Efficient Monitoring and Adaptive Control of Indoor Air Quality Based on IoT Technology and Fuzzy Inference

Abstract

In recent years, more and more occupants have suffered from respiratory illness due to poor indoor air quality (IAQ). In order to address this issue, this paper presents a method to achieve efficient monitoring and adaptive control of IAQ. Firstly, an indoor air quality monitoring and control system (IAQMCS) is developed using IoT technology. Then, based on fuzzy inference, a novel fuzzy air quality index (FAQI) model is proposed to effectively assess IAQ. Furthermore, a simple adaptive control mechanism, called SACM, is designed to automatically control the IAQMCS according to a real-time FAQI value. Finally, extensive experiments are performed by comparing with regular control (time-based control), which show that our proposed method effectively measures various air parameters (CO₂, VOC, HCHO, PM_2.5, PM10, etc.) and has good performance in terms of evaluation accuracy, average FAQI value, and overall IAQ.

This work is licensed under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.

1. Introduction

It is reported that 90% of the world’s population breathes polluted air, causing 7 million deaths annually. Nowadays, poor air quality has emerged as one of the biggest global environmental issues [1, 2]. Compared with outdoor air quality, indoor air quality (IAQ) is more significant because people spend nearly 80% of their time staying in indoor environments (homes, schools, offices, and so on) [3, 4]. Therefore, there is an urgent need to conduct effective IAQ monitoring and control for avoiding respiratory illness and protecting people’s health [5, 6].

In recent years, researchers all over the world have proposed various IAQ monitoring systems. For instance, An and Chung [7] proposed a wavelength division multiplexing optical transmission system for electromagnetic interference-free indoor dust monitoring, which can only collect indoor PM_2.5 concentration. Jo et al. [8] proposed an IoT-based indoor air quality monitoring platform, enabling users to monitor indoor air quality in anywhere and anytime. However, this platform can only perceive the indoor air quality but incapable of adjusting it. Benammar et al. [9] presented an IAQM (indoor air quality monitoring) system to realize the measurement of CO₂, CO, SO₂, NO₂, O₃, Cl₂, temperature, and relative humidity. Nevertheless, the concentration of CO, SO₂, and Cl₂ in the indoor environment is not very high, meaning that air parameters selected by the IAQM system are unreasonable. Arroyo et al. [10] developed a wireless gas sensor network system with low cost, small size, and low energy consumption, which can detect the concentration of VOC (volatile organic compounds), toluene, ethylbenzene, and xylene in the room. However, for ordinary indoor scenario, the level of toluene, ethylbenzene, and xylene is relatively low; hence, this system is only suitable for some specific pollutant detection application. Kim et al. [11] designed and implemented an integrated sensing system for real-time indoor air quality monitoring, which can detect the concentration of 7 common gases in real time and provide overall air quality alert timely. By using an extended fractional-order Kalman filter (EFKF), Ha et al. [12] merged indoor air quality index (IAQI) and humidex into an enhanced indoor air quality index (EIAQI) and proposed an EIAQI-based air quality management system, realizing the accurate prediction of indoor air quality. Zhao et al. [13] designed an IoT-based indoor air quality detector (IAQD) with multiple communication interfaces such as Modbus, LoRa, WiFi, general packet radio service, and NB-IoT. The IAQD allows users to remotely track the IAQ status and supports various communication scenarios such as wired communications, short-range wireless communications, and remote transmission to the cloud. Dhingra et al. [14] proposed an IoT-based mobile air pollution detection system, which mainly aimed at the detection of outdoor air quality, not indoor air. By using distributed deep reinforcement learning, Hu et al. [15] proposed a mobile robots-assisted cooperative indoor air quality sensing system, called AirScope, which can effectively reduce the data latency.

According to above literature review, most of existing works focus on the monitoring/measurement of IAQ but ignore its control and regulation. Up to date, there are a few references investigating the control problem of IAQ. Fermo et al. [16] improves indoor air quality by using air purifiers (AP); AP can effectively reduce the concentration of indoor particulate matter (PM) and VOC but cannot reduce the concentration of CO₂. Ali et al. [17] introduced an open-source hardware and software platform for monitoring buildings, called Elemental. Elemental can adjust indoor CO₂ level, temperature, and humidity by controlling HVAC, but it cannot improve formaldehyde, PM_2.5, VOC, and other air pollutants. Based on multiagent theory, Chen and Chen [18] constructed an indoor air quality control system, which can calculate the collected air data and use agents to make reasonable control decisions according to the prewritten rules. However, this system only stays at the level of theoretical simulation, and its effectiveness and stability need to be further verified in real application scenarios.

In order to improve indoor air quality, it is necessary to evaluate it accurately and scientifically. The evaluation methods of indoor air quality include subjective evaluation method (using human sensory organs to describe and evaluate) and objective evaluation method (using sensors to directly measure pollutant concentration) [19]. Kraus and Nováková [20] introduced a classroom air quality evaluation method that relies on students’ subjective feelings. However, students will have different reactions due to different factors such as their mental state, learning pressure, and gender, resulting in inaccurate evaluation results. Fuzzy comprehensive evaluation is a popular method to objectively evaluate air quality using fuzzy mathematics [21]. Olvera-García et al. [22] proposed an air quality evaluation method based on a weighted fuzzy inference system, which is mainly aimed at the comprehensive evaluation of urban (outdoor) air quality, and is not suitable for air evaluation in indoor scenes such as classrooms, home, and office. On this basis, Dionova et al. [23] proposed an environment indoor air quality index (EIAQI) by combining indoor air quality index (IAQI) and thermal comfort index (TCI), but the membership function of some indoor air pollutants in EIAQI is selected inappropriately.

