KR20210025920A - System and Method for Predicting Indoor Surface Temperature and Indoor Condensation Using Machine Learning - Google Patents

Abstract

Provided are a condensation prediction system for predicting the indoor surface temperature and condensation using machine learning and a method thereof. Machine learning for the surface temperature of the indoors where condensation will occur according to the temperature change between the indoor and outdoor temperatures may be performed. Accordingly, the surface temperature of the inner surface of the window with respect to the indoor temperature and the outdoor temperature may be estimated. The present invention monitors the change pattern over time of the dew point temperature and the estimate of the surface temperature of the inner surface of the window to determine the dew condensation, so as to prevent condensation in advance through pre-ventilation work.

Description

System and Method for Predicting Indoor Surface Temperature and Indoor Condensation Using Machine Learning}

The present invention relates to a condensation prediction system and method, in particular, by machine learning the surface temperature of an indoor room where condensation will occur according to a temperature change between the indoor temperature and the outdoor temperature, and accordingly, the surface temperature of the inner surface of the window with respect to the indoor temperature and the outdoor temperature. The present invention relates to a condensation prediction system and method that estimates indoor surface temperature and condensation using machine learning.

Condensation refers to a kind of wet state caused by water vapor in the air.

Condensation in indoor buildings has persisted as one of the major problems in the field of architecture for the past several decades, and various efforts have been made to minimize the damage caused by condensation.

In indoor buildings, when condensation occurs, water droplets form on the surface of impermeable materials, and moisture in the absorbent material causes damage and unpleasant odor due to the propagation of molds and various fungi. It will hurt.

Fungi are filamentous fungi composed of hyphae and can cause respiratory diseases when inhaled by the human body, and a problem of deteriorating the living environment may occur.

The solution to condensation is to perform construction by considering the materials that are strong against condensation in the interior and exterior materials of the building.

However, condensation in indoor buildings is fundamentally a solution to the condensation problem because condensation occurs due to changes in indoor humidity due to the temperature difference between indoor and outdoor throughout the occupancy and life even if architectural design is performed using the optimal materials and structures. This is not being possible.

The reason why the condensation problem is difficult is that the materials used for each building are different, the size of the structure is different, and the living patterns of the residents are different, so it is difficult to find a comprehensive solution for this.

Korean Patent Registration No. 10-1742059

In order to solve such a problem, the present invention machine learns the indoor surface temperature where condensation will occur according to the temperature change of the indoor temperature and the outdoor temperature, and accordingly estimates the surface temperature of the inner surface of the window with respect to the indoor temperature and the outdoor temperature. Thus, an object of the present invention is to provide a condensation prediction system and method for predicting indoor surface temperature and condensation using machine learning.

According to the features of the present invention to achieve the above object

The condensation prediction system that predicts the occurrence of indoor condensation,

A first sensor node attached to the inner surface of the window where condensation occurs;

A second sensor node attached to one side of the room; And

Receives the measured surface temperature of the inner side of the window measured from the first sensor node through wireless communication, receives indoor temperature and humidity information from the second sensor node through wireless communication, and receives outdoor temperature from an external weather information server And estimating a surface temperature of an inner surface of a window with respect to the indoor temperature and the outdoor temperature by machine learning a correlation between the indoor temperature and the measured surface temperature according to the outdoor temperature for each time period.

,

,

)의 최적값을 찾기 위해서 상기 실내 온도와 상기 실외 온도에 따른 상기 측정 표면 온도의 상관 관계를 기계 학습하고, 상기 기계 학습으로 얻어진 선형 계수값(

,

,

)을 하기의 수학식 1에 대입하여 상기 창문 내측면의 표면 온도 추정치를 계산하는 기계 학습부를 더 포함하는 것을 특징으로 한다.In the client server, the client server uses a linear regression model of Equation 1 below, and a linear coefficient value (

,

,

In order to find the optimum value of ), the correlation between the indoor temperature and the measured surface temperature according to the outdoor temperature is machine-learned, and a linear coefficient value obtained by the machine learning (

,

,

) Is substituted into Equation 1 below to further include a machine learning unit that calculates an estimate of the surface temperature of the inner surface of the window.

[Equation 1]

은 시간(i)에 따른 창문 내측면의 표면 온도 추정치,

은 시간(i)에 따른 실내 온도,

은 시간(i)에 따른 실외 온도임.here,

Is the estimate of the surface temperature of the inner side of the window over time (i),

Is the room temperature over time (i),

Is the outdoor temperature over time (i).

