variable sensing transducer adopts the KBM-44 alternating voltage to turn the RS485 data acquisition module, the input that gathers heating furnace adds the driving voltage of 0～220V, by the RS485 bus, with device (100) decision-making device, be connected, passback alternating voltage digital signal;

1.2.5 controlled device output sensing transducer is selected the Pt100 temperature transmitter that model is YUDIAN-AI-7021.The sensor incoming end of Pt100 temperature transmitter adopts the mode of three-wire system to connect the Pt100 temperature sensor, and the output of Pt100 temperature transmitter is connected with device (100) decision-making device by the RS485 bus, passback temperature digital signal.The power supply of Pt100 temperature transmitter directly connects the 220V power supply;

2. the marriage relation of system and method and implementation procedure (according to the method step order, describe, wherein, because the present embodiment adopts the single device decision-making device, so controller, control parameter on-line tuning device, fallout predictor are the part of decision-making device):

Step (001) initialization is realized by decision-making device; Specifically by control parameter no-load voltage ratio on-line tuning device reception input, set, initialization result is distributed to and controls parameter on-line tuning device and fallout predictor;

Step (002) system is truly exported to gather by controlled quentity controlled variable sensing transducer and controlled device output sensing transducer and is coordinated decision-making device to realize;

Step (003) judges whether to start to be predicted to step (005) prediction output and is realized by fallout predictor;

Step (006) error is calculated to step (008) and controls the parameter transformation-ratio meter parameter on-line tuning device realizes by controlling at last;

Step (009) control parameter transformation-ratio meter is realized by control parameter on-line tuning device at last;

Step (010) controlled quentity controlled variable is calculated and is realized by controller;

Step (011) controlled device produces true output and is realized by controlled device;

Step (012) computation period finishes from increasing to step (014), by decision-making device, is realized.

Claims (3)

1. a vector time series is predicted expert fuzzy no-load voltage ratio Optimization about control parameter method, it is characterized in that:

The method by step (001) to step (014) totally 14 steps form:

Step (001): initialization:

Initialization max calculation cycle Kmax: wherein Kmax=0 or Kmax ∈ N, N means natural number, if Kmax=0 means that control algolithm stops never;

The current computation period k:k=1 of initialization;

Target Rin (k): Rin (k)=f (k) is controlled in initialization, and wherein f (n) means Controlling object function, according to moment n, provides the control target;

The initialization prediction step is L _predict, L _predictfor positive integer, according to engineering experience, its span is referenced as: 1≤L _predict≤ 50;

Initialization prediction Startup time T _predict, according to engineering experience, its span is referenced as: 5%Kmax≤T _predict≤ 10%Kmax, if Kmax=0,10L _predict≤ T _predict≤ 100; Initialization system output yout (k)=0;

Determine expert fuzzy no-load voltage ratio inference rule and initialization expert fuzzy no-load voltage ratio inference rule parameter;

Step (002): system is truly exported collection: when the method for the present invention's proposition is used for real system; yout_real (k) is that the controlled device of necessary being produces; by harvester, gather controlled device and produce true output yout_real (k); The method proposed as the present invention is during for computer simulation experiment, and yout_real (k) is that the calculated with mathematical model by controlled device draws: during current calculatings moment k=1, and indirect assignment yout_real (k)=1; Current calculating is k>1 o'clock constantly, and yout_real (k) was calculated k-1 performance period;

Step (003): judge whether to start prediction: if current time not yet arrives prediction Startup time, i.e. k<T _prsdict, do not start prediction, jump to step (004); If current time arrives prediction Startup time, i.e. k>=T _predict, start prediction, jump to step (005);

Step (004): indirect assignment: not yet start prediction, participate in the yout (k) that error calculates=yout_real (k-1);