As far as we can see, this paper provides the following two contributions: (1) We design an IoT-based indoor air quality monitoring and control system (IAQMCS), which allows users to not only remotely check real-time/historic air quality status but also control it. (2) We propose a novel fuzzy air quality index (FAQI) model to effectively assess indoor air quality. Based on the calculated FAQI value, a simple adaptive control mechanism, called SACM, is designed to automatically control the IAQMCS system to improve indoor air quality.

The remainder of this paper is organized as follows. Section 2 introduces the architecture of the proposed IAQMCS system. Section 3 describes the design of FAQI model and SACM mechanism. Section 4 evaluates the simulation and experimental results, respectively. Finally, Section 5 concludes the paper and presents our future work.

2. Architecture of the Proposed IAQMCS System

Based on the understanding of IoT elements in [24, 25], we design the architecture of an IoT-based indoor air quality monitoring and control system (IAQMCS), as shown in Figure 1. The proposed IAQMCS system can be divided into three layers [26, 27]:

(1) Data acquisition layer: this layer uses various gas sensors to collect common air parameters such as CO₂, VOC, HCHO (formaldehyde), PM_2.5/10, temperature, and humidity. Besides, a ventilation system is used as an actuator to improve indoor air

(2) Data transmission layer: this layer consists of a STM32 embedded system, a 4G module, and an electric relay. The STM32 embedded system transmits indoor air data collected by sensors to remote cloud server through 4G module. And the STM32 also controls the working state of ventilation system through electric relay

(3) Data process and display layer: this layer includes a database, a cloud server, and a computer control system. The database stores massive air quality and other kinds of data, and the cloud server analyzes and processes these big data. The computer control system uses predefined fuzzy logic rules to calculate the evaluation result of air quality in an indoor environment, downloads control decision to STM32 to automatically adjust the ventilation system, and provides the visualization of real-time/historic IAQ data

[figure(s) omitted; refer to PDF]

3. Design of FAQI Model and SACM Mechanism

3.1. Design of FAQI Model

In order to solve the drawbacks such as unreasonable input parameters, poor comprehensiveness, and low accuracy in existing air quality fuzzy evaluation methods, a novel fuzzy air quality index (FAQI) model is designed, as shown in Figure 2.

[figure(s) omitted; refer to PDF]

The inputs of the FAQI model are divided into two categories: (1) air pollutants and (2) inhalable particulate matters. Air pollutants include HCHO, CO₂, and VOC, while inhalable particulate matters include PM_2.5 and PM₁₀. By fuzzification operation, these physical inputs can be translated into fuzzy inputs. Then, based on predefined fuzzy rules, the fuzzy inference system (FIS) determines fuzzy output according to fuzzy inputs. After defuzzification operation, fuzzy output can be converted into sharp output (${\text{AQI}}_1$/${\text{AQI}}_2$). Finally, the total assessment result FAQI is obtained by weighted average of ${\text{AQI}}_1$ and ${\text{AQI}}_2$, that is, $$\begin{array}{}\text{(1)}& \text{FAQI}=K_1\times {\text{AQI}}_1+K_2\times {\text{AQI}}_2,\end{array}$$where ${\text{AQI}}_1=f_1{\text{CO}}_2,\text{VOC},\text{HCHO}$ and ${\text{AQI}}_2=f_2{\text{PM}}_{2.5},{\text{PM}}_{10}$ represent the fuzzy air quality index of air polluntants and inhalable particulate matters, respectively, $K_1$ and $K_2$ are their weight coefficients, and FAQI is the overall fuzzy evaluation result.

3.1.1. Transformation of Input/Output Ranges

Actual input/output should be converted into fuzzy input/output by means of scale transformation [28–30]. Assume that $x$ is the actual input/output and its range is $x_{\min },x_{\max }$, if the corresponding fuzzy input/output range is $X_{\min },X_{\max }$, then the fuzzy input/output $X$ can be calculated by $$\begin{array}{}\text{(2)}& X=\frac{X_{\min }+X_{\max }}{2}+\frac{X_{\max }-X_{\min }}{x_{\max }-x_{\min }}\times x-\frac{x_{\min }+x_{\max }}{2},\end{array}$$where $X\in X_{\min },X_{\max }$ and $x\in x_{\min },x_{\max }$.

According to “Indoor Air Quality Standard (China)” (GB/T 18883-2002), the value range of the abovementioned five input indicators (measured air parameters) and two output evaluation indicators (air quality index) are given, as shown in Table 1.

Table 1

Value range of actual input and output parameters.

Parameter category	Parameters	Value range
Input	HCHO	0 ~ 0.25 mg/m3
	CO2	0 ~ 2000 ppm
	VOC	0 ~ 0.8 mg/m3
	PM2.5	0 ~ 300 μg/m3
	PM10	0 ~ 500 μg/m3
Output	$${\text{AQI}}_1$$	0 ~ 400
	$${\text{AQI}}_2$$	0 ~ 400

By using formula (1), five kinds of actual input (HCHO, CO₂, VOC, PM_2.5, and PM₁₀) and two kinds of actual output (${\text{AQI}}_1$ and ${\text{AQI}}_2$) can be translated into fuzzy variables, that is, $\text{HCHO}$, $\text{C}{\text{O}}_2$, $\text{VOC}$, $\text{P}{\text{M}}_{2.5}$, $\text{P}{\text{M}}_{10}$, $\text{AQ}{\text{I}}_1$, and $\text{AQ}{\text{I}}_2$, respectively.

