0% found this document useful (0 votes)
12 views19 pages

Li 2018

Uploaded by

Avril Pacio
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
12 views19 pages

Li 2018

Uploaded by

Avril Pacio
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/issn/15375110

Review

Automatic carbon dioxide enrichment strategies in


the greenhouse: A review

Yongwei Li a,b,c,d, Ying Ding a,b,c,d, Daoliang Li a,b,c,d,*, Zheng Miao a,b,c,d
a
College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China
b
China-EU Centre for Information and Communication Technologies in Agriculture, China Agricultural University,
Beijing 100083, PR China
c
Key Laboratory of Agricultural Information Acquisition Technology, Ministry of Agriculture, China Agricultural
University, Beijing 100083, China
d
Beijing Engineering and Technology Research Centre for Internet of Things in Agriculture, China Agricultural
University, Beijing 100083, China

article info
Greenhouses constitute a proven solution for coping with environmental degradation and
Article history: for increasing both the quantity and quality of agricultural products. Appropriate carbon
Received 11 September 2017 dioxide (CO2) control not only improves crop yield and quality but also reduces the carbon
Received in revised form footprint of greenhouses. However, CO2 enrichment control in greenhouses is a dynamic,
27 March 2018 interactive and time-delayed process. In practice, optimal CO2 concentrations in the
Accepted 23 April 2018 greenhouse are difficult to maintain because CO2 is confounded with temperature, hu-
midity, light intensity, etc.; therefore, ambient CO2 concentrations in greenhouses are
often suboptimal or excessive. This paper is a review of the current theoretical and applied
Keywords: studies of CO2 enrichment in greenhouses and discusses the advantages and limitations of
Carbon dioxide various methods. The major points addressed are as follows: 1) the five sources of CO2
Greenhouse enrichment in greenhouses; 2) the monitoring and data processing of CO2 concentrations;
Source of carbon dioxide and 3) the various methods for controlling automatic CO2 enrichment. This paper discusses
Data processing new challenges and perspectives and suggests future studies and methods for a green-
Control method house CO2 enrichment system. A new symbiotic greenhouse system requiring sensible CO2
balance is also presented.
© 2018 IAgrE. Published by Elsevier Ltd. All rights reserved.

climatic factors, nutrition, biotic factors and other variables


1. Introduction that affect crop growth and development; thus, optimal con-
ditions can be imposed during different crop growth stages
Economic and social development has increased the re- (Pasgianos, Arvanitis, Polycarpou, Sigrimis, 2003). Moreover,
quirements for abundant, high-quality agricultural products. greenhouses allow the production of high-quality products at
Greenhouses are structures that allow growers to control low cost. Greenhouses have become increasingly important in

* Corresponding author. College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China.
E-mail addresses: [email protected] (Y. Li), [email protected] (Y. Ding), [email protected] (D. Li), [email protected]
(Z. Miao).
https://doi.org/10.1016/j.biosystemseng.2018.04.018
1537-5110/© 2018 IAgrE. Published by Elsevier Ltd. All rights reserved.
102 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

concentration affects the environmental and economic effi-


Nomenclature ciency of the entire system (Yogev, Barnes, Gross, 2016). From
global energy-saving and emission-reduction perspectives, it
ANNs artificial neural networks
is important to control CO2 enrichment within greenhouses
FNN fuzzy neural network
appropriately.
GAs genetic algorithms
Economic benefit is one of the most important metrics for
MIV mean impact value
evaluating greenhouse business aspects (Rodriguez,
MPC model predictive control
Berenguel, Arahal, 2003; van Straten, van Henten, van
NB-IoT narrow band internet of things
Willigenburg, van Ooteghem, 2010). An important cost of
PCA principal component analysis
CO2 enrichment is the source of CO2; thus, choosing an
PID proportional-integral-derivative
appropriate CO2 source is essential. The existing literature
PSO particle swarm optimization
about greenhouse enrichment has primarily focused on five
RBF radical basis function
sources of CO2 for enrichment: compost, exhaust gas from
SVM support vector machine
fossil fuels, exhaust gas from renewable energy, natural or
WSNs wireless sensor networks
forced ventilation and pure liquefied CO2. The choice of the
appropriate method for generating CO2 and supply control in
practice will boost efficiency and profits.
agriculture, and many researchers are dedicated to green- Accurate CO2 concentration data are critical for controller
house climate control. Greenhouse variables that are gener- design and actuator control. In conventional greenhouses,
ally controlled include CO2 concentration, temperature, actuators are operated by growers, depending on experience
humidity, etc. CO2 is an important variable, and the positive (Rodrı́guez, Berenguel, Guzma  n, Ramı́rez-Arias, 2015).
effects of CO2 on plant growth were discovered 200 years ago Although an experienced grower can manage a greenhouse
(Panwar, Kaushik, Kothari, 2011). In addition, CO2 enrichment well, the profits are not always optimal. The development of
is ubiquitous in modern greenhouses. CO2 enrichment im- sensor technology has allowed the objective monitoring of
proves the rate of photosynthetic assimilation (Thongbai, greenhouse parameters. However, accurate measurements of
Kozai, Ohyama, 2010; Xu, Zhu, Li, & Ye, 2014) and as a CO2 concentrations are difficult because CO2 is not homoge-
consequence improves energy and water use efficiency nous in the greenhouse. Wireless sensor networks have been
(Deryng et al., 2016; Frantz, 2011; Sanchez-Guerrero, Lorenzo, widely applied in modern greenhouses for monitoring
Medrano, Baille, Castilla, 2009), nutrient uptake rates (Li, greenhouse climate parameters, and more accurate mea-
Zhou, Duan, Du, Wang, 2007), and product nutritional value, surement results have been achieved.
as well as many other production factors (Mohamed, Jellings, The greenhouse process contains both fast and slow dy-
Fuller, 2013; Weiss, Mizrahi, & Raveh, 2010; Zhang, Liu, Zhang, namics. The greenhouse reacts rapidly to changes in the
Zhang, & Wang, 2014). Plants take up CO2 when illuminated environmental conditions. But crop development responds
and evolve CO2 in the dark, so fluctuations of CO2 concen- slowly to the changes of greenhouse environmental condi-
trations in greenhouses are governed by the photoperiod (Liu, tions (i.e. CO2 concentrations and temperature) and the
Hoogenboom, Ingram, 2000). In general, the optimum CO2 photosynthesis responds relatively quickly to changes in
concentration for plant growth is approximately 1000 ppm ambient conditions. In addition, CO2 level has a long time
(Chen, 2010; Jin et al., 2009); however, the CO2 concentration in constant when increasing and much longer when supply is
the air is approximately 350 ppm and is probably lower than shut off (waiting for photosynthesis to deplete the extra CO2 in
150 ppm in greenhouses during the daytime, which is far from the air). In the morning, the question is whether it is worth
sufficient for optimal crop growth. Additionally, excess CO2 starting the supply process and more importantly when to
concentrations can reduce plant growth, cause leaf injury, and cut-off the supply at an early stage to avoid exhausting CO2 by
increase the loss of CO2 via leakage (Panwar et al., 2011). ventilation, which could make the process economically and
Weather also has a significant influence on CO2 enrichment environmentally negative. Therefore, accurate weather pre-
(Kuroyanagi, Yasuba, Higashide, Iwasaki, Takaichi, 2014). In diction is important to control CO2 concentrations within
sunny chilly weather, the supply of CO2 at a relatively high greenhouses precisely. Many control strategies for automatic
concentration is appropriate. However, CO2 enrichment could CO2 enrichment are based on growers' empirical rules
be economically prohibitive because of the lower photosyn- (Rodrı́guez et al., 2015). The use of computers in greenhouses
thetically active radiation on cloudy days or because of the allows growers to control environment variables and to
loss due to ventilation for cooling in hot weather. Most loca- modify the control parameters automatically. However,
tions, even in winter time, do need ventilation and challenge several important problems related to the control of CO2
is that CO2 supply is justified when enough solar radiation is enrichment exist. First, set-point tracking is affected by con-
reaching the GH, when ventilation is also required. Therefore, trol loops and actuators of other parameters (Titus &
CO2 supply may only be justified in a few instances, or only Wendlberger, 2016). Second, it is difficult to determine scien-
few morning hours in summer, and mostly in well-equipped tifically the optimum set-point value because of interactions
greenhouses with shades internal and external. So the with other associated parameters, which can affect crop
optimal CO2 concentration for greenhouse production de- growth (Pasgianos et al., 2003). Third, the control performance
pends on many factors, such as local economic conditions, is difficult to evaluate because of the uncertainty inherent in
external weather, and ventilation rates. In some symbiotic biological systems. Additionally, modern greenhouses allow a
greenhouses, such as aquaponic systems, the CO2 particularly high degree of freedom and order. These factors
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 103

are responsible for many challenges in the control of green- optimum fermentation environment, which involves the ratio
house CO2 enrichment. of carbon to nitrogen, humidity, initial pH and water content
This article comprehensively reviews CO2 enrichment in (Du, Lin, & Zhang, 2004); in addition, emissions such as NH3
the greenhouse, describing the latest developments in auto- need to be prevented. Furthermore, the amount of CO2 from
matic CO2 enrichment technology in greenhouses. Green- compost will decrease over time, so the CO2 cannot be
house engineers can use this review to aid the design of controlled easily. The common method of CO2 enrichment by
appropriate CO2 enrichment systems considering existing composting is placing a composting unit in the centre of a
conditions, and researchers may be inspired to discover better greenhouse, so the distribution of CO2 is related to the green-
methods for automatic CO2 enrichment. This review is divided house structure. To control CO2 concentrations in greenhouses
into five sections. The first section discusses various sources accurately, CO2 storage equipment that can automatically
of CO2 enrichment in the greenhouse. The second section control the CO2 enrichment via solenoid valves is necessary. An
addresses CO2 concentration monitoring and original sensor enrichment strategy that combines compost and secondary
data processing. The third section specifically focuses on compost products in an optimum fermentation environment
different methods of CO2 control in the greenhouse, which could be widely applied, particularly in greenhouses that have
mainly involves classical control, modern control, intelligent relatively low levels of automation.
control and hybrid control. The fourth section discusses new
challenges and perspectives and includes the latest de- 2.2. Controllable sources
velopments in automatic CO2 enrichment and a new form of
greenhouse cultivation. Our conclusions are presented in the In automatic greenhouses, it is appropriate to automatically
final section. supply CO2 using a controlled source of CO2 enrichment such
as exhaust gas from fossil fuels or renewable energy and pure
liquefied energy. Regarding CO2 from combustion emissions,
2. Sources of carbon dioxide enrichment the diluted gases are distributed throughout the greenhouse
with circulation fans or perforated plastic tubes, and growers
The average optimum CO2 concentration for crops is need to consider the heat produced during CO2 generation; hot
approximately three times (Jin et al., 2009) that of atmospheric water is commonly used for storing heat. Co-generation can
CO2 concentration and the optimal CO2 concentration be a better solution but requires high investment and
changes with different plants as well as environmental con- advanced technology. Pure liquefied CO2 from a hydraulic
ditions such as solar radiation intensity and temperature. storage tank evaporates and is distributed in the greenhouse
Furthermore, the CO2 enrichment source is a considerable via a series of semi-bore plastic tubes. It is convenient to
investment in terms of CO2 enrichment systems. Therefore, control, but its pre-investment is high.
the choice of an appropriate CO2 source is important in
greenhouse cultivation. This section reviews five sources of 2.2.1. Exhaust gases from fossil fuels
CO2 enrichment that are widely used for greenhouse enrich- Enriching greenhouses with CO2 generated from burning
ment (Table 1): compost, exhaust gas from fossil fuels, fossil fuels has been commonly used, particularly in cooler
exhaust gas from renewable energy, natural or forced venti- northern areas. In rudimentary modes of greenhouse culti-
lation and pure liquefied CO2. We review these different vation, growers directly burn coal to heat and fertilise green-
sources of CO2 enrichment from a control perspective. houses during the winter. A more advanced cultivation mode
employs natural gas and other hydrocarbons as heat and CO2
2.1. Uncontrollable sources sources (Dion, Lefsrud, Orsat, 2011). Additionally, it is gener-
ally accepted that natural gas can produce clean CO2 if proper
CO2 concentrations in greenhouses are typically suboptimal. combustion occurs. Chalabi, Biro, Bailey, Aikman, and
Compost is used as a CO2 source in greenhouses because it is Cockshull (2002b) used CO2 from the exhaust gases of boilers
inexpensive. Additionally, compost comprising crop residue burning natural gas to enrich greenhouse atmospheres, and
or animal manure can reduce carbon, nitrogen, and phos- their results showed that this method increased financial
phorus emissions from agriculture. Jin et al. (2009) developed margins after control had been optimised. Bailey (2002)
a CO2 fertilisation strategy using crop residue and animal enriched a tomato greenhouse with CO2 obtained from the
manure compost. The results showed that this strategy can combustion of natural gas by the greenhouse heater and by
improve the yield of celery, leaf lettuce, stem lettuce, etc. combined heat and power units. Both methods produced good
However, copious amounts of composted products are pro- economic benefits, and the financial margins of the latter
duced after compost completion, and the disposal of these increased when heat storage was used.
products is costly. To solve this problem, Liu et al. (2016) used Enrichment with fossil fuels can increase profits; however,
the secondary compost products to provide CO2 fertilisation other gases such as carbon monoxide, which is harmful to
in greenhouses. This process is a more cost-effective method plants and humans, are inevitably present because incom-
because the compost products are reused, and the secondary plete combustion occurs under empirical conditions (Jerzak,
compost products can be used to inhibit tomato bacterial 2014). Therefore, an additional purification system is neces-
wilt. sary to remove particulates and noxious gases from the flue
Waste compost can increase CO2 concentrations in green- gas. Additionally, little heating is required in warm areas, so
houses to levels higher than those in the ambient atmosphere. producing CO2 from fossil fuel emissions is not cost effective
However, optimal CO2 production from composting requires an in such conditions.
104
Table 1 e Different sources of CO2 enrichment.
Sources of CO2 Materials Other coproduct The highest CO2 Advantages Limitations References
enrichment gases concentration
that can be
achieved
Compost Crop residue and NH3 1000-1500 ppm Less waste Provision of an optimum Jin et al. (2009)
animal manure fermentation environment and
disposal of compost products
Compost products e e Reuse of compost products Requires an appropriate Liu et al. (2016)
environment for secondary
composting processes

