JP2651092B2 - Air conditioner load prediction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load estimating apparatus for estimating a load during cooling and heating of an air conditioner.

[0002]

2. Description of the Related Art In recent years, attempts have been made to predict the heat load of the next day or the day from weather data, temperature and humidity sensors, etc., and to plan the operation plan of the air conditioner or the number of heat sources to be operated based on the prediction results. . Such attempts will become more and more important as energy savings become more important in the future, and the accuracy of their predictions will also be increasingly required.

Under such a background, as a conventional load prediction method, for example, a method disclosed in Japanese Patent Application Laid-Open No. 3-195849 or Japanese Patent Application Laid-Open No. 3-50431 has been proposed. It is possible to obtain the load as a fluctuation amount for each time on the next day or as an integrated amount for one day. In the future, when energy saving is considered, it is necessary to set the air conditioner while managing the energy consumed by the air conditioner. For this purpose, it is essential to predict the load on the air conditioner with respect to the set value for operating the air conditioner. In fact, the load of the air conditioner changes depending on the set value for operating the air conditioner. For example, when the set temperature of the air conditioner is set to 25 ° C. or 28 ° C. during cooling, the latter consumes less energy in the latter,
That is, the load on the air conditioner is small. Therefore, if the load of the air conditioner with respect to the set temperature at which the air conditioner is operated can be predicted, for example, "how much energy can be saved by lowering the set temperature by 1 ° C." can be specifically predicted. It is very effective.

[0004]

However, the conventional air conditioner load predicting apparatus has the following problems.

[0005] (1) Since the set value of the air conditioner, particularly the set temperature, is considered to be a constant value, it is impossible to predict how much load is applied to various set temperatures.

The present invention has been made in view of the above problems, and provides an air conditioner load estimating apparatus having the following objects.

1) An air conditioner load estimating device capable of estimating an air conditioner load with respect to a set temperature of the air conditioner. Further, 2) a load prediction device for an air conditioner in which the prediction accuracy of the load on the air conditioner does not deteriorate even at a set temperature not included in the teacher data.

[0008] 3) A load prediction device for an air conditioner that can obtain good load prediction accuracy even if there is a different internal load for each day of the week.

4) A load prediction device for an air conditioner that can obtain good load prediction accuracy even when a comfortable set temperature varies from season to season.

[0010]

In order to solve the above-mentioned problems, a load predicting apparatus for an air conditioner according to the present invention comprises a group of outdoor sensors for detecting a state of an outdoor environment such as an outside air temperature, an outside air humidity, a solar radiation amount, and the like. An outdoor environment detection unit that detects data at the time of load prediction from the outdoor sensor group, an indoor sensor group that detects an indoor environment such as indoor temperature and indoor humidity, and an indoor environment detection unit that detects data at the time of load prediction from the indoor sensor group. ,
An indoor / outdoor environment memory for storing data detected by the outdoor environment detection means and the indoor environment detection means, a forecast value input means for inputting weather forecast data at the time of load prediction, and a value of the forecast value input means. A predicted value memory to be operated, an air conditioner set value input means for inputting set values such as a set temperature, a set humidity, and an operation time for operating the air conditioner, and each set value of the air conditioner set value input means. An air conditioner set value memory, load detecting means for detecting a load of the air conditioner when the air conditioner is operated with the set value set by the air conditioner set value input means, and an air conditioner detected by the load detecting means Load detection memory for storing the load of, the indoor and outdoor memory, the forecast value memory, the air conditioner set value memory and teacher data creating means for creating teacher data for learning from the load detection memory, After calculating the predicted value of the load of the air conditioner from the values respectively detected by the outside environment detecting means and the indoor environment detecting means, and the values respectively inputted by the forecast value input means and the air conditioner setting input means. It has a configuration of a neural model circuit for outputting and learning based on the teacher data created by the teacher data creating means after the operation of the air conditioner is completed.

[0011]

According to the present invention, the above-described configuration allows the neural network to input outdoor environment data from the outdoor environment detecting means, indoor environment data from the indoor environment detecting means, weather forecast values from the forecast value input means, and the air conditioner at that time. By inputting the set temperature from the air conditioner set value input means, the load of the air conditioner on the set temperature can be predicted. Furthermore, the accuracy of prediction can be improved by detecting the actual measured value of the load of the actual air conditioner by the load detecting means and learning the neural model circuit based on the teacher data created based on the detected value.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First, a first embodiment of the present invention will be described.

The purpose of the air conditioner load prediction device of the first embodiment is to predict the load of the air conditioner with respect to the set temperature of the air conditioner.

