Simulation of greenhouse energy use: an application of energy informatics

Abstract

Greenhouse agriculture is a highly efficient method of food production that can greatly benefit from supplemental electric lighting. The needed electricity associated with greenhouse lighting amounts to about 30% of its operating costs. As the light level of LED lighting can be easily controlled, it offers the potential to reduce energy costs by precisely matching the amount of supplemental light provided to current weather conditions and a crop’s light needs. Three simulations of LED lighting for growing lettuce in the south-east of the US using historical solar radiation data for the area were conducted. Lighting costs can be potentially reduced by approximately 60%.

Simulation of greenhouse energy use: an application of energy informatics. Available from: https://www.researchgate.net/publication/326783542_Simulation_of_greenhouse_energy_use_an_application_of_energy_informatics [accessed Aug 11 2018].

Full text

R E S E A R C H Open Access
Simulation of greenhouse energy use: an
application of energy informatics
Richard T. Watson
1*
, Marie-Claude Boudreau
1
and Marc W. van Iersel
2
* Correspondence: rwatson@terry.
uga.edu
1
Department of Management
Information Systems, University of
Georgia, Athens, GA 30602-1575,
USA
Full list of author information is
available at the end of the article
Abstract
Greenhouse agriculture is a highly efficient method of food production that can greatly
benefit from supplemental electric lighting. The needed electricity associated with
greenhouse lighting amounts to about 30% of its operating costs. As the light level
of LED lighting can be easily controlled, it offers the potential to reduce energy costs
by precisely matching the amount of supplemental light provided to current weather
conditions and a crop’s light needs. Three simulations of LED lighting for growing
lettuce in the south-east of the US using historical solar radiation data for the area
were conducted. Lighting costs can be potentially reduced by approximately 60%.
Keywords: Controlled-environment agriculture, Horticulture, Supplemental lighting,
Simulation, Energy efficiency
Background
In January 2007, the first two authors of this article started a project to link informa-
tion systems (IS) scholarship to ecological sustainability. They were concerned that IS
scholars were paying minimal attention to global warming and that they did not
consider it a problem worthy of their attention. While leading scientists (Holdren
2009) were calling for their peers to spend 10% of their time solving the existential
threat resulting from the burning of fossil fuels, IS scholars were generally ignoring the
issue. Consequently, this important problem has remained a marginal IS research topic
(Gholami et al. 2016;Malhotraetal.2013). Since 2010, we have advocated for academic
leaders (journal editors in particular) to embrace environmental sustainability as a core
principle in the research they publish (Watson et al. 2010); the emergence of this journal
constitutes a critical step in engaging more scholars in solving the problem of our time.
The project to engage IS scholars in sustainability scholarship resulted in a proposal to
create Energy Informatics as a new field of IS research (Watson et al. 2010) (Watson and
Boudreau 2011). Leveraging the Energy Informatics framework, we produced a case study
on how one of the world’s major logistic companies was applying the principles to reduce
energy consumption and advance sustainability (Watson and Boudreau 2011). Energy
efficiency, the invisible fuel, reduces the demand for fossil fuels and thus contributes to
carbon emission reduction.
The Energy Informatics framework identifies the central elements of an energy supply
and demand system, including the key stakeholders, the three major eco-goals, and the
major social forces influencing an energy production/consumption system. The framework
E
ner
gy
Informatic
s
© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International
License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and
indicate if changes were made.
Watson et al. Energy Informatics (2018) 1:1
https://doi.org/10.1007/s42162-018-0005-7
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

directs attention to flow networks, sensor networks, and sensitized objects as these are the
source and destination of the necessary data inputs and outputs to advance energy efficiency
through the use of information systems. The framework arose from reviewing many energy
production/consumption systems, such as traffic congestion management and building
energy management.
The principal observation is that economies consist of flow networks (e.g., cars, water,
people, packages, containers) that consume energy. The management of the energy con-
sumed by such flows requires sensor networks to provide digital data streams reporting on
the current status of the flow network so that high quality decisions can be made about the
status of sensitized objects that can control the flow network. Actions might be as simple as
turning off a valve supplying hot air to a work space. They may also be remotely controlled,
such as in the management of hundreds of traffic lights to reduce city traffic congestion.
The original Energy Informatics framework has proved robust because it identifies
the main components of an energy flow network and the means of controlling them. It
highlights the central role of an information system, linking the interdependencies
between supply and demand and the major components (i.e., flow networks, sensor
networks, and sensitized objects).