3.1.2. Fuzzy Language and Membership Functions

Fuzzy languages to describe HCHO are defined as low, medium, and high, that is, $L,M,H$, with universe $0,1,2$. Z-shaped, triangle, and S-shaped membership functions (MF) are used to represent $L$, $M$, and $H$, respectively, as shown in Figure 3(a).

[figure(s) omitted; refer to PDF]

Fuzzy languages to represent $\text{C}{\text{O}}_2$, $\text{VOC}$, $\text{P}{\text{M}}_{2.5}$, and $\text{P}{\text{M}}_{10}$ are defined as very low, low, medium, high, and very high, that is, $\text{VL},L,M,H,\text{VH}$. The corresponding universe of these input variables is all $0,1,2,3,4$, and their membership function curves are similar. The membership function curve of $\text{P}{\text{M}}_{2.5}$, as a representative example, is shown in Figure 3(b).

Fuzzy languages to describe $\text{AQ}{\text{I}}_1$ and $\text{AQ}{\text{I}}_2$ are defined as excellent, good, moderate, bad, very bad, and hazardous, that is $E,G,M,B,\text{VB},H$, and their membership function curves are shown in Figure 3(c).

3.1.3. Fuzzy Logic Rules and Fuzzy Inference

Fuzzy logic rules between the input and output variables greatly affect the performance of the FIS. Therefore, the designed logic rules must be complete and inconsistent rules must be avoided. Based on experts’ knowledge and practical experience, two fuzzy logic rule tables are established for ${\text{FIS}}_1$ and ${\text{FIS}}_2$; one describes the relationship between $\text{HCHO}$, $\text{C}{\text{O}}_2$, $\text{VOC}$, and $\text{AQ}{\text{I}}_1$ (shown in Table 2), and another describes the relationship between $\text{P}{\text{M}}_{2.5}$, $\text{P}{\text{M}}_{10}$, and $\text{AQ}{\text{I}}_2$ (shown in Table 3).

Table 2

Fuzzy logic rules for ${\text{FIS}}_1$.

$\text{AQ}{\text{I}}_1$

$\text{HCHO}=L$

$\text{HCHO}=M$

$\text{HCHO}=H$

$\text{VOC}$

$\text{VOC}$

$\text{VOC}$

$\text{VL}$

$L$

$M$

$H$

$\text{VH}$

$\text{VL}$

$L$

$M$

$H$

$\text{VH}$

$\text{VL}$

$L$

$M$

$H$

$\text{VH}$

$$\text{C}{\text{O}}_2$$

$$\text{VL}$$

$$E^1$$

$$E$$

$$G^2$$

$$G$$

$$M^3$$

$$G$$

$$M$$

$$M$$

$$B^4$$

$$B$$

$$M$$

$$B$$

$$B$$

$$\text{VB}$$

$$\text{VB}$$

$$L$$

$$E$$

$$G$$

$$G$$

$$M$$

$$M$$

$$M$$

$$M$$

$$B$$

$$B$$

$$\text{VB}$$

$$B$$

$$B$$

$$\text{VB}$$

$$\text{VB}$$

$$\text{VB}$$

$$M$$

$$G$$

$$G$$

$$M$$

$$M$$

$$B$$

$$M$$

$$M$$

$$B$$

$$B$$

$$\text{VB}$$

$$B$$

$$\text{VB}$$

$$\text{VB}$$

$$VB$$

$$H$$

$$H$$

$$G$$

$$M$$

$$M$$

$$B$$

$$B$$

$$M$$

$$B$$

$$\text{VB}$$

$$\text{VB}$$

$$\text{VB}$$

$$B$$

$$\text{VB}$$

$$\text{VB}$$

$$H$$

$$H$$

$$\text{VH}$$

$$M$$

$$M$$

$$B$$

$$B$$

$$\text{VB}$$

$$B$$

$$\text{VB}$$

$$\text{VB}$$

$$\text{VB}$$

$$H$$

$$VB$$

$$\text{VB}$$

$$H$$

$$H$$

$$H$$

¹If $\text{HCHO}$ is $L$ (low), $\text{C}{\text{O}}_2$ is $\text{VL}$ (very low), and $\text{VOC}$ is $\text{VL}$ (very low), then $\text{AQ}{\text{I}}_1$ is $E$ (excellent). ²If $\text{HCHO}$ is $L$ (low), $\text{C}{\text{O}}_2$ is $\text{VL}$ (very low), and $\text{VOC}$ is $M$ (medium), then $\text{AQ}{\text{I}}_1$ is $G$ (good). ³If $\text{HCHO}$ is $L$ (low), $\text{C}{\text{O}}_2$ is $\text{VL}$ (very low), and $\text{VOC}$ is $\text{VH}$ (very high), then $\text{AQ}{\text{I}}_1$ is $M$ (moderate). ⁴If $\text{HCHO}$ is $M$ (medium), $\text{C}{\text{O}}_2$ is $\text{VL}$ (very low), and $\text{VOC}$ is $H$ (high), then $\text{AQ}{\text{I}}_1$ is $B$ (bad).

Table 3

Fuzzy logic rules for ${\text{FIS}}_2$.

