JP3233567B2 - Control algorithm automatic acquisition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To acquire a rule for executing correct control by estimating a correction scale factor so that increase/decrease manipulated variable becomes a prescribed variable, multiplying the increase/decrease manipulated variable of existing output by the correction scale factor so as to correct it and successively setting a control deviation and a control deviation change at that time and the corrected increase/decrease manipulated variable to be learning data. SOLUTION: The output dHV(t-1) of a hot water valve is estimated from a temperature deviation e(t-1) and a temperature deviation change de(t) by using a prepared rule. Validity is checked from the temperature deviation e(t), the temperature deviation change de(t) and the acceleration d<2> e(t) of a temperature and inference output is corrected. Then, the operation result of the temperature deviation e(t-1) and the temperature deviation change de(t-1), which are validity-checked, and output dHV(t-1) is given to a learning mechanism as teacher data. The conclusion part real value of a learning-type fuzzy PI controller is learnt by means of teacher data of the learning mechanism. Such operation steps are repeated at every control interval.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention judges a control deviation, which is a difference between a control amount and a set amount, and whether or not the change is ideal. The present invention relates to a control algorithm automatic acquisition device for sequentially learning and evolving a control main body based on the idealized learning data.

[0002]

2. Description of the Related Art An example of fuzzy constant temperature and humidity control will be described. To create a more comfortable living and working environment,
Now, air conditioning control has become an indispensable system. In recent years, the development of industry centering on semiconductors has a great influence on the air conditioning system due to the necessity of advanced environmental maintenance. Representative examples include clean rooms and biohazard countermeasures. In semiconductor production lines, appropriate drying is required to maintain product quality,
Constant temperature and humidity control is used to obtain constant temperature and humidity throughout the year. Conventionally, this has been roughly divided into two types, a V-type temperature and humidity control method and a dew point temperature control method.

The conventional V-type temperature / humidity control method is economical and is often adopted in a case where control in which the fluctuation of humidity is relatively small is desired. The V-type temperature / humidity control device has two controllers, an indoor temperature controller and an indoor humidity controller, to configure two closed loops according to the number of inputs. Determine the manipulated variable. In this control method, an irregular control system of two inputs (temperature, humidity) and three outputs (cold water valve (valve), hot water valve, humidification valve) is divided into one input and one output to facilitate control. Is economical because one valve is always closed,
In the case where the requirements for cooling and dehumidification are overlapped, the response to disturbance may be deteriorated.

[0004] In the dew point temperature control, in order to more accurately control the relative humidity in the constant temperature and humidity apparatus, the dew point temperature is controlled so as to temporarily correspond to a predetermined room temperature and humidity. In the reheating, the shortage of the latent heat load is a control method of adjusting by the rehumidification. In this control method, control is facilitated by dividing an irregular control system of two inputs and three outputs into three control systems of one input and one output. However, in the middle period (spring or autumn) when the outside air is close to the indoor environment, reheating or rehumidification occurs, which may significantly impair economic efficiency.

Next, conventional fuzzy control will be described. In the fuzzy control method described in Japanese Patent Application Laid-Open No. 2-183302, various methods have been proposed for identifying the dynamic characteristics of a process, and it is assumed that the dynamic characteristics can be estimated by these methods. And

The process characteristic estimating means is a means for estimating the process characteristic (L, T, G), and drives a control sampling period and a gain changing function based on the identification result. A fuzzy rule for estimating the control sampling period and the gain from the dead time L, the lead time T and the process gain G is constructed, and the control sampling period and the gain are estimated by fuzzy inference.

The control interval time is adjusted by estimating the control sampling period. Also, the gain of the fuzzy controller is adjusted by estimating the gain.

[0008] The learning type control device and the fuzzy inference device described in Japanese Patent Application Laid-Open No. 5-265512 comprise a fuzzy control model and a fuzzy object model. The fuzzy object model is a learning type fuzzy model that is arranged in parallel with the control object, inputs an output value from the fuzzy control model, and performs learning to match the output value from the control object. In addition, the fuzzy control model determines the amount of change in the error between the output value from the fuzzy object model and the desired output value, and changes the parameters of the fuzzy control model based on this amount of change so as to reduce the error. is there.

