Renewable and Sustainable Energy Reviews 117 (2020) 109480

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: http://www.elsevier.com/locate/rser

Energy efficient operation and modeling for greenhouses: A

literature review
E. Iddio, L. Wang *, Y. Thomas, G. McMorrow, A. Denzer
University of Wyoming, 1000 E University Ave, Laramie, WY, 82071, United States

A R T I C L E I N F O A B S T R A C T

Keywords: With growing food demand worldwide, controlled environment agriculture is an important strategy for crop
Greenhouses production year-round. One of the important types of controlled environment agriculture is greenhouses. Key
Energy efficiency indoor environmental parameters such as carbon dioxide, moisture, lighting, and temperature are required to be
Heating ventilation and air conditioning
maintained for favorable crop growth in greenhouses. Due to lightweight construction and inefficient operation,
Modeling
Controls
greenhouses consume more fossil fuel energy in the operation of mechanical systems than other similar sized
buildings and have larger carbon footprints. In fact, greenhouses are one of the most energy-intensive sectors of
the agricultural industry. Energy consumption in greenhouses is influenced by mechanical systems, indoor
environment, crop growth, and evapotranspiration. Therefore, energy simulations help analyze the complex
thermal processes in greenhouse operation, and contribute to energy efficient greenhouse operation. This paper
reviews existing strategies on energy efficient control operation and state-of-the-art energy simulation for
greenhouses. It first discusses strategies for improving energy efficiency in greenhouse control operation by
summarizing the studies on energy efficient operation strategies, the control of key greenhouse parameters,
sensing network and monitoring systems, along with various control algorithms. Second, the review covers
energy modeling of greenhouses by summarizing existing and developed approaches. Finally, this review
identifies areas in which future research has the potential to reduce greenhouse energy consumption and carbon
footprint.

1. Introduction the growing food demand worldwide. The controlled environment of a

greenhouse allows for crop production in climates and seasons which
The global food production system accounts for one-third of global would otherwise prohibit growth, prolonging the cultivation periods of
greenhouse gas (GHG) emissions [1]. Since food production centers are seasonal crops. This allows for crops to be grown near locations of
commonly far away from the consumers, 30% of the world’s food per­ consumption. CEA increases yield, reduces transportation distances,
ishes in transit [2]. Greenhouse crop production, as one type of decreases water and land consumption, and reduces the need for pesti­
controlled environment agriculture (CEA), is an important way to meet cides. These advantages are significant but, maintaining the

Abbreviations: AJAX, Asynchronous JavaScript and XML; ALCA, Attributional Life Cycle Assessment; AML, Adaptive Machine Learning; ANFIS, Adaptive Neuro-
Fuzzy Interference Systems; ANN, Artificial Neural Network; ARIMA, Autoregressive Integrated Moving Average; BP, Back Propagation; CEA, Controlled Environ­
ment Agriculture; CFD, Computational Fluid Dynamics; CLCA, Consequential Life Cycle Analysis; CLD, Causal Loop Diagram; CO2, Carbon Dioxide; DSS, Decision
Support System; EA, Evolutionary Algorithm; EKF, Extended Kalman Filter; ET, Evapotranspiration; FL, Fuzzy Logic; FLC, Fuzzy Logic Control; FPDF, Fuzzy Pseudo-
Derivative Feedback; GA, Genetic Algorithm; GHG, Greenhouse Gas; GPRS, General Packet Radio Services; GWT, Google Web Toolkit; HVAC, Heating Ventilation
and Air Conditioning; ISO, International Organization for Standard; i-RTG, Integrated Rooftop Greenhouse; LCA, Life Cycle Assessment; LCI, Life Cycle Inventory;
LCIA, Life Cycle Impact Assessment; LED, Light Emitting Diode; MAF, Moving Average Filter; MIMO, Multiple Input Multiple Output; MPC, Model Predictive Control;
PCA, Principal Component Analysis; PDF, Pseudo-Derivative Feedback; PI, Proportional Integral; PID, Proportional Integral Derivative; PPFD, Photosynthetic Photon
Flux Density; Ppm, Parts Per Million; PSO, Particle Swarm Optimization; QFT, Quantitative Feedback Theory; RC, Resistance-Capacitance; RTG, Rooftop Greenhouse;
RH, Relative Humidity; RLS, Recursive Least Square; SAEELM, Self-Adaptive Evolutionary Extreme Learning Machine; SC, Shipping Containers; SD, System Dynamic;
SQP, Sequential Quadratic Programming; SVM, Support Vector Machines; TI, Temperature Integration; UK, United Kingdom; UKF, Unscented Kalman Filter; US,
United States; WIMAX, Worldwide Interoperability for Microwave Access; WSNs, Wireless Sensor Networks.
* Corresponding author.
E-mail address: [email protected] (L. Wang).

https://doi.org/10.1016/j.rser.2019.109480
Received 2 January 2019; Received in revised form 26 April 2019; Accepted 8 October 2019
Available online 1 November 2019
1364-0321/© 2019 Elsevier Ltd. All rights reserved.
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

environmental conditions necessary for plant growth involves regula­ tropics. This section, therefore, seeks to explore how such control and
tion of temperature, humidity, light availability, and water use. Hence, automation systems can be employed in greenhouse operation to reduce
greenhouses are energy-intensive compared to other sectors of the energy use.
agricultural industry. Energy is usually the second largest overhead cost
in greenhouse crop production, after labor cost [3]. It accounts for up to 2.1. Energy savings by control of greenhouse environmental parameters
25% of the operating costs of large vertical farms in the U.S [4].
Reducing high energy demand within agricultural greenhouses, while Temperature, RH, CO2 concentrations, and illumination are inter­
increasing crop yield, is identified as a key industry sustainable devel­ connected environmental parameters that have a significant effect on
opment goal [5]. yield. These parameters can be dynamically controlled to improve
This literature review summarizes recent research related to energy- resource efficiency. The commonly studied strategies in the literatures
efficient environmental operation and energy simulation for green­ include temperature regimes, humidity control through ventilation,
houses. It first discusses strategies for improving energy efficiency in event-based CO2 supplementation, and dynamic control of supplemental
greenhouse operation by summarizing the studies on energy-efficient grow lights based on measured photosynthetic photon flux density
operation control strategies, the control of key environmental parame­ (PPFD) from solar radiation. With each of these controlled environ­
ters, sensing network and monitoring systems, along with various con­ mental parameters, there is a balance between economic viability and
trol algorithms. Second, the review covers energy modeling of high-yield.
greenhouses within existing, developed and advanced approaches. Dy­ Temperature integration (TI) — maintaining an optimal average
namic energy simulations play an important role to better understand temperature over a time period while minimizing energy demand [12]
the complex mechanism in the thermal process of greenhouse operation, —is an effective temperature control technique. The range is determined
and therefore contribute to energy-efficient greenhouse operation. on the low-end by temperatures at which damage will occur and on the
Advanced greenhouse controls require energy models to accurately high-end by temperatures at which growth rate is significantly reduced
predict the thermal process in greenhouses for decision-making. The [13]. One study by van Beveren et al. [14] on Avalanche rose cultivars
goal of this review is to identify areas in which further research needs to found that energy use can be reduced by relaxing temperature and hu­
be conducted where there is a potential to reduce greenhouse energy midity bounds. A dynamic optimization tool was used to determine
consumption and emissions. trends in the transfer of energy, minimizing energy input. van Henten
and Bontsema [15] conducted optimization of open-loop temperature
2. Energy efficiency in greenhouse operation control in greenhouses to minimize energy consumption, maintaining an
average temperature over one day instead of a rigid pre-defined set­
Greenhouse control poses a unique challenge as it requires control point. Other studies have explored the TI concept using dynamic tem­
systems that can respond to the ever-changing environmental conditions perature constraints [16,17] and temperature bandwidths [18].
and microclimatic variables. In northern latitudes and extreme climates, However, care must be taken as plants exposed to low nighttime tem­
the cost of heating and cooling greenhouses can reach 70–85% of the peratures exhibit symptoms such as retarded growth, reduced leaf
total operation cost [6]. Even in warmer areas such as the southwestern number, and deformed fruit [19].
U.S., these costs can amount to 50% of the total [6]. Efficiently operating In many ways, humidity control is closely tied to temperature con­
mechanical systems are crucial to reduce energy consumption. The key trol. Campen [20] found that removal of moisture through condensation
environmental parameters related to greenhouse operation include air can sometimes work against temperature control. TI can lead to unde­
temperature, relative humidity (RH), carbon dioxide (CO2) concentra­ sired RH levels in closed greenhouses with low infiltration [21]. Air
tion, lighting intensity and quality. These parameters present themselves movement can help prevent condensation on leaves and improve air
differently when considering the needs of plants rather than human uniformity. An example is dehumidification by proportional ventilation.
comfort. de Halleux and Gauthier [22] compared energy consumption for three
There are three levels of greenhouse control— equipment, function, scenarios - no dehumidification, dehumidification by on-off ventilation,
and system—defined by the National Greenhouse Manufacturers Asso­ and dehumidification by proportional ventilation. They found propor­
ciation. At the equipment level, actuators control the operation of spe­ tional ventilation to be the most effective albeit with a slight increase in
cific equipment. At the function level, all equipment that pertains to a energy use of about 5.8% compared to on-off ventilation. Humidity
particular function such as humidity control are regulated. The system control in which setpoints are determined based on plant physiology
level control is a supervisory control that monitors and connects all such as plant water stress, crop growth and calcium deficiency is another
equipment and function controls, and determines setpoints and opera­ useful strategy [23].
tion stages from an overall system perspective [7]. CO2 enrichment of the greenhouse environment has been found to
Studies on energy-efficient operation strategies in greenhouses range boost yield [13]. Most plants show an increase in net photosynthesis
from adjustments of vent opening to sequencing the operation modes for with a rise in CO2 levels from 340 to 1000 ppm (parts per million) [24].
mechanical heating/cooling systems. Mashonjowa et al. [8] propor­ Outdoor air CO2 is at about 400 ppm however, it is often the case that
tionally operated the vents when the inside air temperatures exceeded levels of CO2 within the greenhouse are lower [25]. In this case, CO2
the calculated ventilated temperature, using an automatic control sys­ enrichment is beneficial. Kla €ring et al. [26] studied CO2 supply strate­
tem. A study explored variations in the inlet and outlet condenser bypass gies for efficient use of CO2 and maximum yield. They found an increase
and fresh air ratio to come up with six different scenarios for a green­ in yield of about 35% in a greenhouse supplied with CO2 compared to
house in the United Arab Emirates [9]. The most energy-efficient option one without. The results indicate that the most efficient use of CO2 is to
was using an outlet condenser of 50% bypass and 23% fresh air ratio. deliver it to the greenhouse when the sun and internal temperatures are
Another study in a humid subtropical climate considered the perfor­ less intense such as early morning.
mance of evaporative cooling pads alone or in tandem with ventilation Illumination control for high-yield and energy efficiency becomes an
or shading for saving energy [10]. important aspect of greenhouse operation especially with the increasing
In this section, first, we summarize the control of key greenhouse use of supplemental lighting. With automated controls, greenhouse
environmental parameters; second, we discuss sensor networks for lighting systems can automatically react to outdoor conditions trig­
monitoring environmental parameters; and finally, we review various gering supplementary lighting as required. A recent study solved the
control algorithms. Shamshiri and Ismail [11] reviewed greenhouse light intensity optimization problem for greenhouses using parallel
climate control and automation systems with the aim of examining is­ particle swarm optimization (PSO) algorithm [27]. The objective was to
sues encountered in the design of such systems for greenhouses in the find the best locations and number of LED lights to be turned on based on

