JPH09196434A - Air-conditioning control device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an air-conditioning control device to perform setting of a room temperature immediately reflecting the degree of amenity inputted by a resident and realize air-conditioning environment comfortable against various environment conditions through improvement of a learning effect as saving of energy is satisfied. SOLUTION: When the degree of comfort is inputted by operating a degree-of- amenity input part 11 by a resident, a comfortable air-conditioning control part 16 immediately corrects a PMV target value based on the degree-of-amenity thereof for resetting. A fuzzy room temperature computing part 15 computes a room temperature set value through fuzzy inference from a PMV target value after correction and a present neuro PMV. An inputted degree of amenity of a resident is reflected in a room temperature set value at a real time. By urging inputting of the degree of comfort by flickering a lamp at intervals of a specified period, a learning data amount is increased and learning about various environment conditions is practicable.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an air conditioning control device for realizing a comfortable air conditioning environment in a room of various buildings such as a multipurpose intelligent building, an office building, a hotel, a hospital and a living building while satisfying energy saving.

[0002]

2. Description of the Related Art Conventionally, scheduled air conditioning control in buildings has been carried out with a constant room temperature set value, but recently, attention has been paid to the comfort of the occupants, and the comfortable air conditioning is controlled momentarily. Control is starting to be put to practical use.

There are various methods for calculating the comfort index, but PMV (Predicte
"D Mean Vote" is one of the main indexes, and an air conditioning control device using this PMV is known (Japanese Patent Laid-Open No. 5-126380).

PMV is a fanger (F
Anger) 's comfort equation and expressed as a function of six comfort factors (process variables) that affect human thermal sensation: room temperature, humidity, average radiant temperature, air velocity, clothing state, and activity state. .

PMV = {0.352 × exp (−0.042 × M / A _DU ) +0.032} · L (1) Here, M: Metabolism amount (kcal / h) according to activity state A _DU : Human body Body surface area (m ² ) L: Human heat load (kcal / m ² · h) Furthermore, L, which is the difference between heat generation inside the human body and heat release to the outside of the body
Is expressed by the following equation (2).

L = f {room temperature, humidity, average radiant temperature, air velocity, clothing state} (2) where f: function PMV is the comfort that the occupant feels when hot (+3), Numerically quantified as warm (+2), slightly warm (+1), neither, comfortable (0), slightly cool (-1), cool (-2), cold (-3), and PMV- The range of 0.5 to +0.5 is said to be the indoor environment (thermal neutral range) that most people feel comfortable with.

The comfort equation, which is the basis of PMV calculation, is derived from the heat balance of heat generation in the human body and heat release to the outside of the human body, and is based on the results of questionnaires of subjects under various environmental conditions. Since this is an experimental index, it may differ from the comfort of the individual living environment and the residents who are actually there.

Therefore, the device disclosed in the above-mentioned Japanese Patent Laid-Open No. 5-126380 shows the actual comfort level of the occupant based on the six comfort factors of room temperature, humidity, average radiant temperature, air velocity, clothing state and active state. It learns by a neural network, and tries to control the room temperature set value for the air conditioner by fuzzy inference so that the index based on this learning falls within the comfortable range.

[0009]

In recent years, an intelligent air conditioning control device has been developed, and an input unit capable of inputting a high-speed and high-level computing function and comfort level of a occupant has been developed.

However, at most one input unit is installed in the air conditioning control zone, and there are the following problems.

Learning using a neural network is
Since the calculation load becomes large, it cannot be performed during the normal comfortable air conditioning control period. Therefore, when the occupant expresses dissatisfaction with the current environment, he / she cannot perform learning in real time when he / she inputs a degree of comfort, and thus cannot be immediately reflected in the room temperature set value after learning.

Since the resident does not have an input unit in his / her seat, he / she goes to perform the operation input of the air conditioning only when he / she is really dissatisfied. On the other hand, when satisfied, it is natural that the air conditioning operation is not performed and the learning effect cannot be sufficiently obtained.