The Energy Informatics framework has been applied to multiple domains, such as
road pricing, farming, logistics, bicycle sharing, and others. Recently, we have applied
this framework to another major existential problem: food security.
Food security
Food security is a critical problem that demands the attention of IS scholars, as well as
horticulturists, agricultural scientists, and others concerned with food production. Over
the next 30 years, the world’s population is predicted to grow by up to 34% and
urbanization will increase by around 20%. To feed this wealthier and larger population,
food production must increase by an estimated 70% (Food and Agriculture Organization
of the United Nations 2009). Achieving food security by minimizing variations in supply
and adjusting to the growth in food demand presents many challenges that will likely
require a major adjustment in current agricultural practices (Magnin 2016). Food security
could be enhanced by reducing personal meat consumption and shifting to a predominantly
vegetarian diet, but this would require significant behavioral changes to well-established
customs and practices. Our focus in this research is on modifying agricultural practices
rather than promoting behavioral change.
A solution
Controlled-environment agriculture (CEA), such as indoor farms and greenhouses, is a
key path to increasing food production. CEA can produce up to 20 times as much
high-end, pesticide-free produce as a similar-size plot of soil but requires electric lighting
to do so (Science Illustrated 2011). Such facilities will be necessary to meet the future
demand for quality fruits and vegetables, particularly in China with its rapidly growing
and large middle class (Science Illustrated 2011).
In 2016, the CEA market was dominated by Europe, Middle East, and Africa (EMEA),
following by the Americas, and Asia Pacific, with percentage shares and values of 62%,
23%, and 15% of a market valued at USD 20.25 billion. The Netherlands, Spain, and Italy
Watson et al. Energy Informatics (2018) 1:1 Page 2 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

dominate EMEA production; in the Americas, horticulture is mainly in North America;
and China has been growing rapidly with more than one million greenhouses.
1
The smart greenhouse is emerging as a solution, offering various collections of integrated
technologies in a greenhouse, to improve the productivity of CEA. The smart greenhouse
vision is based on sensors, actuators, and monitoring and control system that can optimize
plant growth and quality and automate the growing process. This market was valued at
approximately USD 680.3 million in 2016. With a short-term compound annual growth rate
of around 14.12%, it is expected to reach approximately USD 1.3 billion by 2022.
2
At this
stage, the smart greenhouse market is about 3% of the world market.
The greenhouse industry’s current practices can require considerable energy to power
electric lighting to maintain plant growth on overcast days, so as to meet production
schedules. Electricity for lighting can make up to 30% of the costs for greenhouses (van
Iersel MW, Gianino D: An adaptive control approach for LED lights can reduce the en-
ergy costs of supplemental lighting in greenhouses, Submitted). Currently, many commer-
cial greenhouses use high intensity discharge (HID) lights,
3
which have high output, cover
wide areas, and emit high heat. These lights have timers or automated control systems
that use sensors to turn on all lights at full power when natural light levels drop below a
predetermined intensity, even when only a fraction of the light might be required to reach
acrop’s needs for growth. In the extreme, some growers might leave the lights on full
power for a substantial portion of the day (as most plants need between 12 to 18 h of light
per day, depending on the species). This approach to growing crops is inefficient, resulting
in energy waste, higher operational costs, and often unnecessary carbon emissions.
Prior research to reduce CAE electricity costs by making greenhouses smarter has pro-
duced a variety of ‘branded’solutions. Intelligrow (Aaslyng et al. 2003; Markvart et al. 2007),
DynaLight (Kjaer et al. 2011,2012)(Clausenetal.2015), and DynaGrow (Sørensen et al.
2016) are successive developments of software for optimizing greenhouse production and
minimizing energy costs. DynaGrow, the most recent and advanced of cumulative research
in Danish greenhouses, applies multi-criteria methods to control a greenhouse’s climate.
The solution has been physically implemented through software and associated sensors and
actuators within a greenhouse and produces savings of 64% with LED lighting (Sørensen
et al. 2016), the concern of this article.
The study reported in this article differs in several ways. First it is a simulation rather
than a physical implementation. A key purpose is to identify the savings generated by
adaptive LED lighting and what form of software as a service (SaaS) might be economic-
ally viable. Simulation enables inexpensive consideration of alternatives. Second, there is a
single objective of minimizing electricity cost subject to ensuring that a crop receives
sufficient light each day to meet growth needs for on time contract delivery. This research,
like the Danish stream, is based on current knowledge of plant physiology.
