JPH0460703A - Fuzzy controller - Google Patents

Abstract

PURPOSE:To improve the processing speed of fuzzy control without being restricted by means of a storage part by applying the realizing degree of a membership function in an antecedent part to the control rule of fuzzy inference, weighting the membership function of a consequent part and calculating an output value from the membership function of the consequent part. CONSTITUTION:A difference (e) between the room temperature T of a temperature sensor 3 and a room temperature S which is set in a room temperature setting unit 4 and the value of a change quantity DELTAe between a difference e' which is obtained last time and the difference (e) which is obtained this time are inputted. Then, a table 7 is referred to and the values NB(e), NS(e), ZO(e), PS(e) and PB(e) are obtained. In the same way, a table 7 is referred to and the values of NB(DELTAe), NS(DELTAe), ZO(DELTAe), PS(DELTAe) and PB(DELTAe) are obtained. The membership function of the consequent part is weighted. The position of a center of gravity in the membership functions ZO, PS and PM is calculated and the value of DELTAfg corresponding to the center of gravity G is outputted as the increase/decrease quantity of the capacity of an air-conditioning unit.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a fuzzy control device used to control the operation of equipment so that a detected value becomes a target value.

(b) Conventional technology In general, as a conventional fuzzy control device, Japanese Patent Application Laid-Open No. 1-14
There was something that was completely described in Publication No. 2902. What is described in this publication is, ``In a fuzzy control system for an automobile that is equipped with a fuzzy control section that performs fuzzy control of controlled objects such as the propulsion system of the automobile by a processor installed in the automobile, the fuzzy control section has no memory. and a readout control unit of a storage control device, the storage device is provided with a storage device that calculates the detected amount and the temporal change in the controlled target portion in advance according to a predetermined rule, stores the results as a table, and stores the results in the storage device. The read control section of the device reads out the table at a predetermined time and controls the controlled section.

By configuring the fuzzy controller in this way,
When a processor is given a detected value, it can perform fuzzy control simply by reading out a value corresponding to the detected value from a table. That is, the processor does not need to perform fuzzy inference, which takes processing time, on each detected value, and has the advantage of being able to shorten the processing time that the processor spends on fuzzy control.

(C) Problems to be Solved by the Invention In the conventional technology configured as described above, it is possible to shorten the processing time of the processor by using a table, but in such a conventional technology, the processing time of the processor can be shortened by using a table. When fine control with low resolution is required or when it is necessary to widen the output range of fuzzy control,
It was necessary to make the table larger. For example, in order to double the output resolution or double the output range, the table has to be set four times as thick, and multiple such large tables are required. This means that it is not suitable for controlling general equipment due to the limited capacity of the storage unit, and has the problem of not being able to control a plurality of control objects over a wide range with high precision.

In order to solve these problems, the present invention provides a fuzzy control device that increases the processing speed of fuzzy control without being limited by the storage unit.

(d) Means for Solving the Problems The fuzzy control device of the present invention includes a storage unit that stores a table consisting of detected values and degrees of fulfillment of membership functions of antecedents calculated in advance for the detected values. , a weighting section that weights the membership function of the consequent by applying the degree of establishment of the membership function of the antecedent read from this table to the control rule of fuzzy inference, and a member of the weighted consequent. and an arithmetic unit that calculates an output value from the ship function.

(*) Function The fuzzy control device ct of the present invention configured as described above,
By using a table as part of the membership function, the processing speed of fuzzy control can be improved, and other devices can be controlled by changing the table.

(f) Examples Examples of the present invention will now be described based on the drawings. FIG. 1 is a schematic diagram when a fuzzy control device is applied to temperature control of a conditioned room. In this figure, 1 is a temperature-controlled room, and 2 is an air conditioning unit installed in this room, which cools or heats the room by discharging cooling air or heating air from its discharge port. It is something to do. This air conditioning unit can change its cooling capacity or heating capacity by using an inverter device, which will be described later. 3 is a temperature sensor attached to be able to detect the room temperature of the conditioned room 1, and uses a detection element such as a thermistor whose internal resistance value changes depending on the temperature. 4
is a room temperature setting device, which outputs a signal corresponding to the set temperature. There are various types of room temperature setting devices, such as those that use a variable resistor to change the resistance value, those that use a voltage generator to change the output voltage, and those that output a hood or signal corresponding to the set value. Can be used. Reference numeral 5 denotes a detected value output section, which calculates the corresponding room temperature T from the change in the internal resistance value of the temperature sensor 3, and calculates the difference between this room temperature T and the room temperature S fully set in the room temperature setting device 4, "e=T-". 8" (in addition, when performing heating operation, "e--(T-
5)", and the difference e° found last time and the difference e found this time.
The difference "Δe="e"-e" is calculated and the difference C and ΔC are outputted to the microcomputer-6.