$\text{AQ}{\text{I}}_2$		$\text{P}{\text{M}}_{10}$
		$\text{VL}$	$L$	$M$	$H$	$\text{VH}$
$$P{M}_{2.5}$$	$$\text{VL}$$	$$E^1$$	$$G^2$$	$$G$$	$$M$$	$$M$$
	$$L$$	$$G$$	$$M$$	$$M$$	$$M$$	$$B$$
	$$M$$	$$M$$	$$B$$	$$B$$	$$B$$	$$\text{VB}$$
	$$H$$	$$B$$	$$\text{VB}$$	$$\text{VB}$$	$$\text{VB}$$	$$H$$
	$$\text{VH}$$	$$\text{VB}$$	$$\text{VB}$$	$$H$$	$$H$$	$$H$$

¹If $\text{P}{\text{M}}_{2.5}$ is $\text{VL}$ (very low) and $\text{P}{\text{M}}_{10}$ is $\text{VL}$ (very low), then $AQ{I}_2$ is $E$ (excellent). ²If $\text{P}{\text{M}}_{2.5}$ is $\text{VL}$ (very low) and $\text{P}{\text{M}}_{10}$ is $L$ (low), then $\text{AQ}{\text{I}}_2$ is $G$ (good).

Then, we take the calculation of $\text{AQ}{\text{I}}_1$ as an example to illustrate how ${\text{FIS}}_1$ works. The input variables of ${\text{FIS}}_1$ are $\text{HCHO}H$, $\text{C}{\text{O}}_{2}C$, and $\text{VOC}V$, and the output variable of ${\text{FIS}}_1$ is $\text{AQ}{\text{I}}_1{A}^1$. The total number of logic rules in ${\text{FIS}}_1$ is 75 ($3\times 5\times 5$) because $\text{HCHO}$, $\text{C}{\text{O}}_2$, and $\text{VOC}$ have three, five, and five linguistic levels, respectively. And the fuzzy relationship $R_i$ between the input and output variables of ${\text{FIS}}_1$ can be expressed as $$\begin{array}{}\text{(3)}& R_i={H_j\times C_m\times V_n}^{D_1}\times A_k^1,\end{array}$$where $H_i$, $C_m$, $V_n$, and $A_k^1$ are linguistic levels of $H$, $C$, $V$, and $A^1$, respectively; $i=0,1,2,\cdots ,74$; $j=0,1,2$; $m=n=0,1,2,3,4$; $k=0,1,2,3,4,5$; $D_1$ is the dimension of matrix $H_j\times C_m\times V_n$.

Through union operation of $R_i$, we can obtain a fuzzy relationship matrix consists of 75 fuzzy relationships, as follows: $$\begin{array}{}\text{(4)}& R=\underset{i=0}{\overset{74}{\cup }}{R}_i.\end{array}$$

Finally, the air quality assessment result $\text{AQ}{\text{I}}_1{A}^1$ can be calculated by $$\begin{array}{}\text{(5)}& A^1={H\times C\times V}^{D_2}\times R,\end{array}$$where $D_2$ is the dimension of matrix $H\times C\times V$.

3.1.4. Fuzzy Decision Surface

Applying centroid defuzzification approach, for each input pair ($\text{HCHO}$, $\text{C}{\text{O}}_2$ and $\text{VOC}$), the corresponding output ($\text{AQ}{\text{I}}_1$) is computed by formula (4). Repeating this process, we can obtain an output surface, called fuzzy decision surface. When $\text{VOC}$ is fixed, the fuzzy decision surface of ${\text{FIS}}_1$ is shown in Figure 4. It can be seen from Figure 4 that if $\text{C}{\text{O}}_2$ is very high (belongs to interval [3, 4]) and $\text{HCHO}$ is high (belongs to interval [1.5, 2]), then the estimated $\text{AQ}{\text{I}}_1$ is around 4.2 (see yellow area), meaning that air quality is hazardous.

[figure(s) omitted; refer to PDF]

Similarly, we can also get the fuzzy decision surface of ${\text{FIS}}_2$, as shown in Figure 5. It can be found from Figure 5 that $\text{AQ}{\text{I}}_2$ is positively correlated with both $\text{P}{\text{M}}_{2.5}$ and $\text{P}{\text{M}}_{10}$. However, when $\text{P}{\text{M}}_{2.5}$ is very high (belongs to interval [3, 4]), even if $\text{P}{\text{M}}_{10}$ is very low (belongs to interval [0, 1]), the calculated $\text{AQ}{\text{I}}_2$ value is still more than 3, implying that the considered air environment is assessed as very bad.

[figure(s) omitted; refer to PDF]

3.2. Design of SACM Mechanism

In order to realize automatic and reasonable regulation of IAQ, we design a simple adaptive control mechanism, called SACM, to adaptively and automatically control the working status of IAQMCS system according to real-time FAQI value, which ensures good quality of indoor air.

Assume that the ventilation system (VS) bought from market has rated power $P_e^{vs}=100\text{W}$, then we can obtain its actual power $P_r^{vs}$ by $$\begin{array}{}\text{(6)}& P_r^{vs}=u\times P_e^{vs}=u\times 100\text{W},\end{array}$$where $u$ is the output of SACM.

Generally, existing VS on the market has four tap positions: shutdown (0 W), low speed (30 W), medium speed (60 W), and high speed (100 W). Therefore, $u$ is set to 0, 0.3, 0.6, and 1.0, respectively, corresponding to above four tap positions one by one.