Step (005): prediction output: started prediction, participated in the yout (k) that error calculates=y _predict(k); Computing method adopt the vector time series Forecasting Methodology: the time series that is input as following five variablees of vector time series prediction algorithm: proportional control parameter time series K _p(1), K _p(2) ..., K _p(k), integration control parameter time series K _i(1), K _i(2) ..., K _i(k), differential is controlled parameter time series K _d(1), K _d, (2) ..., K _d(k), controlled quentity controlled variable time series u (1), u (2) ..., u (k), control system output yout_real (1), yout_real (2) ..., yout_real (k); The vector time series prediction algorithm is output as the constantly rear L of k _predictsystem output y constantly _predict(k);

Step (006): error is calculated: calculate the error e (k) of current time according to sampled value yout (k) and the desired value rin (k) of controlled device of controlled device current time, wherein e (k)=yout (k)-rin (k);

Step (007): error rate calculates: the error rate ec (k) between the error e (k) of calculating current time and the error e (k-1) of previous moment, wherein ec (k)=e (k)-e (k-1);

Step (008): control the parameter no-load voltage ratio and calculate: according to Fuzzy Controller Parameters T _fAperhaps Expert Rules table T _eAadopt the method for Fuzzy Calculation or inquiry Expert Rules table, being input as of Fuzzy Calculation method or inquiry Expert Rules table method: current time error e (k) and error rate ec (k), Fuzzy Calculation method or inquiry Expert Rules table method are output as controls parameter no-load voltage ratio R _p(k), R _i(k), R _d(k);

Step (009): control calculation of parameter: according to formula 1, formula 2, formula 3, calculate the control parameter K _p(k), K _i(k), K _d(k);

K _p(k)=R _p(e (k), ec (k)) K _p(k-1) formula 1

K _i(k)=R _i(e (k), ec (k)) K _i(k-1) formula 2

K _d(k)=R _d(e (k), ec (k)) K _d(k-1) formula 3

Step (010): calculate controlled quentity controlled variable: adopt Digital PID Algorithm, the input of Digital PID Algorithm is to control parameter K _p(k), K _i(k), K _d(k), output is controlled quentity controlled variable u (k);

Step (011): controlled device produces true output: when the method for the present invention's proposition is used for real system, controlled quentity controlled variable acts on controlled device, and controlled device produces true output yout_real (k); When the method for the present invention's proposition is used for computer simulation experiment; calculated with mathematical model by controlled device draws yout_real (k); the input of the mathematical model of above-mentioned controlled device is controlled quentity controlled variable u (k), and output is that controlled device produces true output yout_real (k);

Step (012): computation period is from increasing: i.e. current time k=k+1;

Step (013): judge whether to finish: if maximum performance period Kmax>0 and current time k>Kmax jump to step (014): finish; If maximum performance period Kmax=0, jump to step (002), repeat process;

Step (014): finish.

2. a kind of vector time series according to claim 1 is predicted expert fuzzy no-load voltage ratio Optimization about control parameter method, it is characterized in that: wherein determine and are specially expert fuzzy no-load voltage ratio inference rule and initialization expert fuzzy no-load voltage ratio inference rule parameter:

If expert fuzzy no-load voltage ratio inference rule adopts the method for expert's no-load voltage ratio reasoning, adopt method for designing design specialist's no-load voltage ratio rule list of Expert Rules table in expert PID, in expert PID, the input of Expert Rules table and expert's no-load voltage ratio rule list is all moment error e (k) and error rate ec (k), and in expert PID, the output of Expert Rules table is to control parameter K _p(k), K _i(k), K _d, and the output of expert's no-load voltage ratio rule list is to control parameter no-load voltage ratio R (k) _p(k), R _i(k), R _d(k);

If expert fuzzy no-load voltage ratio inference rule adopts the method for fuzzy calculation inference, adopt the method for designing of indistinct logic computer in fuzzy to design fuzzy no-load voltage ratio inference machine, in fuzzy, the input of indistinct logic computer and fuzzy no-load voltage ratio inference machine is all moment error e (k) and error rate ec (k), and in fuzzy, the output of indistinct logic computer is to control parameter K _p(k), K _i(k), K _d, and the output of fuzzy no-load voltage ratio inference machine is to control parameter no-load voltage ratio R (k) _p(k), R _i(k), R _d(k).