Further, since a large amount of energy is required for air conditioning, it is desired to control the occupants with energy that is as small as possible.

Therefore, an object of the present invention is to provide an air conditioning control device capable of energy saving operation, capable of promptly responding to occupant's comfort level input, and capable of control based on a sufficient learning effect.

[0015]

In order to achieve the above object, the present invention inputs a plurality of process variables that affect the human thermal sensation and calculates a comfort index from these process variables by a neural network. The first computing means and the calculated comfort index are input, and the deviation between the currently calculated comfort index and the set comfort index target value, the currently calculated comfort index, and the previously calculated comfort index are calculated. And a second computing means for computing a room temperature set value for the air conditioner by fuzzy inference based on each of the obtained deviations, and an input means for inputting the occupant's comfort level. A storage means for storing the input process variable and the occupant's comfort level as learning data, and using the learning data in a predetermined time zone, inputting the process variable and teaching the occupant's comfort level. Learns neural networks as characterized by having a learning means for determining the weight of said first arithmetic means.

According to the above configuration, since the comfort index learned based on the learning data including the comfort level input by the resident is calculated, different from the comfort index determined by the comfort equation, the individual living environment and the actual The degree of comfort felt by the occupants there is a comfort index for the final comfort adjustment control. In addition, since the comfort index target value can be set within the comfort range, the comfort index target value should be set to the barely comfortable range (on the hot side) according to the summer and winter seasons. Or, by lowering it (on the cold side), the air conditioning load can be easily reduced, and energy can be saved. Further, by correcting the comfort index target value from the time when the comfort level is input by the occupant, the room temperature can be changed quickly, and the thermal dissatisfaction of the occupant can be removed to meet the occupant's request.

In addition, display means may be provided for displaying the comfort level input time at regular intervals. The display of the comfort level input time includes a visual display and an audible display. Such a display prompts the user to input the comfort level, and the resident does not forget to input the comfort level at regular intervals. As a result, it is possible to avoid the problem that the learning data amount increases, the comfort level input data is small, and the learning effect does not improve, and it is possible to perform learning for various environmental conditions.

When the occupant's comfort level is input by the input means, the comfort index target value for the second calculation means is immediately corrected and reset based on the comfort level. A target value setting means may be provided.

If the comfort level is not input for a certain period after the occupant's comfort level is input, the comfort level corresponding to the comfort level is input until the next comfort level is input. It may be considered that the comfort level input is used for learning. Since the learning data is automatically collected, the comfort level after a new room temperature change can also be learned.

Further, when there is a change input of the room temperature set value, control based on the comfort level may be stopped and an air conditioning control mode switching means for outputting an air conditioning control signal based on the room temperature set value may be provided.

Further, the comfort index target value modified based on the input of the comfort level of the occupant may be reset to the comfort index target value before modification before the next air conditioning time period.
As a result, on the next day, the normal comfort index target value that can realize energy saving and comfort is reset according to cooling or heating, so that the control aiming at energy saving and comfort is executed.

Further, the learning may be performed outside the air conditioning time period and during a time period when the calculation load of the air conditioning monitoring control is low. Since learning by the neural network involves a calculation load, the learning does not become a calculation load for the air conditioning monitoring control.

[0023]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of an air conditioning control system showing an application example of an air conditioning control device according to an embodiment of the present invention.

The air-conditioning control system shown in the figure has a steam valve 4a, a hot water valve 4b and a cold water valve 4c connected to the preceding stage of the air conditioner 3 for cooling and heating the building room 2.
An air conditioning controller 1 is provided to control the air conditioner via a DDC (Direct Digital Controller) 5A. In addition, an indoor environment measuring unit 6 for measuring the indoor environment is provided on the indoor 2 side,
The measured value is a DDC configured in the same manner as the DDC 5A.
It is input to the air conditioning control device 1 by 5B. The air conditioning controller 1 obtains a room temperature set value based on the measured value from the DDC 5B and outputs it to the DDC 5A.