2
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

the light requirements of the plant, reducing energy use. Compared to greenhouse to test, validate and optimize the WSN system.
fluorescent and incandescent lamps, this LED system used 82.6% and Greenhouse environmental control systems provided by-companies
54.2% less energy respectively [27]. Pinho et al. [28] proposed the such as Climate Control Systems, Priva, Vaisala, and Autogrow are
dynamic control of supplemental LED lights based on the PPFD setpoint proprietary and inflexible. Hence, there is a need for low-cost, open-
and measured PPFD from incoming solar radiation. Both continuous source options. There is also the need for a complete framework to
dimming and on-off strategies were investigated for energy consumption develop holistic systems that take into account various stages from in­
and optimal plant growth with the former leading to 20% energy sav­ formation acquisition to modeling to decision support and actuation
ings. Real-time pricing demand-response strategy has also been applied [49]. Some problems beset the adoption of sensor networks. These range
to greenhouse lighting control in Denmark, realizing 18–25% savings in from sensor accuracy issues due to spatial limitations, adverse envi­
electricity costs [29]. ronmental conditions [56], sensor drifts [57], data security and resource
The control of these greenhouse environmental parameters plays a constraints. They are also susceptible to security attacks such as hacking,
crucial role in increased yield and energy efficiency. Most existing denial of service and physical tampering, due to their unreliable
studies investigate the impacts of individual environmental parameters communication channels and unattended operation [58]. Future
on crop growth and energy efficiency. However, environmental pa­ research should investigate the accuracy and reliability of low-cost
rameters are interconnected in many ways. Controlling temperatures sensors for greenhouse parameter monitoring, expand resource capa­
within a greenhouse affect humidity levels, similarly, increasing illu­ bilities and fault tolerance. There is also the question of translating in­
mination levels within the greenhouse become unproductive at high formation on individual sampled plants into the big picture of overall
levels without CO2 enrichment. Therefore, it is important that these plant growth information by merging data from various sensors [59].
components are jointly considered so that the chosen systems work
together rather than against each other. Future research on optimal
environmental setpoints should consider the interaction of these envi­ 2.3. Control algorithms
ronmental parameters [30].
Controlling microclimate in greenhouses is difficult because multiple
2.2. Sensor networks interdependent variables are at play [60]. The greenhouse environment
control system is a complex non-linear, multi-input and multi-output
Sensor networks for the greenhouse environment are heterogenous (MIMO) system. Several types of control algorithms have been used in
systems of devices that sense, process and communicate with each other the control of greenhouse systems such as
to monitor environmental parameters. There are ongoing studies on Proportional-Integral-Derivative (PID), non-linear, robust, fuzzy logic
monitoring environmental parameters such as the effect of CO2 depri­ (FL), neural network and hybrid control algorithms [61]. Generally,
vation on yield [26], soil evaluation [31,32], ventilation rates [33], and these control algorithms can be classified as either classical, advanced,
light interception by plants [34]. Sensor networks can be wired or intelligent or hybrid control algorithms [62–65]. The following section
wireless. However, wireless sensor networks (WSNs) offer flexibility and discusses a number of these algorithms.
expandability. Key hardware requirements in WSNs include robust
Wi-Fi/radio technology, low-cost energy-efficient processor, flexible 2.3.1. Classical control algorithms
input/output interface and open-source development platforms [35]. Proportional-Integral-Derivative (PID) control is most commonly
Low-cost open-source hardware options include Arduino [36,37], used in greenhouse control due to its flexibility, simplicity and good
Raspberry Pi [38,39], and Beaglebone [40]. Commonly used wireless performance [66]. A PID controller is a control loop feedback mecha­
communication protocols are ZigBee, WiFi, Bluetooth, Wibree, General nism which basically reads a sensor, computes the desired actuator
Packet Radio Services (GPRS) and Worldwide Interoperability for Mi­ output by collating the proportional, integral and derivative [67].
crowave Access (WiMAX) [41]. Adjusting the controller is a difficult and time-consuming process [68].
WSNs can provide real-time monitoring of key parameters, allowing It tends to overshoot the desired setpoint [69] and has difficulties in
researchers to continuously collect measurements. Liu et al. [42] handling external disturbances. Hu et al. [70] proposed a solution to the
investigated a WSN using Short Message Service (SMS) communication. PID tuning issue with the addition of an Evolutionary Algorithm (EA).
In another study, a WSN, consisting of two nodes were developed: However, this approach is only suitable for an analytical greenhouse
A-node, which collects data on the greenhouse microclimate, and model. PID control is also unsuitable for advanced applications, espe­
C-node, which controls lighting [43]. Another study improved the per­ cially for non-linear, dynamic control. It is also prone to wind-up errors
formance of WSNs using fuzzy-based rules to select a control strategy in which there is a loss of control and recovery delay from a large change
amongst six options based on energy use [44]. It also created a climatic in setpoint [71].
parameter map to aid quick detection of abnormalities. Wang et al. [45]
successfully developed and implemented an application for the remote 2.3.2. Advanced control algorithms
monitoring of the greenhouse climate. Google Web Toolkit was This applies to control problems that “are often characterized by
preferred over AJAX (Asynchronous JavaScript and XML) applications extremely long process delays, a high degree of process interaction, and
because the former could get around browser compatibility issues. Other multiple operating constraints and measured process disturbances”
studies on the use of sensor networks for greenhouses include [46–52]. [65]. They are capable of handling non-linear, MIMO systems with
Growers can also benefit from WSNs to effectively improve disturbances better than conventional control, but require some form of
decision-making and yield using a Decision Support System (DSS) as mathematical models in contrast to intelligent control. Examples of such
demonstrated by a number of studies [53,54]. DSS can supplement algorithms are model predictive control (MPC), feedback/feedforward,
automated greenhouse control systems, making important decisions all adaptive and robust controls.
through the operation process.
Driving down the costs of WSNs is important in increasing its rate of 2.3.2.1. Model predictive control (MPC). MPC makes use of a process
adoption. Groener et al. [55] explored the use of low-cost sensors and model to obtain a manipulated variable, optimizing an objective over a
open-source controllers in the design of a low-cost greenhouse in Kenya. future time interval. MPC offers a better option to PID systems by pre­
The system made use of Arduino due to its minimal cost, adaptability, dicting future events and taking control action accordingly. Ding et al.
and popularity. The system monitors the temperature/humidity and soil [72] presented a comprehensive review of the application of MPC in
moisture and sends an alert via SMS if they are out of the optimal specific areas of agriculture such as irrigation, machine operation and
bounds. Future work includes the construction of a scaled prototype greenhouses. This section compares a wide range of existing studies of

3
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

MPC applied to greenhouse control, and highlights areas for further efficiency and quickly converges to an optimal solution making it
research. Table 1 summarizes research in MPC for greenhouses over the appropriate for real-time, online optimization [103]. PSO has been
last two decades. found to outperform GA and SQP in some studies [87,104]. Even though
A crucial component of MPC is an accurate model that describes the SQP has a shorter computation time, it converges to a local minimum
system’s control variables. There are three types of models applied in [74]. Chen et al. [105] explored a hybrid PSO-GA algorithm for
MPC for greenhouse controls: first-principle, hybrid and data-driven modeling greenhouse energy demand, saving optimization time by more
models. The first-principle models, are based on heat transfer and than 21% compared to the individual algorithms.
mass balance equations. No measurement data is required for parameter MPC is a closed-loop control algorithm that can impose constraints
identification. Due to the complexity and computational time of solving on manipulated and controlled variables. It can handle MIMO, non-
them they are rarely used in MPC. The hybrid models such as state-space linear and dynamic systems due to rolling optimization, allowing for
models are established based on simplified energy balance equations time delays, changing objectives, sensor failure, and inverse response.
and require measurement data to determine key model parameters. However, its effectiveness depends strongly on the accuracy of the
State-space models are the most commonly used models in MPC for mathematical model used and real-time optimization as required in
greenhouses. The data-driven models use input and output data to greenhouse control places a heavy demand on computation resources.
approximate the behavior of the system under study and do not provide Installation cost is also significant.
an explicit physical meaning of the relationship between inputs and Furthermore, many of these studies are simulations and have not
outputs. They include neural networks, autoregressive models, evolu­ been tested for real scenarios. Others have been validated with experi­
tionary regression, and support vector machines (SVM). Oliviera et al. mental data, but still need to be applied in the real-time control of
[73] employed an auto-regressive model with multiple input signals. existing greenhouses. Studies for emerging types of CEA that rely
Fuzzy models [74], mixed logical dynamical models [75], and Volterra heavily on heating, ventilation and air conditioning (HVAC) systems
series model [76] have also been used as data-driven models. A complete such as SC farms are lacking; this could offer an area for further research.
greenhouse model for control operation should include the energy The control of greenhouse supplemental lighting would also be an area
model and a crop growth model [77]. Pohlheim and Heibner [78] used a of focus for such farms. Optimization algorithms with reduced compu­
short time-scale energy-crop growth model for sweet pepper consisting tation time and requirement, capable of real-time optimization are also
of transpiration and CO2 exchange terms. Van Ooteghem et al. [79] used required to improve the robustness of MPC for greenhouses.
an energy-crop model accounting for transpiration and photosynthesis.
In other studies, tomato-seedling crop growth model called the TOM­ 2.3.2.2. Feedback and feedforward control. Feedback control is simple
GRO model was used [80,81]. As the purpose of greenhouse production for greenhouse applications but prone to errors. Deviation from the
is to maximize crop yield, an energy-crop growth model is necessary, setpoint is common during disturbances and it is not suitable for long
both climatic conditions and crop growth are interconnected and sub­ time-delays. Feedforward control offers a proactive, timely response to
sequently affect resource use. disturbance which nullifies its negative effects before it reaches the
Optimization plays a key role in the formulation of an MPC algo­ system. However, its accuracy depends on the model provided. There­
rithm. The aim here is the minimization of a cost function or the fore, the combination of feedback/feedforward control is usually
maximization of an objective function under certain constraint condi­ preferred for better performance [30]. Feedback/feedforward control
tions. When multiple objective/cost functions are used it is referred to as can be combined with MPC or PID to improve performance [108]. In this
multi-objective optimization. A wide range of objective/cost functions case, feedback/feedforward is used to linearize the non-linear problem
has been explored in greenhouse applications. These include minimi­ so that it can be easily solved by linear controllers. Sigrimis et al. [109]
zation of reference deviation [76,82,83], minimization of reference used feedback/feedforward linearization and outer loop PID controllers
deviation and variation of commands [73–75,77,84–88], minimization to simultaneously control temperature and humidity in a greenhouse.
of energy consumption [79,89–91], minimization of energy and water They found it effective in setpoint tracking and disturbance rejection.
use [92], maximization of profits [78,93–95], and maximization of However, in the case of large parameter variations, like plant growth
product quality [81]. The purpose of employing the MPC algorithm is effects on greenhouse thermal capacity and evapotranspiration, feed­
the control of desired variables such as temperature, humidity, CO2 back/feedforward control does not perform well [30]. Setiawan et al.
concentration, etc. through manipulation of window opening percent­ [110] compared Pseudo-Derivative-Feedback (PDF) control to PI. The
ages [76,79,88,89,95], heating hot water temperature [74,77,91,96, PDF control had a better load handling capability and particularly out
97], flow rates of fog systems [82,90,92,98,99] and CO2 injection fluxes preformed PI for systems without time-delay but was also much better
[78,80,94,100]. Energy-efficient manipulation of grow light intensity for systems with time-delay.
and quality for CEA has rarely been studied. This could be attributed to
the fact that our understanding of the impact of light intensity and 2.3.2.3. Adaptive control. The capability of handling unknown model
quality on plants in CEA is insufficient to establish a lighting component uncertainties is what differentiates adaptive control from linear control
model for both energy consumption and plant growth. algorithms [111]. Where MPC adjusts model parameters based on future
The choice of an appropriate optimization algorithm depends on the disturbances, adaptive control adjusts model parameters based on cur­
type of problem involved. Due to the complex nature of greenhouse rent disturbances and are applied to “processes in which the dynamics
models, global optimization algorithms perform better compared to change during normal operating conditions due to stochastic distur­
analytical and classical numerical solutions [82]. Popular algorithms bances” [62]. It is basically of two types: gain scheduling in which
used in literature include Genetic Algorithm (GA) [92,101], control is based on a-priori knowledge and adaptive self-adjusting
Branch-and-Bound Algorithm [74,99], Sequential Quadratic Program­ control where online parameter estimation is performed [112]. Adap­
ming (SQP) [88,100], EA [85,90], Iterative Dynamic Programming tive control obtains instantaneous solutions from both real measured
(IDP) [80] and PSO algorithm [87]. and forecasted parameters [113]. Adaptive neuro-fuzzy interference
Finding an optimization algorithm suitable for all types of optimi­ systems (ANFIS) has been found to outperform PI, FL and artificial
zation problems is difficult. Generally, optimization algorithms are beset neural network (ANN), in the control of greenhouse indoor tempera­
by issues such as slow search speed, trapping in local optimum, tures, boasting a quick response time [114]. Berenguel et al. [115]
convergence issues, poor solution diversity and early termination [102]. developed a mixed-feedforward adaptive controller for diurnal tem­
However, some optimization algorithms try to overcome some of these perature control using vents as control inputs. Arvanitis et al. [116]
issues. GA performs parallel searches from a population of points to applied adaptive multirate-output controllers for the stable and effective
avoid being trapped in local optima. PSO exhibits good computational

4
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

Table 1
Summary of MPC research for greenhouses over the last two decades.
S/ Energy Model Crop Optimization Objective Function Controlled Variable(s) Greenhouse Reference Year
N Growth Algorithm Implementation
Model

1. Energy Balance Yes EA Maximize profit (revenue minus Heating; CO2 Injection and No [78] 1999
Equations cost of energy and CO2 Ventilation.
enrichment).
2. State-space model Yes Maximize profits (the difference Window Opening and CO2 Yes [95] 1999
between revenues from the crop dosage.
and the costs of resources).
3. ARX and ARIMA No SQP Minimize reference deviation, Actuation signals for vents, Yes [100] 2000
models energy consumption and thermal screens, gas
variation of command. heaters, CO2 injectors, and
irrigation System.
4. Integrated ARIMA No GA Minimize reference deviation. Window Opening No [82] 2001
and state-space model percentage; Heating
power; Flowrate of fog
System.
5. State-space model No SQP Minimize reference deviation Window opening Yes [88] 2002
and variation of command. percentage; heating pipe
temperature and CO2
injection flux.
6. State-space model No Minimize reference deviation No [106] 2002
and energy consumption.
7. State-space model Yes Gradient and Maximize profits (the difference Heating valve position; Yes [94] 2002
Sequential Search between revenues from the crop Window Opening and CO2
Algorithm and the costs of resources). dosage.
8. State-space model Yes Indirect Gradient Minimize fossil energy use. Window opening No [79] 2003
Method percentage; on/off for heat
recovery; heating valve
position and CO2 supply
flux.
9. Integrated ARX and No Quadratic Minimize reference deviation, Actuation signals for Yes [98] 2004
transfer function Programming energy consumption and heating; fog system and
model variation of command. window opening.
10. State-space model Yes NA Minimize fossil energy Window opening No [89] 2005
consumption. percentage; on/off for heat
recovery; heating valve
position and CO2 supply
flux.
11. Transfer function No Quadratic Minimize reference deviation Heating pipe temperature Yes [96] 2005
model Programming and energy consumption. and on/off air heaters.
12. Recursive least square No PSO Minimize reference deviation Heating and ventilation Yes [87] 2005
model and variation of command. control signals.
13. State-space model No SQP Minimize reference deviation Heating system Yes [77] 2005
and variation of command. temperature; window
opening percentage and
CO2 injection flux.
14. Single input and No Gradient Minimize the sum of reference Heat flow. No [107] 2005
single output data- Descendent deviation and energy use.
driven model Algorithm
15. State-space model Yes IDP Minimize heating load, CO2 Heating, CO2 Injection and Yes [80] 2006
enrichment and reference Window Opening.
deviation.
16. Integrated ARIMA No GA Minimize energy and water use. Window Opening Yes [92] 2007
and state-space model percentage; Heating
power; Flowrate of fog
System.
17. Non-linear auto- No Branch-and-Bound Minimize energy consumption Actuation signals for Yes [99] 2008
regressive model Algorithm and variation of command. cooling, heating and fog
systems.
18. State-space model No Non-dominated Minimize energy consumption. Actuation signals for No [90] 2009
Sorting Genetic ventilation, heating and
Algorithm-II fog system.
(NSGA-II)
19. Single input and No Quadratic Minimize reference deviation. Window Opening Yes [76] 2011
single output data- Programming percentage.
driven model
20. Step-response No Minimize reference deviation Temperature. No [86] 2011
mathematical model and variation of the command
21. State-space model No EA Minimize reference deviation Window opening No [85] 2011
and variation of the command percentage and actuation
signal of the fogging
system.
22. State-space model Yes SQP Maximize profits (the difference Actuation signals for Yes [81] 2012
between the revenues from the heating, window opening,
(continued on next page)