Hereinafter, a load prediction device for an air conditioner according to a first embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an air conditioner load predicting apparatus according to a first embodiment of the present invention.
In FIG. 1, 101 is an outdoor temperature sensor, 102 is an outdoor humidity sensor, 103 is a solar radiation sensor, 104 is an indoor temperature sensor, 105 is an indoor humidity sensor, 112 is an outdoor environment detecting means, 114 is an indoor environment detecting means, and 116 is an indoor environment detecting means. Indoor and outdoor environment memory, 118 is a forecast value input means, 120 is a forecast value memory, 122 is an air conditioner set value input means, 124 is an air conditioner set value memory, 140 is a neural model circuit, 150 is an air conditioner, and 126 is load detection Means, 128 is a load detection memory,
138 is a prediction / learning selection circuit, 130 is teacher data candidate memory means, 132 is teacher data set memory means, 1
34 is a data discriminating means, 160, 170, 180, 19
0 is a switch.

The operation of the load prediction device for an air conditioner configured as described above will be described below with reference to FIG. The load prediction device according to the present embodiment includes two operations, a prediction operation and a learning operation. In the prediction operation, the weather data, the indoor and outdoor environment data, and the set value of the air conditioner are input to the neural model circuit 140, and the predicted value of the load of the air conditioner with respect to the set value is calculated. In the learning operation, the neural model circuit 140 learns from the teacher data that stores a combination of each data input at the time of the prediction operation and the actual measured value of the actual air conditioner load for the situation, and a circuit that can perform more accurate prediction. That is.

First, the prediction operation will be described. The outdoor environment detecting means 112 detects the state of the outdoor environment immediately before performing the load prediction from the outdoor sensor 110. In this embodiment, the outdoor temperature, the outdoor humidity, and the amount of solar radiation are selected as typical outdoor environments that affect the load of the air conditioner. The outdoor temperature is measured by an outdoor temperature sensor 101, and the outdoor humidity is measured by an outdoor humidity sensor 10.
In step 2, the solar radiation amount is detected by the solar radiation amount sensor 103, respectively. The indoor environment detecting means 114 detects the state of the indoor environment immediately before performing the load prediction from the indoor sensor 111. In the present embodiment, the indoor temperature and the indoor humidity are selected as typical indoor environments that affect the load of the air conditioner. The indoor temperature is detected by an indoor temperature sensor 104, and the indoor humidity is detected by an indoor humidity sensor 105.

The forecast value input means 118 inputs weather forecast data immediately before performing load prediction. In the present embodiment, as weather forecast data affecting the load on the air conditioner, the highest and lowest forecast temperatures on the forecast day and the weather are selected and input. For example, if the weather is fine,
A value of 0.5 if cloudy and a value of 0 if rainy are input as numerical values in advance.

The air conditioner set value input means 122 inputs the set value of the air conditioner 150 on the day of the prediction immediately before the prediction.
In this embodiment, the set temperature of the air conditioner 150 is input as a set value affecting the load of the air conditioner, so that the load of the air conditioner for each set temperature can be predicted.

When the time for performing the prediction (for example, 7:00 AM) comes, the prediction / learning selection circuit 138 selects the prediction mode, turns on the switch 160, and turns off the switches 180 and 190. .

When the switch 160 is turned on, data detected or input by the outdoor environment detecting means 112, the indoor environment detecting means 114, the forecast value input means 118, and the air conditioner set value input means 122 is input to the neural model circuit 140. You. At the same time, the respective data are stored in the indoor / outdoor environment memory 116, the forecast value memory 120, and the air conditioner set value memory 124. Each data stored at this time is used as teacher data when the neural model circuit 140 learns during the learning operation, which will be described later.

The neural model circuit 140 includes an outdoor environment detecting means 112, an indoor environment detecting means 114, and a forecast value input means 11.
8. When the data detected or input by the air conditioner set value input means 122 is input, a predicted value of the load of the air conditioner on that day is output. Here, if the set temperature input by the air conditioner set value input means 122 is variously changed, a predicted value of the load of the air conditioner when the air conditioner is operated at the set temperature can be obtained. The configuration of the neural network 140 will be described later in detail.

The prediction operation has been described above. Next, the learning operation will be described.

The neural model circuit 140 performs back propagation type learning in order to improve load prediction accuracy. The back propagation method will be described later in the description of the configuration of the neural schematic circuit 140.

The teacher data for learning in the back propagation is obtained after the operation of the air conditioner on the predicted day is completed.
It is created by the teacher data creating means 136. Finally, the air conditioner 150 is operated at the set temperature input by the air conditioner set value input unit 122. The load detecting unit 126 detects a load consumed when the air conditioner 150 operates. Specifically, the amount of electric power or gas consumed by the heat source constituting the air conditioner 150, the amount of electric power consumed by the fan for supplying air, and the like are detected by a meter or the like, and converted into one value as consumed energy. The load detection memory 128 stores the load detection means 12
The energy consumption detected in 6 is accumulated every hour and stored. For example, when the operation time of the air conditioner is from AM 8:00 to PM 6:00, a value obtained by integrating the consumed energy of the air conditioner consumed every hour during this period is predicted as the time load. If it is 10 hours
Keep the data.