b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9
Fossil fuel Natural gas e Approximately Can heat the greenhouse on The cost is affected by fluctuations Bailey (2002) and Chalabi et al.
emissions 1500 ppm and lower chilly days; Cheaper than in fossil fuel prices. Not (2002b)
in summer pure CO2 environmentally friendly. Not
suitable in warm areas. Emissions
must be purified before use as a
fertiliser.
Renewable energy Landfill biogas SO2, CO, H2S, NO2, 2500 ppm Environmentally friendly; Not suitable for warm areas; Jaffrin et al. (2003)
emissions NO, C2H4 and other Can achieve carbon Emissions must be purified before
VOCs neutrality with appropriate use as a fertiliser. Additional
control renewable energy-producing
system is necessary.
Biogas from the SOX, H2S, NO2, NOX, e Promote waste to energy Must be purified. Zhang et al. (2015)
anaerobic digestion C2H4, NH3, PMs and and waste to material
of dairy manure other VOCs
Wood pellets CO, NOX, SOX, and e Reduced dependency on Equipment such a precipitator,  nchez-
Roy et al. (2014) and Sa
VOCs fossil fuel energy filter, scrubber and catalytic Molina et al. (2014)
converter is required.
Almond shells, pine CO and unsaturated e Cost savings by using CO2 and hot water storage  nchez-Molina et al. (2014)
Sa
and wood pellets, hydrocarbons chipper biomass fuel; equipment are required.
and olive pits Can heat and enrich
greenhouse
simultaneously
Natural or forced Ambient air e Approximately Low cost and ease of The maximum concentration of Amir et al. (2005), Montero et al.
ventilation 350 ppm control CO2 is only approximately (2009) and Stanghellini et al. (2008)
350 ppm; Not suitable during
winter
Pure liquefied CO2 Pure liquefied CO2 None Any predicted Safe and pure; Can Relatively expensive. Chalabi et al. (2002a, 2002b),
concentration enrich at any time Kuroyanagi et al. (2014) and
 nchez-Guerrero et al. (2005)
Sa

SOX e Sulphur Oxides, NOX e Nitrogen Oxides, PMs e different diameter Particulate Matter.
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 105

2.2.2. Exhaust gases from renewable energy one strategy involves burning gas only until the storage tank is
Renewable energy from sources such as landfill biogas and heated to its maximum temperature, after which additional
biomass (including non-fossilised and biodegradable organic CO2 demand is met using pure liquefied CO2; another strategy
materials) can also be a CO2 source for greenhouse CO2 involves CO2 enrichment and storing heat by burning gas only
enrichment, and biomass has been proven to be more eco- during the period when the crop photosynthetic rate is rela-
friendly than natural gas (Dion et al., 2011). Jaffrin, tively high throughout the whole day. However, the economic
Bentounes, Joan, and Makhlouf (2003) used biogas generated and environmental benefits of these strategies need to be
from a landfill to provide heat and CO2 for a rose horticulture further studied.
greenhouse. Roy et al. (2014) and Sa  nchez-Molina, Reinoso,
 
Acien, Rodrı́guez, and Lopez (2014) developed a biomass 2.2.3. Pure liquefied carbon dioxide
furnace that could recover heat and CO2 from the fuel gas, Pure liquefied CO2 is widely used to fertilise greenhouses.
both of which were then used in the greenhouse. Zhang, Bi, Generally, liquefied CO2 from a storage tank is injected into
and Clift (2015) explored an eco-industrial complex greenhouses by a controlled valve. Many researchers fertilise
composed of a dairy farm and a greenhouse with an anaerobic with bottled pure CO2 to maintain CO2 concentrations in
digestion plant. Anaerobic digestion involves the degradation greenhouses at optimum levels (Kuroyanagi et al., 2014;
of organic matter into biogas, which can then be combusted to Sa nchez-Guerrero et al., 2005). Using an optimum control
produce heat and CO2 for use on the farm and in the green- strategy, Chalabi, Biro, Bailey, Aikman, and Cockshull (2002a,
house. Additionally, the digestate can be used as a fertiliser for 2002b) used pure CO2 to enrich a greenhouse. Simulation re-
crops. A detailed discussion of CO2 production and heat re- sults have shown that this enrichment method was more
covery from biomass emissions can be found in a review by profitable than the method in which CO2 from natural gas
Dion et al. (2011). Dion, Lefsrud, Orsat, and Cimon (2013) pro- combustion was used to enrich a greenhouse without heat
posed a CO2 enrichment method that generates CO2 by com- storage under their circumstances.
busting syngas generated from wood pellet gasification. Chau Pure liquefied CO2 has the potential for widespread use in
et al. (2009) studied a greenhouse that used wood pellets and a modern greenhouses because of its advantages: it is pure,
wood residue boiler to generate heat and CO2 and the results nontoxic, and portable, and it can easily be used both for
showed this greenhouse was more profitable than was a enriching and for maintaining greenhouse CO2 levels at suf-
greenhouse using a natural gas boiler. ficient concentrations. Enrichment with pure CO2 can result in
Renewable energy combustion can increase CO2 concen- good profits if CO2 concentrations in greenhouses are appro-
tration in greenhouses. In addition, reusing both the residues priately controlled at optimum levels based on the weather
of digestion, such as the liquid phase of a digestate, as a fer- routine, local economic conditions and the size of the green-
tiliser and the solid phase as a growth medium is environ- house business.
mentally conserving and cost effective. However, purification
of the gas is also essential and complex, as landfill biogas or 2.3. Other sources
that resulting from biomass combustion contains many other
gases, such as CO, CH4, H2S, SO2, NO2, NO, C2H4 and other Other sources of greenhouse CO2 include ambient air that
volatile organic compounds (VOCs), some of which are toxic enters the interior of the greenhouse because of greenhouse
(Jaffrin et al., 2003; Zhang et al., 2015). The main purification cooling by natural or forced ventilation. During the daytime,
methods for such emissions are physical and chemical CO2 concentrations in greenhouses are probably as low as
methods. For instance, the exhaust gases flow through a 100e150 ppm, which is lower than the concentration outside
washer box with heating water to extract some of the chem- (approximately 400 ppm) (Kozai, Kubota, Takagaki, Maruo,
ical pollutants and VOCs, and then flow through a scrubbing 2015). These low CO2 concentrations will severely affect the
container with chemical reactive liquid to produce CO2-rich rate of photosynthesis of crops in the greenhouse. In green-
gas, and finally air excess is precisely adjusted via a fan to houses lacking artificial supplies of CO2, the CO2 absorbed for
keep some harmful residual gases at a safe low level. Mem- plant photosynthesis ultimately comes from the ambient
brane separation has a great deal of potential for CO2 extrac- environment. Some growers turn the ventilation on or off to
tion from mixed gas emissions (Dion et al., 2011), so the allow natural ventilation to increase the CO2 levels in the
selection of appropriate materials and strategies for sepa- greenhouse (Stanghellini, Incrocci, Gazquez, Dimauro, 2008),
rating CO2 from emissions is worthy of study. and in other greenhouses, forced ventilation systems that
Heating demands during greenhouse cultivation do not control fans have been designed to maintain CO2 levels in
often coincide with CO2 enrichment demands. Furthermore, greenhouses (Amir, Teitel, Shemer, 2005; Montero,
some greenhouses in warm areas do not need to be heated. Stanghellini, Castilla, 2009).
This problem exists for fossil fuel and renewable energy Natural or forced ventilation provides CO2 by exchanging
enrichment systems. The two most common solutions the internal and external air. The distribution of CO2 is rela-
include the use of a heat reservoir as a buffer (Chalabi et al., tively uniform and the concentration of CO2 in the greenhouse
2002b; Seginer, van Straten, & van Beveren, 2017a, 2017b) or is generally similar to that of the outside, but the relatively
enrichment in the morning or late afternoon (Sa  nchez- high concentration areas may be not where crops are grown
Guerrero et al., 2005). However, the efficiency of heating and (Hao, Shen, He, & Feng, 2014). The ventilation is typically used
enrichment in greenhouses with heat storage still needs to decrease the greenhouse temperature in most instances.
improvement. Taking into account environmental aspects, Therefore, ventilation of greenhouses needs to consider the
106 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