The air conditioner 3 comprises an air filter, an air cooler (cooling coil), an air heater (heating coil), a sprayer for humidifying air (spray spray), a blower and the like.

The DDC 5A is a controller for directly controlling a comfort factor (process variable) such as room temperature,
Controlling the air conditioner 3 through the valves 4a, 4b, 4c so that the temperature and humidity of the room 2 match the room temperature set value from the fuzzy room temperature calculator 17 and the humidity set value given from outside. Is.

The indoor environment measuring unit 6 includes a thermometer 6a, a hygrometer 6b, and an average radiation thermometer 6c installed in the building room 2.
And an air velocity meter 6d, and these instruments 6a to 6d
The indoor environment is measured by the above, and the measured values of room temperature, humidity, average radiant temperature, and air velocity are input to the learning data storage unit 12 and the neuro PMV calculation means 13 described later via the DDC 5B as comfort factors related to the indoor environment. It is like this.

In the air conditioning control device 1, the driver of the present system inputs a set value input section 10 for the occupant to input a seasonal clothing condition concerning the occupant and an activity condition according to the indoor use, and the occupant inputs the comfort level. And a comfort level input unit 11.

The neuro PMV calculating means (first calculating means) 13 inputs the measured value from the DDC 5B, which changes moment by moment, and the set value relating to the clothing state or the active state from the set value input section 10, that is, the process variable. Then, the neural PMV is calculated by the neural network. Further, the learning data storage unit 12 accumulates and stores the comfort level input from the comfort level input unit 11 as learning data together with the process variables.

Reference numeral 16 is a learning means by a neural network, which inputs learning data concerning the process variables, and learns the weight of the neuro PMV calculating means 13 by the back propagation method using the learning data concerning the comfort level as a teaching signal.

Reference numeral 15 denotes a fuzzy room temperature calculation unit (second calculation means) which calculates a room temperature set value using a neuro PMV value, a preset PMV target value, a fuzzy table and the like. The PMV target value is preset by the target value setting means 14, and when the comfort level input section 11 inputs the comfort level, the PMV target value setting means 14 corrects and updates the PMV target value.

FIG. 2 is a view showing the operation panel surface of the comfort level input section 11.

On the operation panel surface of the comfort level input section 11, as shown in the figure, "very hot", "hot", "slightly hot", "neutral", "slightly cold", "cold", Comfort level input keys 11a to 11g, which indicate comfort levels in seven levels such as "very cold", are provided in order from top to bottom. Further, a lamp 11h is provided on the upper right of the operation panel surface, and a room temperature setting key 11i for setting the room temperature higher and a room temperature setting for setting the room temperature lower on the lower right of the operation panel surface. Key 11j and these room temperature setting keys 11i, 11
A set temperature indicator 11k for displaying the room temperature set by j is provided. The key 11 that says "It's very hot"
+1.5 for a, +1.0 for key 11b labeled "Hot", +1.0 for key 11c labeled "Slightly Hot"
0.5, 0 for the key 11d labeled "neutral", -0.5 for the key 11e labeled "slightly cold", -1.0 for the key 11f labeled "cold", A comfort level of -1.5 is assigned to each of the keys 11g displayed as "very cold" in advance.

The comfort levels assigned to the comfort level input keys 11a to 11g can be changed by the numerical value to the first decimal place, and it is very hot (+3) to very cold (-3). It may be very hot (+2) to very cold (-2).

Here, as a structure that is conventionally used for setting and inputting a numerical value such as comfort level, a push button for increasing and decreasing numerical values is provided, and the numerical value is incremented by "1" each time this button is pressed. There is a structure of (or subtraction). When such a structure is used for the comfort level input unit, although the comfort level is in the range of, for example, +3 to -3, depending on the number of pushes by the occupant, an extreme value such as +10 (-10) may occur. It is conceivable that the number will be input and the control will be lost.

However, as shown in FIG. 2, each input key 11a
If a numerical value is set to ˜11 g, it can be set within a preset range, for example, +3 to −3, even if the occupant makes a mistake in the operation. In other words, the comfort level input unit 11 is restricted from the specification in terms of the abnormal value input, and the abnormal value input can be reliably blocked.