When the operation of the air conditioner 150 is completed, a prediction /
The learning selection circuit 138 selects the learning mode, first turns off the switch 160, and then turns on the switch 190. When switch 190 is turned on, indoor / outdoor environment memory 1
16, forecast value memory 120, air conditioner set value memory 12
4. Each data stored in the load detection memory 128 is input to the teacher data candidate memory means 130 and becomes a teacher data candidate.

The data discriminating means 134 monitors the state of the outdoor environment and the state of the indoor environment on the day of the prediction during the operation of the air conditioner, and turns on the switch 170 when it judges that the condition is normal. As a result, the data stored in the teacher data candidate memory means 130 is input to the teacher data set memory means 132, and the neural model circuit 140
Becomes teacher data for learning. If it is determined that the state is abnormal, the switch 170 is turned off, and the data on that day is not input to the teacher data set memory unit 132 and is not added to the teacher data. This is because data such as when there is a sudden change in weather conditions, such as in the evening, or when the indoor temperature is significantly disturbed due to an increase or decrease in heat load due to sudden entry or exit of a room, etc. This is to prevent the prediction accuracy of the neural model circuit 140 from lowering when learning is performed. In such a case, the determination of abnormality or normal can be made based on whether the outside air temperature or the room temperature exceeds a preset upper and lower limit value.

When teacher data is set in the teacher data set memory means 132, the prediction / learning selection circuit 138 turns on the switch 180, and the teacher data is stored in the neural circuit 1
Input to 40 for learning. The learning method is back-propagation type learning, and the method is described in detail in the references described below. The learning operation is performed after the operation of the air conditioner is completed. In the present embodiment, the learning operation is performed at, for example, PM 7:00. However, the learning operation is substantially performed until the next prediction is performed on the next day. No problem. The learning operation has been described above.

By repeating the prediction operation and the learning operation described above, the prediction accuracy of the neural model circuit 140 can be improved.

Next, the neural model circuit 140 will be described. FIG. 5 is a configuration diagram showing a configuration of the neural model circuit 140. In this embodiment, a multilayer perceptron-type neural circuit is used as the neural circuit 140. Regarding the multilayer perceptron type neural circuit, for example, PP 60-66, "Computer Roll No. 24, Special Issue / Neurocomputer", "Basic Neurocomputer, Kaoru Nakano, Corona", or "Introductory Neuro" Computers, Toyohiko Kikuchi, Ohmsha, etc. " In FIG. 5, 501 is a neuron,
520 is a teacher data distribution circuit, and 510 and 511 are switches. The number of input layers is equal to the number of data input through the switch 160, nine, the number of intermediate layers is 20, and the number of output layers is the time during which the air conditioner is operating. , Prepare as many as that number. For example, as in this embodiment, AM 8:00 to PM 6: 0.
When operating the air conditioner to 0, if it is desired to predict a value obtained by integrating the consumed energy of the air conditioner consumed every hour as a time load as described above, a predicted value for 10 hours is required. It becomes ten. The middle layer is 20
Although the number of the output layers is about twice as a guide, there is no problem if the number is not 20.

The neural circuit 14 configured as described above
The operation of 0 will be described. First, at the time of the prediction operation, the switches 510 and 511 are connected to the B side. Switch 5
10, the outdoor temperature from the outdoor environment detecting means 112,
The outdoor humidity, the amount of solar radiation, the indoor temperature, the indoor humidity from the indoor environment detecting means 114, the maximum expected temperature, the minimum expected temperature, the weather, and the air conditioner set value input means 122 from the forecast value input means 118
Is input to the neuron 501 in the input layer. Each input data is calculated by each neuron in the input layer and output to the neuron 501 in the intermediate layer. In the same process, the data that has passed through the neuron 501 in the intermediate layer and the neuron 501 in the output layer is
It passes through the B side of No. 11 and is output as a predicted load value of the air conditioner. Next, at the time of learning operation, switches 510 and 5
11 is connected to the A side. During the learning operation, the teacher data set created by the teacher data creation means 136 is input to the teacher data distribution circuit 520 by turning on the switch 180 as described above. Teacher data distribution circuit 520
Then, from the teacher data set, the group of outdoor temperature, outdoor humidity, solar radiation, indoor temperature, indoor humidity, maximum expected temperature, minimum expected temperature, weather, set temperature, and the set of measured values of air conditioner load were first separated. The former is input to neurons in the input layer. This data passes through the input layer, the intermediate layer, and the output layer, performs the same operation as in the prediction operation, and the operation result is input to the switch 511. Since the switch 511 is connected to the A side,
This calculation result is returned to the output layer again as an error from the actually measured value of the load of the air conditioner distributed by the teacher data distribution circuit 520. In back-propagation type learning, the weight of neurons in each layer is corrected based on this error. The method of the correction is described in detail in the above-mentioned reference. By performing the above learning operation on all the data of the teacher data set, the prediction accuracy of the neural model circuit 140 can be improved. As described above, according to the present embodiment, as one of the inputs at the time of the prediction operation of the neural model circuit 140, the set temperature of the air conditioner for that day is input from the air conditioner set value input unit 122, The predicted value of the load of the air conditioner with respect to the temperature can be calculated.