trade-off between ensuring the inflow of CO2 and maintaining Guevaragonzalez, Troncoso, Osorniorios, Torrespacheco,
an adequate temperature. CO2 concentrations in greenhouses 2009; Zosel, Oelssner, Decker, Gerlach, Guth, 2011).
can be at suboptimal levels (approximately 400 ppm) in the A large advantage of WSNs is eliminating the trouble of
daytime by ventilation during mild days, and the temperature wiring, largely owing to the wireless communication tech-
and humidity in the greenhouse can be reduced simulta- nology. Many wireless communication technologies have
neously. However, during chilly days, ventilation can been used in greenhouse environmental monitoring. For
adversely affect the maintenance of adequate temperatures example, some short-range communications technologies
within greenhouses. Stanghellini et al. (2008) reported that include Bluetooth, wireless local area network and ZigBee
compensating for the effect of depletion by artificially sup- technology, and some long-range communication technolo-
plying CO2 is more economically efficient than is making up gies include 2G, 3G, 4G networks. With evolving machine-to-
the loss by heating, under conditions where the crop pro- machine technology, many networking techniques have
duction reduction caused by depletion is comparable to the been developed, including the narrow-band internet of things
reduction caused by lower temperature resulting from venti- (NB-IoT), extended coverage global system mobile communi-
lation. Therefore, natural or forced ventilation methods cation, 5G networks, among others. The NB-IoT has both low
should be used to fertilise greenhouses only in warm-climate latency and low power consumption and can connect enor-
areas. mous devices (Ali, Shah, Farooq, Ghani, 2017); NB-IoT has
great potential in intelligent agriculture. The gateway nodes
transmit the monitored data to remote control centres
3. Monitoring and data processing of carbon through long-range communication technologies. Agricul-
dioxide concentrations tural experts can get access to environment information
through the internet, analyse the relationship between plant
Greenhouse buildings range from small to very large. In large growth status and environmental factors and then provide
buildings or multispan greenhouses, it is important to guidance for scientific decision. Through building reliable
monitor greenhouse environmental parameters to achieve models, this can achieve automatic optimal control or remote
better control. The traditional method of monitoring CO2 control by growers.
concentrations in the greenhouses is with a wired monitoring In CO2 monitoring system within greenhouse, the spatial
system which has many limitations, such as complicated distribution of CO2 concentration, which is relevant to crop
deployment and wiring, poor mobility and a high cost. Sensor growth, need to be considered. In these greenhouses with
technology development has resulted in wireless sensor net- good air mixing, the differences in vertical CO2 distribution are
works (WSN) that can be used for greenhouse environmental not significant. Therefore the CO2 sensors are placed at loca-
monitoring. tions where they are easy to operate. However, in some
A WSN-based monitoring system comprises many parts, complex greenhouses, the CO2 concentration and horizontal
including sensor nodes, gateway nodes and remote-control or vertical changes are complex. In general, the vertical CO2
platform. Sensor nodes are the base layer of a WSN, so the concentration distribution follows the order of the
precision of these sensors should satisfy agricultural re- ground > the canopy > the upper region (Heetae et al., 2015).
quirements. The common CO2 sensors can be divided into The horizontal CO2 concentration at the height of canopy is
optical sensors and electrochemical sensors (Neethirajan, relatively uniform. Therefore the sensor is commonly placed
Jayas, Sadistap, 2009). The optical sensors obtain CO2 con- approximately at the height of the crop canopy. In some
centration by detecting the changes of optical properties in greenhouses, some sensors are not stationary. For example,
response to CO2. The advantages of optical sensors are quick some greenhouses contained movable sensor nodes which
response times and high reliability; they include infrared were fixed in crop pot. These mobile nodes alter the dynamic
sensors (Wang, Niu, Zheng, Zheng, & Wang, 2016), some fibre network topology. Therefore, the data collected from the dy-
optic sensors (Colin, Carter, Wright, 2003) etc. It is important namic topology structure of a WSN cannot properly reflect the
for some infrared sensors to calibrate regularly with CO2 free greenhouse environment. Some researchers have used me-
gas because of source illumination and thermal drift. The fibre dium access control protocols to strengthen the relationships
optic sensor is not cross-sensitive to the presence of other between the dynamic nodes and the entire net for reliable
gases in the background but it needs frequent cleaning of data collection (Demirbas, Arora, Mittal, Kulathumani, 2006;
lenses for continuous operation. Electrochemical sensors Jiang et al., 2016).
detect CO2 based on the chemical reactions that occur on the Additionally, these collected sensor data might be related
sensor surface. They have smaller size, high sensitivity and or invalid. For example, some acquired data exist as redun-
lower power consumption and include metal oxide sensors dant information because of inappropriate layout of sensor
(Zhu et al., 2005), polymer-based sensors (Oho, Tonosaki, nodes, or some sensor data could be errors because of
Isomura, Ogura, 2002) etc. Metal oxide sensors are low cost damaged sensor nodes. To process related or invalid data,
but robustness and repeatability are poor. The polymer-based some methods to simplify redundancy can be applied. Rough
sensors can detect various constituents other than CO2 and set theory is a symbolic method for analysing the correlation
have a large detection range. Development of application- and dependence of data, and this method is easy to perform
specific CO2 sensors for greenhouse is important and new (Li & Zhang, 2002). Ji et al. (2016) and Li, Ji, Zhang, and Li (2017)
materials which can reduce the effects of dust, humidity and used rough set theory to analyse the correlation and de-
heat on detecting CO2 are critical. More information about CO2 pendency of CO2 concentration sensor data and reduced the
sensors can be found in some review articles (Millanalmaraz, redundancy before modelling. The mean impact value (MIV)
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 107

can also be used to reduce redundancy (Li et al., 2012). Wang, also beneficial for the development of precision greenhouse
Zhang, Liu, Li, and Liu (2013) used an MIV analysis to deter- and advanced control methods. In addition, most studies have
mine the relative importance of original input CO2 sensor data focused on optimising the economic benefits of controlling
and to reduce the amount of inputs by filtering inaccurate or CO2 enrichment. Throughout this section, we categorise the
unimportant inputs. Principal component analysis (PCA) is a various control methods proposed in the literature according
data processing method that can reduce the dimensionality of to the control method classification (Fig. 1).
the original data space by building a new set of synthetic
variables based on the original data space. These new vari- 4.1. Conventional control methods
ables are independent and maximally carry useful informa-
tion concerning the original data (Poekaew, Champrasert, & Many control theories and methods have been applied to the
Ieee, 2015). Li, Ji, et al. (2015) and Li, Zhang, et al. (2015) used CO2 enrichment of greenhouses. Conventional control
PCA to process the original sensor data and accepted first four methods have been proposed for the longest and are still the
principal components which contained over 89% of the orig- most widely used methods. Conventional control methods
inal information. can be categorised as classical control and modern control
In modern greenhouses, the use of WSNs makes remote types. It is important to build an accurate mechanistic model
monitoring and controlling of greenhouse environments that represents greenhouse CO2 concentration dynamics in a
possible. The coverage of WSNs is important and challenging, conventional control system. In this section, the mechanistic
particularly in greenhouses containing mobile sensor nodes, models of CO2 concentration in the greenhouse, classical
and the coverage directly affects the quality of monitoring control methods and modern control methods for CO2
data. Additionally, sensor data processing is also an important enrichment in the greenhouse are reviewed.
part for advancing CO2 control.
4.1.1. Mechanistic models of greenhouses for controlling
carbon dioxide
4. Control of carbon dioxide enrichment in Many researchers have specialised in constructing models of
the greenhouse greenhouse environments and crop growth. An accurate
mathematical model that represents greenhouse environ-
Greenhouse environmental control differs greatly from the mental dynamics is helpful for analysing the effects of a
comparable control of other physical systems. The environ- changing environment on crop growth and for adjusting
mental requirements of a greenhouse are typically more microclimate parameters to accurate, predicted levels.
complex and nonlinear (Taylor & Aerts, 2013), and the bio- Regarding greenhouse CO2 control, the goal is to decrease
logical systems inside have many substantial effects on their energy consumption while maintaining or even increasing
physical surroundings. The direct factors that affect the crop production. Therefore, models representing CO2 within
changes of CO2 in greenhouses include photosynthesis, crop greenhouses need to consider at least CO2 concentrations or
and soil respiration and ventilation. Generally, the photo- the rate of change of the CO2 content. In actual production, the
synthesis intensity of plants gradually increases to the highest CO2 concentration is coupled with other environmental pa-
value at noon when the light and temperature conditions are rameters and the crop itself, so models need to account for
best. And then, it decreases over time. The respiration in- these coupled parameters, such as temperature, humidity,
tensity also changes with a single peak curve as photosyn- and illumination as well as crop photosynthesis and respira-
thesis intensity. With the changes of photosynthesis and tion, among others. Further, because the distribution of CO2 in
respiration, the CO2 concentration in greenhouses usually the greenhouse has an influence on crop production, these
shows an irregular “U” shape over the day. The greenhouse models should accurately consider this distribution.
must be controlled to address rapidly changing wind speed, To optimally control CO2 in the greenhouse, greenhouse
solar loads, etc. Therefore, weather predictions and proactive microenvironment models need to consider these coupled
responses are important. The location of a greenhouse also environmental factors. Van Henten (1994, p. 159) described
has various effects on plant morphology and cost (Pasgianos the influence of photosynthesis and respiration of crops as
et al., 2003). In addition, the spatial uniformity of CO2 con- well as other plant processes on the greenhouse environment
centration is important to consider for improving the effect of by constructing a dynamic model that included the CO2 con-
CO2 fertilisation. The distribution of CO2 in the greenhouse is centration, temperature and humidity in the greenhouse.
more uniform when using perforated plastic tubes rather than Albright, Gates, Arvanitis, and Drysdale (2001) developed a
natural diffusion. Circulation fans can also make the CO2 model considering CO2 concentration and light intensity. Lee
distribution more uniform because of the intensity of free et al. (2012) developed a model that considers solar radia-
convection (Boulard, Roy, Pouillard, Fatnassi, Grisey, 2017). tion, relative humidity and CO2 production from the soil. In
Some control methods such as classical controls and some addition, Dayan, Dayan, Strassberg, and Presnov (2004) pro-
robust control methods control the CO2 just by reducing to posed a simple model that represents a rose-growing green-
zero the deviation between set points of variables of interest house as three vertically stacked horizontal segments; this
and measured values. Other optimal control methods that model can be used for calculating ventilation and for con-
consider the requirements of greenhouse behaviour, actuator trolling greenhouse CO2 levels.
capabilities, energy consumption, etc. and that predict envi- Additionally, it is important to develop control strategies to
ronmental parameter dynamics are increasingly interesting. quantitatively model crop growth and describe the relation-
Optimising economic benefits is welcomed by growers and is ship between crop growth and CO2 concentration. Farquhar,
108 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

Fig. 1 e Classification of CO2 control methods in the greenhouse.

Caemmerer, and Berry (1980) constructed a classic CO2 Most greenhouse environmental models are based on an
biochemical model to describe photosynthetic CO2 assimila- energy and mass balance (Pasgianos et al., 2003; Setiawan,
tion in leaves. Classical models such as the TOMGRO and Albright, Phelan, 2000). In the process of building green-
HORTISIM models can be used for CO2 control. The TOMGRO house CO2 models, the first step involves the qualitative
model can be used to analyse the quantitative relationship analysis of both the CO2 transmission status and the influ-
between tomato growth and temperature, CO2 concentra- encing factors inside and outside greenhouse, after which a
tions, radiation, etc. This model considers many factors and hypothesis is proposed. The CO2 balance equation considering
simulates the whole process of tomato growth. The HORTISIM the crop, soil, air, etc. in the greenhouse is then built. Last, the
model integrates environmental controls, greenhouse man- parameters in the equation are determined. For practical
agement, supporting decisions and so on. This model is suit- production, growers should construct suitable models ac-
able for a wide range of climate types. Aaslyng, Lund, Ehler, cording to their practical requirements, environment and
and Rosenqvist (2003) developed a general leaf model that control conditions. For example, the vertical distribution of
can be used to estimate leaf photosynthesis, respiration and CO2 does not need to be considered in a solar greenhouse in
absorption and that is applicable to different greenhouse which the air is thoroughly mixed, but in a precision green-
crops with few modifications. Dong (2005) predicted the CO2 house, even soil respiration needs to be considered. In
concentration using a complicated model based on the CO2 mathematical models, basic photosynthesis data are impor-
transmission mechanism; this model considers plant tran- tant for perfecting the models, so more experiments and hy-
spiration, photosynthesis, soil respiration within the green- pothesis testing are needed.
house, air exchange, temperature, humidity, etc. Kla € ring,
Hauschild, Heißner, and Bar-Yosef (2007) estimated the net 4.1.2. Classical control methods
uptake rate of CO2 by a net photosynthesis model that Classical control methods are typically applied to control in-
considered photosynthetically active radiation, air tempera- dustrial and agricultural systems. These methods are simple
ture, leaf area index and so on. but are evolving; they include on/off, P, PI, and PID control
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 109