The opening or closing status of doors and windows will affect the operation effect of the VS. For example, when haze weather occurs, if the doors and windows are open, no matter how the VS works, the indoor PM_2.5 concentration will exceed standard level. In order to avoid the impact of outdoor air pollution on indoor air environment, we assume that doors and windows are closed throughout, and design six adaptive control rules between input FAQI and output $u$, as follows:

(i) Rule 1: if $0\le \text{FAQI}\le 50$ (excellent), then $u=0$ (shutdown)

(ii) Rule 2: if $50<\text{FAQI}\le 100$ (good), then $u=0.3$ (low speed)

(iii) Rule 3: if $100<\text{FAQI}\le 150$ (lightly polluted), then $u=0.3$ (low speed)

(iv) Rule 4: if $150<\text{FAQI}\le 200$ (moderately polluted), then $u=0.6$ (medium speed)

(v) Rule 5: if $200<\text{FAQI}\le 300$ (heavily polluted), then $u=1.0$ (high speed)

(vi) Rule 6: if $\text{FAQI}>300$ (severely polluted), then $u=1.0$ (high speed).

It is not difficult to understand above control rules. When FAQI belongs to different AQI intervals [31], reflecting that the comprehensive quality of indoor air has reached different pollution levels. Hence, adaptive controller should dynamically regulate the working state of the VS according to the real-time indoor air pollution level.

4. Simulation and Experimental Analysis

In this section, in order to evaluate the performance of the proposed FAQI model and SACM mechanism, theoretical simulations are first performed under MATLAB/Simulink. Then, our proposal is implemented into the proposed IAQMCS system, and some practical experiments are performed. During actual tests, two school offices, equipped with IAQMCS system, are considered experimental subjects, one using SACM, while another using traditional time-based control method.

4.1. Simulation Results and Analysis

Inspired by literature [32, 33], we construct an office air quality Simulink model, consisting of 6 air metrics: PM_2.5, PM₁₀, HCHO, VOC, CO₂, and FAQI. Simulation experiments are performed under three different scenarios: (1) the ventilation system is off (VS off), (2) the VS is under regular control method (turn on the VS at 8 am and turn off the VS at 6 pm), and (3) the VS is under the proposed adaptive control strategy. Simulink simulation parameters are summarized in Table 4.

Table 4

Simulink simulation parameters.

Parameter	Value
Room volume	147m3 ($7\text{m}\times 7\text{m}\times 3\text{m}$)
Removal rate of ${\text{PM}}_{2.5}/{\text{PM}}_{10}$	$$0.2/0.1$$
Indoor source strength of ${\text{PM}}_{2.5}/{\text{PM}}_{10}$	0 μg/(m3·h)
Indoor source strength of ${\text{CO}}_2$, VOC, HCHO	0.015 m3/(h·person), 0.025 mg/(m3·h), 0.005 mg/(m3·h)
Ambient air flow	200 m3/h
Recirculated air flow	100 m3/h
Penetration factor for ${\text{PM}}_{2.5}/{\text{PM}}_{10}$	0.68/0.7
Initial indoor concentration of ${\text{CO}}_2$, VOC, HCHO	500 ppm, 0.2 mg/m3, 0.01 mg/m3
Initial indoor concentration of ${\text{PM}}_{2.5}/{\text{PM}}_{10}$	124.62/134.02 μg/m3
Simulation time	48 h (2 days)

4.1.1. PM_2.5 and PM₁₀

Figure 6 shows the curve of indoor PM_2.5 concentration under three schemes. Please note that the outdoor ${\text{PM}}_{2.5}$ data come from “Moji Weather” website [34] and that the time of interest (ToI) is office time (from 8 am to 6 pm, see gray area in Figure 6).

[figure(s) omitted; refer to PDF]

It can be seen from Figure 6 that during office hours, if the VS is off, the concentration of indoor ${\text{PM}}_{2.5}$ will be the lowest. This is because when we turn off the VS, the ambient air flow from outdoor is greatly reduced, preventing outdoor ${\text{PM}}_{2.5}$ entering indoors. However, compared with regular control, the proposed adaptive control can lower the level of indoor ${\text{PM}}_{2.5}$ within ToI. Please note that the curve of indoor ${\text{PM}}_{10}$ concentration under different schemes is similar to that of indoor ${\text{PM}}_{2.5}$, due to page limitations, further discussion is not given.

4.1.2. HCHO, VOC, and CO₂

Figure 7 shows the curve of indoor HCHO concentration under three scenarios. We can see from Figure 7 that when the VS if off, the indoor HCHO concentration will become higher and higher, due to lack of ventilation. However, during ToI, indoor HCHO concentration under adaptive control is around 0.01 mg/m³, while that under regular control is around 0.005 mg/m³. The reason is that adaptive control considers the balance issue of various air parameters, in order to suppress the increase of indoor ${\text{PM}}_{2.5}$ and ${\text{PM}}_{10}$ levels, ventilation rate is not set very high, resulting in slightly higher indoor HCHO concentration.

[figure(s) omitted; refer to PDF]

Regarding performance metric VOC, the change trend of its concentration is similar to that of HCHO concentration. Under adaptive control, indoor VOC concentration during working hours is around 0.1 mg/m³, which is much lower than the upper limit value (0.8 mg/m³) given by Chinese national standard. Hence, people who work indoors will not feel uncomfortable.

