KR102408444B1 - Method and System of virtually monitoring indoor-air-quality using Artificial Neural Network - Google Patents

Abstract

The virtual indoor air quality monitoring method using an artificial neural network of the present disclosure is a virtual indoor air quality monitoring method using an artificial neural network performed by a computing device, and uses at least one sensor installed in a partial area within a target indoor space to collect indoor air quality data. A first step of detecting, and a second step of predicting indoor air quality data of the entire area of the target indoor space by inputting the detected indoor air quality data into a machine-learned artificial neural network (ANN).

Description

{Method and System of virtually monitoring indoor-air-quality using Artificial Neural Network}

The present invention relates to a virtual indoor air quality monitoring method and system using an artificial neural network.

According to the data of "The National Human Activity Pattern Survey, 2001, U.S. Environmental Protection Agency," which analyzed the behavioral patterns of citizens in modern society, ordinary modern people, other than outdoor workers, spend an average of 89% or more of their time indoors. Considering the special conditions, on average, more than 94.5% of the time is exposed to the indoor environment.

According to the Korea Institute of Construction Technology, indoor air is generally up to 10 times more polluted than the outdoor environment due to pollution sources such as fine dust, carbon dioxide, and VOCs generated by various indoor activities. According to the WHO report, pollutants emitted indoors are 1,000 times more likely to be transmitted to the human lungs than outdoors.

Therefore, in order to create a comfortable indoor environment and maintain the health of occupants, management of indoor air quality is essential. In order to manage indoor air quality in real time through ventilation, air purifiers, and air appliances, and to save energy, the need for indoor air quality monitoring is gradually increasing.

Sensing technology for indoor air quality monitoring has developed rapidly, and more sensors are being deployed in buildings to provide spatial indoor environmental data (eg, IAQ measurement) for systematic analysis, and fluid analysis to predict indoor air technique is being developed.

Computational Fluid Dynamics (CFD) has been widely applied to predict indoor airflow and pollutant distribution, and has the advantage of being able to efficiently predict indoor environments, that is, airflow patterns and pollutant distribution phenomena. However, as ventilation conditions, air volume, and the location and concentration of fine dust frequently change, it is difficult to perform and predict fluid analysis quickly.

Looking at the international research trend, according to the data of “Virtual models of indoor-air-quality sensors, Applied Energy, 87, 2010”, a study on indoor air quality monitoring was conducted. However, the conducted study measured indoor air quality virtually, but Indoor air quality was monitored at only one limited point in the space, and according to the data “Incorporating online monitoring data into fast prediction models towards the development of artificial intelligent ventilation systems, Sustainable Cities and Society, 47, 2019”, CO ₂ in indoor air For concentration control, deep-learning-based indoor air flow prediction and study on the distribution of CO ₂ throughout the room were performed, however, the conducted study assumed a steady-state with no physical change over time, and real-time analysis There was a limit to the flow prediction and control of this necessary unsteady state.

Looking at the domestic research level, H&Company Co., Ltd. developed an integrated system with IoT-based indoor air quality monitoring, purification and control functions optimized for the school environment in 2018. Although an integrated system such as a school indoor air quality monitoring device and purification device was developed, measurements were made based on physical sensors, and monitoring was performed based on limited measurement data.

One aspect of the embodiment is a virtual monitoring of indoor air quality using an artificial neural network that can machine learning the concentration and flow pattern of pollutants in the indoor air using an artificial neural network (ANN) and predict new results with untrained data input It is intended to provide a method and system.

However, the problems to be solved by the embodiments of the present invention are not limited to the above problems and may be variously expanded within the scope of the technical idea included in the present invention.

The virtual indoor air quality monitoring method using an artificial neural network according to an embodiment is a virtual indoor air quality monitoring method using an artificial neural network performed by a computing device, and indoor air quality using at least one sensor installed in a partial area within a target indoor space A first step of detecting data, and a second step of predicting indoor air quality data for all areas of the target indoor space by inputting the detected indoor air quality data into a machine-learning artificial neural network (ANN) do.

The second step is to build an indoor air quality database through computational fluid dynamics (CFD) simulation, pre-processing the data of the indoor air quality database, and using the pre-processed data in a network using an artificial neural network. It may include the step of generating an indoor air quality prediction model that predicts the indoor fine dust distribution and concentration by learning through

The step of building the indoor air quality database may include collecting indoor fine dust distribution and concentration data by analyzing the diffusion path and concentration distribution of fine dust through a fine particle tracking technique and fluid analysis.