A method of predicting indoor surface temperature and condensation using machine learning according to a feature of the present invention,

The cloud server receives the measured surface temperature of the inner side of the window measured from the first sensor node through wireless communication, receives the indoor temperature and humidity information from the second sensor node attached to one side of the room through wireless communication, and external weather information Receiving an outdoor temperature from a server;

,

,

)의 최적값을 찾기 위해서 상기 실내 온도와 상기 실외 온도에 따른 상기 측정 표면 온도의 상관 관계를 기계 학습하는 단계; 및The cloud server uses a linear regression model of Equation 2 below, which is a parameter of Equation 4 below (

,

,

Machine learning a correlation between the indoor temperature and the measured surface temperature according to the outdoor temperature in order to find an optimum value of ); And

,

,

)을 하기의 수학식 4에 대입하여 상기 창문 내측면의 표면 온도 추정치를 계산하는 단계를 포함하는 것을 특징으로 하는 결로 예측 방법.The cloud server is a linear coefficient value obtained by the machine learning (

,

,

And calculating an estimate of the surface temperature of the inner side of the window by substituting) into Equation 4 below.

[Equation 4]

은 시간(i)에 따른 창문 내측면의 표면 온도 추정치,

은 시간(i)에 따른 실내 온도,

은 시간(i)에 따른 실외 온도임.here,

Is the estimate of the surface temperature of the inner side of the window over time (i),

Is the room temperature over time (i),

Is the outdoor temperature over time (i).

According to the above-described configuration, the present invention is an effect of machine learning the indoor surface temperature where condensation will occur according to the temperature change of the indoor temperature and the outdoor temperature, and accordingly, estimating the surface temperature of the inner surface of the window with respect to the indoor temperature and the outdoor temperature There is.

The present invention has an effect of preventing condensation in advance through a pre-ventilation operation by determining condensation by monitoring a pattern of changes over time in the dew point temperature and the estimated surface temperature of the window inner surface.

1 is a diagram showing the overall configuration of a condensation prediction system for predicting the occurrence of indoor condensation according to an embodiment of the present invention.
2 is a block diagram schematically showing an internal configuration of a first sensor node, a second sensor node, and a condensation prediction system according to an embodiment of the present invention.
FIG. 3 is a diagram comparing the measured surface temperature (Tsurf-real) of the inner side of the window measured from the first sensor node according to the present invention and the estimated surface temperature of the inner side of the window (Tsurf-pred) for 6 days. .

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included rather than excluding other components unless specifically stated to the contrary.

In the case of existing buildings, it is difficult to consider the details of the interior and exterior materials of the building, and there is a phenomenon in which the pattern of condensation in the same house varies depending on the user's life pattern.

The present invention collects the environment (indoor temperature, indoor humidity, outdoor temperature, surface temperature on the inner side of the window) that affects condensation in the house where you live in real time through a wireless sensor node, and based on the collected data, the window A surface temperature prediction model that predicts the surface temperature of the inner surface is generated, and a system that predicts the occurrence time of condensation on the inner surface of a window using a surface temperature prediction model and a change in dew point temperature is provided.

When the temperature of the air in a space or place is higher than the temperature of a specific object, dew condenses on the surface of the object. At this time, the temperature is called the dew point, and condensation occurs at the dew point temperature.

1 is a diagram showing the overall configuration of a condensation prediction system for predicting the occurrence of indoor condensation according to an embodiment of the present invention, and FIG. 2 is a first sensor node, a second sensor node, and a condensation prediction system according to an embodiment of the present invention. It is a block diagram briefly showing the internal configuration of.

The condensation prediction system 100 according to an embodiment of the present invention includes a first sensor node 110, a second sensor node 120, and a cloud server 140. In addition, a weather application programming interface (API) server 150 that provides weather information to the cloud server 140 is further included.

The first sensor node 110 includes a first sensor unit 111, a first control unit 112, and a first wireless communication module 113. It can be attached to the inner side or one side of the wall.

It will be described that the first sensor node 110 of the present invention is installed on the inner side of the window for convenience of description.

The first sensor unit 111 measures first sensing information including surface temperature information and humidity information on the inner side of the window in the home 10.

The first control unit 112 generates a sensing control signal and transmits it to the first sensor unit 111, receives first sensing information from the first sensor unit 111 in response to the sensing control signal, and receives the received The first sensing information is transmitted to the cloud server 140 through the communication network 130 through the first wireless communication module 113.

The second sensor node 120 includes a second sensor unit 121, a second control unit 122, and a second wireless communication module 123, and may be installed at one side of the interior of the home 10.

The second sensor unit 121 measures second sensing information including room temperature information and humidity information.