Figure 8 shows the curve of indoor ${\text{CO}}_2$ concentration under three schemes. It can be found that indoor ${\text{CO}}_2$ concentration under “VS off” increases dramatically within ToI but drops a lot during non-ToI. Because people stay in the office during working hours and produce a large amount of ${\text{CO}}_2$ through breathing, when they leave the office after work, indoor ${\text{CO}}_2$ concentration will gradually decrease. Compared with regular control, the proposed adaptive control brings higher ${\text{CO}}_2$ concentration, but its maximum just slightly exceeds the national standard limitation value (1000 ppm). Hence, it is acceptable to tolerate a little bit higher ${\text{CO}}_2$ concentration without suffering from symptoms like dizziness and chest tightness.

[figure(s) omitted; refer to PDF]

4.1.3. Outdoor AQI versus Indoor FAQI

The curve of outdoor AQI and indoor FAQI is shown in Figure 9. Note that outdoor AQI data are from [34] and calculated by traditional method, while indoor FAQI data are from the designed IAQMCS system and computed by the proposed FAQI model. Due to serious air pollution, the outdoor AQI is relatively high during 48 h experiment time. As for indoor FAQI, we can see from Figure 9 that the proposed adaptive control performs better than other methods within ToI. By simultaneously taking into account 5 kinds of air parameters (${\text{PM}}_{2.5}$, ${\text{PM}}_{10}$, HCHO, VOC, ${\text{CO}}_2$), the proposed method avoids the problem of imbalanced concentration of these parameters and reduces comprehensive FAQI value, thus improving overall IAQ level.

[figure(s) omitted; refer to PDF]

Furthermore, we compare the average FAQI under three schemes, as shown in Figure 10. It can be found from Figure 10 that the average FAQI of three schemes in ToI of day 1 (8-18 h) differs a little (their FAQI value is all around 80). However, in ToI of Day 2 (32-42 h), the difference among average FAQI under three control methods is significant. Exactly, the average FAQI of VS-off, regular control and adaptive control is about 115, 61, and 52, respectively. That is to say, compared with conventional regular control, the proposed adaptive control can decrease average FAQI by 14.75% and improve IAQ from “good” to nearly “excellent.”

[figure(s) omitted; refer to PDF]

4.2. Experimental Results and Analysis

In order to further evaluate the practical performance of our proposal, the IoT prototype of IAQMCS system is designed and implemented in a $7\text{m}\times 7\text{m}\times 3\text{m}$ office, as shown in Figure 11.

[figure(s) omitted; refer to PDF]

We can see from Figure 11 that IAQMCS system consists of a ventilation system (see red box), an air quality sensing system (see yellow box), and 4 air exchange holes (see green marks). And the air quality sensing system is made up of an integrated air quality sensor (see red box), a MCU module (see white box), and a GUI module (see yellow box), as shown in Figure 12.

[figure(s) omitted; refer to PDF]

The integrated air quality sensor includes ${\text{CO}}_2$ sensor, VOC sensor, HCHO sensor, PM sensor, and Temp-humi sensor. The MCU module contains STM32 embedded system, 4G wireless communication unit, electric relay, etc. The GUI module is responsible for displaying real-time air pollutants data on the screen and transferring users’ operation commands to STM32 embedded system, which realizes friendly human-machine interaction.

Figure 13 shows the Web GUI of IAQMCS system. Remote users can view real-time IAQ data and set up the working mode of the VS. They can also track historic IAQ data and check other device information such as child lock status, filter time, and operation log.

[figure(s) omitted; refer to PDF]

In order to explore the practicability of the proposed IAQMCS system, FAQI model, and SACM mechanism, comparative experiments were performed during 31 December 2021 to 4 January 2022. After experiments, we extract historic data of various air parameters from database, as shown in Figures 14–17.

[figure(s) omitted; refer to PDF]

Figure 14 depicts the measured outdoor and indoor ${\text{PM}}_{2.5}$ concentrations during 31 December 2021 to 4 January 2022. Obviously, indoor ${\text{PM}}_{2.5}$ concentration is lower than outdoor one. However, compared with regular control, the proposed adaptive control (SACM) significantly decreases indoor ${\text{PM}}_{2.5}$ concentration.

Figure 15 presents the measured outdoor and indoor CO₂ concentrations during 31 December 2021 to 4 January 2022. We can see from Figure 15 that indoor CO₂ concentration is evidently higher than outdoors, owing to the respiration of office occupants. Nevertheless, during most of the experimental time, the indoor CO₂ concentration under adaptive control is higher than that under regular control. It is easy to understand this fact, in order to restrain outdoor PM_2.5/PM₁₀ from penetrating indoors, the ventilation rate of the VS is appropriately reduced, resulting in a rise of CO₂ level. However, the sacrifice of CO₂ is worthy, because rising CO₂ to a bearable level (1000 ppm) can solve the imbalance of various air parameters, thus improving the comprehensive level of IAQ.

Figure 16 shows the official outdoor AQI and the calculated indoor FAQI during 31 December 2021 to 4 January 2022. We can find from Figure 16 that indoor FAQI (using regular or adaptive control) is lower than outdoor AQI, implying that indoor air environment is better and more desirable than outdoors. However, compared with regular control, the proposed adaptive control can further reduce indoor FAQI and improve overall IAQ level.