Further, as teacher data at the time of learning operation,
Since the set temperature of the air conditioner has been input from the air conditioner set value input unit 122, a large amount of actual measurement data of operating the air conditioner in the past at various set temperatures is included in the teacher data. The accuracy of the predicted value of the load at the time of the prediction operation can be further improved.

Next, a second embodiment of the present invention will be described. The purpose of the air conditioner load prediction device of the second embodiment is to improve the prediction accuracy even for a set temperature not included in the teacher data.

Hereinafter, a second embodiment of the present invention will be described with reference to the drawings. FIG. 2 is a block diagram showing a configuration of an air conditioner load prediction device according to a second embodiment of the present invention.

In FIG. 1, reference numeral 101 denotes an outdoor temperature sensor;
102 is an outdoor humidity sensor, 103 is a solar radiation sensor, 10
4 is an indoor temperature sensor, 105 is an indoor humidity sensor, 112
Is an outdoor environment detecting means, 114 is an indoor environment detecting means, 11
6 is an indoor / outdoor environment memory, 118 is a forecast value input means, 12
0 is a forecast value memory, 122 is an air conditioner set value input means, 1
24 is an air conditioner set value memory, 140 is a neural schematic circuit, 1
50 is an air conditioner, 126 is load detecting means, 128 is load detecting memory, 138 is a prediction / learning selection circuit, 130 is teacher data candidate memory means, 132 is teacher data set memory means, 134 is data discriminating means, 160 and 170 , 1
Switches 80 and 190 are the same as those shown in FIG.

The difference from FIG. 1 is that the subtraction circuits 200 and 20
1, 202, and 203, the outdoor temperature from the outdoor environment detection means 112, the indoor temperature from the indoor environment detection means 114, the maximum expected temperature, the minimum expected temperature from the forecast value input means 118, and the air conditioner set value input means 122 The difference is that the temperature difference from the set temperature is input to the neural model circuit 140.

The operation of the air conditioner load predicting apparatus configured as described above will be described below with reference to FIG.
The load estimating apparatus of this embodiment is also composed of two operations, a prediction operation and a learning operation, as in the first embodiment. First
The difference from this embodiment is that, of the data input in the prediction operation and the learning operation, the temperature is input as a temperature difference from the set temperature of the air conditioner.

First, the prediction operation will be described. The operations of the outdoor environment detection means 112, indoor environment detection means 114, forecast value input means 118, and air conditioner set value input means 122 are the same as those in the first embodiment. The difference is that the neural schematic circuit 1
Outdoor environment detecting means 112 of the data input to 40
, The indoor temperature from the indoor environment detecting means 114, and the maximum expected temperature and the minimum expected temperature from the forecast value input means 118 are subtraction circuits 200, 201, and 20, respectively.
2, 203, the temperature difference from the set temperature input by the air conditioner set value input means 122 is input to the neural model circuit 140. Each data is stored in the indoor / outdoor environment memory 116 and the forecast value memory 12.
0, stored in the air conditioner set value memory 124. At this time,
The outdoor temperature, the indoor temperature, the maximum expected temperature, and the minimum expected temperature are stored in the form of a temperature difference from the set temperature of the air conditioner, and are used as teacher data when the neural model circuit 140 learns during the learning operation.

The configuration of the neural model circuit 140 is the same as that of the first embodiment. When the set temperature input by the air conditioner set value input means 122 is variously changed, the air conditioner when the air conditioner is operated at the set temperature is changed. Can be obtained.

The prediction operation has been described above. Next, the learning operation will be described. About learning operation,
Basically, it is the same as the first embodiment. The difference is that the indoor / outdoor environment memory 116, the forecast value memory 1
20, air conditioner set value memory 124, load detection memory 12
8 is that the outdoor temperature, the indoor temperature, the highest expected temperature, and the lowest expected temperature are in the form of a temperature difference from the set temperature of the air conditioner. in this way,
Nerve schematic circuit 1 with data of temperature difference from set temperature of air conditioner
Performing the learning operation of 40 has the effect of improving the accuracy of predicting the load on the air conditioner for a set temperature not included in the teacher data due to the nature of the neural model circuit.

Next, a third embodiment of the present invention will be described. The purpose of the load prediction device for an air conditioner of the third embodiment is to obtain good load prediction accuracy even when there is a different internal load for each day of the week.

Hereinafter, a third embodiment of the present invention will be described with reference to the drawings. FIG. 3 is a block diagram showing a configuration of an air conditioner load prediction device according to a third embodiment of the present invention.