5
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

Table 1 (continued )
S/ Energy Model Crop Optimization Objective Function Controlled Variable(s) Greenhouse Reference Year
N Growth Algorithm Implementation
Model

crop and the costs of resources), shade screen, water, and

product quality, and water use electrical conductivity.
efficiency.
23. State-space model No NA Minimize reference deviation. Heater on/off. Yes [83] 2012
24. Step-response No Rolling Horizon Minimize reference deviation Temperature. No [84] 2013
mathematical model Optimization and variation of the command
25. State-space model Yes The Steepest Maximize profits (the difference Heating power. No [93] 2014
Descent Method between the revenues from the
crop and the operating cost)
26. An energy simulation No SQP Minimize Energy consumption. Hot water supply Yes [91] 2015
model integrated with temperature and FCU
offline CFD results loops.
27. Integrated ARIMA No Branch-and-Bound Minimize reference deviation Switch signals of No [75] 2015
and state-space model Algorithm and variation of command. environment control
devices.
28. Hybrid wavelet fuzzy No Branch-and-Bound Minimize reference deviation Heating and ventilation No [74] 2015
model Algorithm and variation of command. control signals.
29. Auto-regressive No Quadratic Minimize reference deviation Actuation signals of No [73] 2016
integrated moving- Programming and variation of command. ventilator, heater and
average model irrigation system.
30. Energy balance No Exhaustive search Minimize reference deviation Switch signals for Heating No [97] 2018
Equations and energy consumption. and Window Opening.

control of greenhouse temperature, estimating unknown parameters 2.3.3. Intelligent control algorithms
on-line from sequential sampling temperature and heating load data. Intelligent algorithms make use of artificial intelligence techniques
Adaptive hierarchical control has also been applied to the green­ and prove useful in the absence of a mathematical model. However, they
house microclimate to account for the different timescales in greenhouse require a large set of data for model training. These include Fuzzy and
crop growth control [117]. Speetjens et al. [118] used extended Kalman Neural Network Control.
filter (EKF) for the adaptive, online parameter estimation for greenhouse
climate control with better prediction results compared to 2.3.3.1. Fuzzy logic control (FLC). FLC is preferable to digital logic as it
fixed-parameter strategies. However, it only works well on a small data can be cast in terms more understandable to a human operator, inter­
set. Adaptive control makes it possible to quickly change parameters in polating input values as logical variables with values from 0 to 1 [126].
response to changes in the greenhouse environment dynamics [116]. Azaza et al. [127] employed a smart FLC-based system for the control of
They also account for errors in the model formulation and parameter greenhouse temperature and humidity. With ventilation and heating
estimates [118]. However, it still requires the identification of a suitable rates as control variables, 22% and 33% in energy and water savings
model and the design required for implementation is enormous. respectively were realized. Marquez-Vera et al. [128] used a fuzzy
model that controls the greenhouse internal temperature using a pro­
2.3.2.4. Robust control. In contrast to MPC and Adaptive control, robust portional servo valve to activate heating at night and another actuator
control requires assumed process uncertainties beforehand due to non- controls the ventilation rate during the day.
linear or time-varying processes [119,120]. This provides a descrip­ Kolokotsa et al. [129] used two FL controllers for the design of a
tion of the system under all possible operating conditions. The robust smart greenhouse climate and energy control system. However, the
controller is therefore designed to maintain stability and performance energy efficiency of the system could not be evaluated due to a lack of
over this range of operating conditions [61]. Bennis et al. [121] inves­ pre-installation energy consumption data. The time taken to determine
tigated the modeling and control of air temperature and humidity in an the control action for FLC is a restriction. Also, some dynamics are not
experimental greenhouse. The control system was based on H∞ robust well represented, especially when having a mutual exclusivity between
control design which performed better compared to an on/off regulator. control actions is desired [128]. For greenhouse parameters control, FL
Linker et al. [122] applied robust control to the simultaneous ventilation provides good flexibility especially for systems with incomplete model
and CO2 enrichment of a greenhouse using the Horowitz’s method. Due information [130]. However, the number of rules required increases
to the delay in the sensor response time, the CO2 concentration posed a exponentially for the improvement in accuracy [62,131]. Also, since it
bit of a challenge to control. In a later study, quantitative feedback takes into account expert experience in the model formulation,
theory (QFT) was employed to improve performance for the control of comprehensive knowledge of the system’s operating states is required
air temperature and humidity by specifying the air enthalpy and hu­ [132]. There is the need for general purpose, accurate and low-cost
midity ratio [123]. Moreno et al. [124] also developed a robust control fuzzy controllers for greenhouse applications [66].
strategy for greenhouse temperatures based on QFT with the incorpo­
ration of a feedforward compensation and anti-windup action to 2.3.3.2. Neural control. Neural controls and networks are some of the
improve the handling of disturbances and uncertainties. Agmail et al. most complex forms of control. These computing systems are made up of
[125] considered the strong temperature-humidity coupling in the highly connected processing elements which respond dynamically to
robust control of greenhouse ventilation and fogging systems. The external inputs. Neural networks are usually arranged in layers of
robust controller was successful at maintaining the greenhouse tem­ interconnected nodes with an activation function, with patterns pre­
perature and humidity within 2 � C and 10% of the desired setpoints sented to the network by an input layer which signals deeper layers. The
respectively. For greenhouse control, with coupled and time-varying processing is carried out by a system of weighted connections and then
conditions, robust control has proved effective [66]. However, it links to an output layer where the command is initiated. Castaneda-
struggles in conditions with drastic changes, requiring additional pa­ Miranda and Castano [133] used a multi-layer perceptron ANN for
rameters for each different greenhouse system. Sometimes, its robust­ frost control in a greenhouse. It was able to predict temperatures within
ness contradicts the need for dynamic control.

6
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

the greenhouse with a 95% accuracy. In greenhouse application, neural Table 2

control offers the benefit of being easy to implement without the need Summary of the application of Control Algorithms in Greenhouses in literature
for a complex mathematical model a priori. It is also a powerful over the last twenty years.
non-linear estimator and can detect relationships between input and S/ Type of Control Algorithm Number of Citations
output variables even if they are complicated or unknown [66]. How­ N Studies
ever, a downside of this control method is that it requires reasoning 1. PID 6 [70,71,138–141]
based on data-driven or embedded expert knowledge [72]. It also re­ 2. MPC 31 [73–100,103,106,
quires the adjustment of many parameters and is prone to overfitting. 107]
3. Adaptive Control 15 [112–118,142–149]
4. Feedback/Feedforward 8 [30,109,138,
2.3.4. Hybrid control Control 150–153]
Hybrid control algorithms combine two or more classical or intelli­ 5. FLC 21 [127–130,154–169]
gent algorithms. This method takes advantage of the simplicity and 6. Neural Control 10 [133,135,170–177]
stability of classical and advanced control algorithms with the speed and 7. Hybrid Control 7 [134–137,178–180]
8. Robust Control 5 [121–125]
expert experience of intelligent algorithms [66]. Ding et al. [134] used a
triple-layer feed-forward-back propagation (BP) neural network model
for the predictive control of the greenhouse microclimate. This hybrid have shown promising results over the years but recent studies look to
control algorithm had a faster network training time, a better conver­ improve upon the shortcomings and apply these methods to new areas in
gence efficiency and stability. Fourati et al. [135] used multilayer CEA. Dense sensor networks have the ability to measure the spatial
feed-forward neural networks and found the error between the actual variability of greenhouse conditions but to exploit this fully, actuators
and the desired output to be minimal compared to standalone neural need to be developed in such an advanced way to support distributed
control networks. Another study suggested a GA-based Adaptive control [56]. In a recent survey on urban farmers’ willingness to pay for
Neuro-Fuzzy Inference System (ANFIS) controller [136]. Compared to new technologies, climate control systems came second only to data and
conventional Fuzzy Pseudo-Derivative-Feedback (FPDF) controller the analytics technology [4]. Because costs are a major barrier to adoption,
GA-based ANFIS controller matched the setpoint values better. Xiao greenhouse automation and control could be further improved with the
et al. [137] used a PID controller combined with CFD to solve the issue of design of more accurate, low-cost sensors [181].
adjusting controller values on site, as the CFD simulation could cater to There has been rapid development in the control of greenhouse cli­
all kinds of climatic scenarios. Hybrid Control offers better response matic parameters recently, thanks to quantum leaps in computational
speed, accuracy, robustness, optimality and disturbance handling. technology. However, some areas for further research still remain. A
However, this method is difficult to implement and it is affected by the number of these studies are simply simulations and so need to be tested
limitations of its constituent algorithms. and validated under real scenarios [72]. Unlike human comfort,
knowledge of plant comfort is limited. Numerous factors determine crop
growth which makes it difficult to ascertain the environmental condi­
2.4. Discussion
tions for optimal growth [134]. Therefore, a better understanding of
crop responses to their environment would aid the evaluation of
Over the last twenty years, the most popular control algorithm used
greenhouse control technologies. New studies have explored the use of
by researchers has been MPC as shown in Fig. 1. Others such as FLC,
machine vision in evaluating crop growth, incorporating these results
adaptive and neural network control have also been extensively studied.
into control technologies would further improve greenhouse crop yield.
A summary of the application of control algorithms in greenhouse
research over the last twenty years is presented in Table 2.
3. Energy simulation
Energy efficiency in greenhouse operation revolves around the
proper control and management of temperature, humidity, CO2 levels,
A thorough understanding of thermal interactions within a green­
and illumination. However, due to the dynamic and non-linear nature of
house is key in energy-efficient greenhouse operation. Modeling thermal
the greenhouse environment, the proper control of these parameters is a
behaviors of greenhouses is complex due to the low thermal inertia, high
complicated process. The use of sensor networks and control algorithms

Fig. 1. Number of studies on Application of Control Algorithms in Greenhouse Research by Category (PID – Proportional Integral Derivative; MPC – Model Pre­
dictive Control).