[0009]

A conventional problem of fuzzy control is that it is difficult to introduce fuzzy control when there is no expert. This problem indicates that the advantage of fuzzy control may be reversed to the problem. In other words, fuzzy control, which has the advantage of being able to easily reflect the skills of experts and experts, is difficult to apply when there are no skills to be extracted, such as when there are no experts originally.

[0010] Even if there is no expert at all,
Some qualitative (basic) rules are obtained. But, of course, that alone is not perfect and cannot be controlled precisely.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a control algorithm automatic acquisition device capable of acquiring a rule for performing correct control by automatically learning when a basic rule is given. Aim.

[0012]

In order to achieve the above object, a control algorithm automatic acquisition device of the present invention monitors a dynamic characteristic of a control process, and determines an operation unit and an operation direction of the operation unit which affect a control target. From the logic check means to find out, the control deviation and the control deviation change and the acceleration, which are the difference between the control amount and the set amount, estimate the correction magnification such that the already output increase / decrease operation amount becomes the predetermined increase / decrease operation amount, This correction factor is multiplied by the already output increase / decrease operation amount to add a correction, and the control deviation and the control deviation change at that time, and the corrected increase / decrease operation amount are sequentially learned data to learn the control body. Means.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a configuration explanatory view showing an outline of an air conditioner according to the present invention. That is, the cooling coil 2, the heating coil 3, and the steam spray (steam spray humidifier) 4 were incorporated in the air conditioner 1. In addition, the air conditioner 1
It is a general term for the parts in charge of heat exchange equipment excluding heat source equipment and heat medium transport equipment in air conditioning equipment. From the air conditioner 1,
Control target room 6 by internal built-in fan (FAN) 5
Circulating air, which is a mixture of return air and fresh air, is sent out. On the way, the circulating air passes through the two coils 2 and 3 and the steam spray 4 while adjusting the cleanliness. The respective roles of the cooling coil 2 and the heating coil 3 are for adjusting the temperature and humidity by cooling (dehumidifying) and heating, and the steam spray 4 is for adjusting the humidity by spraying steam. FIG.
In the above, T indicates an indoor temperature detecting end of a temperature sensor, H indicates an indoor humidity detecting end of a humidity sensor, and 7 indicates a fuzzy constant temperature and humidity controller. D / A is a digital-to-analog converter built in the CPU, A / D is a analog-to-digital converter built in the CPU, and FuzzyInfer.
ence Board is a fuzzy inference board (PB85
109).

Next, the process of controlling the temperature and humidity inside the air conditioner 1 will be described with reference to a schematic diagram of a psychrometric chart as shown in FIG. That is, inside the air conditioner 1, air conditioning is achieved by various control processes relating to temperature and humidity as shown in FIG. 2 during the passage of air. These control processes are a cooling (dehumidification) control process, a heating control process, and a humidification control process. Each control process is based on physical change rules. Coil 2, 3 mutual arrangement and steam spray 4
If the arrangement of the coils 2 and 3 is changed, the expected control process does not work effectively.

Next, a cooling (dehumidification) control process will be described. That is, the cooling coil 2 is formed by processing a pipe into a coil shape for the purpose of flowing a secondary refrigerant such as a primary refrigerant or brine into the pipe and cooling an object to be cooled such as air or water outside the pipe. As the circulating air passes through the cooling coil 2, if the temperature is higher than the surface temperature of the coil 2, the cooling air is cooled. However, whether or not the moist air is dehumidified depends on whether the air needs to be cooled below the dew point. That is, in the case where it is not necessary, moisture does not condense on the surface of the coil 2 and the amount of moisture in the air does not change. On the other hand, in the case where it is necessary, the air is cooled to a temperature equal to or lower than the dew point temperature, and moisture is condensed on the surface of the coil 2 to be dehumidified.

Next, the heating control process will be described. That is, the heating coil 3 is formed by processing a tube into a coil shape for the purpose of flowing a heat medium such as hot water into the tube and heating an object to be heated such as air or water outside the tube. When the circulating air passes through the heating coil 3, only the dry-bulb temperature rises. In the heating coil 3, if the inlet temperature of hot water, the flow rate of hot water, and the flow rate of air are constant, the amount of heat exchange between hot water and air on the heating surface is constant. In real air conditioning, such a fixed amount of heat exchange process does not always exist,
The indoor temperature is kept constant by changing the heating capacity in accordance with the fluctuation of the indoor load and adjusting the amount of heat supplied to the indoor.