In the figure, 101 is an outdoor temperature sensor,
102 is an outdoor humidity sensor, 103 is a solar radiation sensor, 10
4 is an indoor temperature sensor, 105 is an indoor humidity sensor, 112
Is an outdoor environment detecting means, 114 is an indoor environment detecting means, 11
6 is an indoor / outdoor environment memory, 118 is a forecast value input means, 12
0 is a forecast value memory, 122 is an air conditioner set value input means, 1
24 is an air conditioner set value memory, 150 is an air conditioner, 126 is load detection means, 128 is load detection memory, 138 is a prediction / learning selection circuit, 130 is teacher data candidate memory means,
132 is a teacher data set memory unit, 134 is a data discriminating unit, and 160, 170, 180, and 190 are switches, which have the same configuration as that of FIG.

The difference from FIG. 1 is that the Monday neural circuit 30
0, Tuesday nerve schematic circuit 301, Wednesday nerve schematic circuit 30
2. Thursday neural circuit 303, Friday neural circuit 30
4. Saturday neural schematic circuit 305 (Due to the drawing space,
1 are provided in place of the neural circuit shown in FIG. 1 and are selectively connected to day-of-the-week determination selecting means 310 for determining which neural circuit is to be selected according to the day of the week. Switches 320 and 3
30 is provided.

The operation of the air conditioner load predicting apparatus configured as described above will be described below with reference to FIG.

The load estimating apparatus of this embodiment is also composed of two operations, a prediction operation and a learning operation, as in the first embodiment. The difference from the first embodiment is that a neural schematic circuit is set for each day of the week, and a neural schematic circuit to be used from the day of the prediction day is selected from 300 to 305 by the day of week determination selecting means 310. The point is that 330 is switched.

First, in the prediction operation, the operations of the outdoor environment detecting means 112, indoor environment detecting means 114, forecast value input means 118, and air conditioner set value input means 122 are the same as those in the first embodiment. The day-of-week determination selecting unit 310 determines the day of the week on which the prediction operation is to be performed, and switches the switches 320 and 330. For example, if the day on which the prediction operation is to be performed is Tuesday, the Tuesday neural schematic circuit 301 is selected.

The learning operation is basically the same as in the first embodiment. The difference is that the teacher data creation means 1
That is, the teacher data set created in 36 is sent to the neural schematic circuit switched by the switch 320, and the learning operation is performed.

The effect when the neural circuit is set for each day of the week will be described. In building air-conditioning, the state of work, the number of people entering and leaving the room, etc. change every day, which affects the internal heat load,
As a result, the load on the air conditioner is affected. Therefore,
Detecting the internal heat load directly or indirectly detecting the use status of the workstation and the number of people entering and exiting the room, etc. Invite. However, it is generally difficult to detect such an amount, so that it is difficult to improve the prediction accuracy with only one neural circuit. However, due to its nature, the internal heat load is reproducible for each day of the week. For example, it has been reported that the state of work on Wednesday is very similar every week and is different from Friday. Therefore, the load prediction device for an air conditioner according to the present embodiment has an effect that accurate load prediction can be performed without detecting an internal heat load by having a neural circuit for each day of the week. In other words, if a neural circuit is provided for each day of the week, even if the internal heat load can be detected, the value input to each neural circuit becomes a constant value, so that it becomes practically meaningless to input. Even without inputting, the prediction accuracy of the load does not decrease.

In this embodiment, the operation of the air conditioner is suspended on Sunday, and the Sunday neural circuit is not provided. However, if it is operated, the load can be predicted even on Sunday.

Next, a fourth embodiment of the present invention will be described. The purpose of the load prediction device for an air conditioner of the fourth embodiment is to obtain good load prediction accuracy even when the comfortable set temperature varies from season to season.

Hereinafter, a fourth embodiment of the present invention will be described with reference to the drawings. FIG. 4 is a block diagram showing a configuration of an air conditioner load prediction device according to a fourth embodiment of the present invention.

In the figure, 101 is an outdoor temperature sensor,
102 is an outdoor humidity sensor, 103 is a solar radiation sensor, 10
4 is an indoor temperature sensor, 105 is an indoor humidity sensor, 112
Is an outdoor environment detecting means, 114 is an indoor environment detecting means, 11
6 is an indoor / outdoor environment memory, 118 is a forecast value input means, 12
0 is a forecast value memory, 122 is an air conditioner set value input means, 1
24 is an air conditioner set value memory, 150 is an air conditioner, 126 is load detection means, 128 is load detection memory, 138 is a prediction / learning selection circuit, 130 is teacher data candidate memory means,
132 is a teacher data set memory means, 134 is data discriminating means, 160 is a switch, 160, 170, 180, 190 are switches, 310 is a day of the week determination selecting means, 320 and 330 are switches, and the above is the same as the configuration of FIG. .