In order to further analyze the improvement of FAQI after using our proposal, we compare the average daily FAQI under regular control and adaptive control, as shown in Figure 17. It can be clearly seen from Figure 17 that the average daily FAQI under the proposed adaptive control (see green columns) declines significantly, compared with that under regular control (see red columns). Moreover, the reduction rate of FAQI ranges from 32.97% to 48.96%, with mean value reaching 40.89%.

Based on above analysis, we can believe that our proposal not only stably monitors real-time/historic indoor air environment but also adaptively control the VS to lower the FAQI value and improve overall IAQ. It is worth pointing out that the production cost of the whole system is only 5000 RMB (around 746.5 dollars), which achieves efficient monitoring, assessment, and control of IAQ with low cost.

5. Conclusion

In this paper, we propose a method to achieve efficient monitoring and adaptive control of indoor air environment. Firstly, the IoT architecture of an indoor air quality monitoring and control system (IAQMCS) is designed. Based on fuzzy control theory, a fuzzy air quality index (FAQI) model and a simple adaptive control mechanism (SACM) are proposed to realize the accurate evaluation and reasonable control of indoor air. Theoretical simulations and practical experiments are performed, respectively. The experimental results demonstrate that the proposed method reduces average daily FAQI by 40.89% and improves overall IAQ level.

The next step of our work is to perform long-term experiments to further verify the stability and effectiveness of the proposed method. And deep learning will be adopted to optimize the system, so as to predict future IAQ and provide suitable measures for users in advance.

Acknowledgments

This work is financially supported by the National Natural Science Foundation of China (No. 61872221) and Scientific Research Fund of Hubei Provincial Department of Education.

References

[1] F. Yin, "Practice of air environment quality monitoring data visualization technology based on adaptive wireless sensor networks," Wireless Communications and Mobile Computing, vol. 2022,DOI: 10.1155/2022/4160186, 2022.

[2] Y. Wu, T. Liu, S. H. Ling, J. Szymanski, W. Zhang, S. W. Su, "Air quality monitoring for vulnerable groups in residential environments using a multiple hazard gas detector," Sensors, vol. 19 no. 2,DOI: 10.3390/s19020362, 2019.

[3] W. Sung, S. Hsiao, "Building an indoor air quality monitoring system based on the architecture of the Internet of Things," EURASIP Journal on Wireless Communications and Networking, vol. 2021 no. 1,DOI: 10.1186/s13638-021-02030-1, 2021.

[4] H. Zhou, X. Chen, S. He, J. Chen, J. Wu, "DRAIM: a novel delay-constraint and reverse auction-based incentive mechanism for WiFi offloading," IEEE Journal on Selected Areas in Communication, vol. 38 no. 4, pp. 711-722, DOI: 10.1109/JSAC.2020.2971871, 2020.

[5] K. Zhang, X. Zhang, H. Song, H. Pan, B. Wang, "Air quality prediction model based on spatiotemporal data analysis and metalearning," Wireless Communications and Mobile Computing, vol. 2021,DOI: 10.1155/2021/9627776, 2021.

[6] H. Zhou, K. Jiang, X. Liu, X. Li, V. C. M. Leung, "Deep reinforcement learning for energy-efficient computation offloading in mobile-edge computing," IEEE Internet of Things Journal, vol. 9 no. 2, pp. 1517-1530, DOI: 10.1109/JIOT.2021.3091142, 2022.

[7] J. An, W. Chung, "Wavelength-division multiplexing optical transmission for EMI-free indoor fine particulate matter monitoring," IEEE Access, vol. 6, pp. 74885-74894, DOI: 10.1109/ACCESS.2018.2882576, 2018.

[8] J. Jo, B. Jo, J. Kim, S. Kim, W. Han, "Development of an IoT-based indoor air quality monitoring platform," Journal of Sensors, vol. 2020,DOI: 10.1155/2020/8749764, 2020.

[9] M. Benammar, A. Abdaoui, S. H. M. Ahmad, F. Touati, A. Kadri, "A modular IoT platform for real-time indoor air quality monitoring," Sensors, vol. 18 no. 2,DOI: 10.3390/s18020581, 2018.

[10] P. Arroyo, J. Lozano, J. I. Suárez Marcelo, "Evolution of wireless sensor network for air quality measurements," Electronics, vol. 7 no. 12,DOI: 10.3390/electronics7120342, 2018.

[11] J. Kim, C. Chu, S. Shin, "ISSAQ: an integrated sensing systems for real-time indoor air quality monitoring," IEEE Sensors Journal, vol. 14 no. 12, pp. 4230-4244, DOI: 10.1109/JSEN.2014.2359832, 2014.

[12] Q. P. Ha, S. Metia, M. D. Phung, "Sensing data fusion for enhanced indoor air quality monitoring," IEEE Sensors Journal, vol. 20 no. 8, pp. 4430-4441, DOI: 10.1109/JSEN.2020.2964396, 2020.

[13] L. Zhao, W. Wu, S. Li, "Design and implementation of an IoT-based indoor air quality detector with multiple communication interfaces," IEEE Internet of Things Journal, vol. 6 no. 6, pp. 9621-9632, DOI: 10.1109/JIOT.2019.2930191, 2019.

[14] S. Dhingra, R. B. Madda, A. H. Gandomi, R. Patan, M. Daneshmand, "Internet of Things mobile–air pollution monitoring system (IoT-Mobair)," IEEE Internet of Things Journal, vol. 6 no. 3, pp. 5577-5584, DOI: 10.1109/JIOT.2019.2903821, 2019.

