CN106802554A - Two-Degree-of-Freedom Internal Model PID controller parameter setting method - Google Patents

Abstract

The present invention relates to a kind of Two-Degree-of-Freedom Internal Model PID controller parameter setting method, by the parameter of varying cyclically controlled device first under off-line state, and every group of optimal λ 1 of object parameters correspondence is obtained using genetic algorithm, as the training sample of neutral net；Object parameters are recognized according to least square method under presence again, the parameter of controlled device is obtained according to identification, neutral net is respectively adopted λ 2 and λ 1 are adjusted.In the present invention, the acquisition of λ 1 is obtained by genetic algorithm optimizing in Sample Storehouse, λ 1 now can be optimal the ITAE indexs of system, it is to avoid the defect of artificial experience Selecting All Parameters；Consider the change that occurs in actual motion of object parameters, on-line tuning is carried out to λ 2 and λ 1 according to the parameter that identification obtains controlled device, overcome causes control effect to be paid no attention to think over a problem using the controller parameter of fixation.

Description

Two-Degree-of-Freedom Internal Model PID controller parameter setting method

Technical field

The present invention relates to attitude conirol technical field, more particularly, to the ginseng of Two-Degree-of-Freedom Internal Model PID controller Number setting method.

Background technology

PID controller is due to simple structure, reliability is high and is widely used in various control field, but for ship, The controlled device of the big inertia time lag such as submarine, tends not to obtain satisfied control effect only with PID control.By internal model control It is combined to form Internal Model PID Controller with PID control, can significantly improves the bad shadow that Time Delay in controlled device brings Ring, and the parameter tuning of controller is easier.Conventional PID controller and Internal Model PID Controller are all single-degree-of-freedom controller, Only one group regulation parameter, the parameter of controller needs to carry out compromise choosing according to the anti-interference and target value tracking characteristic of system Select, it is impossible to system is obtained optimum performance.A filtering link is connected on the basis of single-degree-of-freedom Internal Model PID Controller can To constitute Two-Degree-of-Freedom Internal Model PID controller, make controller contain two groups can separately adjustable system anti-interference and desired value The parameter of tracking characteristics, substantially increases the control effect of system, but increased the parameter number that filtering link also causes controller Amount is more, and parameter tuning process is more complicated.At present, still without very complete, effective Two-Degree-of-Freedom Internal Model PID controller Parameter tuning method, especially to the main warp for being also to rely on designer of selection of filtering parameter in the filtering link of series connection Test, and controller also makes it apply in object parameters with running environment, working conditions change typically using fixed parameter It is difficult to obtain superior control effect in the system of change.

The content of the invention

It is difficult to make by designer's experience selection controller parameter and using fixed controller parameter for above-mentioned System obtains the defect of superior function, and applicant is studied and improved, there is provided a kind of Two-Degree-of-Freedom Internal Model PID controller parameter Setting method.

In order to solve the above problems, the present invention uses following scheme：

It is inertia Time Delay for controlled device, its transmission function can be expressed as：

Wherein, K_mIt is gain coefficient；L_mIt is time lag constant；T_miIt is inertia time constant；N is systematic education.For formula (1) In controlled device, use Two-Degree-of-Freedom Internal Model PID controller for：

Wherein, λ₁And λ₂It is controller parameter to be adjusted.A kind of Two-Degree-of-Freedom Internal Model PID controller parameter setting method, Its feature is comprised the following steps：

Step I：In controlled device each parameter variation range, several representative values are respectively listed, combination forms multigroup quilt Control image parameter；

Step II：Under off-line state, Do statement is write in Matlab, cycle-index is equal to controlled device in step I The quantity of parameter.The operation of two steps is completed in single cycle：1. object parameters are changed, while according to given peak response M_SAgain adjust λ₂：

2. genetic algorithm is used again, with the ITAE indexs under system step response as object function, to λ₁Carry out optimizing.Institute After thering is time circulation terminate, obtain controlled device and take different groups of parameters corresponding to optimal λ respectively₁, and it is controlled right by all groups As parameter and the optimal λ of every group of parameter correspondence₁It is saved into Sample Storehouse；

