JP2586965B2 - How to control an electric carpet - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of controlling an electric carpet for realizing a comfortable floor temperature by adjusting the temperature of the electric carpet.

[0002]

2. Description of the Related Art In recent years, electric carpets have been widely used in ordinary households as heating equipment. Conventionally, when controlling the floor temperature with such an electric carpet,
In order to improve the characteristics when the floor temperature rises, the electric carpet is forcedly operated at the maximum scale for the target setting scale for a certain time, and after a certain time, the electric carpet is controlled so that the reference temperature of the scale set by the user is maintained. In general, a method of performing the above has been adopted.

FIG. 7 shows an example of an apparatus employing such a conventional electric carpet control method. In FIG. 7, 81 is an operation panel, 82 is a heating element temperature sensor,
83 is a timer, 84 is a temperature controller, and 85 is a heating element. The temperature control unit 84 receives a temperature control signal based on the temperature of the heating element from the heating element temperature sensor 82 after turning on the electric carpet, the energization time from the timer 83, the user's setting scale value from the operation panel 81, and the like. And the heating element 85
On and off. As an example, in order to quickly raise the floor temperature within 30 minutes after the power is turned on, the forced operation is performed with the set scale at the maximum value.

[0004]

However, in the conventional electric carpet control method as described above, since only the energized time after turning on the power is controlled, the environment of the room where the electric carpet is laid is limited. There is a problem that it cannot be flexibly dealt with the setting scale set by the user without being considered. For example, the user's setting scale,
There are problems that the surface temperature of the carpet becomes too high, or conversely, it takes too much time to reach the set value, depending on the level of the room temperature and the difference in flooring.

Also, a value quantifying the comfort level that the user prefers or a physical quantity (hereinafter referred to as comfort information) of the heating state such as the surface temperature of the electric carpet that determines the comfort level is determined. There was another big problem that we couldn't do. The comfort information is determined as comfort information at that time from indoor environmental conditions such as room temperature, use conditions of the electric carpet such as set temperature, and user conditions based on personal factors such as the amount of clothes and metabolism. However, in actual electric carpets, due to manufacturing and structural constraints, for example, a sensor that measures indoor environmental conditions must be placed close to the heating element, and the temperature of the surface that directly touches the user is measured. Providing a separate means has not been realized because it leads to a complicated structure and an increase in cost. Furthermore, it is difficult to measure the amount of clothing and metabolism of humans, and there has been no means for calculating comfort information including individual deviations and differences due to regions and climates.

The present invention solves such a conventional problem, and a control method of an electric carpet that can always realize a comfortable floor temperature in consideration of an indoor environment where the electric carpet is used and individual differences. The purpose is to provide.

[0007]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides an indoor environment condition detected by a sensor and a temperature scale setting of a user input by a user.
Pet size, driving time, month / day time, cumulative power consumption,
At least the heating element's electrical conductivity and carpet thickness
Conditions of use, as well as the user of the electric carpet is also one
Of clothing estimated from knowledge base based on personal factors of clothing
As input and the user condition is metabolic rate, the neural network model
Estimating the surface temperature of the electric carpet by formula means ,
The electric carpet is controlled based on the estimated value.

[0008] The user conditions include a temperature set by the user,
Estimate the amount of clothing and metabolism of the user based on information such as date, time, area, climate, etc., and use the estimated result, sensor output value as indoor environmental condition, past history of sensor, and usage as usage condition. The user's temperature scale setting, carpet size converted to tatami mats, operation time, month / day time (calendar), cumulative power consumption, heat-generating factor, etc. are input to the neural network model, and the neural network is input. The model estimates the comfort information of the person in the room and controls the temperature of the heating element of the electric carpet from the result.

[0009]

Therefore, according to the present invention, the indoor environment conditions which can be detected by the sensor, the user conditions of the electric carpet such as the temperature scale setting of the user, the amount of clothes, etc. The neural network model infers the user's comfort information from the user's condition and controls the temperature of the electric carpet based on the result, so that a comfortable floor temperature can always be realized in consideration of the indoor environment and individual differences. it can.

[0010]

An embodiment of the present invention will be described below.
FIG. 1 is a conceptual diagram of a control method of an electric carpet showing one embodiment of the present invention. In FIG. 1, reference numeral 11 denotes a temperature sensor provided on the electric carpet, which outputs indoor environment conditions 12. Reference numeral 13 denotes an input device for inputting carpet usage conditions 14, local information 15, date information 16, remote control operation information 17, and the like. Reference numeral 18 denotes a knowledge base for estimating user conditions, which inputs area information 15, date information 16, remote control operation information 17, and the like, and outputs estimated user conditions 19. Reference numeral 20 denotes a neural network model, which inputs the indoor environment condition 12, the use condition 14, and the user condition estimated value 19 and outputs a user comfort information estimated value 21. Reference numeral 22 denotes a control rule, which inputs a comfort information estimated value 21 and outputs a temperature control signal 23. Reference numeral 24 denotes an electric carpet main body, and a temperature control signal 23 is input to a heat-resistant controller thereof.