Further, even when an arbitrary comfort level can be input like a keyboard input, an extreme level of comfort level may be input, but it is difficult to limit due to the above device specifications. , Apply the following restrictions. For example,
Within ± α (α calculated by Whanger's comfort equation
For example, 1.5) is used as the upper and lower limits, and an abnormal value check of the input value is performed, and an extreme value is excluded by the upper and lower limits.

Further, it is preferable that the comfort level input section 11 informs the occupant of the comfort level input time and depresses the corresponding comfort level input keys 11a to 11g. Therefore, when the comfort level is input (a certain fixed period). A blinking drive unit (not shown) that drives the lamp 11h to blink each time for a fixed time (for example, 1 minute)
Equipped inside. Note that the comfort level input time may be notified by a buzzer sound. This allows the occupant to know the comfort level input time without looking at the comfort level input unit 11.

The neuro PMV calculating means 13 calculates the neuro PMV by the neural network shown in FIG. 3, and the indoor environment measuring unit 6 and the setting input unit 10 are operated.
6 comfort factors (process variables) from the input layer are input to the input layer, and the output layer according to the weight of the intermediate layer (only the weight W _i between the final intermediate layer and the output layer is shown in the figure). To output a neuro PMV. In the example of the figure, the neuro PMV is calculated from the value O _i and the weight W _i of the neuron in the final intermediate layer, and the neuro PMV = W ₁ · O ₁ + W ₂ · O ₂ + W ₃ · O ₃ + W ₄ · O ₄ + W ₅ are calculated in _{· O 5 + W 6 · O} 6 + W 7 · O 7 ... (3).

The PMV target value setting unit 14 sets the PMV according to the summer and winter seasons within the range where the occupants feel comfortable.
A target value is set in advance, and a RAM for storing the set PMV target value is provided. For example, a low PMV target value (eg, −0.3) in winter and a high PMV target value (eg, +0.3) in summer are stored in the RAM in advance.

In the fuzzy room temperature calculation unit 15, a 2-input 1-output fuzzy rule table shown in Table 1 described later and a membership function are preset. The fuzzy room temperature calculation unit 15 includes a neuro PMV and PMV target value setting unit 1 calculated by the neuro PMV calculation unit 13.
The PMV target value set in the RAM of No. 4 is input, and the currently calculated neuro PMV is calculated as shown in the equation (4) described later.
Of the fuzzy input by calculating the deviation between the target value and PMV
Find the V value. Further, as shown in equation (5) described later, the deviation between the currently calculated neuro PMV and the previously calculated neuro PMV is calculated to obtain the PMV change amount ΔPMV. PMV value of these fuzzy inputs and PMV change amount Δ
The change amount of the room temperature set value is obtained according to the fuzzy rule table shown in Table 1 and the fuzzy inference using the membership function based on the PMV, and thereafter, the preset room temperature target value is added to the change amount of the room temperature set value. It is designed to find the room temperature set value.

Fuzzy input PMV = (current neuro PMV) − (PMV target value) (4) ΔPMV = (current neuro PMV) − (previous neuro PMV) (5)

[Table 1]

The calculation of the change amount of the room temperature set value by fuzzy inference using the fuzzy rule table is specifically performed as follows. That is, assuming that the fuzzy input PMV is “PM” (positive and medium) and the PMV variation ΔPMV is “PS” (positive and small), “NM” (negative and medium) from the rule table shown in Table 1. Is obtained, the change amount of the room temperature set value corresponding to “NM” is obtained from the membership function. Considering the time of heating, when the warm sensation is "warm" and "PMV is increasing", the room temperature is changed to "middle".

Each of the above-mentioned function realizing means is actually a CPU which controls the entire air conditioning controller 1 and a ROM which stores a control program as shown in the flowchart of FIG.
And a RAM that stores various types of information. CP
The U performs various controls such as lamp blinking control, comfortable air conditioning control and learning control, which will be described below, according to a control program stored in the ROM.