7
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

solar heat loads, internal components and constantly changing micro­ Trejo-Perea et al. [195] employed a multi-layer perceptron ANN model
climate compared to traditional buildings [182]. In this section, we validated with measured data and a non-linear regression model for the
discuss standalone energy simulation approaches, coupled energy prediction of greenhouse electricity use. A study by del Sagrado et al.
simulation with ET or crop growth models, and the state-of-the-art [196] created and tested a Bayesian network to model an existing
modeling techniques for future greenhouse modeling. greenhouse. The variables included indoor and outdoor climate, soil
conditions, and crop conditions. The study concluded that the Bayesian
3.1. Energy simulation models network model can provide good accuracy to capture the dynamic
behavior of greenhouse microclimate. However, the data-driven
Energy modeling for greenhouses can be classified into three cate­ approach does not have a detailed geometry modeling capacity. It
gories: first-principle, data-driven, and hybrid models. These are would also be unsuitable for conceptual designs as it requires mea­
standalone energy modeling approaches that help quantify greenhouse surement data to train the model. It is also incapable of HVAC modeling
energy use through the analysis of the thermal loads involved in cool­ where historic performance data is lacking.
ing/heating spaces.
3.1.3. Hybrid energy models
3.1.1. First-principle energy models First-principle models depend on a complete knowledge of the
First-principle models consist of a mathematical set of equations greenhouse cover properties and HVAC systems and require significant
including physical models as well as empirical equations or performance effort in calibration. Data-driven models, on the other hand, do not
curves that describe the thermal behavior of a greenhouse with respect require such detailed information [197]. However, they require training
to its components, which can then be used to predict its interior climate data and the accuracy of such models depends on the quantity and
and determine energy consumption. The number of components quality of data. The demerits of these models have led to the develop­
considered varies but usually includes at least a combination of these: ment of a hybrid model. It provides a viable option where there is a lack
soil surface/growing media, internal air, cover, floor and crop canopy of data or incomplete knowledge of the physics of the model. It usually
[183]. There are majorly two types of first-principle models - static and consists of a state-space model which can be generated using an RC
dynamic models. Static models are simplified steady-state models and model describing the thermal dynamics via energy balance [198,199].
do not include a solar radiation component and energy storage in air/­ Such formulation is constructed as a first-order differential equation of
plant. Static models are easier to formulate and implement, however, system states, inputs, disturbances and outputs [59]. Parameters of the
their precision is questionable, hence their limited use amongst re­ model are estimated, and the model is trained using the data-driven
searchers. Jolliet et al. [184] created HORTICERN, an improved static approach. Such models help save computation time with good accu­
model comparable to complex dynamic models, is simpler and appli­ racy. Blasco et al. [92] utilized a state-space model in formulating the
cable for longer time periods. energy model for greenhouse climate control. Fifteen parameters were
Dynamic models, on the other hand, provide better results since they selected for model tuning based on the crop characteristics, heat trans­
capture solar energy interactions, energy storage in air/plant but require mission properties, and greenhouse temperature. Chen et al. [200] used
hourly or instantaneous climatic data. There are two types of dynamic an adaptive PSO-GA trained state-space model for the energy prediction
models for predicting energy performance in greenhouses: 1) dynamic of a greenhouse with a prediction accuracy of 95.6%. The state-space
models without detailed HVAC systems through energy balance equa­ coefficient matrix can also be derived in terms of resistors and capaci­
tions and 2) building energy simulation models with HVAC systems tors based on the thermal properties of the greenhouse cover. Li et al.
through building energy simulation tools. [201] investigated the thermal modeling of a greenhouse-aquaculture
For the first type, researchers developed their own codes and solvers raceway system, analyzing the thermal dynamics of components of the
for a set of energy balance equations. Joudi and Farhan [183] used a greenhouse-aquaculture system. The heat balance equation for the
dynamic model to test the viability of greenhouse solar air heaters. In­ model was solved numerically, with simulation results in close agree­
side air and soil temperature were successfully predicted using this ment with measured data. Zhou et al. [202] applied this method for the
mathematical MATLAB model. Tiwari et al. [185] employed dynamic thermal performance analysis of a plastic greenhouse. Hybrid models
models programmed in FORTRAN in their study of energy losses and the are desirable for on-line energy estimation and integration with dy­
thermal efficiency of a greenhouse. Using Simulink MATLAB, Singh namic, coupled models.
et al. [186] created a dynamic energy model for soilless cultivated cu­
cumber in an open greenhouse. 3.2. Coupled energy models
A few researchers employed building energy simulation tools
capable of detailed geometry and HVAC analysis. Semple et al. [187] To properly describe the energy fluxes within a greenhouse, addi­
employed a TRNSYS model for the energy simulation and economic tional models are required to account for important processes such as
analysis of a greenhouse with thermal energy storage with specified crop growth, evapotranspiration, artificial lighting, photosynthesis, etc.
geometry. Using DesignBuilder, Alvarez-Sanchez et al. [188] concluded Therefore, advanced models exist that can handle such interactions. A
that through energy simulation, proper design and material selection recent study reviewed about 30 existing greenhouse energy balance
can be done, reducing energy use and operating costs. Vadiee and Martin models to derive an integrated energy-yield model with crop growth and
[189] carried out energy analysis using TRNSYS and found a combina­ ET models [203]. This model is able to identify the main drivers of
tion of double thermal screen-double glazing and a closed greenhouse greenhouse energy losses and its implication on yield. Vanthoor et al.
concept led to a 60% and 80% reduction in energy use. EnergyPlus has [204] coupled a MATLAB-based energy balance model with a crop
also been used for parametric study on sizing a greenhouse utilizing growth model for tomatoes to study parameters crucial for high-yield. In
waste heat [190] and quantifying its energy savings [191]. Dynamic another study by Hill [205], a dynamic energy-crop growth model was
energy simulations have also been performed to test the viability of used to study energy use for nursery cultivation.
energy retrofit measures [192,193]. KASPRO, a greenhouse energy simulation tool accounts for the
thermal process and incorporates evapotranspiration and photosyn­
3.1.2. Data-driven energy models thesis modules [206]. Elings et al. [207] studied some conservation
Data-driven approaches for the simulation of greenhouse energy use measures with direct effects on crop production, energy consumption,
rely on a large set of measured data to create a predictor model. Taki economics and yield of tomato using KASPRO. Montero et al. [208] used
et al. [194] found the most accurate estimation of energy exchange with KASPRO for the energy simulation of an Integrated Rooftop Greenhouse
Radial Basis Function model type of the ANN compared to SVM. (i-RTG) with adapted boundary conditions. Another study investigated

8
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

the cooling system capacity effects and water use efficiency under reciprocal [222], like the greenhouse microclimate, where climate pa­
various climate conditions on the ventilation requirements of a rameters and the processes that control them are interdependent. SD
semi-closed greenhouse [209]. A downside to KASPRO is that it lacks would prove useful in simulating the effects of different greenhouse
detailed HVAC modeling capabilities and the ability to incorporate al­ energy conservation strategies for the energy-efficient design and
gorithm editors like MATLAB/Simulink. operation of CEAs.
Building energy simulation tools have also been coupled with sub-
models. Benis et al. [210] carried out energy simulations of rooftop 3.3.2. Adaptive Machine Learning (AML) approach
greenhouses (RTG) and shipping containers (SC). In contrast to previous Data-driven models operate in an open-loop system whereby input
studies, this workflow integrates sub-models such as daylight, energy, data is fed in, the model is trained, and output data is generated. There is
crop growth, and water and is used for the early stage validation of a no feedback mechanism and such models struggle when large deviations
project’s economic viability. Lando et al. [182] also carried out a occur. However, an AML model incorporates a closed-loop methodology
simulation of the energy use of a large plastic multi-span greenhouse adjusting the model parameters by directly processing new measure­
using EnergyPlus, with an internal source of vapor representing ments [197].
evapotranspiration from Stanghellini in MATLAB. Common AML algorithms, such as Unscented Kalman Filter (UKF)
Coupled modeling has been used in the energy consumption com­ and Extended Kalman Filter (EKF), has been conducted in previous
parison of traditional greenhouses to high-tech plant factories. Graa­ studies for greenhouses. AML has been applied to the predictive control
mans et al. [211] found plant factories to be resource-efficient but use of greenhouse temperature using UKF [234,235]. van Mourik et al.
more energy for lighting compared to traditional greenhouses. They [236] compared the performance of EKF, UKF and a moving average
integrated a crop transpiration sub-model into DesignBuilder to deter­ filter (MAF), with EKF outperforming both in terms of reducing moni­
mine the fraction of radiation load dissipated as latent heat. Harbick and toring errors. Prediction via data assimilation is improved by EKF and
Albright [212] carried out a similar study using EnergyPlus. Another UKF compared to regular algorithms [237].
study explored energy saving potential in optimizing greenhouse shell AML has also been used in other engineering applications. Wang
properties (thermal and optical) using the energy simulation software et al. [197] used least-mean-squares, recursive-least-square (RLS) and
ESP-r integrated with a tomato crop model [213]. Although not specif­ Gaussian mixture regression models for the hourly prediction of building
ically designed for the elaborate modeling of greenhouses, energy use. Li et al. [238] investigated two recursive Principal Compo­
Fitz-Rodriguez et al. [214] created a web-based application which serves nent Analysis (PCA) algorithms for semiconductor manufacturing.
as an educational tool to demonstrate greenhouse physics and control Compared to machine learning approaches, AML improves prediction
strategies. accuracy, robustness, and efficiency of estimation as demonstrated by
Mahmoud et al. [239] using Self-Adaptive Evolutionary Extreme
3.3. Advanced simulation methods viable for future use in greenhouse Learning Machines (SAEELM) for modeling wind power generation
modeling prediction intervals.
AML approach would be beneficial in a greenhouse control appli­
Some state-of-the-art modeling approaches exist which have not seen cation for event-based control. Data-driven approaches rely on large
widespread use in greenhouse energy modeling. Such methods include static datasets. However, the greenhouse environment is dynamic, sen­
System Dynamics and Adaptive Machine Learning approach. A key sors and actuators are required to rapidly respond to the ever-changing
component of these methods is the presence of a feedback loop and microclimatic parameters. It is therefore difficult to re-train the model
adaptive modeling capabilities, suitable for multi-feedback and real- when environmental conditions change. AML algorithms can quickly
time modeling applications. detect such changes given a small dataset and can react rapidly to adjust
the model parameters accordingly.
3.3.1. System Dynamic (SD) approach
SD is a modeling technique that models the dynamics of a system 3.4. Discussion
consisting of components that represent state variables and their rate of
change. SD involves four key steps: causal loop diagram (CLD), a A great deal of research has been done in greenhouse energy
quantitative simulation model, model validation and utilization [215]. modeling over the last few years aided by advancement in computing
SD has been used in building and environment [215–219], renewable capabilities. However, very little of this application has trickled down to
energy [220,221], construction [222–225], and transportation [226, the everyday urban grower mainly because a number of these modeling
227], but there is still a lack of research in greenhouses. Liu et al. [215] techniques are complex and require specialized software and knowl­
addressed energy consumption gaps for green office buildings in China edge. Therefore, a fully-integrated, user-friendly, all-in-one simulation
using SD, this enabled the development of targeted operations man­ software for greenhouse energy simulation is much needed. Also, the
agement strategies based on the feedback loop. Onat et al. [217] majority of the existing energy models considered either tomato or let­
modeled the residential green building movement in the US using SD, tuce crops, therefore further research needs to be done to adapt and test
investigating the effects of a range of policies on GHG emissions until such models for other crops. Energy modeling for emerging scenarios
2050. such as vertical farm, SC, building-integrated indoor agriculture should
Most importantly, SD is capable of accounting for complex infor­ also be performed [210].
mation feedbacks and time-delay analysis, making it powerful for Going forward, optimization of key greenhouse parameters based on
contextual problems [228]. This gives it an advantage over other simulation results would prove vital. It would help us determine the
modeling approaches, providing feedback on the consequence of system exact climatic conditions for optimum production on a case-by-case
disturbances over several scenarios. However, SD is incapable of basis in contrast to the range of values currently in use. This would
obtaining optimum decisions. The subjective selection of the causal also help in research on alternative greenhouse shells with optimal
relationship between parameters also leads to unrealistic results due to thermal properties. A major contributor to the cooling load for high-
oversimplified or overly-detailed models [229]. Finally, uncertainty tech, closed greenhouses is artificial lighting, therefore, results from
modeling is an issue, it is difficult to easily and accurately model energy models would help minimize its energy use. The effect of CO2
spatially distributed data [230]. One way to overcome this is to combine supplementation on energy consumption due to HVAC and lighting
SD with other modeling tools like agent-based modeling [231], FL systems should also be studied, and control strategies that reduce the
[232], Bayesian Network [233], and Monte Carlo [220]. overall cost of energy through appropriate schedules for HVAC, light
Generally, the underlying mechanism of systems suitable to the SD is and shade operations [212].

9
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

Furthermore, energy simulations can help in the study of the po­ energy efficiency measures, and energy simulations for greenhouses. It is
tential for integrating renewable energy resources, reducing the envi­ aimed at researchers, growers, and policymakers in CEA. Due to the
ronmental impact of CEA. Further research is needed in the areas of SD complex nature of greenhouse microclimate, energy-efficient operation
and AML modeling approach. SD and AML would prove useful in requires the use of appropriate control algorithms and sensors. Optimal
handling parameter interactions and provide feedback in modeling. system control has been achieved primarily using MPC. However, the
cost and accuracy of sensors still pose a barrier to full adoption by
4. Life cycle assessment (LCA) everyday growers. Greenhouse automation is also affected by the lack of
consensus on optimal setpoints for energy-efficient cultivation. Also, a
Agriculture places a heavy demand on resources, greenhouses use number of the studies reviewed are based on simulations and need to be
less land and water; however, its energy use is still significant. There­ tested and validated under real scenarios. For emerging areas and so­
fore, the environmental impact of greenhouse production is an impor­ phisticated CEA projects like the use of vertical farm, SC, and building-
tant assessment that helps quantify sustainability and drive decision- integrated indoor agriculture, few studies have been performed on its
making. LCA quantifies such impacts throughout the entire life cycle environmental impact, energy use, and efficiency. Although it is evident
of a product or service as described by ISO 14040 and 14044 [240,241]. that there is extensive research on many facets of greenhouse operation
Two main LCA methodologies exist: attributional (ALCA) and conse­ and energy modeling, significant gaps still exist in our understanding of
quential (CLCA). ALCA focuses on the differences in the CEA.
physical-engineered features and inherent emissions between products An important aspect of greenhouse sustainability and efficiency is
while CLCA goes further, analyzing the feedback effects driven by eco­ energy-efficient operation for greenhouses. This can be accomplished by
nomic responses [242]. Generally, LCA analysis comprises of four major monitoring or predicting the greenhouse environment and controlling
phases: goal and scope definition, life cycle inventory analysis (LCI), life the mechanical systems accordingly. Temperature, RH, and CO2 are
cycle impact assessment (LCIA), and interpretation. environmental parameters that typically need to be maintained through
Goal and scope definition serves as the initial phase and typically mechanical processes. Energy use due to temperature control can be
include the system boundary, functional unit, impact categories, and minimized using TI which allows for temperature ranges rather than
objectives. The system boundary defines what is included in the strict temperature setpoints. Environmental parameter control is beset
assessment. System boundaries for ALCA are well established, unlike by the interdependence of key parameters and spatial variability of the
CLCA where there are no general principles [243]. Functional units greenhouse microclimate. There are many algorithms and control stra­
represent the quantifiable aspect of a product, impact categories help tegies that have been developed to dynamically control these parame­
characterize the impact of different options, while the objectives of the ters. Classic PID control algorithms are unsuitable for non-linear
LCA is usually user defined. The LCI phase evaluates the total environ­ systems, advanced control algorithms offer better accuracy, however, it
mental loads that are created at each stage of the product’s life cycle. depends on the precision of the mathematical model which is sometimes
Data can be obtained from administered questionnaires, databases like difficult to obtain. Intelligent algorithms do not rely heavily on mathe­
Ecoinvent 3, and peer-reviewed articles. The LCIA step measures envi­ matical models but require large training data. Hybrid controls offer a
ronmental impact by multiplying the LCI results with impact factors. better alternative, combining the strengths of multiple control algo­
The final phase is the interpretation phase which can include feedback rithms. However, each greenhouse offers unique challenges and there is
for goal and scope redefinition. no established “best solution.” It is also difficult to account for human
LCA has seen a wide range of application in greenhouses. Bosona and factors such as harvesting and pruning in the formulation of control
Gebresenbet [244] performed LCA to study the environmental impact of models. Although proprietary control systems exist for indoor agricul­
fresh vs dried tomatoes. They found the latter to have a lower envi­ ture, growers from small to medium-sized greenhouses or indoor growth
ronmental impact through loss reduction although it consumed more facilities require low-cost systems (such as Arduino or Raspberry Pi-
energy. Zarei et al. [245] found greenhouse cucumber production to based monitoring systems).
consume less natural gas for heating compared to tomatoes and Because of the non-linear nature of greenhouse systems, the control
open-field cultivation. Dias et al. [246] found heating with fossil fuels to of these key parameters is a complex task. A lack of consensus on what
account for between half to four-fifths of the overall impact for ozone constitutes the optimal setpoint for these parameters is also a pressing
depletion. Environmental impact gap incorporated with LCA has been issue. Still, research performed over the last two decades has suggested
employed in evaluating the mitigation potential for greenhouse culti­ that control systems that can cope with dynamic conditions within a
vated pepper [247]. Fertilizer use and greenhouse structure materials greenhouse are feasible. Sensor networks and the use of control algo­
were identified as major contributors. The environmental impact of rithms continue to play a key role, and most importantly, farmers are
RTGs and multi-tunnel greenhouses [248] and greenhouse lighting becoming receptive to the use of such technology. But the cost, reli­
[249] has also been studied. ability, and accuracy of these sensors could be further improved. A
LCA studies play a key role in decision-making, identifying energy- number of these control research studies also need to be validated using
efficient production methods [245,247]. A limitation of LCA is that re­ empirical data. However, at the current level of operation, energy sav­
sults from different studies are not easily comparable, they vary based ings can be realized through the intelligent operation of greenhouse
on boundary conditions, production strategy and geographical location equipment to take advantage of periods of surplus energy supply,
[250]. For the LCI phase, estimation based on market information could reducing energy cost at peak demand.
be used in place of unavailable data, however, its accuracy impacts the On the aspect of energy modeling, we found a number of these
results. Further research should explore the establishment of a bench­ modeling techniques to be complex, requiring specialized software and
mark for LCA result comparison. Existing studies have focused on the knowledge, putting them out of the reach of everyday urban growers.
ALCA approach. However, CLCA would be vital for analyzing market Such models were also limited in their approach, applied to only a
effects and policy-making, therefore more studies should focus on CLCA specific type of crop. The adoption of state-of-the-art modeling tech­
for greenhouses. Full LCAs for emerging areas like plant factories is also niques such as SD and AML in greenhouses or CEA is lacking. Hence, the
required to justify its evident benefits over open and semi-closed need to integrate advanced control approaches with such modeling
greenhouses. methods so feedback from the system can be identified and implemented
to improve system performance. As research continues and greenhouse
5. Conclusion design evolves, it is trending towards high-tech facilities such as plant
factories, vertical and SC farms. These require large capital costs;
This literature review on greenhouse environmental control covers therefore, it is important to accurately predict and model their energy