Plant physiology
Plants grow by converting photons (sunshine or supplemental lighting), water, and CO2 to
sugars
4
and oxygen. The environmental conditions and physiology of each plant determine
the rate of photosynthesis. For the purposes of this research, we can think of a plant using
photons to transport electrons. In the light reactions of photosynthesis, photons are
absorbed by photosynthetic pigments, and the energy is used to transport electrons. This
Watson et al. Energy Informatics (2018) 1:1 Page 3 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

electron transport then results in the production of chemicals required for the synthesis of
sugars. The electron transport rate(ETR)isadirectmeasureofthelightreactionsofphoto-
synthesis in response to photosynthetic photon flux (PPF) (Fig. 1). ETR is the driving force
for photosynthesis and ultimately crop growth. Both ETR and PPF are measured in micro-
moles
5
per square meter per second (μmol m
−2
s
−1
of electrons and photons, respectively).
The efficiency of the conversion of the energy of photons into electron transport varies by
plant species, but in general we can represent this relationship as a saturation curve of the
form ETR = a.(1-e
(−b.PPF)
), where a and b vary by species. The fitted saturation curve for
lettuce illustrates the natureofthisrelationship(Fig.1). As the saturation curve shows,
conversion of photons into electron transport is most efficient at low levels of PPF, which is
an important consideration when electric lighting is used. Essentially, low levels of lighting
for long periods are more energy efficient than high levels for short periods. Furthermore,
supplemental light provided when sunlight levels are low will be used more efficiently than
supplemental light provided when sunlight levels are high.
A plant needs to transport a threshold number of moles of electrons per day to optimize
its growth. This is particularly important for commercial crops, which are usually grown
under contract with a scheduled harvesting date and defined delivery volume. Based on
experiments at the University of Georgia, it has been determined that lettuce, for example,
needs to transport approximately 3 mol m
−2
day
−1
of electrons. On a typical day, this rate
of electron transport requires approximately 18 mol of photons m
−2
day
−1
. This is close
to the recommended daily light level for year-round production of high-quality lettuce of
17 mol m
−2
day
−1
(Both et al. 1994).
We can convert daily electron transport into a required level of photons per second,
as follows:
D = ETR in moles m
−2
day
−1
to maximize growth
Fig. 1 Electron transport rate (ETR) of lettuce as a function of the photosynthetic photon flux (PPF). Source:
Unpublished University of Georgia research
Watson et al. Energy Informatics (2018) 1:1 Page 4 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Tetr = threshold ETR in μmoles m
−2
s
−1
to achieve D
Tppf = threshold PPF in μmoles m
−2
s
−1
to achieve D
t = seconds of operation per day of the greenhouse
Tetr ¼D106
t
Tppf ¼1
blog a
a−Tetr

The innovation
Supplementary lighting typically lacks intelligence and can be left on continually during
the hours of operation, if not for an entire day. This means that energy can be wasted
by providing more photons than a crop needs to optimize its daily growth.
A recent development in the context of commercial greenhouses is the replacement of
HID lights with LED technology. LED technology has many advantages, such as smaller
size (and thus easier to mount) than its HID predecessor and it is more energy efficient.
More importantly, LEDs can be designed to produce light in the part of the spectrum that
drives photosynthesis (400–700 nm), without producing infra-red radiation (which is not
used for photosynthesis). LEDs are also fully dimmable, unlike HID lights, thus allowing
growers to precisely control how much supplemental light is provided. For a LED light,
the relationship between energy consumption and photons generated is essentially linear
(National Electrical Manufacturers Association 2015), and we so assume in this research.
Leveraging this kind of lighting, a local company, Phytosynthetix,
6
collaborated with the
University of Georgia’s Horticultural Physiology Laboratory to develop an innovation that
could be transformative to the industry: adaptive LED lighting, which uses a built-in light
intensity sensor to determine how much supplemental lighting to provide when natural
lighting falls below a crop’s threshold needs (van Iersel MW, Gianino D: An adaptive
control approach for LED lights can reduce the energy costs of supplemental lighting in
greenhouses, Submitted). Adaptive lighting provides just enough light to assure optimal
crop growth and reduces electricity use compared to conventional control algorithms.