The building of the indoor air quality database may include performing fluid analysis on a case in which the fine dust is distributed in the entire target indoor space or when the fine dust is locally distributed in the target indoor space.

In each of the above cases, it is possible to obtain indoor fine dust distribution data according to time.

The pre-processing step may include a pre-processing of data integration or data cleansing through partitioning the target indoor space.

The pre-processing step may include uniformly cutting the volume of the target indoor space to make a plurality of cubes, and averaging the simulated indoor air quality data in the space of each cube.

The predicting of the indoor fine dust concentration and distribution may include performing an LSTM (Long Short-Term Memory) model-based indoor air quality monitoring algorithm.

An indoor air quality virtual monitoring system using an artificial neural network according to another embodiment of the present invention is an indoor air quality virtual monitoring system using an artificial neural network performed by a computing device, and uses at least one sensor installed in a partial area within a target indoor space. to include a detector for detecting indoor air quality data, and a predictor for predicting indoor air quality data for all areas of the target indoor space by inputting the detected indoor air quality data into a machine-learning artificial neural network (ANN). can

The prediction unit may include a Long Short-Term Memory (LSTM) algorithm model.

According to the method of generating an indoor air quality prediction model using an artificial neural network according to the embodiment, machine learning the concentration and flow pattern of pollutants in the indoor air using an artificial neural network (ANN), results can be predicted.

By developing a predictive model using an artificial neural network, it is possible to monitor the concentration and distribution of fine dust in the entire indoor space based on a small number of sensors.

Development of a real-time ventilation control system that can manage indoor air quality with less energy consumption by linking indoor air quality monitoring in all spaces and air purifiers (energy recovery ventilation (ERV), air appliances, air purifiers, etc.) It is possible.

Even if the ventilation system in the entire space is not operated when indoor pollutants are generated locally, it is possible to manage indoor air quality using less energy if only the monitored and polluted local parts are removed.

1 is a schematic diagram schematically illustrating an indoor air quality virtual monitoring method using an artificial neural network according to an embodiment of the present invention.
2 is a diagram schematically illustrating an indoor air quality virtual monitoring system using an artificial neural network.
3 is a view showing the simulation result of analyzing the distribution of harmful environmental factors (fine dust) in the indoor space according to time through fluid analysis.
4 is a view showing the indoor space (a) and the initial distribution of various fine dust (b to i) in the indoor space.
5 is a schematic diagram schematically illustrating a raw database obtained by acquiring data for 102 cases within a target indoor space and a database reconstructed for LSTM algorithm training.
6 is a diagram illustrating a preprocessing process for indoor environment data.
7 is a diagram schematically illustrating an indoor air quality virtual monitoring method using an artificial neural network according to an embodiment of the present invention.
8 is a schematic diagram schematically illustrating an example of a Long Short-Term Memory (LSTM) structure.
9 is a diagram illustrating a real-time indoor air quality virtual monitoring result using a virtual indoor air quality monitoring method using an artificial neural network according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art can easily carry out the present invention. In order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and the same reference numerals are given to the same or similar components throughout the specification. In addition, the accompanying drawings are only for easy understanding of the embodiments disclosed in the present specification, and the technical idea disclosed herein is not limited by the accompanying drawings, and all changes included in the spirit and scope of the present invention , should be understood to include equivalents or substitutes.

Terms including an ordinal number such as 1st, 2nd, etc. may be used to describe various elements, but the elements are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

When an element is referred to as being “connected” or “connected” to another element, it is understood that it may be directly connected or connected to the other element, but other elements may exist in between. it should be On the other hand, when it is said that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist in the middle.

Throughout the specification, terms such as “comprises” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification is present, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof. Therefore, when a part "includes" a certain component, this means that other components may be further included, rather than excluding other components, unless otherwise stated.

1 is a schematic diagram schematically illustrating a method for virtual indoor air quality monitoring using an artificial neural network according to an embodiment of the present invention, and FIG. 2 is a diagram schematically illustrating an indoor air quality virtual monitoring system using an artificial neural network.