Next, the humidification control process will be described. That is, the steam spray 4 is a humidifier of a type in which steam generated in a boiler or the like is jetted from a nozzle or a small hole to humidify. The humidification control process increases the room humidity.

Next, an economical temperature and humidity control process will be described. That is, the circulating air handled in the air-conditioning control is simply a mixed gas consisting of air and water vapor.
A wet psychrometric chart is generally used to represent the state of the gas mixture. According to the psychrometric chart, it is understood that the temperature and humidity conditions in the air conditioner can be created in any air condition by a combination of the two control processes.
That is, a state in which three control processes are operating at the same time acts to negate each other's capabilities, and is uneconomical.

Next, a general operation in summer will be described. That is, since the weather conditions in Japan in summer are hot and humid, air conditioning performed in summer, that is, summer air conditioning focuses on cooling and dehumidification of air. In the summer, the cooling coil 2 acts as a device for removing moisture in the air, that is, a dehumidifier, depending on the indoor load of the controlled room 6 and the state of the outside air (fresh air).
The general operation of the heating coil 3 is to act as an air heater for the purpose of reheating, that is, a reheater. In addition,
Japanese pleasure zone in summer (pleasure zone in summer air conditioning)
Is based on a room temperature of 26 ° C. and a relative humidity of 50%.

Next, a general operation in winter will be described. That is, the winter pleasant zone is set to a room temperature of 20 ° C. and a relative humidity of 50%. Although the outdoor air in Japan is dry in winter, depending on the state of indoor load, in air conditioning performed in winter, that is, in winter air conditioning, emphasis is placed on heating the air by the heating coil 3 and humidifying the air by the steam spray 4. .

Next, the operation of the operator of the fuzzy constant temperature / humidity adaptive control device will be described. That is, FIG.
Fig. 4 shows analysis results of an operator's thinking pattern and behavior pattern regarding operation of the fuzzy constant temperature / humidity adaptive control device. There are two aspects to an operator's decision. The first is to determine what to do with temperature and humidity. Secondly, based on the temperature and humidity requirements, the opening degree of each valve is confirmed, and the amount of operation of each valve is determined.

Next, a form of fuzzy inference will be described. That is, FIG. 4 shows a form of fuzzy inference based on the decision of the operator. The fuzzy inference mechanism was composed of two stages and five blocks. The first stage infers a required temperature and a required humidity for making the room temperature and humidity, which are the control amounts, coincide with a target value. In the second step, the amount of increase / decrease of the three valves is inferred from the opening degrees of the three valves (cold water, hot water, and humidification) based on the required amounts obtained in the first step. However, the actual operation is limited to one or two valves. Each fuzzy inference block consists of five parts: temperature required fuzzy intermediate inference, humidity required fuzzy intermediate inference, cold water valve operation amount fuzzy final inference, hot water valve operation amount fuzzy final inference, and humidification valve operation amount fuzzy final inference. .

In the fuzzy intermediate inference, a control deviation, a change in the control deviation, is captured in order to make the temperature / humidity which is a control amount coincide with a target value, and a required amount of whether the control amount is raised, lowered, or nothing is performed. Ask for. Further, the fuzzy final inference consists of the following basic control operations of the operator. [Rule 1] If it is desired to raise the temperature, there is a method of closing the cold water valve or a method of opening the hot water valve. However, closing the cold water valve is more economical than opening the hot water valve. However, if the cold water valve is already fully closed or the cooling coil is acting as a dehumidifier, the method of opening the hot water valve must be selected. [Rule 2] If it is desired to lower the temperature, there is a method of opening a cold water valve or a method of closing a hot water valve. However, closing the hot water valve is more economical than opening the cold water valve. However, if the hot water valve is already fully closed, the method of opening the cold water valve must be selected. [Rule 3] If it is desired to increase the humidity, there is a method of closing the chilled water valve or a method of opening the humidification valve. However, closing the chilled water valve is more economical than opening the humidification valve. However, if the chilled water valve is already fully closed or the cooling coil is acting as a cooler, the method of opening the humidification valve must be selected. [Rule 4] If it is desired to lower the humidity, there is a method of opening a cold water valve or a method of closing a humidification valve. However, closing the humidification valve is more economical than opening the chilled water valve. However, if the humidification valve is already fully closed, the method of opening the chilled water valve must be selected. [Rule 5] If three valves are open, close the three valves little by little.