The amount of supplemental light to provide can be based on the crop’s physiological
ability to use that light efficiently (see Fig. 1). This is an important innovation in green-
house production as none of the available similar solutions take into consideration both
natural light levels and crop-specific light use efficiency.
Enhancing adaptive lighting by applying Energy Informatics
The value of adaptive LED lighting can be increased by the application of Energy Inform-
atics principles (Watson et al. 2010) to manage the lighting system to minimize electricity
costs while meeting schedule constraints. The CEA version of the Energy Informatics
framework (Fig. 2) incorporates all elements of an energy supply and demand system for
greenhouses. We have taken the basic Energy Informatics framework, the yellow section
of Fig. 2, and added details of the digital data streams (energy prices and solar radiation
forecasts) and databases (plant details and production schedule) necessary to control the
sensitized object (LED lighting). These four additional components illustrate how the core
Energy Informatics information system can be extended for CEA.
Watson et al. Energy Informatics (2018) 1:1 Page 5 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Simulations
To understand the advantage of adaptive LED lighting, while leveraging the Energy
Informatics framework adapted to the context of CEA, we ran three simulations with
increasing levels of sophistication. The first simulation leverages LED adaptive lighting.
The second incorporates a daily decision, where the adaptive lights are turned off for the
day when the expected daily solar radiation exceeds the ETR that optimizes growth. As to
the third simulation, it leverages “within day”decision making, where the adaptive lights
are turned off when the target solar radiation for the day has been achieved. Note that
whereas Fig. 3suggests an energy price forecast, we assumed a fixed energy cost for the
time being. Moreover, rather than solar radiation forecasts, we used historical solar
radiation data for a specific location and period. Last, for sake of simplicity in this first set
of simulations, we did not consider the cost of production schedule delays.
The simulations were based on a 5-year period of growing lettuce in Athens, Georgia,
with the following parameters:
D = 3 mol m
−2
day
−1
t= 20 h day
−1
a = 124.3 μmol m
−2
s
−1
b = .002737 m
2
sμmol
−1
Electricity cost = USD .12/kWh (the rate charged by the local utility)
LED light = 600 W
LED light range is 0–200 μmoles m
−2
s
−1
Fig. 2 Energy Informatics framework
Watson et al. Energy Informatics (2018) 1:1 Page 6 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

We assumed 70% transmittance of the received solar radiation into the greenhouse.
We compared the results of all three simulations to a baseline scenario, which involved
non-adaptive LED lighting. This baseline scenario, along with the three simulations, are
further detailed next. The greenhouse lighting layout assumes the use of 1,200 600 W
LED lights per hectare, capable of providing a PPF of 200 μmoles m
−2
s
−1
with all
lights on at full power.
Baseline scenario: non-adaptive LED lighting
Based on a recent survey of growers,
7
we assume a typical greenhouse uses supplemental
lighting for an average of 3.25 h per day (9 h in winter, 2 h in fall and spring, and 0 in
summer). The annual cost for this level of LED lighting is .12*3.25*365.25*600/1000 =
USD 85.47 per light per year. A one hectare greenhouse needs about 1200 lights, so the
cost would be USD 102,562 per year.
Simulation#1: Adaptive lighting
This scenario assumes that the lights can be dimmed to any PPF between 0 and 200
μmoles m
−2
s
−1
, and the relationship between energy use and PPF is linear between 0
and 600 W. At its peak, adaptive lighting consumes the same energy as non-adaptive
lighting. We use these parameters for the simulation.
Using solar radiation data collected in 10-min intervals in Athens, Georgia in 2010–
2014, we simulate the use of adaptive lighting by computing Tppf for lettuce and setting
the lights to maintain the mix of natural and supplemental lighting at this level. As the
threshold for Tppf in this case is 149 μmoles m
−2
s
−1
, the adaptive lights were set to
maintain this level. When there is sufficient natural light, the adaptive lights will consume
0 W and when there is complete darkness, they will consume 149/200*600 = 447 W.
Fig. 3 Average cost per light per day (USD) of adaptive lighting
Watson et al. Energy Informatics (2018) 1:1 Page 7 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

The cost of operating the adaptive lighting in 2010–2014 to grow lettuce for 20 h per
day is estimated to be USD 194.78 per light per year or for a one hectare greenhouse
USD 233,738 per year. As expected, the lights are costlier to operate in winter than in
summer (Fig. 3).