This microcomputer 6 calculates the output using tables 7 to 9 and outputs it to the inverter device 10. This inverter device 10 variably controls the operating capacity of the air conditioning unit 2 based on a signal given from a microcomputer 6. For example, the air conditioning unit 2 uses a compressor whose driving source is an induction motor. In this case, the operating capacity of the air conditioning unit 2 can be changed by changing the frequency of the AC power supplied to the induction motor. (
The following example is used. ) Alternatively, if a compressor using a DC motor as the drive source is used, the operating capacity of the air conditioning unit 2 can be changed by changing the voltage of the DC power supplied to the DC motor.

The operation of the microcomputer 6 will be explained below. The microcomputer 6 performs fuzzy inference using the e and ΔC. First, five membership functions regarding the difference e in the antecedent part are set as shown in FIG. Membership function ZO represents a set where we feel that there is no difference e, membership function PS represents a set where we feel that difference e is slightly on the positive side, and membership function PB represents a set where we feel that difference e is large on the positive side. , the membership function NS represents a set for which the difference C is slightly on the negative side, and the membership function NB
represents a set for which the difference e is felt to be large on the negative side.

In Fig. 2, membership functions NB(e) to FB(
e) is “0” for “emi n≦e≦emax”, respectively.
~Emax". This e
For those exceeding the range of , the value when e"emax" or "e=emin" is set. As can be seen from this figure, when “e=e1”, the membership function ZO(
The value of e) is El, and the value of the membership function PS(e) is E2. Further, when "e=e2", the value of the membership function PS(e) is E4 and the value of the membership function PB(e) is E3. In this way, the values of membership functions NB(e) to PB(e) are determined to be constant values for the value of e, so calculate this relationship in advance and create a table as shown in Figure 3, which is shown in Figure 1. Store in table 7.

Note that emaX and emjn are arbitrarily set according to the object to which fuzzy inference is applied, the five senses of the user, and the like. For example, emax and emin are the differences between the room temperature and the set temperature, so emax is a positive value and emin is a negative value.For example, emax = 5, emin
Let n = -5. The resolution from emin to emax is set in consideration of the processing speed and capability of the microcomputer 6. In this embodiment, for ease of explanation, the resolution, ie, the number of steps up to emin Nemax, is set to 100. (1 step = temperature difference of 0.1) Emax takes a value from 0 to 255 (FF in hexadecimal). It goes without saying that the membership function is not limited to the triangle shown in Figure 2, but can be set to any shape such as a trapezoid or bell shape based on the five senses of the user. E m a
It goes without saying that the value of x may be changed. Therefore, the table shown in FIG. 3 shows the relationship between e and the value determined by an arbitrary form of membership function.

Similarly to the above, FIG. 4 shows five membership functions related to the difference Δe in the antecedent part. This membership function also has a membership function NB-PB representing a set similar to the membership function for the difference e.

In Figure 4, membership functions NB (80) ~ PB
(Δe) is “Δe min≦Δe≦△e m aX
” is set to take a value of “0 to ΔEmax”.For values exceeding this 80 range, “80=Δemax” or “△C=△e m i
n'' is set. The membership function for this Δe is set in the same way as the membership function for C shown in FIG. 2. Since Δe is the change in difference, Δe m The width from im to Δemax is small, Δemax takes a positive value, and 8emax takes a negative value.As an example, Δemax=1, 8e
Let min = −1. This Δe min
The resolution from .DELTA.emax to .DELTA.emax is set in consideration of the processing speed and capability of the microcomputer 6, as described above. In this example, for ease of explanation, the resolution, that is, Δ
The description will be made assuming that the number of steps from emin to ΔCmax is 200.

(1 step = 0.01 change) △E m aX
takes a value from 0 to 255 (FF in hexadecimal).

