JPH04291602A - Fuzzy control system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention]

[Object of the invention]

0002

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuzzy inference method used in a general-purpose fuzzy controller.

0003

2. Description of the Related Art Recently, many attempts have been made to incorporate fuzzy control into control objects whose systems are difficult to model mathematically. Currently, the use of fuzzy rules for control can be roughly divided into two types:

(1) A fuzzy model of the operator is created by making rules based on the operator's experience and analyzing actual operation history data, and realizing control similar to that of the operator.

(2) Using a target model (such as a fuzzy model), predict the future state when a certain temporary operation amount is given, and use fuzzy inference to evaluate the extent to which the predicted value satisfies the target. determine the optimal amount of operation.

[0006] Method (1) is the most commonly used method in fuzzy control and can be easily formulated into rules. However, since the rules include the operator's global control strategy, each rule has complex interactions. For this reason, simply arranging the operator's control knowledge in parallel often does not function effectively as it is;
Often, changes to some rules require the entire rule to be redesigned. Furthermore, if the purpose of control changes, the rules must be completely rewritten.

In this respect, in the control method (2), the fuzzy model rules that indicate the ambiguity of the target system and the goal evaluation rules that indicate the ambiguity included in the part that determines the control strategy are clearly distinguished. This makes it easy to formulate adjustment and design policies. Furthermore, any control target can be easily changed by simply changing the target fuzzy set given to the target evaluation rule. Hereinafter, this will be referred to as a state predictive fuzzy control method.

FIG. 2 shows a block diagram of a conventional state prediction type fuzzy control system. However, a fuzzy model is used here. In FIG. 2, 1 is a part that gives the temporary manipulated variable to the model, 2 is a fuzzy model for predicting the state, 3 is a fuzzy inference part that evaluates the predicted value, and 4 is a composite of the temporary manipulated variable according to the evaluation result. , or the selection section 5 is a section that determines the target value B*, which is the evaluation standard, according to the current state.

[0009] Generally, a fuzzy model has the following rule format.

0010

[Math 1]

[0011] Here, f() represents the included function. Also, x
, u indicate the state vector and input vector, respectively, and t
is the current time, and dt is the sampling period.

[0012] Also,

[0013]

[Math 2]

##EQU1## where A and B are state spaces, and U is a label of a reference fuzzy set defined on the input space. NA, NB, and NU each represent the number of defined reference fuzzy sets.

[0015] The rules for state predictive fuzzy control are as follows:

0
016]

[Math 3]

It is given by: Here, B* j is a reference fuzzy number representing a target on the state space, and u ′|t is a temporary manipulated variable. Uj can be anything as long as it is given as a temporary manipulated variable, and u|t in the conclusion part is the same Uj
All you have to do is become.

However, when trying to control a dynamic system, the state predictive fuzzy control method has the following two problems.

The first problem is that the model calculates predicted values when given each temporary operation amount from the past and present state of the system, and then evaluates each predicted value using the target evaluation rule. Since the controller has a two-stage structure, the inference speed is slow.

The second problem is that since fuzzy inference is divided into two stages: model and target evaluation, it becomes difficult to design and difficult to handle when used simultaneously with normal feedback type fuzzy control rules.

[0021]

[Problems to be Solved by the Invention] As mentioned above, state predictive fuzzy control methods have slow inference speeds and are difficult to design and difficult to handle when used simultaneously with normal feedback fuzzy control rules. There was a problem.

The present invention was devised to solve these problems, and provides a fuzzy control method that can significantly reduce inference time and facilitate simultaneous use with feedback-type fuzzy control rules. The purpose is to provide

[Configuration of the invention]

[0024]

[Means for Solving the Problems] In order to achieve the above-mentioned object, the fuzzy control method of the present invention uses a past manipulated variable input to a controlled system and observed values of past and current system state quantities to perform fuzzy control. This is a fuzzy control method that determines the output operation amount by inference, and the fuzzy rule antecedent part is a fuzzy proposition that evaluates the degree of compatibility between the predicted value of the system state quantity and the control target value by the operation defined in the rule conclusion part. It is characterized by fuzzy inference.