The difference from FIG. 3 is that the neural schematic circuits 300, 301, 302, 303,
Instead of 304, 305, summer season, winter season,
A neural circuit is provided for each of three seasons in which a comfortable set temperature range is different from the spring and autumn seasons, and is provided for each day of the week in each season. That is, in the summer season,
Summer seasonal neural network 400, summer season Tuesday neural network 401, summer season Wednesday neural network 40
2. Summer season Thursday neural network 403, Summer season Friday neural network 404, Summer season Saturday neural network 4
05, in winter season, winter nerve Monday schematic circuit 4
10, the winter season Tuesday neural network 411, the winter season Wednesday neural network 412, the winter season Thursday neural network 413, the winter season Friday neural network 414, the winter season Saturday neural network 415, and the spring / autumn season
Fall season Monday neural circuit 420, spring / fall season Tuesday neural circuit 421, spring / autumn season Wednesday neural circuit 422, spring / fall season Thursday neural circuit 423, spring / fall season Friday neural circuit 424, spring / fall season A fall neural network schematic circuit 425 is provided. (Due to the drawing space, 401, 402, 403, 404, 411,
412, 413, 414, 421, 422, 423, 4
24 is omitted) Further, a seasonal judgment selecting means 430 for determining which neural schematic circuit is to be selected according to the season, and switches 440, 450, 451, 452, and 45 selectively connected according to the seasonal judgment selecting means 430.
3, 454, 455, 460, 461, 462, 46
3, 464, 465 (for convenience of drawing space, 451,
452, 453, 454, 461, 462, 463, 4
64 is omitted from FIG. 3.

The operation of the air conditioner load prediction device configured as described above will be described below with reference to FIG.

The load estimating apparatus of this embodiment is also composed of two operations, a prediction operation and a learning operation, as in the third embodiment. The difference from the third embodiment is that a comfortable set temperature zone is roughly divided into three seasons, a neural schematic circuit is set for each season, and the season determination selection means 430 is used to determine whether the season on the day of the prediction is the summer season. The neural circuit to be used is selected from 400 to 405, the neural circuit to be used is selected from 410 to 415 in the winter season, and the neural circuit to be used is selected from 420 to 415 in the spring and autumn seasons. 425, and switches 440, 450-455, and 460-465.

First, in the prediction operation, the outdoor environment detecting means 112, the indoor environment detecting means 114, the forecast value input means 1
18. The operation of the air conditioner set value input means 122 is the same as that of the third embodiment. The season determination selection means 430 determines the season from the set temperature of the air conditioner on the day on which the prediction operation is performed, and switches the switches 440, 450 to 455, and 460 to 465. As a general standard, the correspondence between the set temperature of the air conditioner and the season is as follows: a set temperature of 25 ° C to 28 ° C is a summer season, a set temperature of 18 ° C to 22 ° C is a winter season, and a set temperature is 22 ° C to 25 ° C. Is the spring / autumn season. For example, the day when the prediction operation is performed is Monday, and the temperature set value of the air conditioner is 26 ° C.
Then, the summer neural network schematic circuit 400 is selected.

The learning operation is basically the same as in the third embodiment. The difference is that the teacher data creation means 1
The teacher data set created in 36 is sent to the neural model circuit in each season switched by the switch 440 to perform a learning operation.

As described above, in the present embodiment, the comfortable set temperatures are roughly classified into three seasons, and the neural schematic circuit is set for each season.
The prediction accuracy can be improved more than when prediction is performed for all set temperatures in one season.

Also in this embodiment, the operation of the air conditioner is stopped on Sunday, and the Sunday neural circuit is not provided. However, if it is installed, the load can be predicted even on Sunday.

[0061]

As described above, according to the present invention, the load of the air conditioner with respect to the set temperature of the air conditioner input by the air conditioner set value input means, that is, the amount of energy consumed by the air conditioner is represented by a neural circuit. For example, if the room temperature set value is raised by 1 ° C. during cooling and the operation is performed, it is possible to know in advance with specific numerical values how much energy saving effect can be obtained in one day, and the set temperature before the operation of the air conditioner can be obtained. When deciding
An operation plan can be made while considering energy saving.

Further, in the present invention, since a neural model circuit is used, an object having strong linearity, such as a Kalman filter, an ARMA model, and a multiple regression curve, which are frequently used in the operation method of the conventional load prediction device. Although it is effective when the method is applied to, the problem that sufficient prediction accuracy cannot be obtained when the method is applied to load prediction with strong nonlinearity can also be solved.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an air conditioner load prediction device according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of an air conditioner load prediction device according to a second embodiment of the present invention.

FIG. 3 is a block diagram showing a configuration of an air conditioner load prediction device according to a third embodiment of the present invention.

FIG. 4 is a block diagram showing the configuration of an air conditioner load prediction device according to a fourth embodiment of the present invention.

FIG. 5 is a configuration diagram showing a configuration of a neural schematic circuit of the air conditioner load prediction device according to the first embodiment of the present invention.