Referring to FIG. 1 , the virtual indoor air quality monitoring method according to the present embodiment includes detecting real-time indoor air quality data using a sensor array and using an artificial neural network (ANN) for the detected indoor air quality data. and outputting a real-time fine dust distribution result in an indoor space by inputting it to the air quality prediction model.

That is, the indoor air quality virtual monitoring method includes a first step of detecting indoor air quality data using at least one sensor installed in a partial area within a target indoor space, and a machine-learning artificial neural network using the detected indoor air quality data. , ANN) input to predict the indoor air quality data of the entire area of the target indoor space may include a second step.

Referring to FIG. 2 , the indoor air quality virtual monitoring system 100 according to the present embodiment includes a detector 10 that detects real-time indoor air quality data using a sensor array and an artificial neural network (ANN) for the detected indoor air quality data. and a prediction unit 20 that predicts the real-time fine dust distribution result in the indoor space by inputting it into the used indoor air quality prediction model.

The method of this embodiment may be evaluated by a computer or processor not shown, or may be processed in the form of a network in which a plurality of computers are connected. Accordingly, in the following method, the subject of processing will not be separately indicated.

The sensor array may include a plurality of physical sensors, and the plurality of physical sensors may be installed at a plurality of locations in an indoor space to be monitored to detect indoor air quality data. The indoor air quality data detected through the physical sensor may include indoor air quality pollutant concentration, temperature, and humidity. In this embodiment, in order to input to the artificial neural network prediction model, the concentration of fine dust, which is one of the main factors affecting indoor air quality, can be set as the acquired data. It can be machine-learning using data obtained through computational fluid dynamics (CFD) simulation analysis.

First, an indoor air quality database can be built by acquiring indoor air quality data, for example, fine dust concentration data, of the indoor space model through CFD verified through an experiment before learning the artificial neural network. CFD is a method to obtain an approximate solution through numerical analysis of the Navier-Stokes equation, which is a transport equation representing the flow of a fluid, in order to simulate the flow of a fluid in a modeled space. In the actual measurement process, physical data (density, velocity, pressure, viscosity, etc. of the fluid) can be obtained only from a specific point where the measuring equipment is installed, whereas CFD has the advantage of obtaining physical data for the entire space. It has the advantage that it takes less time and money.

The structure of the artificial neural network may include an input layer, a hidden layer, and an output layer. Artificial neural network is a statistical learning algorithm inspired by neural networks in biology (especially the brain in the central nervous system of animals) in machine learning and cognitive science. An artificial neural network refers to an overall model that has problem-solving ability by changing the strength of synaptic bonds through learning in which artificial neurons (nodes) formed a network by combining synapses.

Such an artificial neural network model can be used as a statistical prediction tool to predict the physical properties (density, viscosity, pressure, etc.) and flow of a fluid in fluid dynamics. In this case, new results can be predicted with untrained data input.

Therefore, it is possible to acquire physical data of various indoor spaces by combining CFD and artificial neural networks, and to perform indoor air quality monitoring in these various indoor spaces by applying artificial neural network algorithms. In addition, the verification of the artificial neural network can be performed by comparing the data (fine dust concentration data) obtained through CFD verified through actual experiments with the fine dust concentration data predicted through the artificial neural network.

In this embodiment, as an input value of the indoor air quality prediction model using the artificial neural network, the actually detected indoor air quality data of three areas of the target indoor space may be input, for example. In this case, as the output value, indoor air quality data of the entire area of the target indoor space may be output.

The indoor air quality data input and output can use the indoor fine dust concentration value, and the unit is set to μg/cm ³ or divided the predicted concentration value by the initial concentration value (predicted concentration value C/initial concentration value Co) as % It can be set by changing.

The target indoor space is divided into a total of 27 zones, and the input value is the values of fine dust concentration for 60 seconds in 3 or 4 of the zones, and the output value is the remaining zone (24 or The fine dust concentration values in 23) can be output. The fine dust distribution can be predicted with the values of each concentration output in this way.

When the fine dust concentration values in 3 zones were input as input values, the neural network prediction showed 93% accuracy. showed a result of 97%.

On the other hand, it is possible to build an indoor environment database based on flow analysis for artificial intelligence learning of artificial neural networks used in indoor air quality prediction models. For artificial intelligence learning, an indoor environment database can be built based on the flow analysis technique verified through experiments.