methods. Usually, classical control methods emphasise con- rate, the amount of CO2 supplied, the CO2 concentration, etc.
trol performance more than economic profit. Most classical These parameters were used to investigate the efficiency of
controllers modify CO2 concentrations in accordance with a CO2 enrichment in an unventilated greenhouse.
set point. Albright and colleagues made efforts to explore the Robustness is a property of a system and refers to the
optimal CO2 concentration set points under different condi- preservation of certain metrics of control performance despite
tions (Both, Albright, Langhans, 1998; De Villiers, Langhans, variations in the controlled system. Robust control can make a
Both, Albright, Scholl, 1999; Ferentinos, Albright, Ramani, system work well in the presence of a disturbance or when
2000). However, optimal CO2 concentrations are affected by parameters are changing. Linker, Gutman, and Seginer (1999)
many factors, and many controller set points are defined by used two distinct control loops for the simultaneous control
growers' experiences. These classical methods can be inap- of CO2 and temperature: the first was a temperature control
plicable when unfavourable weather is frequent, and these loop that adjusted the ventilation, and the second was a CO2
methods exhibit relatively low robustness against new, future control loop that adjusted the CO2 enrichment. This method
greenhouse construction. can maintain both CO2 and temperature setpoints with low
On/off control regulates an actuator according to an upper errors.
and lower threshold within an appropriate range. Kozai et al. Optimum control is an important part of modern control
(2015) proposed a null-balance CO2 enrichment, which in- theory and is a method that identifies the optimum solution
volves a method that maintains greenhouse CO2 concentra- strategy under constrained conditions. The primary methods
tions at the outside level. This method controls whether the of optimum control include dynamic planning and the min-
CO2 container is on or off (according to the sensor data) and imum principle. Greenhouse control is a multiparametric
can produce expected results. Amir et al. (2005) added CO2 to a process, so optimum control is suitable; many studies have
greenhouse at a constant speed from a bottle with three investigated this (Alscher, Krug, & Liebig, 2001a, 2001b, 2001c;
modes for ventilation fans to decrease the temperature and Van Straten & van Henten, 2010; van Straten et al., 2010). For
humidity. As such, they used various combinations of green- CO2 enrichment in the greenhouse, optimal control should
house temperature, humidity and CO2 balance using an on-off find the balance between greenhouse production and CO2
method based on real-time sensor data. consumption or ecological benefits. Using an integrated
PID control is a feedback control method and is typically model, Pohlheim and Heissner (1997) predicted the CO2
used in many controlled systems because of its flexibility, concentration of a greenhouse. They used two modified
simplicity and good performance. In practical applications, evolutionary algorithms e the multi-strategy competition
PID systems usually select a single element or a combination evolutionary algorithm and the evolution strategy evolu-
of two or three elements according to the requirements of the tionary algorithm e to solve the optimisation problem. This
system. Sa  nchez-Molina et al. (2014) designed a methodology system achieved good results in the short term but needs
in which a CO2 storage system was controlled using a PI improvements for long-term results. Later, the same authors
controller and in which a CO2 enrichment system was proposed a new control method used for real-world green-
controlled using an on/off method. The PI controller can house climate control, but this method, which used online
maintain CO2 tank pressure for capturing CO2. Xiao, Feng, Zhi, optimisation, is not good because it requires a long compu-
and Ieee (2013) controlled the greenhouse climate using a PID tation time (Pohlheim & Heissner, 1999). Chalabi et al. (2002a)
system and adjusted the PID parameters based on a compu- and (2002b) controlled CO2 in a tomato greenhouse based on
tational fluid dynamics simulation. mathematical models and optimised the control strategy
These classical control methods have the merits of using successive quadratic programming (SQP) to maximise
simplicity and stability, but their limits are also apparent, the financial margin. Qian and Wang (2013) controlled CO2
especially in modern greenhouses for which complicated enrichment based on an analytical hierarchy process, and
mechanistic models are difficult to construct and validate. the outputs of the controller sent a signal that activated the
Researchers have published many control methods that on/off switch of the CO2 generator, fans and so on. Sa  nchez-
combine a classical controller with intelligent methods and Molina, Pe  rez, Rodrı́guez, Guzma  n, and Lo pez (2015) pre-
have produced good results (Gurban & Andreescu, 2014; sented a decision support system for the management of a
Gurban, Dragomir, Andreescu, 2014; Hu, Xu, Goodman, Zeng, sweet pepper greenhouse environment. In addition, Li, Ji,
2014; Wang et al., 2016; Zeng, Hu, Xu, Li, 2012). A detailed et al. (2015) and Li, Zhang, et al. (2015) predicted the rate of
description is in Section 4.3. photosynthesis for tomatoes at different growth stages with
partial least square regression and determined the optimal
4.1.3. Modern control methods CO2 concentration during the tomato growing season. This
In this section, we review the modern nonlinear control, optimal CO2 concentration can be used as a set point for
robust control and optimum control methods (Table 2). In the optimal control.
nonlinear control method, the control law drives a nonlinear The primary task of a conventional greenhouse control
system to follow a desired reference signal; the greenhouse system is to reliably model the greenhouse microclimate
CO2 control system is a classic nonlinear system, and (Linker, Seginer, Gutman, 1998). However, they are difficult to
nonlinear theory is used for CO2 control in the greenhouse. construct and validate and it is difficult to solve the compli-
Pasgianos et al. (2003) proposed a feedback/feed-forward cated mathematical model, because the greenhouse parame-
approach for ventilation control in the greenhouse and it ters interact and change as time progresses; in addition, the
could be used to adjust the CO2 in the greenhouse. Kuroyanagi growth processes of many crops still have not been charac-
et al. (2014) used a regression equation to estimate the leakage terised. The classical control method has its limitations in
110
Table 2 e Modern control methods in greenhouse CO2 enrichment.
Control method Specific method Other considered factors Controlled greenhouse Results Reference
Nonlinear control Feedback-feedforward approach Ti , To , S, W, Hr Greenhouse area of 1000m2, height Has good regulatory control Pasgianos et al. (2003)
4 m, in Arizona, USA. performance and set-point

b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9
tracking.
Regression analysis W, Lr Tomato greenhouse with a floor The efficiency of CO2 enrichment is Kuroyanagi et al. (2014)
area of 178m2, eave height of 4 m, on average 45.5%.
in Aichi, Japan.
Robust control Using two distinct control loops Ti , To , S, W Single-span, pitched-roof CO2 concentration and Linker et al. (1999)
greenhouse, no crop, at the temperature can be maintained at
Technion, Haifa, Israel. a set-point by balancing ventilation
and enrichment.
Optimal control Multifactorial compatible control Ti , S, etc. e Light-photosynthesis and Huang, Xu, Wei,
algorithm CO2 -photosynthesis response and Ieee, (2014)
curve fit well to field test data.
Modified controlled random search Ti , To , Hi , Ho , etc. Tomato greenhouse in Almeria, Fulfil the goal to design a Vanthoor et al. (2012)
optimisation algorithm Spain and De Bilt, The Netherlands. greenhouse that can yield the
largest annual Net Financial
Return.
Successive quadratic programming Ti , To , S, W Tomato greenhouse in Cardington, Produce optimum CO2 enrichment Bailey (2002)
method Bedfordshire, UK. schemes with differently priced
pure CO2 and natural gas.
Combining non-linear regression Ti , Li Tomato greenhouse in Produce a combination of an Wang, Zou, and
analysis in the Gauss-Newton Wellesbourne, UK. optimum CO2 concentration and Weerakkody (2014)
procedure temperature in different
environments.

Ti e air temperature in greenhouse,  C; To e air temperature outside the greenhouse,  C; Hi e air humidity in greenhouse, g m3; Ho e air humidity outside the greenhouse; Hr e the relative humidity; S e
the solar radiant energy, W; W e the wind speed, m s1; Lr e the CO2 leakage rate, kg m1; Li e illumination.
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 111

Fig. 2 e System framework for intelligent control CO2 in greenhouses.

modern optimal CO2 control, though an off-the-shelve on-off 4.2. Intelligent control methods
controller or PID controller can be operated relatively easily.
Nonlinear control is effective, but a complex mathematical Intelligent control models involve not only definite mathe-
process is necessary for analysis. Robust control and optimum matical models but also non-mathematical generalised
control are effective control methods for coupled and time- models and mixed models. Intelligent control involves more
varying greenhouse CO2 control systems. However, addi- learning and reasoning by data-driven or embedded expert
tional parameters are required for each different greenhouse knowledge. Therefore, intelligent control is a promising
system. Additionally, the robustness of controllers can method for systems whose mechanistic model is complicated,
contradict their dynamic performance, and most optimum and rapidly increasing research and applications have ensued
controls solve the optimisation problem only in the upper since the development of computers. Some intelligent con-
layer, such as by minimising the CO2 supply, increasing the trollers output their control signals directly, but the majority
grower's income, etc. (Ioslovich, Gutman, Linker, 2009; of intelligent control methods, which build relations between
Rodriguez et al., 2003; Van Henten, 1994). Model predictive CO2 concentrations and control targets or predict outputs
control (MPC) has been used in industrial control systems and corresponding to inputs such CO2 concentrations, introduce
for greenhouse temperature control (Pin ~o n, Camacho, set points to low-level controllers. In addition, intelligent
Kuchen, Pen ~ a, 2005) because MPC is easy to model, exhibits controllers can address unfavourable climatic conditions. If
large information redundancy to improve the robustness of there are additions of future greenhouse construction or
the system being controlled, and has good dynamic properties equipment elements, growers need to change only the control
due to rolling optimisation. Various MPC algorithms have algorithm and do not need to substantially change any hard-
been developed. However, the application of MPC for con- ware. Figure 2 shows a system framework for intelligent
trolling CO2 enrichment in the greenhouse is still lacking. control CO2 in greenhouses: the greenhouse environmental
Additional studies should aim in this direction. model and crop growth model presents the optimal CO2
112 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

concentration for growth; the intelligent management mod- ANNs based on optimisation algorithms, such as genetic
ule provides optimal trajectories for a planning period ahead algorithms (GAs) and particle swarm optimisation (PSO),
by considering comprehensively weather prediction, eco- perform better than do ANNs. GAs exhibit a better ability to
nomic conditions and disturbances; the controller provides solve complex optimisation search problems than do other
control signals based on the optimal CO2 trajectories and methods (Husmann & Tantau, 2001). PSO searches for an op-
measured CO2 concentration; and then actuators stimulate timum solution by tracking an optimum particle in the solu-
the physical system of greenhouse for tracking the optimal tion space, and PSO is easier to perform than are GAs.
trajectory. In this section, the application of intelligent control Integrating an optimisation algorithm into an ANN can
methods to greenhouse CO2 control is discussed, primarily improve the training speed of the ANN algorithm and can
including fuzzy logic control and artificial neural network avoid local minimisation of the ANN algorithm. Zhang, Chen,
control. Wang, and Dong (2007) used a modified neural network based
on a fuzzy genetic algorithm to model the tomato growth
4.2.1. Fuzzy logic control process in the greenhouse and further predict tomato growth
A fuzzy controller involves the fuzzification of inputs, fuzzy with 0.064 forecasting error. This method can also provide a
reasoning, fuzzy decision making, fuzzy judgement and the set point for low-level controllers. Chen (2010) proposed a
defuzzification of outputs. Fuzzy control methods have been greenhouse microclimate control system based on the com-
proposed for greenhouse CO2 control studies. Ardabili, bination of an improved particle swarm optimisation and a
Mahmoudi, Gundoshmian, and Roshanianfard (2016) used BPNN. This correlation coefficient between output value and
fuzzy logic to control a mushroom greenhouse and reported expected value was 0.985.
responses that were better than those without fuzzy logic. Neural network control has been successfully applied to
Compared with on/off controllers, fuzzy controllers exhibit CO2 control. However, some problems still exist. Neural
low bandpass, and the parameter changes follow set points network control is not very popular with engineers because of
very smoothly. Many researchers have controlled other the computational load. Despite some studies having been
greenhouse climatic parameters such as temperature and dedicated to reducing the number of neurons and simplifying
humidity based on fuzzy logic, and those methods can be networks and despite some progress having been made,
referenced for designing a CO2 controller because of the sim- additional studies are still necessary. Furthermore, improved
ilarity of the parameters (Castan ~ eda-Miranda, Ventura- network structures and learning schemes are needed to solve
Ramos, del Rocı́o Peniche-Vera, & Herrera-Ruiz, 2006; Lafont inherent problems (such as local minimums) such as deep
& Balmat, 2002; Trabelsi, Lafont, Kamoun, Enea, 2007). learning and extreme learning (Schmidhuber, 2015), which
Fuzzy logic control has many advantages for greenhouse have recently been very actively addressed.
CO2 control, but there is still room for improvement (Precup &
Hellendoorn, 2011). First, general fuzzy controllers rather than 4.2.3. Others
specific controllers which apply only to certain greenhouses A support vector machine (SVM) is a supervised learning
must be designed. Second, the conventional membership model that is generally used for pattern recognition, classifi-
function is identified by relying on the grower's experience, so cation and regression analysis. Li et al. (2017) developed a
developing adjustment and learning techniques is important. photosynthesis prediction model based on an SVM and used
Third, a low-cost and more accurate control method is improved particle swarm optimisation to optimise the pa-
preferred for greenhouse control. rameters of the SVM model and the correlation coefficient was
0.993. Then a remote management platform sends the control
4.2.2. Artificial neural network control signal to a solenoid valve based on the optimal CO2 concen-
Artificial neural networks (ANNs) constitute a method that tration calculated by the model. Biomimetic pattern recogni-
simulates the human brain's microstructure and function. tion is a better method and can be used to solve problems that
ANNs have strong self-learning ability and have been used for conventional methods cannot, such as recognising one object
controlling greenhouse CO2 concentrations. Some researchers from an infinite number of samples (Chen et al., 2016). This
have predicted photosynthetic rates and calculated optimum method will be useful for increasingly complicated green-
CO2 concentrations using a BPNN for CO2 fertilisation; re- house control systems.
searchers have used different input data preprocessing Reinforcement learning is a learning algorithm. In this
methods, such as mean impact value analysis or principal method, an agent can improve an action according to a strong
component analysis, or no preprocessing methods (Jiang, Li, environmental signal. Chen, Tang, and Shen (2011) controlled
Zhang, Sha, Ji, 2015; Li, Ji, et al., 2015, Li, Zhang, et al., 2015; CO2 concentrations in the greenhouse based on Q-learning
Wang et al., 2013). Fitz-Rodriguez and Giacomelli (2009) and integrated case-based reasoning into the Q-learning pro-
modelled the weekly fluctuations of fruit-related parameters cess to reduce the amount of computation and systematic
in a tomato greenhouse with a dynamic neural network and trial-and-error risk. This method drives greenhouse climatic
reported that the accurate prediction results of the models conditions to the approximate desired conditions by coordi-
could be used for greenhouse climate control. Salazar, Rojano, nating coupled control actions. Reinforcement learning can be
and Lopez (2009) developed two linked ANN models for pre- considered optimal adaptive control. This method resolves
dicting CO2 and photosynthesis; one of the inputs of the the problem of general optimum controllers that are designed
photosynthesis prediction model arose from the output of the offline and must solve the difficult Hamilton-Jacobi-Bellman
CO2 prediction model. In future, a better control can be applied equation; this method also simultaneously addresses adap-
according to the prediction of photosynthesis. tive controllers, which are often not considered optimal
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 113