[Explanation of symbols]

Reference Signs List 101 outdoor temperature sensor 102 outdoor humidity sensor 103 solar radiation sensor 104 indoor temperature sensor 105 indoor humidity sensor 112 outdoor environment detecting means 114 indoor environment detecting means 116 indoor and outdoor environment memory 118 forecast value input means 120 forecast value memory 122 air conditioner setting value input Means 124 Air conditioner set value memory 140 Neural schematic circuit 150 Air conditioner 126 Load detection means 128 Load detection memory 138 Prediction / learning selection circuit 130 Teacher data candidate memory means 132 Teacher data set memory means 134 Data determination means 160, 170, 180, 190 Switch 200, 201, 202, 203 Subtraction circuit 300-305 Monday neural circuit-Saturday neural circuit 310 Day of week determination selecting means 320, 330 Switch 400-405 Monday seasonal neural circuit -Summer season Saturday neural network 410-415 Winter season Monday neural network-Winter season Saturday neural network 420-425 Spring / fall season Monday neural network-Spring / fall season Saturday neural network 430 Season determination selection means 440, 450-455,460-465 switch

Continuing on the front page (72) Inventor Masataka Iwasaki 6-2-61 Imafukunishi, Joto-ku, Osaka City, Osaka Prefecture Inside Matsushita Seiko Co., Ltd. (72) Inventor Yoshiyuki Morishita 1006 Kazuma Kazuma, Kadoma City, Osaka Denshi Matsushita Industries (56) References JP-A-4-15441 (JP, A) JP-A-3-204536 (JP, A)

Claims (6)