<Lamp blinking control> A fixed cycle (for example, 30
The blinking drive unit of the comfort level input unit 11 is operated every one minute to one hour, and the lamp 11h is driven to blink for a fixed time (for example, one minute). By this operation, the lamp blinking control is performed so that the occupant periodically presses the comfort level input keys 11a to 11g. The display screen may be blinked to prompt the user to input the comfort level. <Comfortable air conditioning control> If the current time is in the comfortable air conditioning control time zone, the neuro PMV calculation unit 1
3, the neuro PMV is calculated, the room temperature set value is calculated by the fuzzy room temperature calculator 15 using the preset PMV target value, and the comfortable air conditioning control is performed to give the room temperature set value to the DDC 5A. The setting of the room temperature is performed every cycle dt. When the room temperature setting keys 11i and 11j are pressed during this comfortable air conditioning control time period, it is considered that there is a direct intervention in the room temperature setting, and the comfort control being executed by the air conditioning control mode switching means 17 shown in FIG. Stop the air conditioning control and shift to the air conditioning control based on the normal set temperature.
Send the room temperature setting to 5A.

When the resident presses any of the comfort level input keys 11a to 11g corresponding to the blinking of the lamp 11h, the comfort level assigned to the comfort level input keys 11a to 11g is used as learning data as a learning data storage unit. 1
Stored in 2. On the day when the comfort level is input by pressing the comfort level input keys 11a to 11g, learning cannot be done in time as will be described later.
The PM target value is corrected by the V target value setting means 14, and the RAM is rewritten. That is, the change of the comfort level is reflected in the control by changing the PMV target value by fuzzy inference. If the number of times the comfort level is input by the occupant is more than a predetermined number of times within a certain time, it is determined to be abnormal and the comfort level is not accepted.

In the correction of the PMV target value, the value obtained by adding the current PMV target value to the difference between the current PMV target value and the comfort level input value of the occupant is corrected as shown in the following equation (6). PMV
Set the target value.

PMV target value after correction = current PMV target value + (current PMV target value−resident comfort input value) (6) When the PMV target value is corrected, the fuzzy room temperature calculation unit 15
Then, the fuzzy input PMV is newly replaced by (current neuro PMV-corrected PMV target value), and the comfortable air conditioning control for the day is executed.

On the day after the comfort level is input, PMV
The target value is reset to the original PMV target value that saves energy from the start of air conditioning the next morning.

<Learning Control> When the current time is in the learning possible time zone, the learning means 16 executes learning by the neural network. During learning, even if the comfort level is entered, it will not be accepted. A time zone outside the air conditioning time zone (for example, 8:00 to 22:00) and having a low calculation load of the air conditioning monitoring control is determined as a learning possible time zone. If the comfort level is not input even once on that day, it is considered that the thermal environment is satisfied all day and learning is not executed.

The comfort level input key 11a is set by the resident.
If the comfort level input keys 11a to 11g are not pressed for a certain period of time (for example, 30 minutes) from the time when the comfort level input keys 11a to 11g are pressed, "comfort" is set until the next time the comfort level input keys 11a to 11g are pressed. It is considered that the corresponding "neutral" comfort level input key 11d is pressed, and the data necessary for learning during this period are accumulated in the learning data storage unit 12. The comfort level teaching signal at this time is "thermally neutral".
Considering the energy saving effect instead of zero, PMV during heating
= -0.1-0.5, PMV = + 0.1 + during cooling
0.5.

The learning method will be described in detail. Comfort of room temperature, humidity, average radiant temperature, air velocity, clothing state, and activity state within a certain past period accumulated in the learning data storage unit 12 up to that day. The learning process is performed based on the learning data of the sexual element and the learning data of the comfort level of the resident. That is, the learning data of the comfort factor within a certain period in the past is given to the input layer of the neural network shown in FIG. 3, the learning data of the occupant's comfort level is given to the neural network as a teaching signal, and the back propagation method is used. This is performed by updating the weight W of each layer. The backpropagation method is a method in which PMV, which is an output value, is given as a teaching signal of the comfort level of the occupant, back propagation learning is performed from the output layer toward the middle layer, and the weight W is determined. This learning result is reflected in the comfortable air conditioning control of the next day.