Simulation#2: daily decision making
We simulated a model that would inform the grower whether to turn the adaptive lights
off for the day when the forecast solar radiation for the day exceeds the total required to
achieve an ETR of D moles m
−2
day
−1
. Such an approach requires minimal investment
and under a Software as a Service (SaaS) business model, the grower could be sent a text
message on the recommended status of the adaptive lighting early each day.
The cost of operating the adaptive lighting under daily decision making in 2010–2014
to grow lettuce with 20 h of light per day is estimated to be USD 86.92 per light per year
or for a one hectare greenhouse USD 104,302 per year, slightly above the baseline
scenario. As to be expected, on many days there is no need to turn on the lights (Fig. 4).
Simulation#3: within day decision making
A major shortcoming of a daily decision making model is that the forecasted radiation for
a day could be just below the threshold, but the lights are turned on for the entire day.
Ideally, once the target for the day has been achieved, the lights should be turned off for
the remainder of the day. We simulated such a model (Fig. 5), which shows that on some
days the lights come on, but only for a short period with a corresponding lower cost.
The cost of operating the adaptive lighting under this approach in 2010–2014 to
grow lettuce for 20 h per day is estimated to be USD 32.28 per light per year or for a
one hectare greenhouse USD 38,732 per year, about one third of the baseline scenario.
Fig. 4 Average cost per light per day (USD) of adaptive lighting with daily decision making
Watson et al. Energy Informatics (2018) 1:1 Page 8 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Under a Software as a Service (SaaS) business model, the within day approach
requires that the grower invests in a lighting control system that can be controlled
remotely. The controller would receive periodically a message to set the status (on or
off) of the lights in the greenhouse. Thus, some of the energy savings will be lost to the
operation and maintenance of a lighting control system, but we expect these to be
minor compared to the energy use of the lights themselves.
Execution
The simulations were written in R, and the code is in four modules (see Additional file 1).
The main module (simulation.R) loads modules to read the parameters for a simulation
(parameters.R) and prepares data for the simulation (prepare.R). Another module
(report.R) reports the results for each of the simulations discussed previously.
The simulations take advantage of R’s vector-oriented operations for operating on
data frames, and a typical run with 5 years of data takes a few seconds. In comparison,
an earlier loop-based version of the model took 10 or so minutes to run.
The simulations require two binary input files, coded in R’s feather format,
8
which
are described in Appendix.
Return on investment
The findings of the analysis show that only within day decision making will result in a
reduction in total lighting cost compared to current manual practices (Table 1). Partial
solutions, while they might deliver light precisely to ensure optimum plant growth, are
likely to be subject to grower resistance because there is no cost advantage. Growers
can save around 60% of the current cost of lighting by retrofitting their current LED
lighting systems. Note that in the within day decision simulation, the lights are on half
the time of the current manual or timer practice, but the electricity saving is higher
Fig. 5 Average cost per light per day (USD) of adaptive lighting under within-day decision making
Watson et al. Energy Informatics (2018) 1:1 Page 9 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

because the adaptive lighting generally does not operate the lights at full power. The
energy savings are similar in magnitude to those reported for DynaGrow (Sørensen
et al. 2016), however the two sites are at different latitudes, with Denmark in the mid
50s° and Athens, Georgia close to 34°, thus resulting in different light environments.
To compute the ROI, we conservatively estimate that the cost of equipping a green-
house with a single solar radiation sensor, control software, and connecting existing LED
lighting to the control software is USD 100 per light with a lifetime of 10 years, or USD
120,000 for a one-hectare greenhouse. The computed ROI is 68% for within day decision
making. If we triple the cost estimate to USD 300 per light, then the ROI is 14%. In effect,
anything below USD 300 per light for retrofitting a greenhouse is financially sound and
likely to provide additional benefits from exact contract fulfillment.
Although the prior simulations are based on a fixed energy cost of USD 0.12 per kWh,
we wrote the code so that it can also process a file of historical real-time electricity prices.
Such pricing is typically available only to large consumers (e.g., a factory or a university).
In this case, the within day simulation determines which time periods during the day are
the least expensive for operating the lights in order to meet the daily ETR goal.
Limitations
Solar radiation and electricity prices are dependent on local conditions. In this situation,
based on 5 years of solar radiation data measured at the University of Georgia, the average
solar radiation is 5.22 kWh m
−2
day
−1
. The University is in the sunny south-east of the
US, and thus our results show the advantage of this additional solar energy. Also, the
region has some of the cheaper electricity in the US. The US average is USD .13 kWh,
with a range of USD .10 to .29 kWh. Georgia, at USD .125, is towards the lower end of
this range.