10
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

consumption. LCA studies for such facilities are required to compare [10] Xu J, Li Y, Wang RZ, Liu W, Zhou P. Experimental performance of evaporative
cooling pad systems in greenhouses in humid subtropical climates. Appl Energy
their environmental impact to open and semi-closed greenhouses. This
2015;138:291–301.
would also provide vital information on which areas are the most [11] Shamshiri R, Ismail WIW. A review of greenhouse climate control and automation
energy-intensive and how alternatives could be explored via energy systems in tropical regions. J Agric Sci Appl 2013;2(3):176–83.
simulation. Despite the obvious need, there is no commonly-used plat­ [12] Adams SR. The physiology of flowering: quantifying the effects of photo-thermal
environment. In: III international symposium on models for plant growth,
form that supports greenhouse energy modeling. Many disparate models environmental control and farm management in protected cultivation. vol. 718;
have been created over the years; they each have their strengths and 2006. p. 557–66.
weaknesses and are not generally comparable with one another. This [13] Vox G, Teitel M, Pardossi A, Minuto A, Tinivella F, Schettini E. Sustainable
greenhouse systems. Sustainable agriculture: technology, planning and
review concludes there is a great need for a readily available and tech­ management. New York, NY, USA: Nova Science Publishers, Inc.; 2010. p. 1–79.
nically accessible modeling tool that can accurately represent the full [14] Van Beveren PJM, Bontsema J, Van Straten G, Van Henten EJ. Minimal heating
sophistication of the greenhouse environment including energy, crop and cooling in a modern rose greenhouse. Appl Energy 2015;137:97–109.
[15] Van Henten EJ, Bontsema J. Open-loop optimal temperature control in
growth, and evapotranspiration. This would be beneficial to individual greenhouses. Int Symp High Technol Greenh Syst Manag: Greensys2007 2007;
projects and serve to bring some consistency to research efforts. Future 801:629–36.
energy modeling studies should consider a wide variety of crops for [16] K€orner O, Challa H. Design for an improved temperature integration concept in
greenhouse cultivation. Comput Electron Agric 2003;39(1):39–59.
indoor agriculture and emerging indoor farming scenarios such as ver­ [17] K€orner O, Bakker MJ, Heuvelink E. Daily temperature integration: a simulation
tical farm, SC, and building-integrated indoor agriculture, and the study to quantify energy consumption. Biosyst Eng 2004;87(3):333–43.
integration of renewable energy sources. Energy model and greenhouse [18] Rijsdijk AA, Vogelezang JVM. Temperature integration on a 24-hour base: a more
efficient climate control strategy. Acta Hortic (Wagening) 2000;519:163–70.
operation control could intersect in the formulation of appropriate
[19] Pressman E, Shaked R, Firon N. Exposing pepper plants to high day temperatures
schedules for HVAC, light and shade operations. Indeed, there is some prevents the adverse low night temperature symptoms. Physiol Plant 2006;126
immense energy saving potential in these areas. (4):618–26.
CEA is becoming increasingly important, therefore a detailed [20] Campen JB. Dehumidification of greenhouses. Wageningen, NL: Wageningen
University; 2009.
knowledge of its impact on energy use is vital for energy policy makers [21] K€orner O, Challa H. Process-based humidity control regime for greenhouse crops.
and urban planners. Compared to traditional agriculture, energy use can Comput Electron Agric 2003;39(3):173–92.
be significantly reduced through CEA by reducing the distance between [22] De Halleux D, Gauthier L. Energy consumption due to dehumidification of
greenhouses under northern latitudes. J Agric Eng Res 1998;69(1):35–42.
production and consumption centers. However, due to the energy- [23] K€orner O, Challa H. Temperature integration and process-based humidity control
intensive nature of CEA, a bit of skepticism exists about whether they in chrysanthemum. Comput Electron Agric 2004;43(1):1–21.
really achieve energy efficiency compared to traditional methods. [24] Ministry of agriculture, food and rural affairs. Carbon dioxide in greenhouses.
http://www.omafra.gov.on.ca/english/crops/facts/00-077.htm. [Accessed 21
Therefore, this creates the need for energy simulation and modeling, full July 2017].
LCA as well as the energy-efficient operation and control of CEA pro­ [25] Baille A. Trends in greenhouse technology for improved climate control in mild
jects. In fact, greenhouse control depends on energy models to precisely winter climates. Acta Hortic (Wagening) 2001;559:161–8.
[26] Kl€aring HP, Hauschild C, Heibner A, Bar-Yosef B. Model-based control of CO2
monitor and predict greenhouse microclimate. With growing energy and concentration in greenhouses at ambient levels increases cucumber yield. Agric
food demands, greenhouses have the potential to become a much larger For Meteorol 2007;143(3):208–16.
industry than they are today. If designed and operated with efficiency in [27] He F, Zeng L, Li D, Ren Z. Study of LED array fill light Based on parallel particle
swarm Optimization in greenhouse planting. Information processing in
mind, greenhouses can respond to the growing trend toward local, year-
agriculture. 2018.
round, commercial-scale farming. To make this feasible, researchers and [28] Pinho P, Hytonen T, Rantanen M, Elomaa P, Halonen L. Dynamic control of
designers should continue to investigate the unique characteristics of supplemental lighting intensity in a greenhouse environment. Light Res Technol
greenhouses and conduct interdisciplinary research that aims to maxi­ 2013;45(3):295–304.
[29] Clausen A, Maersk-Moeller HM, Soerensen JC, Joergensen BN, Kjaer KH,
mize energy, water efficiency, and crop quality. Future studies should Ottosen CO. Integrating commercial greenhouses in the smart grid with demand
also extend such approach to emerging indoor farming scenarios which response based control of supplemental lighting. International Conference on
further shrinks the distance from production to consumption while Industrial Technology and Management Science. Atlantis Press; 2015. 2015.
[30] Pasgianos GD, Arvanitis KG, Polycarpou P, Sigrimis N. A nonlinear feedback
offerring a great level of environmental control. However, energy effi­ technique for greenhouse environmental control. Comput Electron Agric 2003;40
ciency of such systems should be given more attention. (1):153–77.
[31] Reece C. Evaluation of a line heat dissipation. Sens for Meas Doil Matric Potential
Soil Sci Soc America J 1996;60(4):1022–8.
References [32] Flint A, Campbell GS, Ellett KM, Calissendorff C. Calibration and temperature
correction of heat dissipation matric potential sensors. Soil Sci Soc Am J 2002;66
[1] Nature International Weekly Journal of Science. One-third of our greenhouse gas (5):1439–45.
emissions come from agriculture. 2012. https://www.nature.com/news/one-thir [33] Stanghellini C, Bontsema J. A soft sensor for on-line estimation of ventilation of a
d-of-our-greenhouse-gas-emissions-come-from-agriculture-1.11708. [Accessed 6 greenhouse. 2007.
April 2019]. [34] Janssen HJJ, Sarlikioti V, Gieling TH, Meurs EJJ, Marcelis LFM, van Dugteren R.
[2] Food and Agricultural Organization of the United Nations (FAO). Global food A prototype sensor for estimating light interception by plants. In: A greenhouse.
losses and food waste - extent. Rome: Causes, and Prevention; 2011. International symposium on high technology for greenhouse system
[3] Taki M, Rohani A, Rahmati-Joneidabad M. Solar thermal simulation and management: greensys2007, vol. 801; 2007. p. 621–8.
applications in greenhouse. Inf Process Agric 2018;5:83–113. [35] Wang N, Zhang N, Wang M. Wireless sensors in agriculture and food
[4] State Agrilyst. Of indoor farming. 2017. https://www.agrilyst. industry—recent development and future perspective. Comput Electron Agric
com/stateofindoorfarming2017/#cta. [Accessed 25 September 2018]. 2006;50(1):1–14.
[5] Vadiee A, Yaghoubi M, Sardella M, Farjam P. Energy analysis of fuel cell system [36] Bajer L, Krejcar O. Design and realization of low cost control for greenhouse
for commercial greenhouse application – a feasibility study. Energy Convers environment with remote control. IFAC-PapersOnLine 2015;48(4):368–73.
Manag 2015;89:925–32. [37] Enokela JA, Othoigbe TO. An automated greenhouse control system using
[6] Rorabaugh P, Jensen M, Giacomelli G. Introduction to controlled environment Arduino prototyping platform. Aust J Eng Res 2015;1(1):64–73.
agriculture and hydroponics. Controlled Environment Agriculture Center; 2002. [38] Anire RB, Cruz FRG, Agulto IC. Environmental wireless sensor network using
p. 1–130. raspberry pi 3 for greenhouse monitoring system. 9th international conference on
[7] National Greenhouse Manufacturers Association (N.G.M.A). Environmental humanoid, nanotechnology, information technology, communication and
controls. 2010. https://ngma.com/wp-content/uploads/2018/05/Environmenta control, environment and management (HNICEM). IEEE; 2017.
l2010.pdf. [Accessed 12 September 2018]. [39] Cabaccan CN, Cruz FRG, Agulto IC. Wireless sensor network for agricultural
[8] Mashonjowa E, Ronsse F, Milford JR, Pieters JG. Modelling the thermal environment using raspberry pi based sensor nodes. 9th international conference
performance of a naturally ventilated greenhouse in Zimbabwe using a dynamic on humanoid, nanotechnology, information technology, communication and
greenhouse climate model. Sol Energy 2013;91:381–93. control, environment and management (HNICEM). IEEE; 2017.
[9] Hassan GE, Salah AH, Fath H, Elhelw M, Hassan A, Saqr KM. Optimum [40] Shaffer G, Achieng C, Quezada F, Soenksen C, Gilmour T, Song J, Ready D.
operational performance of a new stand-alone agricultural greenhouse with A development of a PV-powered aquaponics system. PhD dissertation, division of
integrated-TPV solar panels. Sol Energy 2016;136:303–16. engineering. Siloam Springs, R, USA: John Brown University; 2014.