(Ruano et al., 2014). Deep learning has been used to find op- experiences uncertainty factors, such as external distur-
timum solutions for systems with multiple degrees of bances or randomly delayed measurements. By adopting the
freedom (Rezaee & Golpayegani, 2012). Liu et al. (2015) pre- self-learning ability of a neural network, a neural network PID
dicted the temperature in the greenhouse using an extreme control system has better adaptability and robustness.
learning approach, and the resultant training speed was Haigen, Lihong, and Ruihua (2010) and Zeng et al. (2012)
significantly better than that of the back propagation neural designed a neural PID controller for temperature and humid-
network, Elman neural network and support vector machine ity control in the greenhouse and it had good adaptability.
methods. Greenhouse CO2 control systems can therefore be These types of controllers can also be used to control CO2
controlled using reinforcement learning methods and addi- enrichment.
tional studies are needed. A fuzzy neural network (FNN) can combine the advantages
An apparent advantage of intelligent control methods is of fuzzy logic and neural networks. FNN methods modify the
that they are powerful nonlinear estimators when the func- membership functions and inference rules of a fuzzy infer-
tional relationships between outputs and inputs are compli- ence system using an ANN, and using fuzzy inference rules,
cated or unknown. Artificial intelligent models can predict FNN build forward propagation networks. The learning ability
outcomes without any need to understand the underlying of the ANN can avoid the subjectivity of control rules and the
mechanism. Therefore, intelligent control methods can yield membership functions of fuzzy control such that FNN is more
better results for greenhouse CO2 enrichment control, partic- adaptive in nonlinear and time-varying greenhouse systems.
ularly in modern, complex greenhouses. However, there are Zhang, Wang, Huang, and Guo (2011) designed a fuzzy neural
limits to intelligent control. Compared with PID control, fuzzy network control method to directly control CO2 concentration,
control has a faster response and better robustness. However, so it could modify the CO2 level at 1200 ppm within 5 s.
the control precision of fuzzy logic is not good, so this method Lachouri, Lafifi, Mansouri, and Belmeguenai (2016) proposed
can only approximate results (Li & Du, 2004). Artificial neural an adaptive neural-fuzzy inference system for the prediction
network control can self-organise, self-learn and use of CO2 concentrations during tomato seedling growth, the
nonlinear dynamic processing. However, stability analysis is predictions had 98% accuracy compared to 94% using only
difficult because the neural network is nonlinear, and the neural network. Qaddoum, Hines, and Iliescu (2013) proposed
convergence speed is slow if the initial weights are inappro- an evolving fuzzy neural network (EFuNN) that incorporated
priate. In practical application, intelligent control methods are fuzzy logic and a neural network, which further evolved ac-
often combined with other control methods. cording to the evolving connectionist systems principle for
tomato yield prediction. This method had a low root mean
4.3. Hybrid control methods square error and less computational load than did ANN
methods.
Although intelligent control methods can control greenhouses Hybrid control methods benefit from the advantages of
very well, these methods still have some inherent limitations each integrated method; however, these methods also display
when used individually. Hybrid control methods can benefit the inherent limitations of each component method. For
from the qualities of the two integrated methods. The fusion example, intelligent control methods require experience or a
of intelligent methods at high levels and conventional con- particular amount of data to train themselves, and the
trollers at low levels is common; examples include fuzzy PID amount of computation is considerable. In a cost-limited
and neural network PID control methods. Additionally, neural greenhouse, designers must choose other appropriate con-
networks integrated with fuzzy logic have also been used for trol methods. Therefore, modifying these intelligent control
controlling CO2 enrichment in the greenhouse. algorithms and improving the calculation ability of current
A fuzzy PID control method combines the advantages of a computers is a very interesting field.
fuzzy logic control method and a PID control method. PID
control uses a simple algorithm and can track the set point
accurately. Fuzzy logic control methods can exploit expert 5. New challenges and perspectives
experience and exhibit fast modification speed (Esfandyari,
Fanaei, & Zohreie, 2013). The fuzzy PID control method not We have thoroughly reviewed automatic CO2 enrichment
only eliminates the cycle oscillation limitation but also fully processes within greenhouses. However, modern green-
eliminates systematic error, and this method has been houses present new challenges. This section reviews these
incorporated into controllers for CO2 adjustment. Kolokotsa, challenges and proposes both new trends in automatic CO2
Saridakis, Dalamagkidis, Dolianitis, and Kaliakatsos (2010) enrichment processes in the greenhouse and a new symbiosis
developed a fuzzy proportional-derivative (fuzzy-PD) control form for greenhouse cultivation.
method to control the CO2 concentration and radiation inside In greenhouse control, it is difficult to mathematically ex-
a greenhouse and the CO2 concentration could maintain at press human factors such as pruning, harvesting and so on.
1000 ppm for most of the day. Similarly, Wang et al. (2016) Therefore, designing or evaluating CO2 controllers based on
designed an automatic CO2 fertilisation system based on a mathematical models is difficult. In addition, accurate math-
fuzzy PID method and the CO2 concentration was maintained ematical models of greenhouses include many variables. For
at desired levels from 9:00 AM to 4:00 PM. optimal online control of CO2 enrichment, additional model-
Similar to the fuzzy PID method, neural network PID ling research should focus on the applicability in controllers.
schemes alter PID gain parameters using neural networks. Intelligent control methods have great potential for optimal
Classical PID controllers are limited when the real system CO2 control. Compared to fuzzy logic control methods, neural
114 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

network methods have self-learning ability and can optimise 2017; Wongkiew, Hu, Chandran, Lee, Khanal, 2017). Our team
parameters in the control process. Neural network-based has designed a fish, vegetable and mushroom symbiotic sys-
methods can predict the weather to overcome environ- tem. This symbiotic greenhouse is composed of an aquacul-
mental influences and are more appropriate for modern ture unit, mushroom farming area, vegetable cultivation area,
greenhouse CO2 enrichment control. With the development of water processing unit, etc. The fish and mushrooms release
neural network algorithms such as deep learning, computa- CO2 via respiration, and the plants absorb CO2 by photosyn-
tional speed still needs to be improved for typical applications thesis. The carbon cycle joins the symbiotic system. The CO2 is
with respect to the control of greenhouse CO2 enrichment. an important form within the carbon cycle. The mushroom
Analysing the spatial distribution of CO2 concentration farming area is isolated from the other farming area because
that may be relevant to crop growth is complex. Computa- of its critically different growth environment requirements.
tional fluid dynamics (CFD) involves the science of simulating The only link is the gas exchange between the two areas by a
flow fields based on the conservation equations of mass and ventilator. Mushrooms are sensitive to CO2 concentrations,
energy. CFD is suitable for analysing non-uniform data and and different life stages of mushrooms have different re-
has been successfully applied to designs for industrial sys- quirements and tolerances for CO2 concentrations. CO2 is also
tems, building construction, aerospace and so on. CFD can an important parameter for hydroponic vegetable growth, so
also be used to analyse the CO2 distribution within a green- the CO2 concentration needs to be controlled accurately. An
house for optimal control (Boulard et al., 2017). However, these appropriate CO2 control method can make this symbiotic
types of study are still rare. In the future, regarding precision system produce both large profits and environmental benefits.
multispan greenhouse production, it is important to consider In symbiotic greenhouses, CO2 concentrations impact plant
the CO2 distribution within a greenhouse for optimal control. and mushroom growth, and the optimal CO2 balance between
In current greenhouse production, some advanced controllers different areas is sensitive. We need to control CO2 concen-
such as neural and fuzzy controllers are still based on PID. CFD trations by considering the different growth stages of both
can simulate greenhouse microclimate situations and can mushrooms and plants, the interplay between different
further guide the modification of controller parameters. For planting areas, temperature, relative humidity, ventilation,
PID controllers, CFD can reduce the losses caused by modi- etc. In addition, the distribution of CO2 also needs to be
fying PID values on site. CFD can also be used to analyse the considered for precision cultivation. The ultimate goal is to
control effect by other controllers such as intelligent control- achieve a CO2 balance such that the need for additional CO2
lers for improving the performance of CO2 controllers in the enrichment is limited or even non-existent. Additional studies
greenhouse. include designing systems that produce minimum carbon
Like all general automated systems, a CO2 enrichment emissions by building a biomass energy transition plant con-
control system requires a measurement system, a processing taining appropriate controls. These endeavours will greatly
system and actuators. It is more difficult to evaluate the pro- contribute to sustainable greenhouse cultivation.
cess of automatic CO2 enrichment in the greenhouse than in
industrial systems because of the uncertainty of biological
factors. Actually, more feed-back from the crops is needed for 6. Conclusions
optimal control. The crop physiology and morphology need to
be considered in order to control CO2 and other climate vari- Greenhouse climate control is currently one of the most
ables according to real crop needs. Modifying CO2 concentra- interesting fields in precision agriculture. CO2 concentration is
tion based on the available information on crop processes can an important parameter and is associated with temperature,
save energy. Until now, the benchmarking of controllers has humidity, light intensity, etc. Furthermore, CO2 enrichment is
generally involved a mathematical control index; setting op- not harmonious with ventilation. Therefore, it is difficult to
timum environmental values and evaluating greenhouse control the process of CO2 enrichment because a trade-off
control systems has relied mostly on experience or a subjec- exists between all of these parameters. However, CO2 is an
tive feeling instead of the “comfort” of the organism cultivated important carbon source for plant photosynthesis in the
in the greenhouse in most cases, although some greenhouse greenhouse. By properly regulating greenhouse CO2 concen-
production evaluation methods based on machine vision have trations, the quality of production can be ensured, and
been published (Niedbala et al., 2016; Zaborowicz, Boniecki, precious energy can be conserved.
Koszela, Przybylak, Przybyl, 2017). It is important to under- In highly automated greenhouses, the exhaust gas from
stand crop growth processes and crop responses to the the combustion of fossil fuels or from renewable energy can
ambient environment to better evaluate and control green- be used as a supply of CO2, but this supply should be used in
houses. Although there are particular growth chamber ex- conjunction with heat storage for optimal economic and
periments which give us the responses of the plants to environmental benefits. Pure CO2 is also popular due to its
particular inputs, the information about crop growth is convenience. Compost is a suitable choice for increasing
limited. Additional studies on the correlations between production in traditional greenhouses. The development of
physiological plant processes and the growth environment WSNs has promoted the automatic and optimal control of CO2
are necessary to achieve the objective of sustainable green- in the greenhouse. More accurate and low-cost sensors can
house control. further improve the automation level of greenhouses. Intelli-
Recently, aquaponic systems have received significant gent control methods have great potential for optimising CO2
attention in the scientific community (Cerozi & Fitzsimmons, enrichment control because of their self-learning abilities, but
2017; Ru et al., 2017; Suhl et al., 2016; Van Ginkel, Igou, & Chen, these methods also have relatively high requirements of
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 115