As an example, the diffusion path and concentration distribution analysis of harmful environmental factors (fine dust) that impair indoor air quality were conducted through the fine particle tracking technique and fluid analysis, and the results are shown in FIG. 2 .

3 is a view showing the simulation result of analyzing the distribution of harmful environmental factors (fine dust) in the indoor space according to time through fluid analysis.

(a) of FIG. 3 shows the distribution of fine dust in the indoor space according to time when the fine dust is distributed in the entire space, and (b) shows the distribution of fine dust in the indoor space according to time when distributed locally within 1/64 of the entire space. It shows the distribution of fine dust.

The fine dust concentration data obtained through the analysis of the fine dust diffusion path and concentration distribution can be obtained by using the volume average method in each of the 27 regions to obtain the average value of fine dust per unit time within the region.

4 is a view showing the indoor space (a) and the initial distribution of various fine dust (b to i) in the indoor space.

Referring to Figure 4, the indoor space (a) was set as a rectangular parallelepiped space with horizontal, vertical, and height of 4, 5, and 2.7 m in the x, y, and z-axis directions, respectively, and an inlet ( Inlet), outlets are formed in the lower left and right sides, respectively.

When fine dust is distributed within the entire space (b) to obtain indoor fine dust distribution data based on various locations and concentrations of fine dust within the set indoor space (a) (b), when distributed within 1/2 of the volume (c and d), distribution within 1/4 (e, f and g), distribution within 1/8 (h), and distribution within 1/64 (i) were performed. This assumes a variety of fine dust distributions, such as when particles are distributed over the entire area, locally generated in a specific area, introduced through a door or window, or particulate matter (PM) is present on the ceiling or floor.

5 is a schematic diagram schematically illustrating a raw database obtained by acquiring data for 102 cases in a target indoor space and a database reconstructed for LSTM algorithm training.

Referring to FIG. 5 , data for a total of 102 cases of various fine dust distributions were acquired, and the acquired data is fine dust concentration data every 4 seconds within 27 zones, and the acquired time is from 60 seconds to 1200 Acquisitions were taken for 1140 seconds until the second.

In order to train the LSTM algorithm, one dataset was constructed by grouping them in 60-second units. In addition, in order to increase the prediction accuracy of the LSTM, a data set was added by shifting the time every 4 seconds to reflect the change in fine dust concentration over time in each zone. The constructed database consisted of 28,968 datasets (total of 102 distributions x 284 datasets per one fine dust distribution).

6 is a diagram illustrating a preprocessing process for indoor environment data.

Since the data obtained through fluid analysis is high-resolution grid-based data, an indoor environment data preprocessing process is required to learn from the artificial neural network prediction model. This pre-processing process includes data integration (Data integration) and data cleansing (Data cleansing) pre-processing through indoor space division.

CFD is a finite volume method (FVM), which divides the analyzed geometry into countless elements (finite volume) in the form of a grid (or mesh), and is a transport equation representing the flow within the grid. An approximate solution of the Navier-Stokes equation is obtained through a numerical method.

In this embodiment, data was obtained through 3,979,692 grids, and fine dust concentration data was generated for each grid. Since it takes a lot of training time to directly input this into the LSTM, the Discretized Volume Model (DVM) method was used for data preprocessing in order to shorten the training time.

The DVM method consists of two steps: first, multiple cubes are created by evenly cutting the volume of a room, and then the simulation data is averaged in each cubic space.

As can be seen in Fig. 5, the volume of a room of length L can be divided into several uniform cubes of length Li (Li < L). The total number of cubes (27 zones, 27 zones) is much smaller than the total number of grids (3,979,692). The cubic high-resolution grid data, i.e. PM concentration data, were calculated using a volume average approach. A room with high-resolution grid data was replaced with multiple cubes with average values. As a result, the DVM method can significantly reduce the number of input variables of the artificial neural network model, and since the DVM method is based on spatial decomposition, it can be easily applied to indoors with various geometric spaces.