Next, the automatic acquisition of the control algorithm will be described. If the output of the model is clearly poor, it is presumed that the model itself is incomplete. In such a case, it is necessary to correct an incomplete (non-ideal) model to an ideal model.

In an actual system, there are many cases where ideal data sufficient for constructing a model cannot be obtained. In addition, even if there is no expert (expert) in the target control, even a control rule cannot be obtained. However, at least some qualitative (basic) rules can be obtained.
Then, let them learn some of the qualitative rules, create an initial model, and control it with the model. However, because the algorithm is simple, there is a high probability that the desired control will not be achieved. This time, the incomplete model is learned and the algorithm is modified.

The control algorithm automatic acquisition device of the present invention includes a logic check unit and a validity check unit. The logic check unit has a function of monitoring dynamic characteristics of a control process and finding an operation unit and an operation direction of the operation unit that affect a control target. Further, the validity checking means estimates a correction magnification from the control deviation, the control deviation change, and the acceleration, which are the difference between the control amount and the set amount, such that the increase / decrease operation amount already output becomes a predetermined increase / decrease operation amount, The correction magnification is multiplied by the already output increase / decrease operation amount to add a correction, and the control deviation and the control deviation change at that time and the corrected increase / decrease operation amount are sequentially learned data to provide learning to the control body. .

That is, a learning algorithm for realizing the automatic acquisition of the fuzzy control algorithm is constructed, and the model obtained by the basic rules is automatically learned. It became possible to make corrections.

The temperature and humidity control will be described below as an example. [1] Role of valve (valve) In temperature / humidity control, all three valves (cold water, hot water, and humidification) are not opened when energy saving is considered. The two open valves (possibly one) have their respective roles, one controlling the temperature and one controlling the humidity. It is determined whether the output of the valve associated with each is appropriate or not, and if inappropriate, the model is modified. The correspondence between the valves is as follows. When the cold water valve (CV) and hot water valve (HV) are open ... H
V → Temperature, CV → Humidity When the cold water valve (CV) and humidification valve (SV) are open ... C
V → Temperature, SV → Humidity When hot water valve (HV) and humidification valve (SV) are open ... H
V → Temperature, SV → Humidity [2] Methodology for Algorithm Correction Methodology for algorithm correction is composed of two steps: logic check means and validity check means. However,
The correction and the re-learning need to be performed sequentially, and “sequential” is appropriate for each control interval.

[2-1] Logic Check Means As shown in FIG. 5, there are nine types of temperature / humidity response patterns in all. According to these nine patterns and the current opening degree of the valve (three divisions whether it is not fully open or not fully open),
Which valve should control temperature or humidity?
It is identified by the x-Min composite maximum height method. Although the number of rules was 243 rules in total, it was simplified by the Quine-McCluskey method.
As shown in FIG. 7, there are 28 rules. 6 and 7,
dCV indicates the amount of increase / decrease of the cold water valve, dHV indicates the amount of increase / decrease of the hot water valve, dSV indicates the amount of increase / decrease of the humidification valve, T indicates the temperature, and H indicates the humidity. A circle in T in the dCV column indicates that the chilled water valve is operating temperature. For example, Rule 10 indicates that the cold water valve controls humidity and the hot water valve controls temperature.
In the Max-Min combined maximum height method, the rule that best matches the current state is found, and the output of the inference is set as the conclusion of the rule. For example, when increasing both temperature and humidity, when the cold water valve is open, the hot water valve is closed, and the humidification valve is open, the temperature is controlled by the cold water valve and the humidity is controlled by the humidification valve. I understand. This can be represented as in FIG.

Thereafter, it is necessary to confirm whether the operating polarity of the operating valve is correct, and correct it if it is incorrect. The operating polarity of the valve has already been determined,
Anything contrary to this is considered to be wrong. The operating polarity of the valve can be expressed as follows.

[0031]

(Equation 1)

[2-2] Validity Checking Means By monitoring the behavior of the controlled variables (temperature and humidity), the valve which is erroneously moving is gradually guided in the correct direction. By the logic check,
Determine which valve is controlling the temperature (humidity)
The output of the valve will be corrected by a validity check.