The indoor environment condition 12 is a sensor voltage indicating a temperature of a heating element from a sensor provided on the electric carpet and a room temperature. The use condition 14 includes the temperature scale setting of the user, the size of the carpet converted into the number of tatami mats, the operation time,
The date and time, the accumulated power consumption, the energization rate of the heating element, the thickness of the carpet, and the like. In addition, regional information 15, date information 16,
By inputting the remote control operation information 17 and the like to the knowledge base 18, the user condition estimated value 19 is output as, for example, the amount of clothes and the amount of metabolism. The user condition estimation value 19 may be estimated from a temperature scale value set by the user, or may be estimated by operating a remote control device and indicating an intention on the electric carpet.

Neural network schematic means 2 to which indoor environment condition 12, use condition 14 and user condition estimated value 19 are inputted.
0 is the estimated value 21 of the comfort information of the user of the electric carpet.
Is output. In the control rule 12 to which the comfort information estimated value 21 is input, for example, when the value of “cold” quantifies the user's comfort, a temperature control signal 23 is generated so as to increase the surface temperature of the carpet, When the value of "hot" quantifies comfort, a temperature control signal 23 is generated so as to lower the surface temperature of the carpet, and the heating element controller of the electric carpet body 24 is controlled. When the surface temperature of the electric carpet that is directly transmitted to the user is selected as the physical quantity of the heating state of the electric carpet that determines comfort, the neural network schematic means 20 outputs the input indoor environment condition 12, The surface temperature of the electric carpet that the user feels comfortable is estimated from the use condition 14 and the user condition 19, and similar control is performed based on the estimated value.

FIG. 2 is a conceptual diagram showing a learning method in the neural network model means 20 of FIG. In FIG. 2, 3
1 is the environmental condition of the room where the electric carpet is used, 32
Indicates the use condition of the electric carpet, and 33 indicates the user condition. The indoor environment condition 31 indicates the indoor temperature or the temperature of the heating element of the electric carpet, and the use condition 32 indicates the temperature scale setting of the electric carpet, the number of tatami mats for standard use, and the like. The comfort information 35 of the user 34 in the room under these conditions is quantitatively shown to the user 34. For example, the following expression method is used as a standard of this quantification. -3: cold -2: cool -1: slightly cool 0: good +1: slightly warm +2: warm +3: hot

The user 34 also inputs his / her clothing and metabolism as user conditions 33 to the neural network model 36. The neural network model 36 to which the indoor environment condition 31, the use condition 32, and the user condition 33 have been input outputs the comfort information estimated value 37 to the comparator 38. The output of the neural network model 36 that has not learned at all appears almost randomly, but when the output comfort information estimated value 37 and the comfort information 35 of the user 34 are compared by the comparator 38,
The comparison error 39 is fed back to the neural network model 36,
Learning is repeated so that the state inside the neural network model 36 is adapted to the comfort information 35 of the user 34.

[0015] The fixed knowledge representation that estimates the standard environment desired by the user collects data on how people control the comfort in various environments, and combines environmental data with human control data. Can be obtained by learning the correlation in a neural network model. Reference 1 (PDP model,
D. E. FIG. Ramelhart and 2 others, translated by Shunichi Amari, 1989
), Reference 2 (Basic Neurocomputer, Kaoru Nakano et al., 1990, 1990), and JP-B-63-55106.

Hereinafter, the structure and operation of a specific neural network will be described by taking a multilayer perceptron using an error back propagation method as an example of the most well-known learning algorithm described in Document 1.

FIG. 3 is a conceptual diagram of a neural network unit which is a structural unit of a neural network model. In FIG. 3, 41
Numerals -1 to 41-N denote pseudo-synapse converters that simulate synaptic connections of nerves, and reference numeral 42 denotes a pseudo-synapse converter 41-N.
An adder for adding the outputs from 1 to 41-N;
Is a set nonlinear function, for example, a sigmoid function with a threshold value h, f (y, h) = 1 / (1 + exp (−y + h)) (3) It is a converter. Although an illustration is omitted because the drawing becomes complicated, an input line for receiving a correction signal from a correction means (not shown) not shown is connected to the output side of the nonlinear converter 43 and the pseudo-synapse converters 41-1 to 41-N. Connected to the input side. This neural network unit has two types of operation modes, a signal processing mode and a learning mode.