The second control unit 122 generates a sensing control signal and transmits it to the second sensor unit 121, receives second sensing information from the second sensor unit 121 in response to the sensing control signal, and receives the received The second sensing information is transmitted to the cloud server 140 through the communication network 130 through the second wireless communication module 123.

The weather API server 150 is a server that generates and stores outdoor temperature information by location, time, and date of each region.

The cloud server 140 according to an embodiment of the present invention includes a server control unit 141, a condensation prediction unit 142, a temperature and humidity information database unit 143, a display unit 144, a wireless communication unit 145, a machine learning unit ( 146), a memory unit 147, and a storage unit 148.

The cloud server 140 is illustrated as being installed outdoors, but is not limited thereto, and may be installed indoors as well.

The server control unit 141 periodically receives outdoor temperature information from the weather API server 150 and stores it in the temperature/humidity information database unit 143 by time and by date.

The server control unit 141 receives first sensing information including surface temperature information and humidity information on the inner side of the window from the first sensor node 110, and receives indoor temperature information and humidity in the indoor space from the second sensor node 120. Second sensing information including information is received and stored in the temperature/humidity information database unit 143 by time and date.

In general, in architecture, the thermal conductivity of the structure is calculated in advance based on the material, thickness, and thermal resistance of the structure, and based on this, the optimal structure is used to design and construct a building. However, the most powerful factors in the occurrence of condensation in a living house are the temperature difference between indoors and outdoors and an increase in indoor humidity. This depends on the lifestyle pattern you live in.

This condensation phenomenon brings great difficulty in preventing condensation simply based on the characteristics of the structure. Accordingly, the present invention can estimate the surface temperature of the inner side of the window by using indoor temperature information and outdoor temperature information in an indoor space where condensation is expected to occur regardless of the type of structure.

The machine learning method to implement the surface temperature prediction model applies a linear regression model, a random forest, and a deep learning method alone or in combination of one or more. The machine learning method of the present invention applies linear regression analysis as an example.

Linear regression analysis is a statistical method that explains or predicts the relationship between the independent variable and the dependent variable while the basic assumption of linearity is satisfied.In regression analysis, the value of the dependent variable changes in a certain pattern according to the independent variable. It means a case where the regression line representing the relationship appears close to a straight line.

In order to create a surface temperature prediction model, the indoor temperature, the outdoor temperature, and the measured surface temperature of the inner side of the window stored in the temperature/humidity information database unit 143 may be used as a learning data set for machine learning.

In the machine learning process, feature vectors are extracted from many data stored in the past, a learning data set is created based on the extracted feature vectors, and the learning data set generates a surface temperature prediction model based on a machine learning algorithm. Learning data is a set of data used to extract desired information in machine learning.

The machine learning unit 146 may perform machine learning using linear regression analysis. The machine learning model may be a well-known deep convolutional neural network, but is not limited thereto.

The machine learning unit 146 represents a linear regression model of a machine learning method for estimating the surface temperature of the inner surface of the window as shown in [Equation 1] below.

은 시간(i)에 따른 창문 내측면의 표면 온도 추정치,

은 시간(i)에 따른 실내 온도,

은 시간(i)에 따른 실외 온도를 나타낸다.here,

Is the estimate of the surface temperature of the inner side of the window over time (i),

Is the room temperature over time (i),

Represents the outdoor temperature over time (i).

The machine learning unit 146 may load the linear regression model stored in the storage unit 148 into the memory unit 147 to estimate the surface temperature of the inner side of the window using the indoor temperature and the outdoor temperature as input data.

The machine learning unit 146 machine learns the correlation between the indoor temperature for each time period and the measured surface temperature of the inner side of the window according to the outdoor temperature.

,

,

)의 최적값을 찾기 위해서 실내 온도와 실외 온도에 따른 창문 내측면의 측정 표면 온도를 기계 학습한다.In other words, the machine learning unit 146 receives the indoor temperature, the outdoor temperature, and the measured surface temperature of the inner side of the window from the temperature/humidity information database unit 143, and a linear coefficient value (

,

,

In order to find the optimum value of ), machine learning is performed on the measured surface temperature of the inner side of the window according to the indoor temperature and the outdoor temperature.

,

,

)을 수학식 1에 대입하여 창문 내측면의 표면 온도 추정치를 계산한다.The machine learning unit 146 is a linear coefficient value obtained by machine learning (

,

,

) Is substituted into Equation 1 to calculate an estimate of the surface temperature of the inner side of the window.