11
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

[41] Ojha T, Misra S, Raghuwanshi NS. Wireless sensor networks for agriculture: the [71] Goodchild MS, Jenkins MD, Whalley WR, Watts CW. A novel dielectric
state-of-the-art in practice and future challenges. Comput Electron Agric 2015; tensiometer enabling precision PID-based irrigation control of polytunnel-grown
118:66–84. strawberries in coir. Biosyst Eng 2018;165:70–6.
[42] Liu H, Meng Z, Cui S. A wireless sensor network prototype for environmental [72] Ding Y, Wang L, Li Y, Li D. Model predictive control and its application in
monitoring in greenhouses. International conference on wireless agriculture: a review. Comput Electron Agric 2018;151:104–17.
communications. Networking and Mobile Computing; 2007. p. 2344–7. [73] Oliveira JB, Boaventura-Cunha J, Moura Oliveira P. A feasibility study of sliding
[43] Yoo SE, Kim JE, Kim T, Ahn S, Sing J, Kim D. Automated agriculture system based mode predictive control for greenhouses. Optim Control Appl Methods 2016;37
on WSN. IEEE Int Symp Consumer Electron 2007:1–5. (4):730–48.
[44] Martinovi�c G, Simon J. Greenhouse microclimatic environment controlled by a [74] Ramdani M, Hamza A, Boughamsa M. Multiscale fuzzy model-based short term
mobile measuring station. NJAS - Wageningen J Life Sci 2014;70:61–70. predictive control of greenhouse microclimate. In: Industrial informatics (INDIN).
[45] Wang J, Zhou J, Gu R, Chen M, Li P. Manage system for internet of things of IEEE 13th International Conference on. 2015; 2015. p. 1348–53.
greenhouse based on GWT. Inf process agric 2018;5(2):269–78. [75] M. Guoqi, Q. Linlin, L. Xinghua and W. Gang. Modeling and predictive control of
[46] Vermeulen K, Steppe K, Linh NS, Lemeur R, De Backer L, Bleyaert P, et al. greenhouse temperature-humidity system based on MLD and time-series. in
Simultaneous response of stem diameter, sap flow rate and leaf temperature of Control conference (CCC), 2015 34th Chinese. 2015. IEEE. pp. 2234-2239.
tomato plants to drought stress, vol. 801. International Symposium on High [76] Gruber JK, Guzman JL, Rodriguez F, Bordons C, Berenguel M, Sanchez JA.
Technology for Greenhouse System Management: Greensys2007; 2008. Nonlinear MPC based on a Volterra series model for greenhouse temperature
p. 1259–66. control using natural ventilation. Contr Eng Pract 2011;19(4):354–66.
[47] Jansen R, Hofstee JW, Verstappen F, Bouwmeester H, Posthumus HM, van [77] El Ghoumari MY, Tantau HJ, Serrano J. Non-linear constrained MPC: real-time
Henten E. A method to detect baseline emission and plant damage induced implementation of greenhouse air temperature control. Comput Electron Agric
volatile emission in a greenhouse, vol. 801. International Symposium on High 2005;49(3):345–56.
Technology for Greenhouse System Management; 2008. p. 1415–22. [78] Pohlheim H, Heibner A. Optimal control of greenhouse climate using real-world
Greensys2007. weather data and evolutionary algorithms. In: Proceedings of the 1st annual
[48] Rodriguez S, Gualotuna T, Grilo C. A system for the monitoring and predicting of conference on genetic and evolutionary computation. vol. 2. Morgan Kaufmann
data in precision agriculture in a rose greenhouse based on wireless sensor Publishers Inc; 1999. p. 1672–7.
networks. Procedia Comput Sci 2017;121:306–13. [79] van Ooteghem RJC, Stigter JD, van Willigenburg LG, van Starten G. Optimal
[49] Abbasi AZ, Islam N, Shaikh ZA. A review of wireless sensors and networks’ control of a solar greenhouse. In: European control conference (ECC). IEEE; 2003.
applications in agriculture. Comput Stand Interfac 2014;36(2):263–70. p. 2739–46.
[50] Jiang JA, Wang CH, Liao MS, Zheng XY, Liu JH, Chuang CL, et al. A wireless [80] Pucheta JA, Schugurensky C, Fullana R, Patino H, Kuchen B. Optimal greenhouse
sensor network-based monitoring system with dynamic convergecast tree control of tomato-seedling crops. Comput Electron Agric 2006;50(1):70–82.
algorithm for precision cultivation management in orchid greenhouses. Precis [81] Ramírez-Arias A, Rodriguez F, Guzman JL, Berenguel M. Multiobjective
Agric 2016;17(6):766–85. hierarchical control architecture for greenhouse crop growth. Automatica 2012;
[51] Wang C, Zhao C, Qiao X, Zhang X, Zhang Y. The design of wireless sensor 48(3):490–8.
networks node for measuring the greenhouse’s environment parameters. [82] Blasco X, Martinez M, Senent J, Sanchis J. Model Based Predictive control using
International Conference on Computer and Computing Technologies in Genetic algorithms. Application to greenhouses Climate control. In: International
Agriculture. Boston, MA: Springer; 2007. p. 1037–46. work-conference on artificial neural networks. Springer; 2001. p. 457–65.
[52] Yuhan J, Yiqiong J, Ting L, Man Z, Sha S, Minzan L. An improved method for [83] Ito K. Greenhouse temperature control with wooden pellet heater via model
prediction of tomato photosynthetic rate based on WSN in greenhouse. Int J Agric predictive control approach. In: Control & automation (MED), 2012 20th
Biol Eng 2015;9(1):146–52. mediterranean conference on. IEEE; 2012. p. 1542–7.
[53] Aiello G, Giovino I, Vallone M, Catania P, Argento A. A decision support system [84] Xu ZT, Yao ZY, Chen LJ, Du SF. Greenhouse air temperature predictive control
based on multisensor data fusion for sustainable greenhouse management. using the dynamic matrix control. In: Intelligent control and information
J Clean Prod 2018;172:4057–65. processing (ICICIP) fourth international conference on. IEEE; 2013. p. 349–53.
[54] Ca~nadas J, Sanchez-Molina JA, Rodriguez F, del Aguila IM. Improving automatic [85] Hu H, Xu L, Wei R, Zhu B. Multi-objective control optimization for greenhouse
climate control with decision support techniques to minimize disease effects in environment using evolutionary algorithms. Sensors 2011;11(6):5792–807.
greenhouse tomatoes. Inf Process Agric 2017;4(1):50–63. [86] Chunfeng Z, Yonghui C. Applications of DMC-PID algorithm in the measurement
[55] Groener B, Knopp N, Korgan K, Perry R, Romero J, Smith K, et al. Preliminary and control system for the greenhouse environmental factors. In: Control and
Design of a low-cost greenhouse with open source control systems. Procedia decision conference (CCDC), 2011 Chinese IEEE; 2011. p. 483–5.
EngInt Workshop Greenh Environ Control Crop Prod in Semi-Arid Reg 2015;107: [87] Coelho JP, de Moura Oliveira PB, Cunha JB. Greenhouse air temperature
470–9. predictive control using the particle swarm optimisation algorithm. Comput
[56] Ferentinos KP, Katsoulas N, Tzounis A, Bartzanas T, Kittas C. Wireless sensor Electron Agric 2005;49(3):330–44.
networks for greenhouse climate and plant condition assessment. Biosyst Eng [88] El Ghoumari MY, Tantau HJ, Megas D, Serrano J. Real Time non linear
2017;153:70–81. constrained model predictive control of a greenhouse. IFAC Proc. Vol. 2002;35
[57] Bontsema J, van Henten EJ, Gieling TH, Swinkels GLAM. The effect of sensor (1):61–5.
errors on production and energy consumption in greenhouse horticulture. [89] van Ooteghem R, van Willigenburg L, van Straten G. Receding horizon optimal
Comput Electron Agric 2011;79(1):63–6. control of a solar greenhouse. In: International conference on sustainable
[58] Jain MK. Wireless sensor networks: security issues and challenges. Int J Comput greenhouse systems-greensys2004. vol. 691; 2004. p. 797–806.
Inf Technol 2011;2(1):62–7. [90] Xu L, Hu H, Zhu B. Energy-saving control of greenhouse climate based on MOCC
[59] van Straten G, van Willigenburg G, van Henten E, van Ooteghem R. Optimal strategy. In: Proceedings of the first ACM/SIGEVO summit on genetic and
control of greenhouse cultivation. CRC press; 2010. evolutionary computation ACM; 2009. p. 645–50.
[60] Voogt J, van Weel P. Climate control based on stomatal behavior in a semi-closed [91] Chen J, Xu F, Tan D, Shen Z, Zheng L, Ai Q. A control method for agricultural
greenhouse system ’aircokas’. International Workshop on Greenhouse greenhouses heating based on computational fluid dynamics and energy
Environmental Control and Crop Production in Semi-Arid Regions 2008;797: prediction model. Appl Energy 2015;141:106–18.
151–6. [92] Blasco X, Martinez M, Herrero JM, Ramos C, Sanchis J. Model-based predictive
[61] L�opez-Cruz IL, Fitz-Rodriguez E, Torres-Monsivais JC, Trejo-Zuniga EC, Ruiz- control of greenhouse climate for reducing energy and water consumption.
Garcias A, Ramirez-Arias A. Control strategies of greenhouse climate for Comput Electron Agric 2007;55(1):49–70.
vegetables production, biosystems engineering: biofactories for food production [93] Gonz� alez R, Rodriguez F, Guzman JL, Berenguel M. Robust constrained economic
in the century XXI. Cham: Springer; 2014. p. 401–21. receding horizon control applied to the two time-scale dynamics problem of a
[62] Behrooz F, Mariun n, Marhaban M, Mohd Radzi M, Ramli A. Review of control greenhouse. Optim Control Appl Methods 2014;35(4):435–53.
techniques for HVAC systems—nonlinearity approaches based on Fuzzy cognitive [94] van Straten G, van Willigenburg LG, Tap RF. The significance of crop co-states for
maps. Energies 2018;11(3):495. receding horizon optimal control of greenhouse climate. Contr Eng Pract 2002;10
[63] Dounis AI, Caraiscos C. Advanced control systems engineering for energy and (6):625–32.
comfort management in a building environment—a review. Renew Sustain [95] van Straten G, Tap RF, van Willigenburg LG. Sensitivity of on-line RHOC of
Energy Rev 2009;13(6–7):1246–61. greenhouse climate to adjoint variables for the crop. IFAC Proc. Vol. 1999;32(2):
[64] Mirinejad H, Sadati SH, Ghasemian M, Torab H. Control techniques in heating, 5547–51.
ventilating and air conditioning (HVAC) systems 1. 2008. [96] Ramírez-Arias A, Rodriguez F, Guzman JL, Arahal MR, Berenguel M, Lopez JC.
[65] Blevins T, Wojsznis WK, Nixon M. Advanced control foundation: tools, techniques Improving efficiency of greenhouse heating systems using model predictive
and applications. International Society of Automation (ISA); 2013. control. In: Proceedings of the 16th IFAC world congress; 2005.
[66] Li Y, Ding Y, Li D, Miao Z. Automatic carbon dioxide enrichment strategies in the [97] Liang M-H, Chen L-J, He Y-F, Du S-F. Greenhouse temperature predictive control
greenhouse: a review. Biosyst Eng 2018;171:101–19. for energy saving using switch actuators. IFAC-PapersOnLine 2018;51(17):
[67] Instruments National. PID theory explained. 2018. http://www.ni.com/wh 747–51.
ite-paper/3782/en/#top. [Accessed 23 September 2018]. [98] Lecomte A, Flaus JM, Brajeul E, Brun R, Reich P, Barthelemy L. Multivariable
[68] Wang Y-G, Shi Z-G, Cai W-J. PID autotuner and its application in HVAC systems. greenhouse control: applications to fertigation and climate management. In:
In: Am control conf IEEE. vol. 3; 2001. p. 2192–6. International conference on sustainable greenhouse systems-greensys2004 691;
[69] Afram A, Janabi-Sharifi F. Theory and applications of HVAC control systems–A 2004. p. 249–58.
review of model predictive control (MPC). Build Environ 2014;72:343–55. [99] Ferreira P, Ruano A. Discrete model-based greenhouse environmental control
[70] Hu H, Xu L, Wei R, Zhu B. Multi-objective control optimization for greenhouse using the branch & bound algorithm. IFAC Proc. Vol. 2008;41(2):2937–43.
environment using evolutionary algorithms. Sensors 2011;11(6):5792–807.