Step III：Three layers of BP neural network are built, the input of neutral net is the parameter of controlled device, is output as λ₁.From Under wire state, the sample obtained in step II is used to be trained the weights of neutral net in MATLAB；

Step IV：Under presence, real-time identification is carried out to object parameters using least square method, obtain controlled right The parameter of elephant, adjusts C in conjunction with formula (3)₂λ in (s)₂；

Step V：Under presence, the object parameters that obtain as the input of neutral net will be recognized, due to nerve Network has certain generalization ability, now λ₁The output valve of neutral net is taken, can make system that there is superior ITAE indexs.

By above scheme as can be seen that Two-Degree-of-Freedom Internal Model PID controller parameter setting method of the invention, first from By the parameter of varying cyclically controlled device under wire state, and it is optimal to obtain every group of object parameters correspondence using genetic algorithm λ₁, as the training sample of neutral net；Object parameters are distinguished according to least square method under presence again Know, the parameter of controlled device is obtained according to identification, the neutral net after formula (3) being respectively adopted and training carries out whole to λ 2 and λ 1 It is fixed.

The technical effects of the invention are that：

1st, the acquisition of λ 1 is obtained by genetic algorithm optimizing in Sample Storehouse, and λ 1 now can reach the ITAE indexs of system To optimal, it is to avoid the defect of artificial experience Selecting All Parameters；

2nd, consider the change that object parameters occur in actual motion, the parameter pair of controlled device is obtained according to identification λ 2 and λ 1 carry out on-line tuning, and overcome causes control effect to be paid no attention to think over a problem using fixed controller parameter；

3rd, the parameter lambda 1 of controller is adjusted using the neutral net after training, and its process shows as simple four fundamental rules fortune Calculate, easily, real-time and practicality are all stronger for programming realization.

Brief description of the drawings

Fig. 1 is Control system architecture block diagram of the present invention using Two-Degree-of-Freedom Internal Model PID controller

Fig. 2 is theory diagram of the present invention using genetic algorithm optimization λ 1

Fig. 3 is the flow chart that Sample Storehouse of the present invention is obtained

Fig. 4 is three layers of BP neural network structure chart of the invention

Fig. 5 is Two-Degree-of-Freedom Internal Model PID controller parameter on-line tuning principle frame of the present invention.

Specific embodiment

Specific embodiment of the invention is described further below in conjunction with the accompanying drawings.

The present embodiment application the inventive method is to controlled device in system in the two degrees of freedom of second-order inertia Time Delay Mould PID controller carries out parameter tuning.If the transmission function of controlled device is in system：

Wherein, K_mIt is gain coefficient；L_mIt is time lag constant；T_m1, T_m2It is inertia time constant.

Fig. 1 is using the Control system architecture block diagram of Two-Degree-of-Freedom Internal Model PID controller.C in Fig. 1₁(s) and C₂S () is Two-Degree-of-Freedom Internal Model PID controller, its transmission function is：

Wherein, λ₁And λ₂It is controller parameter to be adjusted.

Parameter tuning is carried out to Two-Degree-of-Freedom Internal Model PID controller in formula (2) using the inventive method, its feature include with Lower step：

Step I：Assuming that the parameter K of controlled device_m、L_m、T_m1、T_m2Excursion be respectively [K_mmin,K_mmax]、 [L_mmin,L_mmax]、[T_m1min,T_m1max]、[T_m2min,T_m2Max], separated respectively etc. in every group of parameter variation range o, p, Q, h value, these values are combined to form the unduplicated object parameters of M groups, M=o*p*q*h；

Step II：Under off-line state, Do statement is write in matlab, cycle-index is M.Make two steps in single cycle Operation：1. in the M group object parameters for being produced in step I, one group of new parameter is taken in order as object parameters, Peak response M_S1.6 are taken, is adjusted again λ according to given peak response₂：

2. using genetic algorithm to λ₁Carry out optimizing, λ₁Scope take [λ₂, 10 λ₂], genetic algorithm is responded with system step Under ITAE indexs be object function, taking fitness function is