[0025]

[Operation] In the fuzzy control method of the present invention, when a fuzzy model is used in state predictive fuzzy inference, the number of rules used in the fuzzy model to predict one provisional manipulated variable is only a small part of all the rules. Utilizing this property, by grouping the fuzzy model rules and evaluation rules for each temporary operation amount, and further integrating the model rules and evaluation rules, it is possible to perform the same type of state predictive fuzzy control using a single rule. effect, significantly shortens inference time, and facilitates simultaneous use with feedback-type fuzzy control rules.

[0026]

Embodiments Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram for explaining a fuzzy control method according to an embodiment of the present invention.

In the rule shown in Equation 3 of the conventional fuzzy control method, the restrictions of Equations 4 and 5 are added to the temporary operation amount u ′ given to the fuzzy model, and the temporary operation amount is equal to the amount of operation used in the antecedent part of the fuzzy model. Assume that we use the same reference fuzzy sets and that they are disjoint. Then, the fuzzy model rules and the target evaluation rules (state prediction type fuzzy control rules) can be easily grouped. Condition 5 is automatically satisfied when the temporary manipulated variable is a non-fuzzy value.

[0029]

[Math 4]

[0030]

[Math 5]

That is, when Ui is added to the fuzzy model, fuzzy model rules that do not have the proposition (u|t = Ui) in the antecedent part have a fitness of 0 and are not used. A rule with (u|t = Ui) can also be considered excluding that proposition since its fitness is 1.0. By utilizing this fact, it is possible to extract and group only the rules related to each provisional operation amount from among the rules of the fuzzy model. As an example, consider the following fuzzy model.

[0032]

[Math 6]

Here, from condition number 3

[0034]

[Math 7]

When set as follows, the following two state prediction type fuzzy control rules exist.

[0036]

[Math. 8]

Considering the fuzzy model rules related to RR1, R3 and R4, which do not have (u'|t = UA) as a condition, can be ignored, and for R1 and R2, (u'|t = UA) can be ignored. It can be deleted and associated as shown below.

<Group of rules M |u′=UA when UA is used as the temporary operation amount u′|t>

00
39]

[Math. 9]

Even if u'=UB is considered, a similar rule group M|u'=UB can be obtained. These rule groups do not interfere with each other at all, and the values to be given as temporary manipulated variables are given in advance, so if you give the current state quantity, past state quantity, and manipulated quantity to the model room part, it will work as it is. The ratio at which (u|t = UA) should be output is determined. In other words, one rule group can be handled in the same way as one conventional state feedback type fuzzy control rule.

Furthermore, by utilizing the above-mentioned properties, each rule group can be integrated into one fuzzy control rule. In other words, by providing a fuzzy proposition in the rule antecedent part that evaluates the degree of compatibility between the predicted value of the state quantity based on the operation defined in the rule conclusion part and the current control purpose, the prediction by the fuzzy model is not directly performed. 1
It is possible to achieve the same effect as state-predictive fuzzy control using staged fuzzy inference alone.

In the above example, the rule group M|u′=UA
can be integrated as shown in the number 10.

[0043]

[Math. 10]

In the above fuzzy inference, if a method other than Max-Min operation is used as the fuzzy operation, or if the conclusion part is a fuzzy model rule with a mathematical formula, it is not necessarily possible to obtain exactly the same inference result, but almost the same effect can be obtained. is obtained.

[0045]

[Effects of the Invention] According to the present invention, fuzzy model rules and evaluation rules are grouped for each temporary operation amount,
Furthermore, by integrating the prediction rule and the target evaluation rule into one fuzzy control rule, it is possible to achieve the same effect as state prediction type control with a single rule, significantly shorten the inference time, and achieve the same effect as the feedback type fuzzy control rule. It also becomes easier to use them simultaneously.

[Brief explanation of the drawing]

FIG. 1 is a block diagram for explaining a fuzzy control method according to an embodiment of the present invention.

FIG. 2 is a block diagram for explaining a conventional state prediction type fuzzy control method.

Claims (1)

[Claims] [Claim 1] Past operation amount input to the controlled system;
This is a fuzzy control method that determines the output manipulated variable by fuzzy inference using observed values of past and current system state quantities, and the control target value and the predicted value of the system state quantity based on the operation defined in the rule conclusion section. A fuzzy control method characterized in that fuzzy inference is performed by providing a fuzzy proposition for evaluating the fitness of a fuzzy rule in the antecedent part of the fuzzy rule.