[15] Z. Hu, S. Cong, T. Song, K. Bian, L. Song, "AirScope: mobile robots-assisted cooperative indoor air quality sensing by distributed deep reinforcement learning," IEEE Internet of Things Journal, vol. 7 no. 9, pp. 9189-9200, DOI: 10.1109/JIOT.2020.3004339, 2020.

[16] P. Fermo, B. Artíñano, G. D. Gennaro, "Improving indoor air quality through an air purifier able to reduce aerosol particulate matter (PM) and volatile organic compounds (VOCs): Experimental results," Environmental Research, vol. 197, article 111131,DOI: 10.1016/j.envres.2021.111131, 2021.

[17] A. S. Ali, C. Coté, M. Heidarinejad, B. Stephens, "Elemental: an open-source wireless hardware and software platform for building energy and indoor environmental monitoring and control," Sensors, vol. 19 no. 18,DOI: 10.3390/s19184017, 2019.

[18] S. Chen, C. Chen, "Use of multi-agent theory to resolve complex indoor air quality control problems," Sensors, vol. 19 no. 5,DOI: 10.3390/s19051206, 2019.

[19] K. Jabłoński, T. Grychowski, "Fuzzy inference system for the assessment of indoor environmental quality in a room," Indoor and Built Environment, vol. 27 no. 10, pp. 1415-1430, DOI: 10.1177/1420326X17728097, 2018.

[20] M. Kraus, P. Nováková, "Assessment of indoor air quality in university classrooms," MATEC Web of Conferences, vol. 279, article 03012,DOI: 10.1051/matecconf/201927903012, 2019.

[21] J. Debnath, D. Majumder, A. Biswas, "Air quality assessment using weighted interval type-2 fuzzy inference system," Ecological Informatics, vol. 46, pp. 133-146, DOI: 10.1016/j.ecoinf.2018.06.002, 2018.

[22] M. A. Olvera-García, J. J. Carbajal-Hernández, L. P. Sánchez-Fernández, I. Hernández-Bautista, "Air quality assessment using a weighted fuzzy inference system," Ecological Informatics, vol. 33, pp. 57-74, DOI: 10.1016/j.ecoinf.2016.04.005, 2016.

[23] B. W. Dionova, M. N. Mohammed, S. Al-Zubaidi, E. Yusuf, "Environment indoor air quality assessment using fuzzy inference system," ICT Express, vol. 6 no. 3, pp. 185-194, DOI: 10.1016/j.icte.2020.05.007, 2020.

[24] H. Zhou, T. Wu, H. Zhang, J. Wu, "Incentive-driven deep reinforcement learning for content caching and D2D offloading," IEEE Journal on Selected Areas in Communication, vol. 39 no. 8, pp. 2445-2460, DOI: 10.1109/JSAC.2021.3087232, 2021.

[25] L. Zhao, S. Qu, J. Zeng, Q. Zhao, "Energy-saving and management of telecom operators’ remote computer rooms using IoT technology," IEEE Access, vol. 8, pp. 166197-166211, DOI: 10.1109/ACCESS.2020.3022641, 2020.

[26] H. Zhou, Z. Wang, N. Cheng, D. Zeng, P. Fan, "UAV-aided computation offloading in mobile edge computing networks: a Stackelberg game approach," IEEE Internet of Things Journal,DOI: 10.1109/JIOT.2022.3197155, 2022.

[27] H. Zhou, Z. Zhang, D. Li, Z. Su, "Joint optimization of computing offloading and service caching in edge computing-based smart grid," IEEE Transactions on Cloud Computing,DOI: 10.1109/TCC.2022.3163750, 2022.

[28] L. Zhao, S. Qu, W. Zhang, Z. Xiong, "An energy-saving fuzzy control system for highway tunnel lighting," Optik, vol. 180, pp. 419-432, DOI: 10.1016/j.ijleo.2018.11.123, 2019.

[29] F. Li, J. Cao, X. Wang, Y. Sun, T. Pan, X. Liu, "Applying buffer to SDN switches: benefits analysis and mechanism design," IEEE Transactions on Cloud Computing, vol. 9 no. 1, pp. 54-65, DOI: 10.1109/TCC.2018.2846620, 2021.

[30] H. Zhou, T. Wu, X. Chen, S. He, D. Guo, J. Wu, "Reverse auction-based computation offloading and resource allocation in mobile cloud-edge computing," IEEE Transactions on Mobile Computing,DOI: 10.1109/TMC.2022.3189050, 2022.

[31] Z. Idrees, Z. Zou, L. Zheng, "Edge computing based IoT architecture for low cost air pollution monitoring systems: a comprehensive system analysis, design considerations & development," Sensors, vol. 18 no. 9,DOI: 10.3390/s18093021, 2018.

[32] T. Marsik, R. Johnson, "HVAC air-quality model and its use to test a PM 2.5 control strategy," Building and Environment, vol. 43 no. 11, pp. 1850-1857, DOI: 10.1016/j.buildenv.2007.11.001, 2008.

[33] F. Li, Z. Guo, C. Zhang, W. Li, Y. Wang, "ATM: an active-detection trust mechanism for VANETs based on blockchain," IEEE Transactions on Vehicular Technology, vol. 70 no. 5, pp. 4011-4021, DOI: 10.1109/TVT.2021.3050007, 2021.

[34] "Webpage of outdoor air quality real-time data," , . https://tianqi.moji.com/aqi/china/hubei/xiling-district