7 is a diagram schematically illustrating an indoor air quality virtual monitoring method using an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 7 , the virtual indoor air quality monitoring method using an artificial neural network according to the present embodiment includes the steps of detecting actual indoor air quality data using sensors installed in a plurality of, for example, three zones in a target indoor space, and the artificial neural network and inputting the detected indoor air quality data into an indoor air quality prediction model using

The target indoor space is set as a rectangular parallelepiped space having horizontal, vertical, and height in the x, y, and z-axis directions, and an inlet is formed on the upper surface of this rectangular parallelepiped indoor space, and an outlet is formed at the lower left and right sides, respectively. can be The sensor, for example, may be installed on one side of the upper surface (Sensor 1), one side of the right side (Sensor 2) of the drawing reference, and the front left corner (Sensor 3) of the drawing reference in the target indoor space to detect actual indoor air quality data.

In the input layer of the artificial neural network according to this embodiment, actual indoor air quality data detected from three sensors installed in the target indoor space, that is, actually monitored data (eg, fine dust concentration data over time) can be entered. Based on this, in the output layer, when the target indoor space is divided into N zones, the predicted indoor air quality data of each zone, that is, data on the distribution and concentration of indoor fine dust in the entire space can be predicted.

Here, the artificial neural network may be an indoor air quality virtual monitoring algorithm based on Long Short-Term Memory (LSTM) for real-time indoor fine dust concentration and distribution prediction. To reduce the error rate and learning time of the indoor air quality virtual monitoring algorithm and to prevent overfitting, the optimization process of hidden layer, hidden node, and activation function can be performed.

8 is a schematic diagram schematically illustrating an example of a Long Short-Term Memory (LSTM) structure.

A recurrent neural network (RNN) is a structure of an artificial neural network in which hidden nodes are connected by directed edges to form a directed cycle. It has a cyclic structure in which the result of the past hidden layer is connected to the input of the hidden layer of the next state. is the structure Recurrent neural networks are being used to process ordered arrays, sequentially appearing data, and time-correlated sequence data. However, when the distance between related information in the past and the state in which the information is used is long, a long-term dependency problem occurs due to the vanishing gradient, and the learning ability is greatly reduced in the cyclic neural network.

LSTM (Long Short Term Memory) is an algorithm designed to solve this problem of the cyclic neural network, and by adding a cell state to the existing cyclic neural network, it is possible to control how much data from the distant past is reflected.

Referring to FIG. 8 , LSTM is a gate mechanism method, and how much of the previous cell state (C _t-1 ) will be reflected is calculated by a forget gate (f _t ), and the current input value (X _t ) and the previous output value ( h _t-1 ) is controlled by the input gate (i _t ). After updating the old cell state (C _t-1 ) with the new cell state (C _t ), the output gate (o _t ) adjusts which output value to output and outputs it.

To reduce the error rate and learning time of the indoor air quality virtual monitoring algorithm and to prevent overfitting, the optimization process of hidden layer, hidden node, and activation function can be performed. In order to increase the accuracy of training, the activation function (sigmoid, tanh, Relu, Leaky Relu, Gaussian error linear unit, Exponential linear unit (ELU)) used for training can be changed within each gate, and as shown in FIG. Similarly, sigmoid, tanh, and Relu functions with high prediction accuracy can be selected. As the optimization function used for training, Adaptive Moment Estimation (ADAM) was used.

9 is a diagram illustrating a real-time indoor air quality virtual monitoring result using a virtual indoor air quality monitoring method using an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 9 , the concentration and distribution of fine dust predicted by the virtual indoor air quality monitoring method using the artificial neural network according to this embodiment were compared with the results predicted by computational fluid dynamics (CFD).

That is, it is assumed that the initial (0s) position of fine dust is on the floor, and the concentration and distribution of fine dust over time were predicted using the indoor air quality virtual monitoring method using the artificial neural network according to this embodiment. Here, the target indoor space was divided into a total of 27 zones, and actual sensors were installed in zones 1, 12, 17, and 27 to detect the actual concentration and distribution of fine dust. The data detected by these real sensors were input into the LSTM algorithm to predict the concentration and distribution of fine dust in the entire remaining area.

In (a) of FIG. 9, the fine dust concentration and distribution when 240s, 480s, 720s, and 960s have passed are shown together with the target indoor space model and graph. It can be seen that the data predicted by the LSTM algorithm and the result calculated by the CFD are generally close to each other.

9(b) shows that the data predicted by the LSTM algorithm and the result calculated by CFD are close to the fine dust concentration in the entire space over time. It is possible to obtain the average fine dust concentration data in the entire space, which was difficult to obtain experimentally, in real time through 4 sensors and LSTM algorithm.