In order to measure validity, one must first know how the controlled variables (temperature and humidity) are changing. The change of the control amount (deviation, deviation change and acceleration) can be derived as follows using the set value and the measured value.

[0034]

(Equation 2)

Since it has already been determined by a logicality check which valve is adjusting the temperature and the humidity, in the validity check, the valve is determined by calculating the deviation of the temperature and the humidity, the deviation change, and the acceleration. The role is to get to normal operation. Specifically, as shown in FIG. 9, the output is changed by multiplying the actual output of the valve by an appropriate value, so as to gradually approach the ideal output.

What is important here is how to determine the magnification to be applied to the output. Here, the mechanism of determining the magnification by inference using the Max-Min composite centroid method is described. adopt.

In the validity check, three input variables of deviation, deviation change, and acceleration are used. In principle, each scaling factor uses the same numerical value as that of the control body. However, since the acceleration is not in the control main body, it is necessary to determine it arbitrarily. From these facts, the rule block of the validity check is 4 as shown in FIG.
Divided into blocks. However, in order to correct the directionality of the output from the model in the B block and the C block, the absolute value of the output from the model is taken in advance. FIG. 11 is a conceptual diagram of a validity check rule block. Since the result of the validity check greatly differs depending on the polarity, it is necessary to sufficiently consider the polarity. For this reason, the membership function is divided into -ZR and + ZR as shown in FIG.

FIG. 13 shows a control rule table for the validity check. FIG. 14 shows the block division of the control rule table. The label defining the fuzzy set representing acceleration is ｛N
B, NS, ZR, PS, PB}, where PB
3 is a control rule table for. This represents (5) in FIG.

However, when actually performing the inference for determining the magnification, the numerical value entered in FIG. 13 cannot be used as it is, so the following conversion is made and the exponent is used.

{1/4, 1/2, 1, 2, 4} → ｛2 ^-2 , 2
^{-1, 2 0, 2 1,} 2 2} {-4, -2, -1, -1 / 2, -1 / 4} → {-2 2, -2 1, -2 0, -2
^-1, -2 ^-2} when performing validation, there is a point that must be another noted other polarity. FIG. 15 shows a behavior of a certain control amount and an initial control rule table. That is, when the actual output is as close as possible to zero (ZR), the behavior of the control amount is assumed to be either a or b. If the ZRs are not classified by polarity, the magnification will be the same for a and b. For example, if the prepared value of the magnification is a positive value, no particular problem occurs in the case of a, but the output is immediately corrected to the negative direction in b. Since the automatic acquisition of the control algorithm evolves by learning sequentially, giving a largely wrong learning guide will lead the learning itself in the wrong direction, and that must be avoided.

[0041]

DESCRIPTION OF THE PREFERRED EMBODIMENTS There are various types of automatic control algorithm acquisition systems. In this example, a learning type fuzzy-PI controller based on a learning type fuzzy inference method and a hierarchical fuzzy model are described. The automatic acquisition device will be described.

[1] Automatic Acquisition Device for Control Algorithm of Learning Fuzzy-PI Controller FIG. 16 is an automatic acquisition device for a control algorithm of a learning fuzzy-PI controller. The learning rule of the learning type fuzzy-PI controller is to provide a plurality of ideal input / output data to a predetermined fuzzy model, capture the difference between the output data from the model and the ideal output data, and use the steepest descent method. An ideal model is obtained by learning sequentially based on. The learning type fuzzy-PI controller uses a real value instead of a fuzzy set in the conclusion.

Here, the temperature control using a cooling coil or a heating coil will be described as an example. The control algorithm of the learning type fuzzy-PI controller is as follows. [Step 1] A rule (for example, 25 rules) necessary for temperature control of a learning type fuzzy-PI controller is prepared. The learning type fuzzy-PI controller has two inputs e (t-1), de
(T-1) One output dHV (t-1), the input is divided into five, and the output (conclusion real value) is a random number.

[Step 2] Using the prepared rules,
The output dHV (t-1) of the hot water valve is inferred from the temperature deviation e (t-1) and the temperature deviation change de (t-1). [Step 3] A logicality check is performed (confirm the operation polarity of the hot water valve: normal operation = + 1. When a cold water valve is used, the reverse operation = −1).

[Step 4] Validity is checked from the temperature deviation e (t), the temperature deviation change de (t), and the temperature acceleration d ² e (t), and the inference output is corrected. [Step 5] The calculation results of the temperature deviation e ^* (t-1), the temperature deviation change de ^* (t-1), and the output dHV (t-1) are given to the learning mechanism as teacher data.