Using the membership functions set in this way, the values of the membership functions NB-PB corresponding to 80 are calculated in advance in the same manner as in FIG. Stored along with the table shown in

Therefore, table 7 stores the values of the membership function of the antecedent part based on e and ΔC.

Figure 5 is a tuning relationship diagram (control rule used in fuzzy inference) that weights the membership function of the consequent. This weights the membership function of the division. Seven membership functions are set for the consequent part, from functions NB to FB. The membership function zO represents a set that does not change the operating capacity of the air conditioning unit 2 shown in FIG. 1, the membership function PS represents a set that slightly increases the operating capacity of the air conditioning unit 2, and the membership function PM The membership function PB represents a set that increases the operating capacity of the air conditioning unit 2 to a medium degree, the membership function NS represents a set that increases the operating capacity of the air conditioning unit 2 to a large extent, and the membership function NS represents a set that slightly decreases the operating capacity of the air conditioning unit 2. , the membership function NM represents a set that reduces the operating capacity of the air conditioning unit 2 to a medium level,
The membership function NB represents a set that can significantly reduce the operating capacity of the air conditioning unit 2. Incidentally, the horizontal axis of these membership functions sets the increase/decrease in the operating capacity of the air conditioning unit 2. The following 17 control rules are set as fuzzy inference control rules.

(1) If the membership interval number NB regarding e holds true and the membership function NB regarding ΔC holds true, then the membership function NB of the consequent part holds true.

(2) If the membership number NS regarding e holds true and the membership function NS regarding Δe holds true, then the membership function NM of the consequent part holds true.

(17) If the membership function PB regarding e holds true and the membership function FB regarding ΔC holds true, then the membership function PB of the consequent part holds true.

FIG. 6 is a characteristic diagram showing the arrangement of membership functions in the consequent part, and these membership functions are weighted based on the control rule shown in FIG. The output of this consequent represents the increase/decrease in the operating capacity (frequency) of the air conditioning unit. Thereafter, the position of the center of gravity of the weighted membership function is calculated using the relationship diagram shown in FIG. 5, and the value of Δf corresponding to this center of gravity is determined and output to the inverter device shown in FIG. In the membership function shown in FIG. 6, Δf'max takes a positive value, and Δf'min takes a negative value. Δf'min~
The width of Δf'max is 150 steps, and one step corresponds to an increase/decrease signal of ±1 hertz.

Next, a flowchart showing operations using these tables is shown in FIG. 7, and will be explained together with actual operations. First, fuzzy inference is started in step SO. This fuzzy inference is performed at predetermined intervals (about 30 seconds). As a result, the capacity of the air conditioning unit shown in FIG. 1 is increased or decreased at each predetermined period. Step S
1, first enter the values of e and Δe. Next step S2
NB(e), NS(e) with reference to Table 7.

Find the values of ZO(e), PS(e), and PB(e). e
When the value of is el (see No. 3150), NB(el
) = O, NS (e 1) - 0, zO (e 1) - E
l.

PS (e 1) = E2, PB (e 1) = 0.

In step S3, NB(Δ
e), NS (Δe), 20 (Δe), PS (Δe).

Find FB(Δe). When the value of Δe is Δe1, NB(Δel)=O, NS(Δel)=O, zO(Δe
1)=ΔEl, PS(ΔC1)=ΔE2, PB(e
l)=O. Next, in step S4, the membership function of the consequent part is weighted, that is, the membership function for which the membership function is valid (the value is not 0) is zO
(el)-El, PS(el)-E2. zO(Δel
) - ΔEl, PS (ΔC1〉=ΔE2. Therefore, according to the set rule in Figure 5, 20(el) x ZO
(Δe1) is 20(el)xPS(Δel
) is PS, PS(el)XZO(Δel) is PS
On the other hand, PS(el) and XPS(Δel) are respectively weighted with respect to PM. Note that when weighting is applied to the same membership functions in the consequent part, the smaller weighting value is taken as the effective value. Therefore, the value of the membership function zO of the consequent part is EIXΔE1, and the value of the membership function PS is E1×ΔE2 (EIXΔE2≦E2XΔ
), and the value of the membership function PM is E
2×ΔE2. This relationship between ZO, PS, and PM can be expressed as
As shown in the figure. Next, the process proceeds to step S5, where the membership function 20 shown in FIG. The center of gravity position G of PS and PM is calculated, and the value of Δrg corresponding to this center of gravity G is determined. This value of 8 fg is output as an increase/decrease in the capacity of the air conditioning unit.