computational speed. Under unfavourable climatic condi- strategies, growth models and input data.
tions, intelligent algorithms can balance the cost of CO2 gen- Gartenbauwissenschaft, 66(3), 105e114.
erators and greenhouse production, providing optimal control Alscher, G., Krug, H., & Liebig, H. P. (2001b). Optimisation of CO2
and temperature control in greenhouse crops by means of
strategies.
growth models at different abstraction levels II. Growth
Machine vision technology has been studied for evaluating models and parameter generation for lettuce crops.
plant growth in the greenhouse. Integrating machine vision Gartenbauwissenschaft, 66(4), 153e163.
evaluation results into controllers would allow better balance Alscher, G., Krug, H., & Liebig, H. P. (2001c). Optimisation of CO2
between greenhouse production and CO2 consumption. In and temperature control in greenhouse crops by means of
precision greenhouse cultivation, the distribution of CO2 growth models at different abstraction levels III. Simulation
concentration needs to be considered because CO2 can influ- and optimisation with the combined model.
Gartenbauwissenschaft, 66(5), 213e218.
ence crop growth as well as economic benefits. CFD has been
Amir, R., Teitel, M., & Shemer, L. (2005). CO2 enrichment in a fan-
used to simulate temperature dynamics within the green- ventilated greenhouse under different ventilation modes. In
house and constitutes a promising method for accurate G. van Straten, G. P. A. Bot, W. T. M. van Meurs, &
analysis of the distribution of CO2. Symbiotic greenhouses L. M. F. Marcelis (Eds.), Proceedings of the International conference
have great economic and environmental significance for in- on sustainable greenhouse systems (Vol. 1 and 2, pp. 619e624).
dustrial greenhouses, and precious control of CO2 in symbiotic Leuven 1: International Society Horticultural Science.
Ardabili, S. F., Mahmoudi, A., Gundoshmian, T. M., &
greenhouses is worthy of in-depth investigations.
Roshanianfard, A. (2016). Modeling and comparison of fuzzy
and on/off controller in a mushroom growing hall.
Measurement, 90, 127e134. https://doi.org/10.1016/
Conflicts of interest j.measurement.2016.04.050.
Bailey, B. J. (2002). Optimal control of carbon dioxide enrichment
None. in tomato greenhouses. In S. Chen, & T. T. Lin (Eds.), Acta
horticulturae: Vol. 578. Proceedings of the International symposium
on design and environmental control of tropical and subtropical
greenhouses (pp. 63e69). Leuven 1: International Society
Acknowledgments Horticultural Science. https://doi.org/10.17660/
ActaHortic.2002.578.6.
The work was supported by EU FP7 Framework Program Both, A. J., Albright, L. D., & Langhans, R. W. (1998). Coordinated
“Innovative model and demonstration based water manage- management of daily PAR integral and carbon dioxide for
ment for resource efficiency in integrated multitrophic agri- hydroponic lettuce production. In L. F. M. Marcelis (Ed.), Second
International symposium on models for plant growth, environmental
culture and aquaculture systems” (619137) and International
control and farm management in protected cultivation (pp. 45e51).
technology cooperation of China (2015DFA00090). The authors
Leuven 1: International Society Horticultural Science.
appreciate the reviewer's constructive comments and impor- Boulard, T., Roy, J.-C., Pouillard, J.-B., Fatnassi, H., & Grisey, A.
tant suggestion of Professor S. F. Du. The authors would like to (2017). Modelling of micrometeorology, canopy transpiration
thank American Journal Experts for providing an English- and photosynthesis in a closed greenhouse using
language edit of this article. computational fluid dynamics. Biosystems Engineering, 158,
110e133. https://doi.org/10.1016/j.biosystemseng.2017.04.001.
Castan~ eda-Miranda, R., Ventura-Ramos, E., del Rocı́o Peniche-
Appendix A. Supplementary data Vera, R., & Herrera-Ruiz, G. (2006). Fuzzy greenhouse climate
control system based on a field programmable gate array.
Biosystems Engineering, 94(2), 165e177. https://doi.org/10.1016/
Supplementary data related to this article can be found at
j.biosystemseng.2006.02.012.
https://doi.org/10.1016/j.biosystemseng.2018.04.018. Cerozi, B. S., & Fitzsimmons, K. (2017). Phosphorus dynamics
modeling and mass balance in an aquaponics system.
references Agricultural Systems, 153, 94e100. https://doi.org/10.1016/
j.agsy.2017.01.020.
Chalabi, Z. S., Biro, A., Bailey, B. J., Aikman, D. P., &
Cockshull, K. E. (2002a). Optimal control strategies for carbon
Aaslyng, J. M., Lund, J. B., Ehler, N., & Rosenqvist, E. (2003). dioxide enrichment in greenhouse tomato crops e Part 1:
IntelliGrow: A greenhouse component-based climate control Using pure carbon dioxide. Biosystems Engineering, 421e431.
system. Environmental Modelling & Software, 18(7), 657e666. https://doi.org/10.1006/bioe.2001.0069.
https://doi.org/10.1016/s1364-8152(03)00052-5. Chalabi, Z. S., Biro, A., Bailey, B. J., Aikman, D. P., & Cockshull, K. E.
Albright, L. D., Gates, R. S., Arvanitis, K. G., & Drysdale, A. E. (2001). (2002b). Optimal control strategies for carbon dioxide
Environmental control for plants on Earth and in space. enrichment in greenhouse tomato crops, part II: Using the
[Article]. IEEE Control Systems Magazine, 21(5). https://doi.org/ exhaust gases of natural gas fired boilers. Biosystems Engineering,
10.1109/37.954518, 28-þ. 81(3), 323e332. https://doi.org/10.1006/bioe.2001.0020.
Ali, A., Shah, G. A., Farooq, M. O., & Ghani, U. (2017). Technologies Chau, J., Sowlati, T., Sokhansanj, S., Preto, F., Melin, S., & Bi, X.
and challenges in developing machine-to-machine (2009). Techno-economic analysis of wood biomass boilers for
applications: A survey. [Review]. Journal of Network and the greenhouse industry. Applied Energy, 86(3), 364e371.
Computer Applications, 83, 124e139. https://doi.org/10.1016/ https://doi.org/10.1016/j.apenergy.2008.05.010.
j.jnca.2017.02.002. Chen, S. (2010). Greenhouse environmental monitoring system
Alscher, G., Krug, H., & Liebig, H. P. (2001a). Optimisation of CO2 with IPSO e BPNN. Computer Measurement & Control, 18(3),
and temperature control in greenhouse crops by means of 601e604. https://doi.org/10.16526/j.cnki.11-4762/
growth models at different abstraction levels e I. Control tp.2010.03.063, 610.
116 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

Chen, Y., Jia, H., Li, W. J., Zhou, X. Q., Dong, X. L., Zhang, L. P., et al. algorithms for greenhouse climate control. In 2014 IEEE 9th
(2016). Progress in research and application of biomimetic International Symposium on Applied Computational Intelligence and
pattern recognition technology. CAAI Transactions on Intelligent Informatics (Saci) (pp. 79e83). https://doi.org/10.1109/
System, 11(1), 1e14. https://doi.org/10.11992/tis.201506011. SACI.2014.6840039.
Chen, F., Tang, Y.-N., & Shen, M.-Y. (2011). Coordination control Gurban, E. H., Dragomir, T. L., & Andreescu, G. D. (2014). Greenhouse
of greenhouse environmental factors. International Journal of climate control enhancement by using genetic algorithms.
Automation and Computing, 8(2), 147e153. https://doi.org/ Control Engineering and Applied Informatics, 16(3), 35e45.
10.1007/s11633-011-0567-3. Haigen, H., Lihong, X., & Ruihua, W. (2010). Nonlinear adaptive
Colin, F., Carter, T. J. N., & Wright, J. D. (2003). Modification of a neuro-PID controller design for greenhouse environment
piezo-optical gas dosimeter system towards continuous gas based on RBF network. [Conference paper]. In 2010
sensing: A feasibility study with carbon dioxide. Sensors and International Joint Conference on Neural Networks (IJCNN 2010).
Actuators B: Chemical, 90(1), 216e221. https://doi.org/10.1016/ https://doi.org/10.1109/ijcnn.2010.5596982, 7 pp. e 7 pp.
S0925-4005(03)00031-5. Hao, F., Shen, M., He, Y., & Feng, L. (2014). Three dimensional
Dayan, J., Dayan, E., Strassberg, Y., & Presnov, E. (2004). steady simulation of microclimate pattern inside single plastic
Simulation and control of ventilation rates in greenhouses. greenhouse using computational fluid dynamics. [article].
Mathematics and Computers in Simulation, 65(1e2), 3e17. https:// Nongye Jixie Xuebao ¼ Transactions of the Chinese Society for
doi.org/10.1016/j.matcom.2003.09.017. Agricultural Machinery, 45(9), 297e304.
Demirbas, M., Arora, A., Mittal, V., & Kulathumani, V. (2006). A Heetae, K., Chiho, K., Eungyu, C., Jin, B. O., Yongcheol, Y., &
fault-local self-stabilizing clustering service for wireless ad Hyeontae, K. (2015). The effect of thermal conditions on CO2
hoc networks. [Article]. IEEE Transactions on Parallel and distribution in a greenhouse. Tropical Agricultural Research,
Distributed Systems, 17(9), 912e922. https://doi.org/10.1109/ 26(4), 714.
tpds.2006.113. Huang, X. Z., Xu, L. H., Wei, R. H., & Ieee. (2014). An optimized control
Deryng, D., Elliott, J., Folberth, C., Muller, C., Pugh, T. A. M., method of multifactor for greenhouse microclimate based on
Boote, K. J., et al. (2016). Regional disparities in the beneficial crop photosynthesis rate. In 2014 15th International conference on
effects of rising CO2 concentrations on crop water sciences & techniques of automatic control & computer engineering (pp.
productivity. Nature Climate Change, 6(8). https://doi.org/ 482e487). https://doi.org/10.1109/STA.2014.7086755.
10.1038/nclimate2995, 786-þ. Husmann, H. J., & Tantau, H. J. (2001). Integrated optimization of
De Villiers, D. S., Langhans, R. W., Both, A. J., Albright, L. D., & energy supply systems in horticulture using genetic
Scholl, S. S. (1999). Intermittent CO2 enrichment for algorithms. Computers and Electronics in Agriculture, 31(1),
hydroponic lettuce production. [Meeting]. Hortscience, 34(3), 47e59. https://doi.org/10.1016/S0168-1699(00)00173-3.
532. Hu, H. G., Xu, L. H., Goodman, E. D., & Zeng, S. W. (2014). NSGA-II-
Dion, L. M., Lefsrud, M., & Orsat, V. (2011). Review of CO2 recovery based nonlinear PID controller tuning of greenhouse climate
methods from the exhaust gas of biomass heating systems for for reducing costs and improving performances. Neural
safe enrichment in greenhouses. Biomass and Bioenergy, 35(8), Computing & Applications, 24(3e4), 927e936. https://doi.org/
3422e3432. https://doi.org/10.1016/j.biombioe.2011.06.013. 10.1007/s00521-012-1312-8.
Dion, L. M., Lefsrud, M., Orsat, V., & Cimon, C. (2013). Biomass Ioslovich, I., Gutman, P. O., & Linker, R. (2009). Hamilton-Jacobi-
gasification and syngas combustion for greenhouse CO2 Bellman formalism for optimal climate control of greenhouse
enrichment. [Article]. Bioresources, 8(2), 1520e1538. https:// crop. [Article]. Automatica, 45(5), 1227e1231. https://doi.org/
doi.org/10.15376/biores.8.2.1520-1538. 10.1016/j.automatica.2008.12.024.
Dong, R. T. (2005). Simulation model of CO2 concentration in a sunlight Jaffrin, A., Bentounes, N., Joan, A. M., & Makhlouf, S. (2003).
greenhouse. Dissertation. China Agricultural University. Landfill biogas for heating greenhouses and providing carbon
Du, J., Lin, X. Y., & Zhang, Y. S. (2004). Affecting factors of CO2 dioxide supplement for plant growth. [Article]. Biosystems
evolution from biodegradation of agricultural organic wastes. Engineering, 86(1), 113e123. https://doi.org/10.1016/s1537-
Chinese Journal of Applied Ecology, 15(3), 501e505. https:// 5110(03)00110-7.
doi.org/10.13287/j.1001-9332.2004.0109. Jerzak, W. (2014). Emissions of NOx and CO from natural gas
Esfandyari, M., Fanaei, M. A., & Zohreie, H. (2013). Adaptive fuzzy combustion with adding CO2 at varying distances from the
tuning of PID controllers. [Article]. Neural Computing & Applications, burner. [Article]. Rocznik Ochrona Srodowiska, 16, 148e160.
23, S19eS28. https://doi.org/10.1007/s00521-012-1215-8. Jiang, Y. Q., Li, T., Zhang, M., Sha, S., & Ji, Y. H. (2015). WSN-based
Farquhar, G. D., Caemmerer, S. V., & Berry, J. A. (1980). A control system of CO2 concentration in greenhouse. Intelligent
biochemical-model of photosynthetic CO2 assimilation in Automation & Soft Computing, 21(3), 285e294. https://doi.org/
leaves of C-3 species. [Article]. Planta, 149(1), 78e90. https:// 10.1080/10798587.2015.1015782.
doi.org/10.1007/bf00386231. Jiang, J. A., Wang, C. H., Liao, M. S., Zheng, X. Y., Liu, J. H.,
Ferentinos, K. P., Albright, L. D., & Ramani, D. V. (2000). Optimal Chuang, C. L., et al. (2016). A wireless sensor network-based
light integral and carbon dioxide concentration combinations monitoring system with dynamic convergecast tree algorithm
for lettuce in ventilated greenhouses. [Article]. Journal of for precision cultivation management in orchid greenhouses.
Agricultural Engineering Research, 77(3), 309e315. https:// [Article]. Precision Agriculture, 17(6), 766e785. https://doi.org/
doi.org/10.1006/jaer.2000.0593. 10.1007/s11119-016-9448-7.
Fitz-Rodriguez, E., & Giacomelli, G. A. (2009). Yield prediction and Ji, Y. H., Jiang, Y. Q., Li, T., Zhang, M., Sha, S., & Li, M. Z. (2016). An
growth modes characterization of greenhouse tomatoes with improved method for prediction of tomato photosynthetic
neural networks and fuzzy logic. [Article]. Transactions of the rate based on WSN in greenhouse. International Journal of
Asabe, 52(6), 2115e2128. https://doi.org/10.13031/2013.29200. Agricultural and Biological Engineering, 9(1), 146e152. https://
Frantz, J. M. (2011). Elevating carbon dioxide in a commercial doi.org/10.3965/j.ijabe.20160901.1243.
greenhouse reduced overall fuel carbon consumption and Jin, C. W., Du, S. T., Wang, Y., Condon, J., Lin, X. Y., & Zhang, Y. S.
production cost when used in combination with cool (2009). Carbon dioxide enrichment by composting in
temperatures for lettuce production. [Editorial material]. greenhouses and its effect on vegetable production. Journal of
Hortechnology, 21(5), 647e651. Plant Nutrition and Soil Science-Zeitschrift Fur Pflanzenernahrung
Gurban, E. H., & Andreescu, G. D. (2014). Comparison of modified Und Bodenkunde, 172(3), 418e424. https://doi.org/10.1002/
smith predictor and PID controller tuned by genetic jpln.200700220.
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 117