[Step 6] The real value of the conclusion part of the learning type fuzzy-PI controller is learned by the teacher data of the learning mechanism. [Step 7] [Step 2] to [Step 6]
Repeat at every control interval.

[2] Hierarchical Fuzzy Model Control Algorithm Automatic Acquisition Apparatus FIG. 17 shows a hierarchical fuzzy model control algorithm automatic acquisition apparatus. As for the learning type fuzzy-PI controller, as the problem becomes larger, the total number of rules becomes enormous, and application becomes difficult. Therefore, if the fuzzy model has a hierarchical structure, the number of parameters is reduced, and the fuzzy model can be applied to a large-scale problem. Moreover, the structure and learning rules of the hierarchically connected network in the neural network are replaced with a hierarchical fuzzy model.

The hierarchical fuzzy model constitutes a two-stage, five-block (fuzzy intermediate inference, fuzzy final inference) fuzzy inference mechanism used in the fuzzy constant temperature and humidity control system of FIG. In the fuzzy intermediate inference, in order to match the temperature and humidity, which is the control amount, with the target value, the temperature deviation and the change in the temperature deviation were captured, and the required amount of whether the control amount was raised, lowered, or nothing was determined. . The fuzzy rules used in the fuzzy final inference are based on the basic control operations of the operator as described in the above [Rule 1] to [Rule 5].

Here, the fuzzy constant temperature and constant humidity control will be described as an example. The control algorithm of the hierarchical fuzzy model is as follows. [Step 1] 7 inputs (cold water valve CV (t-1), hot water valve HV (t-1), humidification valve SV (t-1) opening degree, temperature deviation e _T , temperature deviation change de _T , humidity deviation e _H , humidity deviation change de _H ) 3 outputs (chill water valve opening change dCV (t),
Hot water valve opening change dHV (t), humidification valve opening change dSV
A hierarchical fuzzy model of (t)) is prepared. The hierarchical fuzzy model includes a middle layer first model to a middle layer third model, and an output layer first model to an output layer third model.

[Step 2] Using the prepared hierarchical fuzzy model, the cold water valve dCV (t) and the hot water valve dHV
(T) Infer each output of the humidification valve dSV (t). [Step 3] The required temperature and the required humidity and the current cold water valve CV (t-1), hot water valve HV (t-1), and humidification valve SV
A logicality check is performed from the opening of (t-1).

[Step 4] Temperature and humidity deviations e _T and e _H , deviation changes de _T and de _H , and acceleration d ²
A validity check is performed from e _T and d ² e _H , and the inference output is corrected.

[Step 5] Seven inputs (cold water valve CV)
(T-1), hot water valve HV (t-1), humidification valve SV (t-
1) opening, temperature deviation e _T , temperature deviation change de _T , humidity deviation e _H , humidity deviation change de _H ) and three outputs (chilled water valve opening change dCV (t), hot water valve opening change dHV ( t),
The calculation result of the humidification valve opening change dSV (t)) is given to the learning mechanism as teacher data.

[Step 6] Hierarchical fuzzy model learning is performed using teacher data given to the learning mechanism. [Step 7] [Step 2] to [Step 6]
Repeat at every control interval.

[0054]

As described above, according to the present invention, a learning algorithm for realizing the automatic acquisition of a control algorithm is constructed, and a model obtained by basic rules is automatically learned. It became possible to correct an incomplete model to a model suitable for control.

[Brief description of the drawings]

FIG. 1 is a configuration explanatory view showing an outline of an air conditioner according to the present invention.

FIG. 2 is a schematic diagram showing an example of a psychrometric chart used in the present invention.

FIG. 3 is an explanatory diagram showing an example of an operator's thinking pattern and behavior pattern used in the present invention.

FIG. 4 is an explanatory diagram showing an example of a fuzzy inference mode used in the present invention.

FIG. 5 shows the temperature and temperature used for the logic check according to the present invention.
It is explanatory drawing which shows an example of the response pattern of humidity.

FIG. 6 is an explanatory diagram showing an example of rules 1 to 24 used for a logicality check according to the present invention.

FIG. 7 is an explanatory diagram showing an example of rules 25 to 28 used for a logicality check according to the present invention.