12
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

[100] Boaventura-Cunha J, Couto C, Ruano A. A greenhouse climate multivariable [133] Casta~neda-Miranda A, Casta~ no VM. Smart frost control in greenhouses by neural
predictive controller. In: International conference and British-Israeli workshop on networks models. Comput Electron Agric 2017;137:102–14.
greenhouse techniques towards the 3rd millennium 2000 534; 2000. p. 269–76. [134] Ding J-T, Yao H-Y, Zang Z-L, Huang M, Zhou S-J. Precise control and prediction of
[101] Hasni A, Draoui B, Bounaama F, Tamali M, Boulard T. Evolutionary algorithms in the greenhouse growth environment of Dendrobium candidum. Comput Electron
the optimization of natural ventilation parameters in a greenhouse with Agric 2018;151:453–9.
continuous roof vents. In: Int Symp Greenh Cool. vol. 719; 2006. p. 49–56. [135] Fourati F. Multiple neural control of a greenhouse. Neurocomputing 2014;139:
[102] Si B, Tian Z, Jin X, Zhou X, Shi X. Ineffectiveness of optimization algorithms in 138–44.
building energy optimization and possible causes. Renew Energy 2019;134: [136] Mohamed S, Hameed I. A GA-based adaptive neuro-fuzzy controller for
1295–306. greenhouse climate control system. Alexandria Eng J June 2018;57(2):773–9.
[103] Chen L, Du S, He Y, Liang M, Xu D. Robust model predictive control for [137] Xiao H, Feng L, Zhi Y. Tuning the PID parameters for greenhouse control based on
greenhouse temperature based on particle swarm optimization. Inf Process Agric CFD simulation. In: Agro-geoinformatics (Agro-Geoinformatics) second
2018;5(3):329–38. international conference. IEEE; 2013.
[104] Hasni A, Talibi R, Draoui B, Boulard T. Optimization of greenhouse climate model [138] Gurban EH, Andreescu G-D. Comparison study of PID controller tuning for
parameters using particle swarm optimization and genetic algorithms. Energy greenhouse climate with feedback-feedforward linearization and decoupling. In:
Procedia 2011;6:371–80. System theory, control and computing (ICSTCC), 16th international conference
[105] Chen J, Zhao J, Xu F, Hu H, Ai Q, Yang J. Modeling of energy demand in the on. IEEE; 2012. p. 1–6.
greenhouse using PSO-GA hybrid algorithms. Math Probl Eng 2015;2015. [139] Hu H, Xu L, Goodman ED, Zeng S. NSGA-II-based nonlinear PID controller tuning
[106] Kyriannakis E, Arvanitis K, Sigrimis N. On–line improvement for the of greenhouse climate for reducing costs and improving performances. Neural
decentralized predictive control of the heat dynamics of a greenhouse. IFAC Proc. Comput Appl 2014;24(3–4):927–36.
Vol. 2002;35(1):37–42. [140] Wang W, Xu L, Hu H. Neuron adaptive PID control for greenhouse environment.
[107] Pi~
no�n S, Camacho EF, Kuchen B, Pena M. Constrained predictive control of a J Ind Prod Eng 2015;32(5):291–7.
greenhouse. Comput Electron Agric 2005;49(3):317–29. [141] Gurban EH, Dragomir T-L, Andreescu G-D. Greenhouse climate control
[108] Haas NT, Ierapetritou M, Singh R. Advanced model predictive feedforward/ enhancement by using genetic algorithms. J Control Eng Appl Inf 2014;16(3):
feedback control of a tablet press. J Pharm Innov 2017;12(2):110–23. 35–45.
[109] Sigrimis N, Arvanitis KG, Ferentinos KP, Anastasiou A. An intelligent [142] L�opez-Cruz I, Hern�andez-Larragoiti L. Neuro-fuzzy models for air temperature
noninteracting technique for climate control of greenhouses. In: Proceedings of and humidity of a greenhouse. In: XXVIII international horticultural congress on
the 15th triennial world congress; 2002. p. 21–6. science and horticulture for people (IHC2010): international symposium on. vol.
[110] Setiawan A, Albright LD, Phelan RM. Application of pseudo-derivative-feedback 927; 2010. p. 611–7.
algorithm in greenhouse air temperature control. Comput Electron Agric 2000;26 [143] eddine Lachouri C, Mansouri K, mourad Lafifi M, Belmeguenai A. Adaptive neuro-
(3):283–302. fuzzy inference systems for modeling greenhouse climate. Int J Adv Comput Sci
[111] Cao C, Ma L, Xu Y. Adaptive control theory and applications. J Control Sci Eng Appl 2016;7(1):96–100.
2012;2012:827353. [144] Song Y, Huang X, Feng Y. A kind of temperature and humidity adaptive predictive
[112] Udink ten Cate AJ. Modeling and (adaptive) control of greenhouse climatesvol. decoupling method in wireless greenhouse environmental test simulation system.
43. PhDThesis. Wageningen Agricultural University. Institute of Agricultural and Adv J Food Sci Technol 2013;5(10):1395–403.
Environmental Engineering and Physics The Netherlands; 1983. [145] Luan X, Shi P, Liu F. Robust adaptive control for greenhouse climate using neural
[113] Nicolosi G, Volpe R, Messineo AJE. An innovative adaptive control system to networks. Int J Robust Nonlinear Control 2011;21(7):815–26.
regulate microclimatic conditions in a greenhouse. Energies 2017;10(5):722. [146] Hu H-G, Xu L-H, Wei R-H, Zhu B-K. RBF network based nonlinear model reference
[114] Atia DM, El-madany HT. Analysis and design of greenhouse temperature control adaptive PD controller design for greenhouse climate. Int J Adv Comput Technol
using adaptive neuro-fuzzy inference system. J Electr Syst Inf Technol 2017;4(1): 2011;3:357–66.
34–48. [147] Zeng S, Hu H, Xu L, Li G. Nonlinear adaptive PID control for greenhouse
[115] M. Berenguel, L. Yebra and F. Rodríguez. Adaptive control strategies for environment based on RBF network. Sensors 2012;12(5):5328–48.
greenhouse temperature control. in European control conference (ECC), 2003. [148] Su Y, Xu L, Li D. Adaptive fuzzy control of a class of MIMO nonlinear system with
IEEE. pp. 2747-2752. actuator saturation for greenhouse climate control problem. IEEE Trans Autom
[116] Arvanitis KG, Paraskevopoulos PN, Vernardos AA. Multirate adaptive Sci Eng 2016;13(2):772–88.
temperature control of greenhouses. Comput Electron Agric 2000;26(3):303–20. [149] Hu H, Xu L, Wei R. Nonlinear adaptive Neuro-PID controller design for
[117] Rodríguez F, Guzman JL, Berenguel M, Arahal MR. Adaptive hierarchical control greenhouse environment based on RBF network. In: Neural networks (IJCNN), the
of greenhouse crop production. Int J Adapt Control Signal Process 2008;22(2): 2010 international joint conference on. IEEE; 2010. p. 1–7.
180–97. [150] Hu H, Xu L, Zhu B, Wei R. A compatible control algorithm for greenhouse
[118] Speetjens S, Stigter J, van Straten G. Towards an adaptive model for greenhouse environment control based on MOCC strategy. Sensors 2011;11(3):3281–302.
control. Comput Electron Agric 2009;67(1–2):1–8. [151] Pi~no
�n S, Soria C, Garcia C, Kuchen B. Optimal Control of a greenhouse by
[119] Willis MJ, Tham MT. Advanced process control. UK: Department of Chemical and feedback linearization. Procc. IEEE 98 INDUSCON 1998:261–5.
Process Engineering, University of Newcastle Upon Tyne; 1994. [152] Cheng W, Yang X, Wang L. Feedback feedforward linear decoupling control of
[120] van Straten G, van Henten E. Optimal greenhouse cultivation control: survey and temperature and humidity in greenhouse. J Southeast Univ (Nat Sci Ed) 2012;42:
perspectives. 2010. 5–10.
[121] Bennis N, Duplaix J, Enea G, Haloua M, Youlai H. Greenhouse climate modelling [153] Berenguel M, Rodriguez F, Guzman JL, Lacasa D, Perez-Parra J. Greenhouse
and robust control. Comput Electron Agric 2008;61(2):96–107. diurnal temperature control with natural ventilation based on empirical models.
[122] Linker R, Gutman P, Seginer I. Robust controllers for simultaneous control of In: International symposium on greenhouse cooling. vol. 719; 2006. p. 57–64.
temperature and CO2 concentration in greenhouses. Contr Eng Pract 1999;7(7): [154] Cepeda P, Ponce P, Molina A, Lugo E. Towards sustainability of protected
851–62. agriculture: automatic control and structural technologies integration of an
[123] Linker R, Kacira M, Arbel A. Robust climate control of a greenhouse equipped intelligent greenhouse. IFAC Proc. Vol. 2013;46(7):366–71.
with variable-speed fans and a variable-pressure fogging system. Biosyst Eng [155] Yaofeng H, Meihui L, Lijun C, Xiaohui Q, Shangfeng D. Greenhouse modelling and
2011;110(2):153–67. control based on TS model. IFAC-PapersOnLine 2018;51(17):802–6.
[124] Moreno JC, Berenguel M, Rodriguez F, Banos A. Robust control of greenhouse [156] Lafont F, Balmat J-F. Fuzzy logic to the identification and the command of the
climate exploiting measurable disturbances. IFAC Proc. Vol. 2002;35(1):271–6. multidimensional systems. Int J comput cognit 2004;2(2):21–47.
[125] Agmail WIR, Linker R, Arbel A. Robust control of greenhouse temperature and [157] Casta~neda-Miranda R, Ventura-Ramos Jr E, del Rocio Peniche-Vera R, Herrera-
humidity. IFAC Proc. Vol. 2009;42(6):138–43. Ruiz G. Fuzzy greenhouse climate control system based on a field programmable
[126] Babuska R, Mamdani E. Fuzzy control. Scholarpedia 2008;3(2):2103. gate array. Biosyst Eng 2006;94(2):165–77.
[127] Azaza M, Tanougast C, Fabrizio R, Mami A. Smart greenhouse fuzzy logic based [158] Nachidi M, Rodriguez F, Tadeo F, Guzman JL. Takagi–Sugeno control of
control system enhanced with wireless data monitoring. ISA Trans 2016;61: nocturnal temperature in greenhouses using air heating. ISA Trans 2011;50(2):
297–307. 315–20.
[128] M�arquez-Vera MA, Ramos-Fernandez JC, Cerecero-Natale LF, Lafont F, Balmat JF, [159] Salgado P, Boaventura-Cunha J. Greenhouse climate hierarchical fuzzy
Esparza-Villanueva JI. Temperature control in a MISO greenhouse by inverting its modelling. Contr Eng Pract 2005;13(5):613–28.
fuzzy model. Comput Electron Agric 2016;124:168–74. [160] Maher A, Kamel E, Enrico F, Atif I, Abdelkader M. An intelligent system for the
[129] Kolokotsa D, Saridakis G, Dalamagkidis K, Dolianitis S, Kaliakatsos I. climate control and energy savings in agricultural greenhouses. Energy Effic
Development of an intelligent indoor environment and energy management 2016;9(6):1241–55.
system for greenhouses. Energy Convers Manag 2010;51(1):155–68. [161] Zhang D, Wu X, Zhang C. The application of fuzzy control in greenhouse
[130] Lafont F, Balmat J-F. Optimized fuzzy control of a greenhouse. Fuzzy Sets Syst environment control. Applied Mechanics and Materials. 2014.
2002;128(1):47–59. [162] Iliev L, Zakeri A, Sazdov P, Baytelieva AM. A fuzzy logic based approach for
[131] Yam Y, Nguyen HT, Kreinovich V. Multi-resolution techniques in the rules-based integrated control of protected cultivation. World Appl Sci J 2013;24(5):561–9.
intelligent control systems: a universal approximation result. In: Proceedings of [163] Revathi S, Sivakumaran N. Fuzzy based temperature control of greenhouse. IFAC-
the 1999 IEEE international symposium on intelligent control intelligent systems PapersOnLine 2016;49(1):549–54.
and semiotics. IEEE; 1999. p. 213–8 (Cat. No. 99CH37014). [164] Javadikia P, Tabatabaeefar A, Omid M, Alimardani R, Fathi M. Evaluation of
[132] Afroz Z, Shafiullah GM, Urmee T, Higgins G. Modeling techniques used in intelligent greenhouse climate control system, based fuzzy logic in relation to
building HVAC control systems: a review. Renew Sustain Energy Rev 2018;83: conventional systems. In: Artificial intelligence and computational intelligence,
64–84. 2009. AICI’09. International conference on, vol. 4. IEEE; 2009. p. 146–50.