9( c ) shows the data predicted by the LSTM algorithm and the CFD calculation result of the fine dust concentration over time in a location where it is difficult to install a sensor in real life, and it can be seen that the results are close. In real life, there is a limit to the installation of the sensor in the air rather than on the wall. It is possible to acquire the fine dust concentration in real time at a location that could not be measured through the LSTM algorithm.

Therefore, it was confirmed that the accuracy of the virtual indoor air quality monitoring based on the evaluation data set not used for learning is excellent. In this way, it is possible to implement a virtual indoor air quality monitoring method using an artificial neural network, and to verify indoor air quality virtual monitoring performance through time-dependent indoor fine dust concentration and distribution data.

As described above, it is possible to develop a system capable of controlling ventilation in real time by linking the indoor air quality virtual monitoring system using the artificial neural network according to the present embodiment to the air purifier. In addition, it will be possible to develop a moving air purifier to efficiently remove local parts or generated pollutants by monitoring the entire space when an indoor pollutant is generated.

A program for executing the method according to an embodiment of the present invention may be recorded in a computer-readable recording medium.

The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The medium may be specially designed and configured, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include hard disks, magnetic media such as floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floppy disks, and ROMs, RAMs, flash memories, and the like. Hardware devices specially configured to store and execute the same program instructions are included. Here, the medium may be a transmission medium such as an optical or metal wire or a waveguide including a carrier wave for transmitting a signal designating a program command, a data structure, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

Although preferred embodiments of the present invention have been described above, the present invention is not limited thereto, and it is possible to carry out various modifications within the scope of the claims, the description of the invention, and the accompanying drawings, and this also It is natural to fall within the scope.

Claims (10)

In the virtual indoor air quality monitoring method using an artificial neural network performed by a computing device,
A first step of detecting indoor air quality data using at least one sensor installed in a partial area within the target indoor space; and
A second step of predicting indoor air quality data of all areas of the target indoor space by inputting the detected indoor air quality data into a machine-learned artificial neural network (ANN)
including,
The second step is
building an indoor air quality database through computational fluid dynamics (CFD) simulation;
pre-processing the data of the indoor air quality database; and
Creating an indoor air quality prediction model for predicting indoor fine dust distribution and concentration by learning the preprocessed data through a network using an artificial neural network
includes,
The pre-processing step is
Including data integration or data cleansing pre-processing through the target indoor space division,
uniformly cut the volume of the target indoor space to make a plurality of cubes,
A virtual indoor air quality monitoring method using an artificial neural network, comprising averaging the simulated indoor air quality data in the space of each cube. delete The method of claim 1,
The step of building the indoor air quality database comprises:
A virtual indoor air quality monitoring method using an artificial neural network, which includes collecting indoor fine dust distribution and concentration data by analyzing the diffusion path and concentration distribution of fine dust through a fine particle tracking technique and fluid analysis. 4. The method of claim 3,
The step of building the indoor air quality database comprises:
A virtual indoor air quality monitoring method using an artificial neural network, wherein fluid analysis is performed when the fine dust is distributed in the entire target indoor space or when the fine dust is locally distributed in the target indoor space. 5. The method of claim 4,
A virtual indoor air quality monitoring method using an artificial neural network to acquire indoor fine dust distribution data over time in each of the above cases. delete delete The method of claim 1,
Predicting the indoor fine dust concentration and distribution comprises:
A virtual indoor air quality monitoring method using an artificial neural network to perform an LSTM (Long Short-Term Memory) model-based indoor air quality monitoring algorithm. In the system for performing a virtual indoor air quality monitoring method using an artificial neural network according to any one of claims 1, 3 to 5, and 8, which is performed by a computing device,
a detection unit configured to detect indoor air quality data using at least one sensor installed in a partial area within the target indoor space; and
A prediction unit for predicting indoor air quality data of all areas of the target indoor space by inputting the detected indoor air quality data into a machine-learned artificial neural network (ANN)
Indoor air quality virtual monitoring system using an artificial neural network, including 10. The method of claim 9,
The prediction unit,
Indoor air quality virtual monitoring system using artificial neural network, including LSTM (Long Short-Term Memory) algorithm model.

Images