FIG. 8 is an explanatory diagram showing a part of the rule in FIG. 6;

FIG. 9 is an explanatory diagram showing an example of a correction method in the validity check according to the present invention.

FIG. 10 is an explanatory diagram showing an example of a validity check rule block according to the present invention.

FIG. 11 is an explanatory diagram showing an example of a concept of a rule block of a validity check according to the present invention.

FIG. 12 is an explanatory diagram showing an example of a membership function for validity check according to the present invention.

FIG. 13 is an explanatory diagram showing an example of a control rule table for validity check according to the present invention.

FIG. 14 is an explanatory diagram showing an example of block division of a control rule table for validity check according to the present invention.

FIG. 15 is an explanatory diagram showing an example of a behavior of a certain control amount for validity check and an initial control rule table according to the present invention.

FIG. 16 is an explanatory diagram showing an example of a control algorithm automatic acquisition device of a learning type fuzzy-PI controller according to the present invention.

FIG. 17 is an explanatory view showing an example of a control algorithm automatic acquisition device for a hierarchical fuzzy model according to the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Air conditioner, 2 ... Cooling coil, 3 ... Heating coil, 4 ... Steam spray (steam spray humidifier), 5 ... Fan (FA
N), 6: controlled room, 7: fuzzy constant temperature and humidity controller,
T: Indoor temperature detecting end of temperature sensor, H: Indoor humidity detecting end of humidity sensor, CPU: Measurement control computer, D / A
... Converter for converting digital value to analog value, A / D ... Converter for converting analog value to digital value, Fuzzy I
nreference Board: Fuzzy inference board (PB85109).

Continuation of the front page (56) References JP-A-5-173604 (JP, A) JP-A-5-99482 (JP, A) JP-A-63-29142 (JP, A) Toshikazu Tobi, two others, " Fuzzy Constant Temperature and Humidity Control of Clean Room Part 2: Attempt of Automatic Adjustment of Fuzzy Control Parameters ”, Proc. Of the 11th Fuzzy System Symposium, Fuzzy Society of Japan, July 12, 1995, p. 775-778 Kazuo Nishimura, “Trends in Neural Network Technology”, Toshiba Review, Toshiba Corporation, December 1, 1991, Vol. 46, No. 12, p. 924-930 Toshikazu Tobi and 2 others, "Fuzzy constant temperature and humidity control in clean room, Part 1-Trial of automatic adjustment of fuzzy control parameters", Proc. Of the 11th Fuzzy System Symposium, Fuzzy Society of Japan, July 1995 March 12, p. 773-774 (58) Field surveyed (Int. Cl. ⁷ , DB name) G05B 13/02 F24F 11/02 102 JICST file (JOIS)

Claims (1)

(57) [Claims] 1. A hot and humid condition based on a control rule or a control model.
The cold water valve, hot water valve,
A fuzzy controller for calculating the amount of increase or decrease of the humidifying valve is provided, and a teacher is provided based on the temperature / humidity input to the fuzzy controller and the amount of increase / decrease of the cold water valve, hot water valve, and humidifying valve output from the fuzzy controller. In the control algorithm automatic acquisition device that generates data and gives this teacher data to the learning mechanism of the fuzzy controller to modify the control rules and control model of the fuzzy controller, the fuzzy controller uses the first rule. Entered temperature
Humidity and chilled water valve output from fuzzy controller ,
A cold water valve, a hot water valve, a humidifying valve and a cold water valve, a hot water valve,
Using the logic check means for confirming the operation polarity of the wet valve and the second rule, the processing result by the logic check means, the control deviation calculated from the temperature and humidity input to the fuzzy controller, and the control deviation change Based on the acceleration and the amount of increase / decrease of the chilled water valve, hot water valve, and humidification valve output from the fuzzy controller, the amount of increase / decrease of the chilled water valve, the hot water valve, and the humidification valve becomes a predetermined amount of increase / decrease. The correction magnification is estimated, and the correction magnification is multiplied by the amount of increase / decrease of the cold water valve, hot water valve, and humidification valve output from the fuzzy controller, and a correction is made.
A control algorithm automatic acquisition device comprising: a validity check means for giving an increase / decrease operation amount of a cold water valve, a hot water valve, and a humidification valve as teacher data to a learning mechanism of a fuzzy controller.