13
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

[165] Oduk MN, Allahverdi N. The advantages of fussy control over traditional control [197] Wang L, Kubichek R, Zhou X. Adaptive learning based data-driven models for
system in greenhouse automation. ICGST-AIML-11 conference. Dubai: UAE; 2012. predicting hourly building energy use. Energy Build 2018;159:454–61.
[166] Salgado P, Boaventura-Cunha J, Couto C. A fuzzy identification and controller for [198] Siroký
� J, Oldewurtel F, Cigler J, Privara S. Experimental analysis of model
the agricultural greenhouse. In: 7th international conference on computers in predictive control for an energy efficient building heating system. Appl Energy
agriculture; 1998. 2011;88(9):3079–87.
[167] Dhamakale S, Patil S. Fuzzy logic approach with microcontroller for climate [199] Lee KH, Braun JE. Development and application of an inverse building model for
controlling in green house. Int J Emerg Technol 2011;2(1):17–9. demand response in small commercial buildings. Proc SimBuild 2016;1(1).
[168] Koutb MA, El-Rabaie NM, Awad HA, El-Hamid IA. Environmental control for [200] Chen J, Yang J, Zhao J, Xu F, Shen Z, Zhang L. Energy demand forecasting of the
plants using intelligent control systems. IFAC Proc. Vol. 2004;37(2):101–6. greenhouses using nonlinear models based on model optimized prediction
[169] Souissi M, Chaabane M, Toumi A. Greenhouse climate discrete control with method. Neurocomputing 2016;174:1087–100.
guaranteed H∞ performance. Int J Comput Cognit 2005;3(4):27–33. [201] Li S, Willits DH, Browdy CL, Timmons MB, Losordo TM. Thermal modeling of
[170] Linker R, Seginer I, Gutman PO. Optimal CO2 control in a greenhouse modeled greenhouse aquaculture raceway systems. Aquacult Eng 2009;41(1):1–13.
with neural networks. Comput Electron Agric 1998;19(3):289–310. [202] Zhou N, Yu Y, Yi J, Liu R. A study on thermal calculation method for a plastic
[171] Fourati F, Chtourou M. A greenhouse control with feed-forward and recurrent greenhouse with solar energy storage and heating. Sol Energy 2017;142:39–48.
neural networks. Simul Model Pract Theory 2007;15(8):1016–28. [203] Golzar F, Heeren N, Hellweg S, Roshandel R. A novel integrated framework to
[172] He F, Ma C. Modeling greenhouse air humidity by means of artificial neural evaluate greenhouse energy demand and crop yield production. Renew Sustain
network and principal component analysis. Comput Electron Agric 2010;71: Energy Rev 2018;96:487–501.
S19–23. [204] Vanthoor BHE, van Henten EJ, Stanghellini C, De Visser PHB. A methodology for
[173] Rodríguez F, Arahal M, Berenguel M. Application of artificial neural networks for model-based greenhouse design: Part 3, sensitivity analysis of a combined
greenhouse climate modelling. In: Control conference (ECC), 1999 european. greenhouse climate-crop yield model. Biosyst Eng 2011;110(4):396–412.
IEEE; 1999. p. 2096–101. [205] Hill JM. Dynamic modeling of tree growth and energy use in a nursery
[174] Ferreira PM, Faria E, Ruano A. Neural network models in greenhouse air greenhouse using Matlab and Simulink. 2006.
temperature prediction. Neurocomputing 2002;43(1–4):51–75. [206] De Zwart H. Analyzing energy-saving options in greenhouse cultivation using a
[175] Fourati F, Chtourou M. A greenhouse neural control using generalized and simulation model. 1996.
specialized learning. Energy 2011;5:7. [207] Elings A, Kempkes FL, Kaarsemaker RC, Ruijs MN, van de Braak NJ, Dueck TA.
[176] Patil S, Tantau H, Salokhe V. Modelling of tropical greenhouse temperature by The energy balance and energy-saving measures in greenhouse tomato
auto regressive and neural network models. Biosyst Eng 2008;99(3):423–31. cultivation. In: International conference on sustainable greenhouse systems-
[177] Frausto HU, Pieters JG. Modelling greenhouse temperature using system greensys2004, vol. 691; 2004. p. 67–74.
identification by means of neural networks. Neurocomputing 2004;56:423–8. [208] Montero JI, Baeza E, Heuvelink E, Rieradevall J, Munoz P, Ercilla M,
[178] Hu H, Luo C, Guan Q, Li X, Chen S, Zhou Q. A fast online multivariable Stanghellini C. Productivity of a building-integrated roof top greenhouse in a
identification method for greenhouse environment control problems. Mediterranean climate. Agric Syst 2017;158:14–22.
Neurocomputing 2018;312:63–73. [209] Katsoulas N, Sapounas A, De Zwart F, Dieleman JA, Stanghellini C. Reducing
[179] Rodríguez C, Guzman JL, Rodriguez F, Berenguel M, Arahal MR. Diurnal ventilation requirements in semi-closed greenhouses increases water use
greenhouse temperature control with predictive control and online constrains efficiency. Agric Water Manag 2015;156:90–9.
mapping. IFAC Proc. Vol. 2010;43(1):140–5. [210] Benis K, Reinhart C, Ferr~ ao P. Building-integrated agriculture (BIA) in urban
[180] Montoya AP, Guzman JL, Rodriguez F, Sanchez-Molina JA. A hybrid-controlled contexts: testing A simulation-based decision support workflow. In: Building
approach for maintaining nocturnal greenhouse temperature: simulation study. simulation; 2017. 1972-1951.
Comput Electron Agric 2016;123:116–24. [211] Graamans L, Baeza E, van Den Dobblesteen A, Tsafaras I, Stanghellini C. Plant
[181] Srbinovska M, Gavrovski C, Dimcev V, Krkoleva A, Borozan V. Environmental factories versus greenhouses: comparison of resource use efficiency. Agric Syst
parameters monitoring in precision agriculture using wireless sensor networks. 2018;160:31–43.
J Clean Prod 2015;88:297–307. [212] Harbick K, Albright L. Comparison of energy consumption: greenhouses and plant
[182] Lando I, Serale G, Fabrizio E. Dynamic thermal modelling of a large plastic multi- factories. In: VIII international symposium on light in horticulture, vol. 1134;
span greenhouse: calibrated simulation and energy retrofit. In: 2018 ASABE 2016. p. 285–92.
annual international meeting. American Society of Agricultural and Biological [213] Lee C, Costola D, Loonen RC, Hensen JL. Energy saving potential of long-term
Engineers; 2018. climate adaptive greenhouse shells. In: Proceedings of BS2013: 13th conference of
[183] Joudi KA, Farhan AA. A dynamic model and an experimental study for the international building performance simulation association, vol. 26; 2013. p. 28.
internal air and soil temperatures in an innovative greenhouse. Energy Convers [214] Fitz-Rodríguez E, Kubota C, Giacomelli GA, Tignor ME, Wilson SB, McMahon M.
Manag 2015;91:76–82. Dynamic modeling and simulation of greenhouse environments under several
[184] Jolliet O, Danloy L, Gay J-B, Munday GL, Resit A. HORTICERN: an improved scenarios: a web-based application. Comput Electron Agric 2010;70(1):105–16.
static model for predicting the energy consumption of a greenhouse. Agric For [215] Liu P, Lin B, Wu X, Zhou H. Bridging energy performance gaps of green office
Meteorol 1991;55(3–4):265–94. buildings via more targeted operations management: a system dynamics
[185] Tiwari GN, Sharma PK, Goyal RK, Sutar RF. Estimation of an efficiency factor for approach. J Environ Manag 2019;238:64–71.
a greenhouse: a numerical and experimental study. Energy Build 1998;28(3): [216] Fazeli R, Davidsdottir B. Energy performance of dwelling stock in Iceland: system
241–50. dynamics approach. J Clean Prod 2017;167:1345–53.
[186] Singh MC, Singh JP, Singh KG. Development of a microclimate model for [217] Onat NC, Egilmez G, Tatari O. Towards greening the US residential building stock:
prediction of temperatures inside a naturally ventilated greenhouse under a system dynamics approach. Build Environ 2014;78:68–80.
cucumber crop in soilless media. Comput Electron Agric 2018;154:227–38. [218] Dyner I, Smith RA, Pe~ na GE. System dynamics modelling for residential energy
[187] Semple L, Carriveau R, Ting DS-K. A techno-economic analysis of seasonal efficiency analysis and management. J Oper Res Soc 1995;46(10):1163–73.
thermal energy storage for greenhouse applications. Energy Build 2017;154: [219] Oladokun M, Motawa I, Banfill P. Understanding and improving household
175–87. energy consumption and carbon emissions policies-A system dynamics approach.
[188] Alvarez-S�anchez E, Leyva-Retureta G, Portilla-Flores E, Lopez-Velazquez A. 2012.
Evaluation of thermal behavior for an asymmetric greenhouse by means of [220] Jeon C, Shin J. Long-term renewable energy technology valuation using system
dynamic simulations. Dyna 2014;81(188):152–9. dynamics and Monte Carlo simulation: photovoltaic technology case. Energy
[189] Vadiee A, Martin V. Energy management strategies for commercial greenhouses. 2014;66:447–57.
Appl Energy 2014;114:880–8. [221] Aslani A, Helo P, Naaranoja M. Role of renewable energy policies in energy
[190] McMorrow G, Wang L, Denzer A. Engineering study: using waste heat from the dependency in Finland: system dynamics approach. Appl Energy 2014;113:
western sugar plant to heat a greenhouse in the winter. 2015. https://static1. 758–65.
squarespace.com/static/5510bd02e4b04012df2abe73/t/580cfefd893fc0894 [222] Han S, Love P, Pe~ na-Mora F. A system dynamics model for assessing the impacts
0988066/1477246719838/BERGþWhiteþPaperþ1501þGreenhouse.pdf. of design errors in construction projects. Math Comput Model 2013;57(9–10):
[Accessed 25 August 2018]. 2044–53.
[191] Thomas Y, Wang L, Denzer A. Energy savings analysis of a greenhouse heated by [223] Shen LY, Wu YZ, Chan EH, Hao JL. Application of system dynamics for assessment
waste heat. In: Proceedings of the 15th IBPSA conference; 2017. p. 7–9. San of sustainable performance of construction projects. J Zhejiang Univ - Sci 2005;6
Francisco, CA, USA. (4):339–49.
[192] Bambara J, Athienitis A. Experimental evaluation and energy modeling of a [224] Pe~na-Mora F, Han S, Lee S, Park M. Strategic-operational construction
greenhouse concept with semi-transparent photovoltaics. Energy Procedia 2015; management: hybrid system dynamics and discrete event approach. J Constr Eng
78:435–40. Manag 2008;134(9):701–10.
[193] Fabrizio E. Energy reduction measures in agricultural greenhouses heating: [225] Lyneis JM, Ford DN. System dynamics applied to project management: a survey,
envelope, systems and solar energy collection. Energy Build 2012;53:57–63. assessment, and directions for future research. Syst Dyn Rev: J Syst Dyn Soc 2007;
[194] Taki M, Mehdizadeh SA, Rohani SA, Rahnama A, Rahmati-Joneidabad M. Applied 23(2–3):157–89.
machine learning in greenhouse simulation; new application and analysis. Inf [226] Egilmez G, Tatari O. A dynamic modeling approach to highway sustainability:
process agric 2018;5(2):253–68. strategies to reduce overall impact. Transp Res A Policy Pract 2012;46(7):
[195] Trejo-Perea M, Herrera-Ruiz G, Rios-Moreno J, Miranda RC, Rivasaraiza E. 1086–96.
Greenhouse energy consumption prediction using neural networks models. [227] Shepherd S. A review of system dynamics models applied in transportation.
Training 2009;1(1):2. Transportmetrica B: Transp Dyn 2014;2(2):83–105.
[196] del Sagrado J, Sanchez JA, Rodriguez F, Berenguel M. Bayesian networks for
greenhouse temperature control. J Appl Log 2016;17:25–35.

14
E. Iddio et al. Renewable and Sustainable Energy Reviews 117 (2020) 109480

[228] Zomorodian M, Lai SH, Homayounfar M, Ibrahim S, Fatemi SE, El-Shafie A. The [239] Mahmoud T, Dong Z, Ma J. An advanced approach for optimal wind power
state-of-the-art system dynamics application in integrated water resources generation prediction intervals by using self-adaptive evolutionary extreme
modeling. J Environ Manag 2018;227:294–304. learning machine. Renew Energy 2018;126:254–69.
[229] Walters JP, Archer DW, Sassenrath GF, Hendrickson JR, Hanson JD, Vadas P, [240] International Standard Organization ISO 14040. Environmental
Alarcon VJ. Exploring agricultural production systems and their fundamental management—Life cycle assessment—Principles and framework. 2006.
components with system dynamics modelling. Ecol Model 2016;333:51–65. [241] International Standard Organization ISO 14044. environmental
[230] Mallampalli VR, Mavrommati G, Thompson J, Duveneck M, Meyer S, Ligmann- management—life cycle assessment—requirements and guidelines. 2006.
Zielinska A, Druschke CG, Hychka K, Kenney MA, Kok K, Borsuk ME. Methods for [242] Martin EW, Chester MV, Vergara SE. Attributional and consequential life-cycle
translating narrative scenarios into quantitative assessments of land use change. assessment in biofuels: a review of recent literature in the context of system
Environ Model Softw 2016;82:7–20. boundaries. Curr Sustain/Renew Energy Rep 2015;2(3):82–9.
[231] Swinerd C, McNaught KR. Design classes for hybrid simulations involving agent- [243] Hertwich E. Understanding the climate mitigation benefits of product systems:
based and system dynamics models. Simul Model Pract Theory 2012;25:118–33. comment on“using attributional life cycle assessment to estimate climate-change
[232] de Salles DC, Neto ACG, Marujo LG. Using fuzzy logic to implement decision mitigation…”. J Ind Ecol 2014;18(3):464–5.
policies in system dynamics models. Expert Syst Appl 2016;55:172–83. [244] Bosona T, Gebresenbet G. Life cycle analysis of organic tomato production and
[233] Wang F, Ding L, Love PE, Edwards DJ. Modeling tunnel construction risk supply in Sweden. J Clean Prod 2018;196:635–43.
dynamics: addressing the production versus protection problem. Saf Sci 2016;87: [245] Zarei MJ, Kazemi N, Marzban A. Life cycle environmental impacts of cucumber
101–15. and tomato production in open-field and greenhouse. J Saudi Soc Agric Sci July
[234] Lijun C, Shangfeng D, Meihui L, Yaofeng H. Adaptive feedback linearization- 2019;18(3):249–55.
based predictive control for greenhouse temperature. IFAC-PapersOnLine 2018; [246] Dias GM, Ayer NW, Khosla S, van Acker R, Young SB, Whitney S, Hendricks P. Life
51(17):784–9. cycle perspectives on the sustainability of Ontario greenhouse tomato production:
[235] Hameed I, Sorensen C. A more energy efficient controller for the greenhouses benchmarking and improvement opportunities. J Clean Prod 2017;140:831–9.
climate control system. Appl Eng Agric 2010;26(3):491–8. [247] Wang X, Liu B, Wu G, Sun Y, Guo X, Jin Z, Xu W, Zhao Y, Zhang F, Zou C, Chen X.
[236] van Mourik S, van Beveren PJ, Lopez-Cruz IL, van Henten EJ. Improving climate Environmental costs and mitigation potential in plastic-greenhouse pepper
monitoring in greenhouse cultivation via model based filtering. Biosyst Eng 2019; production system in China: a life cycle assessment. Agric Syst 2018;167:186–94.
181:40–51. [248] Sany�e-Mengual E, Oliver-Sola J, Montero JI, Rieradevail J. An environmental and
[237] L�
opez-Cruz IL, van Beveren PJ, van Mourik S, van Henten EJ. Performance of economic life cycle assessment of rooftop greenhouse (RTG) implementation in
extended and unscented Kalman filters for state and parameter estimation of a Barcelona, Spain. Assessing new forms of urban agriculture from the greenhouse
greenhouse climate model. In: International symposium on new technologies and structure to the final product level. Int J Life Cycle Assess 2015;20(3):350–66.
management for greenhouses. vol. 1170. GreenSys2015; 2015. p. 175–82. [249] Zhang H, Burr J, Zhao F. A comparative life cycle assessment (LCA) of lighting
[238] Li W, Yue HH, Valle-Cervantes S, Qin SJ. Recursive PCA for adaptive process technologies for greenhouse crop production. J Clean Prod 2017;140:705–13.
monitoring. J Process Control 2000;10(5):471–86. [250] Clune S, Crossin E, Verghese K. Systematic review of greenhouse gas emissions for
different fresh food categories. J Clean Prod 2017;140:766–83.

15