9
If the simulations had been based on solar radiation in the north-east of the
US, for example, savings would be different because of less solar radiation, more required
supplemental lighting, and higher electricity prices. In essence, our results should be inter-
preted as being appropriate for a relatively sunny region with relatively cheap electricity.
Our simulation framework can easily be used to simulate different locations.
The main purpose of these simulations is to show the potential of using solar radiation
and electricity cost forecasts to reduce the cost of operating a greenhouse. However,
rather than forecasts, we have simulated the perfect information situation where electri-
city prices and solar radiation can be precisely predicted. As most greenhouses are too
small to presently avail themselves of real-time pricing, treating the price of electricity as
fixed is appropriate. Thus, in moving beyond perfect information, the focus needs to be
on solar radiation forecasting.
It can be argued that simulations represent a simplified reality. However, such simplicity
means that they can be readily explained to a grower, which in turn is likely to increase
the likelihood of adoption. The simple decision rules are very effective in reducing cost,
Table 1 Simulation summary
Model Cost per LED light per year (USD) Average hours of LED light per day
Current manual practice 85.47 3.25
Adaptive 194.78 11.2
Adaptive with daily decision 86.92 5.09
Adaptive within day decision 32.28 1.69
Watson et al. Energy Informatics (2018) 1:1 Page 10 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

and additional complexity is unlikely to be required or produce significant additional
savings. Furthermore, the second and third simulations can be easily implemented from a
central source using a SaaS model, which lowers the cost for adoption and maintenance.
Future research
As mentioned earlier, by using historical data, we assume perfect information. Thus, the
next step is to simulate the effects of solar radiation forecasts with varying degrees of
accuracy to determine the sensitivity of the findings to forecast precision. We speculate
that sensitivity will be low for two reasons: (1) the goal is to determine whether or not to
turn on adaptive lighting for a set period, such as 10 min; and (2) a multi-period forecast
for the remainder of day can be continually adjusted throughout the day.
The essential problem, whether to turn the lights on or off for a set period would
seem to be a simpler problem than trying to forecast the level of radiation for the same
period. Intelligent adaptive LED lighting provides the precision to ensure plants receive
the necessary photons to meet growth goals, provided the LED light is on. We plan to
build two forecasting models, one for the binary decision of LED light on or off and
the second to forecast radiation levels. Based on prior research, it would appear that a
neural network (Wollsen and Jørgensen 2015) with allowance for autocorrelation is
appropriate for radiation level forecasting, whereas logistic regression might be suitable
for the binary decision.
The goal is to deliver sufficient photons each day for a plant to reach its daily growth
needs. Throughout a day, the forecast system must continually adapt to meet this target.
As the forecast period becomes shorter, then adaption can be faster. Thus, one of the
issues is to simulate the precision of the forecasting system to the length of the decision
period (e.g., 5 versus 20 min). As we expect the computing cost of each forecast to be
quite small and the communication cost of the on/off decision to each light to also be
small, the selected time period might make little financial difference for an individual
grower given the gains achieved by efficient management of LED usage. Nevertheless,
these assumptions need to be validated.
Since natural light conditions vary greatly among locations, it will also be important
to run simulations for a range of regions, especially those where greenhouses most
commonly use supplemental light, like Canada, the Northern US, and Northwestern
Europe.
We need to elaborate the simulation model to embrace more decision parameters, such
as those considered in prior research (e.g., Clausen et al. 2015; Kjaer et al. 2011,2012;
Mærsk-Møller and Jørgensen 2011; Markvart et al. 2007; Rytter et al. 2012;Sørensenetal.
2011;Sørensenetal.2016), and apply photosynthesis theories that take account of light
wavelength and the capability of controlling the wavelength of LED lighting.
There is much to learn about LED lighting because it is a relatively recent introduction
to the greenhouse environment. It enables control over spectral power distribution,
optical intensity distribution, form factor, and active color tuning. It can be used, as in this
study, to tailor light to a specific crop to improve productivity by controlling precisely the
photons delivered to ensure threshold needs for growth are met. Other aspects of plant
growth such as height, bushiness, and color or nutritional content are potential areas for
research (Pattison 2017). This research simulates LED usage approaches to minimize
electricity cost by applying prior research on threshold curves. It assumes that if a plant
Watson et al. Energy Informatics (2018) 1:1 Page 11 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

receives sufficient photons, or the right amount of light, to meet daily growth needs, the
quality of the crop will be of commercial standard. Assessment of this assumption is a
topic for future research, and the power of simulation is to help identify opportunities for
fruitful research.