The operation of each mode of the neural network will now be described with reference to FIG. First, the operation in the signal processing mode will be described. The neural network unit is N
It receives _one input X _{1 to} X _N and outputs one output. i-th input signal X _i is converted in the pseudo synaptic converters 41-1 to 41-N shown by squares in W _i X _i. The N signals W ₁ X _{1 to} W _N X _N converted by the pseudo-synapse converters 41-1 to 41- _N are added to adders 42 indicated by circles.
And the addition result y is a non-linear converter 43 indicated by an ellipse.
To be the final output f (y, h).

Next, the operation in the learning mode will be described.
In the learning mode, the pseudo synapse converters 41-1 to 41-
The transformation parameters W ₁ to _W-N with the threshold h of N non-linear converter 43 receives the correction signal representing a correction amount ΔW ₁ ~ΔW _N and Δh of transformation parameters from the correction means (not shown), W _i + ΔW _i : i = 1, 2,..., N h + Δh (Equation 4)

FIG. 4 is a conceptual diagram of a signal conversion means in which four neural network units are connected in parallel. Needless to say, the number of neural network units constituting the signal conversion means is not limited to four.

In FIG. 4, reference numerals 511 to 544 denote pseudo-synapse converters, and reference numerals 51 to 54 denote add-nonlinear converters in which the adder 42 and the non-linear converter 43 are combined.
In FIG. 4, similarly to FIG. 3, an input line receiving a correction signal from a correction means (not shown) is connected to an addition nonlinear converter 5.
And the pseudo synapse converters 511 to 544
Connected to the input side.

The operation of the signal conversion means is such that the same input is given to the four neural network units, and the operations of the neural network units described in FIG. 3 are performed in parallel.

FIG. 5 is a block diagram showing the configuration of the signal processing means when the signal conversion means of FIG. 4 is used and the back propagation method is adopted as a learning algorithm. In FIG. 5, reference numeral 61 denotes the signal conversion means described with reference to FIG. 4. Here, it is assumed that M neural network models receiving N inputs are arranged in parallel. Reference numeral 62 denotes a correction unit that calculates a correction amount of the signal conversion unit 61 in the learning mode.

The operation of learning the signal processing means will be described below with reference to FIG. Signal conversion means 61
Receives N input signals Sin (X) and outputs M output signals Sout (X). The correction means 62 outputs the input signal S
receiving the in (X) and the output signal Sout (X), an error calculating means (not shown) or a signal converting means (not shown) at the subsequent stage
Wait until there are M error signals δ _i (X) input from. When the error signal δ _i (X) is input, the correction amount is calculated as _follows : ΔW _ij = δ _i (X) · S _i out (X) · (1−S _i out (X)) · S _j in (X) (I = 1 to N, j = 1 to M) (Equation 5) and send the correction signal to the signal conversion means 61. The signal conversion means 61 corrects the conversion parameters of the internal neural network model according to the operation of the learning mode described above.

FIG. 6 is a block diagram showing the configuration of a multilayer perceptron in which the signal processing means shown in FIG. 5 is stacked in three stages. In FIG. 6, 71X, 71Y and 71Z are three stages of signal processing means, and 72X, 72Y and 72Z are K
, L and M signal conversion means consisting of a neural network model, 73X, 73Y and 73Z are correction means, and 74 is an error calculation means.

The operation of the multilayer perceptron configured as described above will be described with reference to FIG.
In the signal processing unit 71X, the signal conversion unit 72X
Input S _i in (X) (i = 1,..., N) is received and output S _j o
ut (X) (j = 1,..., K) is output. Correction means 7
3X receives the signal S _i in (X) and the signal S _j out (X), and waits until an error signal δ _j (X) (j = 1,..., K) is input. Hereinafter, the same processing is performed by the signal processing unit 71.
Y, 71Z, and a final output S _k out (Z) (k = 1,..., M) is output from the signal conversion means 72Z. The final output S _k out (Z) is also sent to the error calculation means 74.

In the error calculating means 74, the following equation (number
Ideal based on evaluation function such as square error using 1)
Is calculated from the output T (T1,..., TM).
Signal δ _k(Z) is sent to the correction means 73Z. In addition,
Μ in (Equation 1) is the learning speed of the multilayer perceptron.
It is a parameter to be determined.

[0028]

(Equation 1)

When the evaluation function is a square error, the error signal is as follows: δ _k (Z) = − μ · (S _k out (Z) −T _k ) (Equation 6) The correction means 73Z calculates the correction amount ΔW of the conversion parameter of the signal conversion means 72Z according to the above-described procedure.
_k (Z) is calculated, and an error δ _j (Y) signal to be sent to the correcting means 73Y is calculated based on the following equation (Equation 2), and a corrected signal ΔW _k (Z) is sent to the signal converting means 72Z. The signal δ _j (Y) is sent to the correction means 73Y. Where W _ij
(Z) is a conversion parameter of the pseudo synapse conversion unit of the signal conversion unit 72Z.