Kla€ ring, H. P., Hauschild, C., Heißner, A., & Bar-Yosef, B. (2007). 17th International conference on advanced communication
Model-based control of CO2 concentration in greenhouses at technology (pp. 133e137). New York: IEEE. https://doi.org/
ambient levels increases cucumber yield. Agricultural and 10.1109/ICACT.2015.7224772.
Forest Meteorology, 143(3e4), 208e216. https://doi.org/10.1016/ Li, M., & Zhang, H.-G. (2002). Research on the method of neural
j.agrformet.2006.12.002. network modeling based on rough sets theory. Acta Automatica
Kolokotsa, D., Saridakis, G., Dalamagkidis, K., Dolianitis, S., & Sinica, 28(1), 27e33. https://doi.org/10.16383/j.aas.2002.01.004.
Kaliakatsos, I. (2010). Development of an intelligent indoor Li, T., Zhang, M., Ji, Y. H., Sha, S., Jiang, Y. Q., & Li, M. Z. (2015).
environment and energy management system for Management of CO2 in a tomato greenhouse using WSN and
greenhouses. Energy Conversion and Management, 51(1), BPNN techniques. International Journal of Agricultural and
155e168. https://doi.org/10.1016/j.enconman.2009.09.007. Biological Engineering, 8(4), 43e51. https://doi.org/10.3965/
Kozai, T., Kubota, C., Takagaki, M., & Maruo, T. (2015). j.ijabe.20150804.1572.
Greenhouse environment control technologies for improving Li, J., Zhou, J.-M., Duan, Z.-Q., Du, C.-W., & Wang, H.-Y. (2007).
the sustainability of food production. Acta Horticulturae, 1107, Effect of CO2 enrichment on the growth and nutrient uptake of
1e14. https://doi.org/10.17660/ActaHortic.2015.1107.1. tomato seedlings. Pedosphere, 17(3), 343e351. https://doi.org/
Kuroyanagi, T., Yasuba, K., Higashide, T., Iwasaki, Y., & 10.1016/s1002-0160(07)60041-1.
Takaichi, M. (2014). Efficiency of carbon dioxide enrichment in Millanalmaraz, J. R., Guevaragonzalez, R. G., Troncoso, R. R.,
an unventilated greenhouse. Biosystems Engineering, 119, Osorniorios, R. A., & Torrespacheco, I. (2009). Advantages
58e68. https://doi.org/10.1016/j.biosystemseng.2014.01.007. and disadvantages on photosynthesis measurement
Lachouri, C. E., Lafifi, M. M., Mansouri, K., & Belmeguenai, A. techniques: A review. African Journal of Biotechnology, 8(25),
(2016). Adaptive neuro-fuzzy inference systems for modeling 7340e7349.
greenhouse climate. International Journal of Advanced Computer Mohamed, S. J., Jellings, A. J., & Fuller, M. P. (2013). Positive effects
Science and Applications, 7(1), 96e100. of elevated CO2 and its interaction with nitrogen on safflower
Lafont, F., & Balmat, J. F. (2002). Optimized fuzzy control of a physiology and growth. [Article]. Agronomy for Sustainable
greenhouse. Fuzzy Sets and Systems, 128(1), 47e59. pii: S0165- Development, 33(3), 497e505. https://doi.org/10.1007/s13593-
0114(01)00182-8. 013-0137-x.
Lee, D., Lee, K., Cho, Y. J., Choi, J.-M., Kim, H.-J., & Chung, S.-O. Montero, J. I., Stanghellini, C., & Castilla, N. (2009). Greenhouse
(2012). Optimal CO2 enrichment considering emission from technology for sustainable production in mild winter climate
soil for cucumber greenhouses. Korean Journal of Horticultural areas: Trends and needs. In Y. Tuzel, G. B. Oztekin, &
Science and Technology, 30(5), 501e508. https://doi.org/10.7235/ M. K. Meric (Eds.), International symposium on strategies towards
hort.2012.12084. sustainability of protected cultivation in mild winter climate (Vol.
Li, Y. X., & Du, S. F. (2004). Advances of intelligent control 807, pp. 33e44). Leuven 1: Int Soc Horticultural Science.
algorithm of greenhouse environment in China. Transactions of https://doi.org/10.17660/actahortic.2009.807.1.
the CSAE, 02, 267e272. https://doi.org/10.3321/j.issn:1002- Neethirajan, S., Jayas, D. S., & Sadistap, S. (2009). Carbon dioxide
6819.2004.02.064. (CO2) sensors for the agri-food industry-a review. [Review].
Li, T., Ji, Y. H., Zhang, M., & Li, M. Z. (2017). Determining optimal Food and Bioprocess Technology, 2(2), 115e121. https://doi.org/
CO2 concentration of greenhouse tomato based on PSO-SVM. 10.1007/s11947-008-0154-y.
[Article]. Applied Engineering in Agriculture, 33(2), 157e166. Niedbala, G., Mioduszewska, N., Mueller, W., Boniecki, P.,
https://doi.org/10.13031/aea.11578. Wojcieszak, D., Koszela, K., et al. (2016). Use of computer
Li, T., Ji, Y. H., Zhang, M., Sha, S., & Li, M. Z. (2015). Comparison of image analysis methods to evaluate the quality topping sugar
photosynthesis prediction methods with BPNN and PLSR in beets with using artificial neural networks. In C. M. Falco, &
different growth stages of tomato. Transactions of the CSAE, S2, X. Jiang (Eds.), Eighth International conference on digital image
222e229. https://doi.org/10.11975/j.issn.1002-6819.2015.z2.034. processing (Vol. 10033). https://doi.org/10.1117/12.2244875.
Linker, R., Gutman, P. O., & Seginer, I. (1999). Robust controllers Oho, T., Tonosaki, T., Isomura, K., & Ogura, K. (2002). A CO2 sensor
for simultaneous control of temperature and CO2 operating under high humidity. Journal of Electroanalytical
concentration in greenhouses. Control Engineering Practice, 7(7), Chemistry, 522(2), 173e178. https://doi.org/10.1016/S0022-
851e862. https://doi.org/10.1016/S0967-0661(99)00042-8. 0728(02)00712-X.
Linker, R., Seginer, I., & Gutman, P. O. (1998). Optimal CO2 control Panwar, N. L., Kaushik, S. C., & Kothari, S. (2011). Solar
in a greenhouse modeled with neural networks. Computers and greenhouse an option for renewable and sustainable farming.
Electronics in Agriculture, 19(3), 289e310. https://doi.org/10.1016/ Renewable and Sustainable Energy Reviews, 15(8), 3934e3945.
S0168-1699(98)00008-8. https://doi.org/10.1016/j.rser.2011.07.030.
Li, H. Z., Tao, W., Gao, T., Li, H., Lv, Y. H., & Su, Z. M. (2012). Pasgianos, G. D., Arvanitis, K. G., Polycarpou, P., & Sigrimis, N. (2003).
Improving the accuracy of DFT calculation for hemolysis bond A nonlinear feedback technique for greenhouse environmental
dissociation energies of Y-NO bond via back propagation control. Computers and Electronics in Agriculture, 40(1e3), 153e177.
neural network based on mean impact value. Chemical Journal https://doi.org/10.1016/s0168-1699(03)00018-8.
of Chinese University., 33(2), 346e352. https://doi.org/10.3969/ ~o
Pin  n, S., Camacho, E. F., Kuchen, B., & Pen ~ a, M. (2005).
j.issn.0251-0790.2012.02.024. Constrained predictive control of a greenhouse. Computers and
Liu, L., Hoogenboom, G., & Ingram, K. T. (2000). Controlled- Electronics in Agriculture, 49(3), 317e329. https://doi.org/10.1016/
environment sunlit plant growth chambers. Critical Reviews in j.compag.2005.08.007.
Plant Sciences, 19(4), 347e375. https://doi.org/10.1016/S0735- Poekaew, P., Champrasert, P., & Ieee. (2015). Adaptive-PCA: An
2689(00)80008-8. event-based data aggregation using principal component
Liu, L. J., Sun, C. L., He, X. L., Liu, X. X., Wu, H., Liu, M., et al. (2016). analysis for WSNs. [Proceedings paper]. In 2015 International
The secondary compost products enhances soil suppressive conference on smart sensors and application e Icssa 2015 (pp.
capacity against bacterial wilt of tomato caused by Ralstonia 50e55). https://doi.org/10.1109/ICSSA.2015.7322509.
solanacearum. European Journal of Soil Biology, 75, 70e78. Pohlheim, H., & Heissner, A. (1997). Optimal control of
https://doi.org/10.1016/j.ejsobi.2016.04.005. greenhouse climate using a short time climate model and
Liu, Q., Zhang, Y. Y., Shen, J., Xiao, B., Linge, N., & Ieee. (2015). A evolutionary algorithms. Mathematical and Control Applications
WSN-based prediction model of microclimate in a in Agriculture and Horticulture, 113e118. https://doi.org/10.1016/
greenhouse using an extreme learning approach. In 2015 S1474-6670(17)41255-9.
118 b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9