Conclusion
This initial study of the application of Energy Informatics to CEA illustrates the poten-
tial to considerably reduce the energy required to operate a greenhouse. Our plans are
to extend the research to cover other commercial crops (e.g., red leaf lettuce, spinach,
basil, tomatoes, and cucumbers) as well as ornamental species, including high-light
crops (roses, petunias, and marigolds) and low light crops (impatiens, begonia, and
hellebore). Moreover, we plan to consider four diverse regions of the continental US,
because solar radiation varies greatly across this region (i.e., from an average of 3.0 to
over 6.5 kWh m
−2
day
−1
).
10
This research shows the value of the Energy Informatics framework for addressing
practical problems. It helps scientists communicate with growers by focusing on the three
elements that are necessary for an information system to manage an energy flow. In the
case of CEA, engaging growers will be critical to adoption of information systems
augmented adaptive lighting. It also demonstrates the potential for using simulation to
explore the effects of energy management options.
The Lawrence Livermore National Laboratory publishes annually an infographic on
energy use in the U.S. The 2016 version estimates that about two thirds of the energy
generated is wasted.
11
The application of Energy Informatics thinking to CEA illustrates
that in some areas of the economy, significant energy efficiency gains can be achieved
with simple solutions. We expect that CEA is not an exception and the launching of
Energy Informatics can potentially alert more scholars and practitioners to the synergis-
tic gains of integrating engineering advances, such as adaptive lighting, and information
systems to achieve higher levels of energy efficiency. Creating a sustainable society is
dependent on engineering and IS scholars combining minds and knowledge.
Endnotes
1
https://www.businesswire.com/news/home/20170713006114/en/
Global-Greenhouse-Horticulture-Market-2017-2021-Key-Geographies
2
https://www.zionmarketresearch.com/sample/smart-greenhouse-market
3
Which can be further subdivided into two types, metal halide and high-pressure
sodium.
4
Compounds of carbon, hydrogen, and oxygen. Sucrose is C
12
H
22
O
11
.
5
1 mol = 6.022 × 10
23
(Avogadro’s constant) photons or electrons.
6
http://phytosynthetix.com
7
http://stateofindoorfarming.agrilyst.com
8
https://blog.rstudio.com/2016/03/29/feather/
9
https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_6_a
10
https://azsolarcenter.org/images/articles/az/national_photovoltaic_map_2012-01.jpg
11
https://flowcharts.llnl.gov/commodities/energy
Watson et al. Energy Informatics (2018) 1:1 Page 12 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Appendix
Data availability
electricity.feather
https://www.dropbox.com/s/dxvep9kglspy47n/electricity.feather?dl=0
A binary file containing a timestamp and electricity price in cents per kWh. Measure-
ments are hourly.
radiation10.feather
https://www.dropbox.com/s/ozvtrfudnymmug8/radiation10.feather?dl=0
A binary file containing a timestamp and incident solar radiation in Watts m
−2
s
−1
.
Measurements are for every 10 min, and derived from a file of measurements made
every 2 min.
These files must be stored in an R project working directory along with the R modules.
The location of this working directory must be specified in the second line of parameter.R
Additional file
Additional file 1: R code. (TXT 6 kb)
Acknowledgements
Only the authors listed on the manuscript contributed towards the article.
Funding
This research was supported by a University of Georgia Presidential seed grant.
Authors’contributions
Drs. M-CB and RTW were responsible for the Energy Informatics simulation. Dr. MWvI provided the data on the
saturation model for lettuce and details of plant physiology relevant for the simulation. All authors read and
approved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Publisher’sNote
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Author details
1
Department of Management Information Systems, University of Georgia, Athens, GA 30602-1575, USA.
2
Department
of Horticulture, University of Georgia, Athens, GA 30602-7273, USA.
Received: 31 January 2018 Accepted: 11 June 2018
References
Aaslyng JM, Lund JB, Ehler N, Rosenqvist E (2003) IntelliGrow: a greenhouse component-based climate control system.