[0030]

(Equation 2)

The signal conversion means 72Z outputs the correction signal ΔW
Modify the internal conversion parameters based on _k (Z). Hereinafter, similar processing is performed in the signal processing means 71Y and 71X.

By repeatedly performing the above-mentioned procedure, also called learning, the multilayer perceptron, when given an input, produces an output that closely approximates the ideal output T.

In this way, the neural network model learns the relationship between environmental data and human control data, and describes how to control comfort information desired by a user who cannot easily be described by simple rules. Be able to express in a natural way. Then, learning is performed until the comfort information desired by the user can be estimated by the neural network model means 20, and when the learning is completed, the comfort information estimated value 21 output from the neural network model means 20, Control rule 22
However, when the current surface temperature of the electric carpet is lower than the temperature of the setting scale set by the user, the temperature control signal 23 is generated in a direction to warm the electric carpet, and conversely, the surface temperature is set by the user. When the temperature is higher than the set graduation temperature, the temperature control signal 23 is generated in the direction opposite to the warming direction.

As described above, according to the present embodiment, the user condition is estimated from the date, the area, and the remote control operation information, and the result is compared with the indoor environment condition such as room temperature and the use condition of the electric carpet such as the set temperature. Is input to the neural network schematic means, the user's comfort information is estimated by learning of the neural network schematic means, and a temperature control signal is generated based on the result to control the electric carpet, whereby the indoor environment and Comfortable control can always be performed in consideration of individual differences.

[0035]

As will be apparent from the above embodiment, the present invention can be applied to the indoor environment conditions detected by the sensor, the user's temperature scale input by the user, and the size of the carpet.
Size, operation time, date and time, cumulative power consumption,
At least one of electrical conductivity and carpet thickness
Personal cornerstone of use conditions, as well as the user of the electric carpet that
Of clothing and metabolism estimated from elementary knowledge base
A certain user condition is input and the neural network
Since the surface temperature of the electric carpet is estimated and the electric carpet is controlled based on the estimated value,
A comfortable floor temperature can always be realized in consideration of the indoor environment and individual differences.

[Brief description of the drawings]

FIG. 1 is a conceptual diagram of an electric carpet control method according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram of a learning method in the embodiment.

FIG. 3 is a conceptual diagram of a neural network unit in the embodiment.

FIG. 4 is a conceptual diagram of a signal conversion unit in which four neural network units are connected in parallel;

FIG. 5 is a schematic block diagram showing a configuration of a signal processing unit when an error back propagation method is adopted as a learning algorithm using the signal conversion unit;

FIG. 6 is a schematic block diagram showing a configuration of a multilayer perceptron using the signal processing means.

FIG. 7 is a schematic block diagram showing an example of a conventional electric carpet control device.

[Explanation of symbols]

 Reference Signs List 11 Room temperature sensor 12 Indoor environment condition 13 Input device 14 Usage condition 15 Regional information 16 Month and day information 17 Remote control operation information 18 User condition estimation knowledge base 19 User condition estimation value 20 Neural network model means 21 Comfort information estimation value 22 Condition Rule 23 Temperature Control Signal 24 Electric Carpet Body

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Kazunari Nishii 1006 Kadoma Kadoma, Osaka Prefecture Inside Matsushita Electric Industrial Co., Ltd. (72) Inventor Takashi Iwasa 1006 Kadoma Kadoma, Osaka Prefecture Matsushita Electric (72) Kazutoshi Nagai, Inventor 1006 Kadoma, Kadoma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. (56) References JP-A-54-47352 (JP, A) JP-A 64-64 41187 (JP, A) JP-A-64-28424 (JP, A) JP-A-59-47992 (JP, U)

Claims (4)

(57) [Claims] 1. An indoor environment condition detected by a sensor and a temperature scale setting of a user input by a user,
Pet size, driving time, month / day time, cumulative power consumption,
At least the heating element's electrical conductivity and carpet thickness
Conditions of use, as well as the user of the electric carpet is also one
Of clothing estimated from knowledge base based on personal factors of clothing
As input and the user condition is metabolic rate, the neural network model
Estimating the surface temperature of the electric carpet by formula means ,
A method for controlling an electric carpet, comprising controlling the electric carpet based on the estimated value. 2. The terms of use are based on month, day, time and area.
The method for controlling an electric carpet according to claim 1 , further comprising a monthly average temperature obtained by measurement . 3. The user condition is a user's set temperature scale value.
The method for controlling an electric carpet according to claim 1 or 2, wherein the amount of clothing and the amount of metabolism are estimated from the amount . 4. An electric carpet , wherein user conditions are indicated on the electric carpet by a remote controller.
4. An estimated value of the surface temperature of the target is obtained.
Method of controlling the electric carpet.