3. application rights requires the system of 1 described a kind of vector time series prediction expert fuzzy no-load voltage ratio Optimization about control parameter method, it is characterized in that: system is by 5 installation compositions: decision-making device (100), controller (101), controlled device (102), controlled quentity controlled variable sensing transducer (103), controlled device output sensing transducer (104);

Decision-making device (100) is that computing machine or Programmable Logic Controller or be input as controlled target, controlled quentity controlled variable, system and truly exported, and is output as the device of controlling parameter; Wherein decision-making device is by input equipment input control target; Decision-making device passes through RS232 interface or Ethernet interface or capture card interface and is connected with the controlled quentity controlled variable sensing transducer, receives up-to-date controlled quentity controlled variable; Decision-making device passes through RS232 interface or Ethernet interface or capture card interface and is connected with controlled device output sensing transducer, receives up-to-date system and truly exports;

Wherein, decision-making device is a device, or by 3 sub-installation compositions: control parameter no-load voltage ratio on-line tuning device, control parameter on-line tuning device, fallout predictor; These 3 devices are all computing machine or Programmable Logic Controller or the device that meets subordinate's annexation and input/output relation: control parameter no-load voltage ratio on-line tuning device by its input equipment input control target, by Ethernet or communication bus, with the device fallout predictor, be connected, receiving control system prediction output, be connected output controlled quentity controlled variable no-load voltage ratio value with device control parameters on-line tuning device by Ethernet or communication bus; Control parameter on-line tuning device and be connected with controller with fallout predictor by Ethernet or communication bus, export up-to-date control parameter; Fallout predictor is connected with controlled quentity controlled variable sensing transducer and controlled device output sensing transducer by RS232 interface or Ethernet interface, capture card transmission interface, and reception controlled quentity controlled variable and system truly are output as input;

Controller is with the PLC of programing function and driver or produces with the programing function frequency converter or according to pid control parameter the device that controlled quentity controlled variable drives controlled device; Controller is connected with control parameter on-line tuning device by industrial bus, data line, accepts to control parameter; Controller output controlled quentity controlled variable, is connected with controlled device by physical construction or conduction, transmission medium, to controlled device generation driving;

Controlled device is the object of the controlled quentity controlled variable impact of electric motor and driving object or pressure control equipment and object or the generation of controlled device;

The controlled quentity controlled variable sensing transducer is voltage collecting device, pressure acquisition device, rotating speed harvester or analog acquisition device; The input of controlled quentity controlled variable sensing transducer and the output of controller are connected by wire or gearing or conduction device, obtain the analog quantity of controller output; The controlled quentity controlled variable sensing transducer is translated into controlled quentity controlled variable, then is connected the pipage control amount with decision-making device by RS232 interface or Ethernet interface, capture card transmission interface;

Controlled device output sensing transducer is voltage collecting device, pressure acquisition device, rotating speed harvester or analog acquisition device; The input of controlled device output sensing transducer and controlled device are connected by wire or gearing or conduction device, the analog quantity that controlled device is exported; Controlled device output sensing transducer is translated into system and truly exports, then is connected with decision-making device by RS232 interface or Ethernet interface, capture card transmission interface, and induction system is truly exported;

Marriage relation and the implementation procedure of system and method are as follows:

Step (001) initialization is realized by decision-making device; Specifically by control parameter no-load voltage ratio on-line tuning device reception input, set, initialization result is distributed to and controls parameter on-line tuning device and fallout predictor;

Step (002) system is truly exported to gather by controlled quentity controlled variable sensing transducer and controlled device output sensing transducer and is coordinated decision-making device to realize;

Step (003) judges whether to start to be predicted to step (005) prediction output and is realized by fallout predictor;

Step (006) error is calculated to step (008) and controls the parameter transformation-ratio meter parameter on-line tuning device realizes by controlling at last;

Step (009) control parameter transformation-ratio meter is realized by control parameter on-line tuning device at last;

Step (010) controlled quentity controlled variable is calculated and is realized by controller;

Step (011) controlled device produces true output and is realized by controlled device;

Step (012) computation period finishes from increasing to step (014), by decision-making device, is realized.

Images