Environ Model Softw 18(7):657–666
Both AJ, Albright LD, Langhans RW, Reiser RA, Vinzant BG (1994) Hydroponic lettuce production influenced by
integrated supplemental light levels in a controlled environment agriculture facility: experimental results. Paper
presented at the III International Symposium on Artificial Lighting in Horticulture. p. 418
Clausen A, Maersk-Moeller HM, Soerensen JC, Joergensen BN, Kjaer KH, Ottosen CO (2015) Integrating commercial
greenhouses in the smart grid with demand response based control of supplemental lighting. Paper presented at
the International Conference Industrial Technology Management Science (ITMS 2015)
Food and Agriculture Organization of the United Nations (2009) Global Agriculture Towards 2050. Retrieved 19 Aug
2016, from Food and Agriculture Organization of the United Nations: http://www.fao.org/fileadmin/templates/wsfs/
docs/Issues_papers/HLEF2050_Global_Agriculture.pdf. Accessed 18 June 2018
Gholami R, Watson RT, Molla A, Hasan H, Bjørn-Andersen N (2016) Information systems solutions for environmental
sustainability: how can we do more? J Assoc Inf Syst 17(8):521
Holdren JP (2009) Energy for change: introduction to the special issue on Energy & Climate. Innov Technol
Gov Glob 4(4):3–12
Kjaer KH, Ottosen CO, Jørgensen BN (2011) Cost-efficient light control for production of two campanula species. Sci
Hortic 129(4):825–831
Watson et al. Energy Informatics (2018) 1:1 Page 13 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Kjaer KH, Ottosen CO, Jørgensen BN (2012) Timing growth and development of Campanula by daily light integral and
supplemental light level in a cost-efficient light control system. Sci Hortic 143:189–196
Mærsk-Møller HM, Jørgensen BN (2011) A software product line for energy-efficient control of supplementary lighting
in greenhouses. Paper presented at the The International Conference on Green Computing
Magnin C (2016) How Big Data Will Revolutionize the Global Food Chain. Retrieved 19 Aug 2016, from McKinsey &
Company: http://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/how-big-data-will-
revolutionize-the-global-food-chain?cid=digistrat-eml-alt-mip-mck-oth-1608. Accessed 18 June 2018
Malhotra A, Melville NP, Watson RT (2013) Spurring impactful research on information systems for environmental
sustainability. MIS Q 37(4):1265–1274
Markvart J, Kalita S, Nørregaard Jørgensen B, Mazanti Aaslyng J, Ottosen CO (2007) IntelliGrow 2.0–A greenhouse
component-based climate control system. Paper presented at the International Symposium on High Technology
for Greenhouse System Management: Greensys 2007 801
National Electrical Manufacturers Association (2015) Energy Savings with Fluorescent and LED Dimming.). LSD 73–2015.
http://www.nema.org/Standards/SecureDocuments/NEMALSD%2073-2015%20WATERMARKED.pdf. Accessed 18
June 2018
Pattison (2017) Foreword. In: Gupta SD (ed) Light emitting diodes for agriculture: smart lighting. Springer, Singapore
Rytter M, Sørensen JC, Jørgensen BN, Körner O (2012) Advanced model-based greenhouse climate control using multi-
objective optimization. Paper presented at the IV International Symposium on Models for Plant Growth,
Environmental Control and Farm Management in Protected Cultivation- 957
Science Illustrated (2011) CEA. Science Illustrated, Feb. https://en.wikipedia.org/wiki/Controlled-environment_agriculture.
Accessed 18 June 2018
Sørensen JC, Jørgensen BN, Klein M, Demazeau Y (2011) An agent-based extensible climate control system for
sustainable greenhouse production. Paper presented at the International Conference on Principles and
Practice of Multi-Agent Systems
Sørensen JC, Kjaer KC, Ottosen CO, Jørgensen BN (2016) DynaGrow-Multi-Objective Optimization for Energy Cost-
efficient Control of Supplemental Light in Greenhouses. Paper presented at the IJCCI (ECTA)
Watson RT, Boudreau MC (2011) Energy informatics. Green ePress, Athens
Watson RT, Boudreau MC, Chen AJ (2010) Information systems and environmentally sustainable development: energy
informatics and new directions for the IS community. MIS Q 34(1):23–38
Wollsen MG, Jørgensen BN (2015) Improved local weather forecasts using artificial neural networks. Paper presented at
the Distributed Computing and Artificial Intelligence, 12th International Conference
Watson et al. Energy Informatics (2018) 1:1 Page 14 of 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com