,

,

)을 계산하여 나타낸다.As shown in Table 1, the machine learning unit 146 calculates a root mean square error (RMSE) of each training data and test data in the temperature and humidity information database unit 143 using a linear regression model. And a linear coefficient value obtained by learning with different learning data from the data set received from the temperature/humidity information database unit 143 (

,

,

) Is calculated and expressed.

As can be seen from the results in Table 1, the linear regression model with 70% training data shows the best performance for both the training data and the test data.

Table 1 shows that the linear coefficient value and performance (ie, RMSE) change as the number of training data changes.

The machine learning unit 146 performs optimization to minimize the root mean square error (RMSE), and when the root mean square error reaches a minimum value, selects it as an estimate of the surface temperature of the inner side of the window.

FIG. 3 is a comparison of the measured surface temperature (Tsurf-real) of the inner side of the window measured from the first sensor node 110 and the estimated surface temperature of the inner side of the window (Tsurf-pred) for 6 days. As a result of the comparison, it was confirmed that the linear regression model learned through the data set of the temperature and humidity information database unit 143 can find a surface temperature estimate (Tsurf-pred) close to the measured surface temperature (Tsurf-real).

As shown in FIG. 3, the server control unit 141 calculates the measured surface temperature (Tsurf-real) of the inner side of the window measured by the first sensor node 110 and the estimated surface temperature of the inner side of the window (Tsurf-pred). Calculate the difference, and calculate the average error and the maximum error.

,

,

)을 선택한다.The machine learning unit 146 analyzes the difference value between the measured surface temperature of the inner surface of the window and the estimated surface temperature of the inner surface of the window over time, and when the difference value is less than or equal to a preset reference value, a linear coefficient value (

,

,

Select ).

The server control unit 141 determines the change over time of the measured surface temperature (Tsurf-real) of the inner surface of the window measured by the first sensor node 110 and the estimated surface temperature (Tsurf-pred) of the inner surface of the window measured by the first sensor node 110 over time, Each graph is generated for each date and output to the display unit 144, and the calculated average error and maximum error are output to the display unit 144 in text form.

The server control unit 141 calculates the dew point temperature by using the Barenbrug Formula of [Equation 2] and [Equation 3] below to estimate the surface temperature of the inner surface of the window and the indoor humidity of the indoor space. Here, the indoor humidity of the indoor space can be obtained by searching the temperature/humidity information database unit 143.

Here, a is 17.27°C and b is 237.3°C. T is the estimated surface temperature of the inner side of the window, 0°C <T <60°C, RH is the indoor humidity of the indoor space measured by the second sensor node 120, 1% <RH <100%, Td is the dew point temperature. .

,

,

)을 계산하여 시간에 따른 창문 내측면의 표면 온도를 추정할 수 있다. 이후에는 창문 내측면에 설치된 제1 센서노드(110)를 제거하고, 제2 센서노드(120)만을 설치하여 주기적으로 실내 온도와 실외 온도를 이용하여 창문 내측면의 표면 온도를 추정할 수 있다.The machine learning unit 146 is a linear coefficient value of Equation 1 (

,

,

) Can be calculated to estimate the surface temperature of the inner side of the window over time. Thereafter, the first sensor node 110 installed on the inner side of the window is removed, and only the second sensor node 120 is installed, so that the surface temperature of the inner side of the window may be estimated using the indoor temperature and the outdoor temperature periodically.

According to the present invention, if the surface temperature of the inner surface of the window is estimated by using a linear coefficient value, then indoor condensation can be predicted using only the indoor temperature and the outdoor temperature.

The server controller 140 periodically receives the indoor temperature from the second sensor node 120 by time and date, and periodically receives the outdoor temperature from the weather API server 150 by time and date.

The server controller 140 estimates the surface temperature of the inner side of the window over time by substituting the indoor temperature and the outdoor temperature into Equation 1 described above.

The condensation prediction unit 142 may estimate the occurrence of indoor condensation by using the estimated surface temperature of the inner side of the window and the dew point temperature. Here, for convenience of explanation, the indoor condensation exemplifies the occurrence of condensation on the inner side of the window, but is not limited thereto, and may represent condensation occurring in various locations of the room such as an indoor wall surface.

The condensation prediction unit 142 monitors an estimate of the surface temperature of the inner surface of the window over time and a pattern that changes with respect to the dew point temperature over time, and calculates the difference between the dew point temperature and the estimated surface temperature of the inner surface of the window to determine the condensation. Can be judged.