Next, the operation of this system will be described with reference to the flowchart of FIG.

First, the comfortable air conditioning control unit 16 determines that the current time is not the learning possible time zone (S1) and the comfortable air conditioning control time zone (S2), and that there is no intervention of the room temperature set value ( S3) If there is no occupant comfort level input (S4), the comfortable air conditioning control is performed, the neuro PMV calculation unit 13 calculates the neuro PMV (S5), and the set PMV target value is used. The fuzzy room temperature calculation unit 15 obtains a room temperature set value (S6).

The processes of steps S2 to S6 are performed every cycle dt.

In step S1, if the learning is possible, the learning means 16 determines the learning data storage unit 1
Learning is performed by a neural network based on the learning data accumulated in 2 (S7).

If it is not in the comfortable air conditioning control time zone in step S2 and if there is an intervention of the room temperature set value in step S3, the comfortable air conditioning control unit 16 stops the comfortable air conditioning control and executes the normal air conditioning control. Perform (S8).

If the comfort level is input in step S4, the PMV target value is corrected, and the corrected PMV is corrected.
The fuzzy room temperature calculation unit 15 calculates a room temperature set value corresponding to the comfort level using the V target value (S9), and the comfort level input by pressing the comfort level input keys 11a to 11g is used as learning data as a learning data storage unit. Accumulate in 12 (S
10).

According to this system as described above, the following effects can be obtained.

(1) A low PMV target value (eg -0.3) in winter and a high PMV target value (eg ++) in summer.
0.3) is set, so rather than operating the air conditioner 3 in the range of -0.5 to +0.5, which is called the thermal neutral range, it operates at a slightly lower temperature during heating and slightly higher during cooling. Energy saving can be achieved because temperature operation is performed and air conditioning load is reduced.

(2) Since "cold", "hot", etc. are displayed on the comfort level input keys 11a to 11g, it is possible to input with the sensory expression of the input person (resident). In addition, the conventional method of expressing a thermal sensation is, for example, simply “cool”, and thus has a drawback that the input person is confused as to whether it is cold and uncomfortable, cool and comfortable. According to the expression method shown in FIG. 2, expressions such as "cold" are made according to the input person's actual feeling, so that the key 11a to 11g corresponding to the actual feeling of the inputting person can be pressed without confusion of the inputting person. You can Further, by assigning comfort levels to the comfort level input keys 11a to 11g, it is possible to limit the comfort level input range and prevent the comfort level from being input at an extreme value. Can handle.

(3) If the occupants are satisfied with the air conditioning of the day, generally, the comfort level is not often input. However, the lamp 11h blinks at regular intervals to give the occupant a comfortable level. Since the input is urged, the comfort level of the occupant is positively input every time the air conditioning conditions change, and as a result,
The learning effect is enhanced, and more comfortable air conditioning conditions can be settled.

(4) If only the comfort level input data from the resident is learned, the unsatisfied "satisfaction state" is not learned. However, according to this system, the occupant enters the comfort level for a certain period after the comfort level is input. When there is no input, it is considered as "comfortable" and the data of the satisfaction state is accumulated and used for learning, so even if the air conditioning state changes due to the comfort level input of the occupant, the automatic data Can be accumulated and learning for various environmental conditions becomes possible.

(5) When the occupant's comfort level is input, the PMV target value is immediately corrected based on the comfort level to obtain the room temperature set value. Room temperature settings can be made.

(6) When the comfort level is input on the previous day, if the PMV target value corrected by the input is left as it is, there is no guarantee that comfort will be saved but energy saving will not be guaranteed. According to the system, since the original PMV target value is reset before the air conditioning time zone on the day after the comfort level is input, comfort and energy saving can be achieved.