Incidentally, in the above explanation, the values of the membership functions are directly calculated, but these values may be converted into numerical values from O to 1 and then subjected to calculation processing.

When the present invention is used as described above, the size of the data table corresponding to the membership function of the antecedent part is 500 (=100×5) for the difference C between the temperature and the set value.
, for the deviation ΔC, 1000 (=200X5), a total of 1500 bytes (one data is 1 byte), is the approximate size of Table 7, and the data of the set rule and the membership function of the consequent part are combined. It can be composed of about 100 bytes.

On the other hand, if fuzzy control is performed using a conventional table in which the output value of the consequent corresponds to the input value of the antecedent, the size of the table according to the conventional technology is the same as that of the present invention. 20000 (=100X20
0) A byte-sized table is required. Therefore, by using the present invention, highly accurate fuzzy control can be obtained with a small storage capacity. In addition, by using a table to calculate the membership function value of the antecedent part, which takes a particularly long time to calculate, the time required for fuzzy inference can be shortened, and the control cycle of the fuzzy device can be shortened to improve the control characteristics of the object. can be improved.

In addition, better fuzzy control can be achieved by pre-memorizing tables with different membership function settings for the antecedent part, in other words, by preparing multiple tables with different membership functions and switching the table to be used according to the conditions. According to this embodiment, a table using the membership function of the antecedent part set for the cooling operation of the air conditioning unit and a membership function of the antecedent part set for the heating operation of the air conditioning unit are created. It is desirable to memorize a table using the following table, and to switch the table to be used depending on whether the air conditioning unit is set to cooling operation or heating operation.

(G) Effects of the Invention As described above, the fuzzy control device of the present invention includes a storage unit that stores a table consisting of a detected value and the degree of fulfillment of a membership function of an antecedent part calculated in advance for this detected value. , a weighting section that weights the membership function of the consequent by applying the degree of establishment of the membership function of the antecedent read from this table to the control rule of fuzzy inference, and a member of the weighted consequent. Since it is equipped with a calculation section that calculates the output value from the ship function,
Fuzzy inference with different characteristics can be performed by changing the data in the table.

Therefore, by changing the data in the table, fuzzy inference can be easily performed for other controlled objects. Furthermore, by storing multiple tables for a single controlled object, the characteristics of fuzzy inference can be changed depending on the operating state of the controlled object, allowing optimal fuzzy inference to be performed at all times. .

[Brief explanation of drawings]

FIG. 1 is a schematic diagram showing the outline of the present invention, FIG. 2 is a characteristic diagram of the membership function of the antecedent part, and FIG. 3 is a diagram showing the input values and output values of the membership function shown in FIG. 2. Table diagram, Figure 4 is a characteristic diagram of membership functions of other antecedent parts, Figure 5 is a relationship diagram showing the relationship when weighting membership functions of consequent parts, and Figure 6 is a diagram of consequent parts. FIG. 7 is a flowchart showing the operation when performing fuzzy inference, and FIG. 8 is a diagram of a figure obtained by synthesizing weighted membership functions. 1... Conditioned room, 2... Air conditioning unit, 3.
...Temperature sensor, 4...Room temperature setting device, 5...
・Output section, 6...Microcomputer-7 to 9...
-Storage unit, 10...inverter device.

Claims (1)

[Claims] (1) When controlling the operation of equipment so that the detected value of the controlled object becomes the target value, the antecedent part consists of multiple membership functions based on the detected value, and the membership function of the antecedent part is It has a weighting part that weights the membership function of the consequent part according to the degree of fulfillment, and controls the amount of operation of the equipment and the capacity of the equipment based on the value obtained by combining the weighted membership functions. The fuzzy control device is equipped with a storage unit that stores a table consisting of a detected value and the degree of fulfillment of a membership function of an antecedent part calculated in advance for this detected value, and reads out from this table. A weighting unit that weights the membership function of the consequent by applying the degree of establishment of the membership function of the antecedent to the fuzzy inference control rule, and an output value from the weighted membership function of the consequent. What is claimed is: 1. A fuzzy control device comprising: an arithmetic unit that calculates .