In the above, embodiments of the present invention are not implemented only through an apparatus and/or a method, but may be implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention, a recording medium in which the program is recorded, or the like. And, this implementation can be easily implemented by an expert in the technical field to which the present invention belongs from the description of the above-described embodiment.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

100: condensation prediction system 110: first sensor node
111: first sensor unit 112: first control unit
113: first wireless communication module 120: second sensor node
121: second sensor unit 122: second control unit
123: second wireless communication module 130: communication network
140: cloud server 141: server control unit
142: condensation prediction unit 143: temperature and humidity information database unit
144: display unit 145: wireless communication unit
146: machine learning unit 147: memory unit
148: storage unit 150: weather API server

Claims (8)

A first sensor node attached to the inner surface of the window where condensation occurs;
A second sensor node attached to one side of the room; And
Receives the measured surface temperature of the inner side of the window measured from the first sensor node through wireless communication, receives indoor temperature and humidity information from the second sensor node through wireless communication, and receives outdoor temperature from an external weather information server And a client server estimating a surface temperature of an inner surface of a window with respect to the indoor temperature and the outdoor temperature by machine learning a correlation between the indoor temperature and the measured surface temperature according to the outdoor temperature for each time period. Condensation prediction system. The method of claim 1,
The client server uses a linear regression model of Equation 1 below, and uses a linear coefficient value, which is a parameter of Equation 1 below.

,

,

In order to find the optimum value of ), the correlation between the indoor temperature and the measured surface temperature according to the outdoor temperature is machine-learned, and a linear coefficient value obtained by the machine learning (

,

,

And a machine learning unit that calculates an estimate of the surface temperature of the inner surface of the window by substituting) into Equation 1 below.
[Equation 1]

here,

Is the estimate of the surface temperature of the inner side of the window over time (i),

Is the room temperature over time (i),

Is the outdoor temperature over time (i). The method of claim 2,
The machine learning unit determines the time-dependent variation of the measured surface temperature (Tsurf-real) of the inner side of the window measured by the first sensor node and the calculated Tsurf-pred of the inner side of the window. Each graph is generated for each, and when the difference between the measured surface temperature and the estimated surface temperature is less than or equal to a preset reference value, the linear coefficient value (

,

,

) To select the condensation prediction system. The method of claim 2,
The client server includes a server control unit that calculates the dew point temperature by using the calculated estimated surface temperature of the window inner surface and the humidity information of the indoor space using the Barenbrug Formula of Equations 2 and 3 below. Condensation prediction system, characterized in that it further comprises.
[Equation 2]

[Equation 3]

Where a is 17.27°C, b is 237.3°C, T is the estimated surface temperature of the inner side of the window, 0°C <T <60°C, RH is the indoor humidity of the indoor space measured by the second sensor node, 1% <RH <100%, Td is the dew point temperature. The method of claim 4,
The server control unit monitors a pattern of changes over time between the calculated dew point temperature and the calculated surface temperature estimate of the window inner surface, and determines condensation by calculating a difference between the dew point temperature and the surface temperature estimate of the window inner surface. Condensation prediction system, characterized in that. In the method for predicting condensation consisting of one or more sensor nodes and a client server,
The cloud server receives the measured surface temperature of the inner side of the window measured from the first sensor node through wireless communication, receives the indoor temperature and humidity information from the second sensor node attached to one side of the room through wireless communication, and external weather Receiving an outdoor temperature from an information server;
The cloud server uses a linear regression model of Equation 2 below, which is a parameter of Equation 4 below (

,

,

Machine learning a correlation between the indoor temperature and the measured surface temperature according to the outdoor temperature in order to find an optimum value of ); And
The cloud server is a linear coefficient value obtained by the machine learning (

,

,

) By substituting Equation 4 below to calculate an estimate of the surface temperature of the inner side of the window.
[Equation 4]

here,

Is the estimate of the surface temperature of the inner side of the window over time (i),

Is the room temperature over time (i),

Is the outdoor temperature over time (i). The method of claim 6,
After the step of calculating an estimate of the surface temperature of the inner side of the window,
The client server further calculates the dew point temperature by using the Barenbrug Formula of Equations 5 and 6 below to calculate the estimated surface temperature of the inner surface of the window and the humidity information of the indoor space. Method for predicting condensation, characterized in that it comprises.
[Equation 5]

[Equation 6]

Where a is 17.27°C, b is 237.3°C, T is the estimated surface temperature of the inner side of the window, 0°C <T <60°C, RH is the indoor humidity of the indoor space measured by the second sensor node, 1% <RH <100%, Td is the dew point temperature. The method of claim 7,
And determining condensation by monitoring a change pattern of the calculated dew point temperature and the calculated surface temperature estimate value of the window inner surface over time, and calculating a difference between the dew point temperature and the surface temperature estimate value of the window inner surface Method for predicting condensation, characterized in that to.

Images