(7) When there is a direct intervention of the occupant's room temperature set value, this setting is automatically given priority over the comfortable air conditioning setting. Also, the resident may set the room temperature setting key 11i,
Since the comfortable air conditioning control is switched to the normal air conditioning control by directly setting the room temperature at 11j, it is possible to arbitrarily use the normal air conditioning control and the comfortable air conditioning control.

As another embodiment of the present invention,
There may be more than one comfort level input device in the zone. Thereby, more teaching data for learning can be accumulated. In this case, when a plurality of different comfort levels are input at the same time, there is no problem if arithmetic comfort is performed to capture one comfort level.

[0069]

According to the present invention described in detail above, an energy saving effect can be obtained by using a comfort target target value that can be arbitrarily set. In addition, by inputting the occupant's comfort level, the comfort index target value is immediately corrected based on the comfort level to obtain the room temperature set value, so the real-time room temperature setting that reflects the entered occupant's comfort level is set. Therefore, a comfortable air conditioning environment can be realized while satisfying energy saving.

Further, by displaying the comfort level input time for each fixed cycle, the occupant's comfort level input action can be promoted for each fixed cycle, and the learning effect cannot be improved because the input data of the comfort level is small. Can be avoided, and learning for various environmental conditions can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram of an air conditioning control system showing an application example of an air conditioning control device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an operation panel surface of a comfort level input section in the present apparatus.

FIG. 3 is a diagram for explaining a neural network and learning.

FIG. 4 is a flowchart for explaining the operation of this device.

[Explanation of symbols]

 1 Air-conditioning control device 2 Building room 3 Air-conditioner 5A, 5B DDC 6 Indoor environment measuring unit 10 Set value input unit 11 Comfort level input unit 12 Learning data storage unit 13 Neuro PMV calculation unit 14 PMV target value setting unit 15 Fuzzy room temperature calculation unit 16 Learning means

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Nobutaka Nishimura, Inventor 1-1-1 Shibaura, Minato-ku, Tokyo Inside Toshiba Corporation Head Office

Claims (7)

[Claims] 1. A first calculation means for inputting a plurality of process variables that affect a human thermal sensation and calculating a comfort index from these process variables by a neural network, and the calculated comfort index. Input and calculate the deviation between the comfort index that is currently calculated and the set comfort index target value, and the deviation between the comfort index that is currently calculated and the comfort index that was calculated last time. The second calculation means for calculating the room temperature set value for the air conditioner by fuzzy inference based on the deviation, the input means for inputting the comfort level of the occupant, and the input process variables and the comfort level of the occupant as learning data. Memorizing means for memorizing and using the learning data in a predetermined time period, inputting process variables thereof, and learning with a neural network using the comfort level of the occupant as a teaching signal, Air conditioning control device, characterized in that it comprises a learning means for determining the weight of the first arithmetic means. 2. The air-conditioning control device according to claim 1, further comprising display means for displaying the comfort level input time at regular intervals. 3. When the occupant's comfort level is input by the input means, the comfort index target value for the second computing means is immediately corrected and reset based on the comfort level. The air conditioning control device according to claim 1, further comprising a target value setting means. 4. The comfort level input means, when the comfort level is not input for a certain period after the occupant's comfort level is input, the comfort level corresponding to the comfort level is input until the next comfort level is input. The air conditioning control device according to any one of claims 1 to 3, characterized in that it is considered that a degree input has been made, and this comfortable assumed input is stored in the storage means as learning data. 5. An air conditioning control mode switching means for stopping the control based on the comfort level and outputting an air conditioning control signal based on the room temperature set value when the room temperature set value is changed and inputted. Item 5. An air conditioning control device according to any one of items 1 to 4. 6. The comfort target value setting means sets the comfort index target value modified based on the occupant's comfort level input to the comfort index target value before modification before the next air conditioning time period. The air conditioning control device according to claim 3, wherein the air conditioning control device is corrected. 7. The air conditioning control device according to claim 1, wherein the learning means performs learning outside the air conditioning time period and during a time period when the calculation load of the air conditioning monitoring control is low.