Genetic algorithm is sentenced by parameter initialization, initialization of population, genetic operator operation, new population generation, end condition It is disconnected to wait execution to produce λ optimal corresponding to this group of object parameters after terminating₁, genetic algorithm optimization λ₁Theory diagram see figure 2.After M circulation terminates, M corresponding with M group object parameters optimal λ is obtained₁, by M groups object parameters and M Optimal λ₁Correspondence combination forms M sample, saves as Sample Storehouse.The flow chart that Sample Storehouse is obtained is shown in Fig. 3；

Step III：Build three layers of BP neural network, the input number of nodes of neutral net takes 4, node in hidden layer and takes 4, defeated Go out node layer number and take 1, neural network structure figure is shown in Fig. 4.Under off-line state, M obtained in step II is used in MATLAB Sample is trained to the weights of neutral net；

Step IV：Under presence, the parameter for using least squares identification to obtain controlled device is K_mo、L_mo、T_m1o、 T_m2o, adjusted C in conjunction with formula (3)₂λ in (s)₂, obtain λ₂=0.9665L_mo；

Step V：The K for obtaining will be recognized_mo、L_mo、T_m1o、T_m2oAs the input of neutral net after training, due to neutral net With certain generalization ability, now the output valve of neutral net is λ₁Setting valve.

Above content is to combine specific embodiment further description made for the present invention, it is impossible to assert the present invention Specific implementation be confined to these explanations.For general technical staff of the technical field of the invention, do not departing from On the premise of present inventive concept, some simple deduction or replace can also be made, should all be considered as belonging to protection model of the invention Enclose.

Claims (1)

1. a kind of Two-Degree-of-Freedom Internal Model PID controller parameter setting method, it is characterised in that：

It is inertia Time Delay for controlled device, its transmission function is：

$P\left(s\right)=\frac{K_m{e}^{-L_{m}s}}{\underset{i=1}{\overset{n}{\Π}}(T_{mi}s+1)},T_{mi}>0---\left(1\right)$

Formula (1) wherein, K_mIt is gain coefficient；L_mIt is time lag constant；T_miIt is inertia time constant；N is systematic education；For formula (1) In controlled device, use Two-Degree-of-Freedom Internal Model PID controller for：

$\left\{\begin{array}{c}{C}_1\left(s\right)=\frac{{\mathrm{\λ}}_{2}s+1}{{\mathrm{\λ}}_{1}s+1}\\ C_2\left(s\right)=\frac{\underset{i=1}{\overset{n}{\Π}}(T_{mi}s+1)}{K_m({\mathrm{\λ}}_2+L_m)s}\end{array}\right.---\left(2\right)$

In formula (2), λ₁And λ₂It is controller parameter to be adjusted；

The parameter tuning method, comprises the following steps：

Step I：In controlled device each parameter variation range, several representative values are respectively listed, combination forms multigroup controlled right As parameter；

Step II：Under off-line state, Do statement is write in Matlab, cycle-index is equal to object parameters in step I Quantity；The operation of two steps is completed in single cycle：The first step to change object parameters, while according to given maximum sensitive Degree MS adjusts λ 2 again：

${\mathrm{\λ}}_2=\frac{1.508-0.451M_s}{1.451M_s-1.508}{L}_m---\left(3\right)$

Second step uses genetic algorithm again, with the ITAE indexs under system step response as object function, optimizing is carried out to λ 1；Institute After thering is time circulation terminate, obtain controlled device and take different groups of parameters corresponding to optimal λ 1 respectively, and it is controlled right by all groups As parameter and the optimal λ of every group of parameter correspondence₁It is saved into Sample Storehouse；

Step III：Three layers of BP neural network are built, the input of neutral net is the parameter of controlled device, is output as λ₁；Offline Under state, the sample obtained in step II is used to be trained the weights of neutral net in MATLAB；

Step IV：Under presence, real-time identification is carried out to object parameters using least square method, obtain controlled device Parameter, adjusts C in conjunction with formula (3)₂λ in (s)₂；

Step V：Under presence, the object parameters that obtain as the input of neutral net will be recognized, now λ₁Take nerve net The output valve of network.