(57) [Claims] 1. An outdoor sensor group for detecting a state of an outdoor environment such as an outdoor air temperature, an outdoor air humidity, and a solar radiation amount; and an outdoor environment detecting means for detecting data from the outdoor sensor group when predicting a load.
An indoor sensor group for detecting an indoor environment such as an indoor temperature and an indoor humidity, an indoor environment detecting unit for detecting data from the indoor sensor group at the time of load prediction, and data detected by the outdoor environment detecting unit and the indoor environment detecting unit. Indoor and outdoor environment memory for storing weather forecast data at the time of load prediction, forecast value input means, forecast value memory for storing the value of the forecast value input means, a set temperature for operating the air conditioner, Air conditioner set value input means for inputting each set value such as set humidity and operation time, air conditioner set value memory for storing each set value of the air conditioner set value input means, and setting with the air conditioner set value input means Load detecting means for detecting a load on the air conditioner when the air conditioner is operated at the set value, a load detecting memory for storing the load on the air conditioner detected by the load detecting means, A teacher data creating unit that creates teacher data for learning from the forecast value memory, the air conditioner set value memory, and the load detection memory, and the outdoor environment detecting unit and the indoor environment detecting unit. And calculating and outputting the predicted value of the load of the air conditioner from the values respectively inputted by the forecast value input means and the air conditioner setting input means, and produced by the teacher data creating means after the operation of the air conditioner is completed. And a neural circuit for performing learning based on the learned teacher data. 2. The load prediction apparatus for an air conditioner according to claim 1, wherein the forecast value input means inputs the highest and lowest forecast temperatures of the forecast day and the weather from the weather forecast data. 3. Before the teacher data creation means starts learning.
From the forecast value memory, the indoor / outdoor environment memory, the air conditioner set value memory, and the load detection memory, the data in the teacher data candidate memory means and the teacher data candidate memory means constituting the teacher data candidates are determined, and abnormalities such as sudden weather change are determined. 2. The air conditioner according to claim 1, wherein the data is removed if the data is normal data, and if the data is normal data, the data is stored in the teacher data set memory means, and is used as teacher data for learning a neural circuit. Load prediction device. 4. An outdoor sensor group for detecting a state of an outdoor environment such as an outdoor air temperature, an outdoor air humidity, and an amount of solar radiation; and an outdoor environment detecting means for detecting data from the outdoor sensor group when predicting a load.
An indoor sensor group for detecting indoor environment conditions such as indoor temperature and indoor humidity, an indoor environment detecting means for detecting data from the indoor sensor group at the time of load prediction, and a forecast value input means for inputting weather forecast data at the time of load prediction Air conditioner set value input means for inputting each set value such as set temperature, set humidity, and operation time for operating the air conditioner; and an air conditioner set value for storing each set value of the air conditioner set value input means. Among the values detected by the memory and the outdoor environment detecting means, the outdoor air temperature is in the form of a difference from the set temperature set by the air conditioner setting input means, and the other values are intact, and Among the values detected by the indoor environment detecting means, the room for storing data in the form of a difference from the set temperature set in the air conditioner setting input means for the indoor temperature, and for the other values in the same form. Outside environment memory, of the values input by the forecast value input means, the maximum and minimum expected temperatures are in the form of a difference between the set room temperature set by the air conditioner setting input means, and the other values are not changed. A forecast value memory for storing data in a form, a load detecting means for detecting a load on the air conditioner when the air conditioner is operated with the set value set by the air conditioner set value input means, and a load detecting means for detecting the load on the load detecting means. Load detection memory for storing the load of the air conditioner, and teacher data creating means for creating teacher data for learning from the indoor / outdoor memory, the forecast value memory, the air conditioner set value memory, and the load detection memory. ,
Of the values detected by the outdoor environment detecting means, the outdoor air temperature is in the form of a difference from the set room temperature set by the air conditioner setting input means, and the other values are intact.
Among the values detected by the indoor environment detecting means, the indoor air temperature is in the form of a difference from the set room temperature set by the air conditioner setting input means, and the other values are intact.
Of the values input by the forecast value input means, the maximum and minimum predicted temperatures are input in the form of a difference from the set room temperature set by the air conditioner setting input means, and the other values are input as they are. A neural model circuit for calculating and outputting the load of the air conditioner from the obtained value, and performing learning based on the teacher data created by the teacher data creating means after the operation of the air conditioner is completed. Air conditioner load prediction device. 5. An outdoor sensor group for detecting a state of an outdoor environment such as an outdoor air temperature, an outdoor air humidity, and a solar radiation amount; and an outdoor environment detecting means for detecting data from the outdoor sensor group when predicting a load.
An indoor sensor group that detects the state of the indoor environment such as indoor temperature and indoor humidity, an indoor environment detecting unit that detects data at the time of load prediction from the indoor sensor group, and an outdoor environment detecting unit and the indoor environment detecting unit that detect the data. Indoor / outdoor environment memory for storing weather forecast data, forecast value input means for inputting weather forecast data at the time of load forecasting, forecast value memory for storing the value of the forecast value input means, and setting for operating the air conditioner. Air conditioner set value input means for inputting each set value such as temperature, set humidity, operation time, etc., an air conditioner set value memory for storing each set value of the air conditioner set value input means, and the air conditioner set value input means A load detection unit that detects a load on the air conditioner when the air conditioner is operated with the set value set in the above, a load detection memory that stores a load of the air conditioner detected by the load detection unit, Teacher data creating means for creating teacher data for learning from the external memory, the forecast value memory, the air conditioner set value memory, and the load detection memory; and Day-of-the-week determination selecting means for selecting one of the neural schematic circuits for the day from among the neural schematic circuits, and the outdoor environment detecting means and the neural network using the neural schematic circuit selected by the day-of-the-week determination selecting means. After calculating the load of the air conditioner from the values respectively detected by the indoor environment detecting means and the values respectively input by the forecast value input means and the air conditioner setting input means, and outputting the calculated value, after the operation of the air conditioner is completed, A load prediction device for an air conditioner, wherein learning is performed based on teacher data created by teacher data creation means. 6. An outdoor sensor group for detecting a state of an outdoor environment such as an outside air temperature, an outdoor air humidity, and a solar radiation amount; and an outdoor environment detecting means for detecting data from the outdoor sensor group when predicting a load.
An indoor sensor group that detects a state of an indoor environment such as indoor temperature and indoor humidity, an indoor environment detecting unit that detects data at the time of load prediction from the indoor sensor group, and an indoor environment detecting unit that is detected by the outdoor environment detecting unit and the indoor environment detecting unit. Indoor / outdoor environment memory for storing weather forecast data, forecast value input means for inputting weather forecast data at the time of load forecasting, forecast value memory for storing the value of the forecast value input means, and setting for operating the air conditioner. Air conditioner set value input means for inputting each set value such as temperature, set humidity, operation time, etc., an air conditioner set value memory for storing each set value of the air conditioner set value input means, and the air conditioner set value input means A load detection unit that detects a load on the air conditioner when the air conditioner is operated with the set value set in the above, a load detection memory that stores a load of the air conditioner detected by the load detection unit, Teacher data creating means for creating teacher data for learning from the internal / external memory, the forecast value memory, the air conditioner set value memory, and the load detection memory, for each day of the week, and further for each season, A neural network circuit provided, a day of the prediction day, a day determination selecting means for determining the day of the week on the day of the prediction, and selecting a neural circuit for the day from a plurality of neural circuit circuits provided for each day, and a season on the day of the prediction. And a seasonal judgment selecting means for selecting a neural schematic circuit for the season from a plurality of neural schematic circuits installed for each season, and the day of the week judgment selecting means and the seasonal judgment selecting means to one. Using the selected neural schematic circuit, values detected by the outdoor environment detecting means and the indoor environment detecting means are respectively input by the forecast value inputting means and the air conditioner setting inputting means. Calculating the load of the air conditioner from the obtained value, outputting the calculated result, and performing learning based on the teacher data created by the teacher data creating means after the operation of the air conditioner is completed. apparatus.