Pohlheim, H., & Heissner, A. (1999). Optimal control of weather. Biosystems Engineering, 161, 174e187. https://doi.org/
greenhouse climate using real-world weather data and 10.1016/j.biosystemseng.2017.06.024.
evolutionary algorithms. In Gecco-99: Proceedings of the Genetic Seginer, I., van Straten, G., & van Beveren, P. J. M. (2017b). Day-to-
and Evolutionary Computation Conference (pp. 1672e1677). night heat storage in greenhouses: 2 sub-optimal solution for
Precup, R. E., & Hellendoorn, H. (2011). A survey on industrial realistic weather. Biosystems Engineering, 161, 188e199. https://
applications of fuzzy control. Computers in Industry, 62(3), doi.org/10.1016/j.biosystemseng.2017.06.023.
213e226. https://doi.org/10.1016/j.compind.2010.10.001. Setiawan, A., Albright, L. D., & Phelan, R. M. (2000). Application of
Qaddoum, K., Hines, E. L., & Iliescu, D. D. (2013). Yield prediction pseudo-derivative-feedback algorithm in greenhouse air
for tomato greenhouse using EFuNN. ISRN Artificial Intelligence, temperature control. [Article; proceedings paper]. Computers
2013, 1e9. https://doi.org/10.1155/2013/430986. and Electronics in Agriculture, 26(3), 283e302. https://doi.org/
Qian, G., & Wang, X. D. (2013). Greenhouse parameter control 10.1016/s0168-1699(00)00081-8.
strategy based on the AHP (analytical hierarchy process). In Stanghellini, C., Incrocci, L., Gazquez, J. C., & Dimauro, B. (2008).
2013 25th Chinese control and decision conference (Ccdc) (pp. Carbon dioxide concentration in mediterranean greenhouses:
1589e1594). How much lost production? Acta Horticulturae, 801(801),
Rezaee, A., & Golpayegani, M. K. (2012). Intelligent control of 1541e1549.
cooling-heating systems by using emotional learning. Suhl, J., Dannehl, D., Kloas, W., Baganz, D., Jobs, S., Scheibe, G.,
Elektronika Ir Elektrotechnika, 120(4), 26e30. https://doi.org/ et al. (2016). Advanced aquaponics: Evaluation of intensive
10.5755/j01.eee.120.4.1446. tomato production in aquaponics vs. conventional
Rodriguez, F., Berenguel, M., & Arahal, M. R. (2003). A hierarchical hydroponics. [Article]. Agricultural Water Management, 178,
control system for maximizing profit in greenhouse crop 335e344. https://doi.org/10.1016/j.agwat.2016.10.013.
production. In European control conference ECC 2003 (pp. Taylor, C. J., & Aerts, J.-M. (2013). Control of nonlinear biological
2788e2793). systems by non-minimal state variable feedback. Statistics in
Rodrı́guez, F., Berenguel, M., Guzma  n, J. L., & Ramı́rez-Arias, A. Biosciences, 6(2), 290e313. https://doi.org/10.1007/s12561-013-
(2015). Modeling and control of greenhouse crop growth. Springer 9098-5.
International Publishing. Thongbai, P., Kozai, T., & Ohyama, K. (2010). CO2 and air
Roy, Y., Lefsrud, M., Orsat, V., Filion, F., Bouchard, J., Nguyen, Q., circulation effects on photosynthesis and transpiration of
et al. (2014). Biomass combustion for greenhouse carbon tomato seedlings. Scientia Horticulturae, 126(3), 338e344.
dioxide enrichment. Biomass & Bioenergy, 66, 186e196. https:// https://doi.org/10.1016/j.scienta.2010.07.018.
doi.org/10.1016/j.biombioe.2014.03.001. Titus, J. E., & Wendlberger, S. J. (2016). Gauging submersed plant
Ruano, A. E., Ge, S. S., Guerra, T. M., Lewis, F. L., Principe, J. C., & response to CO2 enrichment: Pot size matters. Aquatic
Colnaric, M. (2014). Computational intelligence in control. Botany, 134, 82e86. https://doi.org/10.1016/j.aquabot.2016.
[Review]. Annual Reviews in Control, 38(2), 233e242. https:// 06.007.
doi.org/10.1016/j.arcontrol.2014.09.006. Trabelsi, A., Lafont, F., Kamoun, M., & Enea, G. (2007). Fuzzy
Ru, D., Liu, J., Hu, Z., Zou, Y., Jiang, L., Cheng, X., et al. (2017). identification of a greenhouse. Applied Soft Computing, 7(3),
Improvement of aquaponic performance through micro- 1092e1101. https://doi.org/10.1016/j.asoc.2006.06.009.
and macro-nutrient addition. Environmental Science and Van Ginkel, S. W., Igou, T., & Chen, Y. (2017). Energy, water and
Pollution Research International. https://doi.org/10.1007/s11356- nutrient impacts of California-grown vegetables compared to
017-9273-1. controlled environmental agriculture systems in Atlanta, GA.
Salazar, R., Rojano, A., & Lopez, I. (2009). A neural network model Resources, Conservation and Recycling, 122, 319e325. https://
for photosynthesis prediction. In 2009 Eighth Mexican doi.org/10.1016/j.resconrec.2017.03.003.
International conference on artificial intelligence, proceedings (pp. Van Henten, E. J. (1994). Greenhouse climate management: An optimal
140e143). https://doi.org/10.1109/Micai.2009.40. control approach. Ph.D. thesis. Wageningen: Wageningen
 nchez-Guerrero, M. C., Lorenzo, P., Medrano, E., Baille, A., &
Sa Agriculture University.
Castilla, N. (2009). Effects of EC-based irrigation scheduling Van Straten, G., & van Henten, E. J. (2010). Optimal greenhouse
and CO2 enrichment on water use efficiency of a greenhouse cultivation control: Survey and perspectives. IFAC Proceedings
cucumber crop. Agricultural Water Management, 96(3), 429e436. Volumes, 43(26), 18e33. https://doi.org/10.3182/20101206-3-JP-
https://doi.org/10.1016/j.agwat.2008.09.001. 3009.00004.
 nchez-Guerrero, M. C., Lorenzo, P., Medrano, E., Castilla, N.,
Sa Van Straten, G., van Willigenburg, L. G., van Henten, E. J., &
Soriano, T., & Baille, A. (2005). Effect of variable CO2 van Ooteghem, R. J. C. (2010). Optimal control of greenhouse
enrichment on greenhouse production in mild winter cultivation. Boca Raton, FL, USA: CRC Press, Taylor &
climates. Agricultural and Forest Meteorology, 132(3e4), 244e252. Francis.
https://doi.org/10.1016/j.agrformet.2005.07.014. Vanthoor, B. H. E., Stigter, J. D., van Henten, E. J., Stanghellini, C.,
 nchez-Molina, J. A., Pe
Sa rez, N., Rodrı́guez, F., Guzma  n, J. L., & de Visser, P. H. B., & Hemming, S. (2012). A methodology for
Lo pez, J. C. (2015). Support system for decision making in the model-based greenhouse design: Part 5, greenhouse design
management of the greenhouse environmental based on optimisation for southern-Spanish and Dutch conditions.
growth model for sweet pepper. Agricultural Systems, 139, Biosystems Engineering, 111(4), 350e368. https://doi.org/10.1016/
144e152. https://doi.org/10.1016/j.agsy.2015.06.009. j.biosystemseng.2012.01.005.
 nchez-Molina, J. A., Reinoso, J. V., Acie
Sa n, F. G., Rodrı́guez, F., & Wang, J. N., Niu, X. T., Zheng, L. J., Zheng, C. T., & Wang, Y. D.
Lo pez, J. C. (2014). Development of a biomass-based system for (2016). Wireless mid-infrared spectroscopy sensor network for
nocturnal temperature and diurnal CO2 concentration control automatic carbon dioxide fertilization in a greenhouse
in greenhouses. Biomass and Bioenergy, 67, 60e71. https:// environment. Sensors, 16(11). https://doi.org/10.3390/
doi.org/10.1016/j.biombioe.2014.04.015. s16111941.
Schmidhuber, J. (2015). Deep learning in neural networks: An Wang, W. Z., Zhang, M., Liu, C. H., Li, M. Z., & Liu, G. (2013). Real-
overview. [Review]. Neural Networks, 61, 85e117. https:// time monitoring of environmental information and modeling
doi.org/10.1016/j.neunet.2014.09.003. of the photosynthetic rate of tomato plants under greenhouse
Seginer, I., van Straten, G., & van Beveren, P. J. M. (2017a). Day-to- conditions. Applied Engineering in Agriculture, 29(5), 783e792.
night heat storage in greenhouses: 1 optimisation for periodic https://doi.org/10.13031/aea.29.9743.
b i o s y s t e m s e n g i n e e r i n g 1 7 1 ( 2 0 1 8 ) 1 0 1 e1 1 9 119

Wang, J., Zou, Z., & Weerakkody, W. A. P. (2014). Optimization of RBF network. Sensors, 12(5), 5328e5348. https://doi.org/
leaf photosynthesis of tomato through micro-climate control. 10.3390/s120505328.
Acta Horticulturae, 1037, 593e597. https://doi.org/10.17660/ Zhang, S. D., Bi, X. T., & Clift, R. (2015). Life cycle analysis of a
ActaHortic.2014.1037.75. biogas-centred integrated dairy farm-greenhouse system in
Weiss, I., Mizrahi, Y., & Raveh, E. (2010). Effect of elevated CO2 on British Columbia. Process Safety and Environmental Protection, 93,
vegetative and reproductive growth characteristics of the CAM 18e30. https://doi.org/10.1016/j.psep.2014.02.017.
plants Hylocereus undatus and Selenicereus megalanthus. Zhang, J., Chen, J., Wang, S., & Dong, L. (2007). Modeling of the
Scientia Horticulturae, 123(4), 531e536. https://doi.org/10.1016/ growing process of tomato based on modified Elman network
j.scienta.2009.11.002. and FGA. [Conference paper]. In 2007 IEEE International
Wongkiew, S., Hu, Z., Chandran, K., Lee, J. W., & Khanal, S. K. conference on control and automation, ICCA 2007 (pp. 2377e2380).
(2017). Nitrogen transformations in aquaponic systems: A https://doi.org/10.1109/icca.2007.4376787.
review. Aquacultural Engineering, 76, 9e19. https://doi.org/ Zhang, Z. M., Liu, L. H., Zhang, M., Zhang, Y. S., & Wang, Q. M.
10.1016/j.aquaeng.2017.01. (2014). Effect of carbon dioxide enrichment on health-
Xiao, H., Feng, L., Zhi, Y., & Ieee. (2013). Tuning the PID parameters promoting compounds and organoleptic properties of tomato
for greenhouse control based on CFD simulation. In 2013 Second fruits grown in greenhouse. Food Chemistry, 153, 157e163.
International Conference on Agro-Geoinformatics (pp. 483e487). https://doi.org/10.1016/j.foodchem.2013.12.052.
https://doi.org/10.1109/Argo-Geoinformatics.2013.6621968. Zhang, R. B., Wang, L. H., Huang, X. L., & Guo, J. J. (2011). Design of
Xu, S. P., Zhu, X. S., Li, C., & Ye, Q. S. (2014). Effects of CO2 greenhouse wireless sensor network control system based on
enrichment on photosynthesis and growth in Gerbera fuzzy neural network. In L. Chen, Y. K. Zhang, A. X. Feng,
jamesonii. Scientia Horticulturae, 177, 77e84. https://doi.org/ Z. Y. Xu, B. Q. Li, & H. Shen (Eds.), Functional manufacturing
10.1016/j.scienta.2014.07.022. technologies and Ceeusro II (Vol. 464, pp. 318e321). Stafa-Zurich:
Yogev, U., Barnes, A., & Gross, A. (2016). Nutrients and energy Trans Tech Publications Ltd. https://doi.org/10.4028/
balance analysis for a conceptual model of a three loops off www.scientific.net/KEM.464.318.
grid, aquaponics. [Article]. Water, 8(12), 16. https://doi.org/ Zhu, Q., Qiu, F., Quan, Y., Sun, Y., Liu, S., & Zou, Z. (2005). Solid-
10.3390/w8120589. electrolyte NASICON thick film CO2 sensor prepared on small-
Zaborowicz, M., Boniecki, P., Koszela, K., Przybylak, A., & volume ceramic tube substrate. Materials Chemistry and Physics,
Przybyl, J. (2017). Application of neural image analysis in 91(2), 338e342. https://doi.org/10.1016/j.matchemphys.
evaluating the quality of greenhouse tomatoes. [Article]. 2004.11.036.
Scientia Horticulturae, 218, 222e229. https://doi.org/10.1016/ Zosel, J., Oelssner, W., Decker, M., Gerlach, G., & Guth, U. (2011).
j.scienta.2017.02.001. The measurement of dissolved and gaseous carbon dioxide
Zeng, S. W., Hu, H. G., Xu, L. H., & Li, G. H. (2012). Nonlinear concentration. [Review]. Measurement Science and Technology,
adaptive PID control for greenhouse environment based on 22(7), 45. https://doi.org/10.1088/0957-0233/22/7/07200.

You might also like