Controlling Greenhouse Light To A Consistent Daily Integral - 2000 Albright PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

se 1672 ms 7/9/01 7:46 AM Page 421

CONTROLLING GREENHOUSE LIGHT TO


A CONSISTENT DAILY INTEGRAL
L. D. Albright, A.-J. Both, A. J. Chiu

ABSTRACT. Lettuce growth data are presented that show the importance of the daily light integral for predictable
vegetative growth. Dry mass accumulation is shown to be proportional to the light integral, and a consistent daily light
integral is proposed to be central to consistent production. Supplemental lighting control rules are defined and described
and a computer implementation is used in conjunction with ten years of hourly weather data to test (by simulation)
adequacy of the rules to control supplemental lights and movable shades in greenhouses to achieve a consistent daily
integral of Photosynthetically Active Radiation (PAR), mol-m–2-day–1, on days of either insufficient or excess solar
irradiation, which are most days. The rules require neither historical data bases of weather characteristics nor daily
weather forecasts. Control decisions are suggested to be made hourly, based on the current day’s accumulating solar PAR
integral inside the greenhouse. The model is sensitive to time-of-day electricity rates, changing seasons, weather,
greenhouse and component characteristics, and greenhouse location (latitude and longitude). The rules contain
parameters with values suggested for northeastern United States solar conditions but which may be adjusted for local
solar climates that are significantly different.
Keywords. Greenhouse control, Greenhouse lighting, Greenhouse climate control, PAR, PPFD, Light control, Plant
lighting.

A
primary reason to grow crops in a greenhouse is
to achieve better control over the environments
affecting plant growth and development.
Temperature and Photosynthetically Active
Radiation (PAR, radiation within the 400 to 700 nm wave
band) are the two most important parameters of the plant
environment in this regard. Greenhouse air temperature is
usually controlled within a narrow range with the help of
heating and cooling (ventilation or evaporative cooling)
systems. It is not unusual for a control system to maintain
temperature inside a greenhouse to within a few degrees of
the set point for many hours of a year. On the other hand,
light is rarely controlled to within a narrow range. The
daily light integral, typically, is never controlled.
Figures 1 and 2 show, respectively, daily average
temperatures and daily outdoor PAR integrals for Ithaca,
New York, for 1983-1996. Variations of average daily
Figure 1–Average daily outdoor air temperature (°C) for Ithaca, N.Y.,
temperatures are clearly smaller than variations of daily 1983-1996.
light integrals. For any day of the year, there is at least an
order of magnitude difference between the largest and
smallest PAR integrals. Without control of daily light experienced by plants grown in greenhouses. Such large
integrals, these large variations will, after transmission variations in magnitude are seldom tolerated for
losses due to glazing and structural elements, also be greenhouse temperature control. It is a thesis of this work
that accurate light control, in combination with careful
control of other environment parameters, is essential for
Article has been reviewed and approved for publication by the consistent crop vegetative growth and timing in
Structures & Environment Division of ASAE. greenhouses.
The authors are Louis D. Albright, ASAE Fellow Engineer,
Professor, Arend-Jan Both, ASAE Member Engineer, Assistant
A basic assumption of the work reported here is that a
Professor, Bioresource Engineering Dept., Rutgers University, New plant lighting system should be designed to provide its
Brunswick, N.J. (former Research Associate) Department of Agricultural maximum benefit. Supplemental lighting systems are
and Biological Engineering, Cornell University, Ithaca, N.Y.; and Alice J. expensive to install and operate. However, once a
Chiu, Field Engineer, Schlumberger, Houston Offshore, Webster, Texas. commitment is made to supplemental lighting, the system
Corresponding author: Louis D. Albright, Cornell University,
Department of Agricultural and Biological Engineering, 204 Riley-Robb should be designed and operated for highest benefit.
Hall, Ithaca, NY 14853-5701, phone: 607.255.2483, fax: 607.255.4080, e- Lighting intensity should not be underdesigned and
mail: <lda1@cornell.edu>. operating hours should not be fewer than optimum.

Transactions of the ASAE


VOL. 43(2): 421-431 © 2000 American Society of Agricultural Engineers 0001-2351 / 00 / 4302-421 421
se 1672 ms 7/9/01 7:46 AM Page 422

lighting and movable shades in commercial greenhouses


can be controlled to achieve a user-specified, daily target of
integrated PPFD in a consistent and cost-effective way. The
supplemental lighting algorithm should use off-peak
electricity (where available) as much as possible and
control should not result in frequent changes between on
and off states of the lights or shades during any single day.
Neither should supplemental lighting be required to
compensate for excessive use of shade, nor the reverse.

CONSISTENT LIGHT FOR PLANT


PRODUCTION
Butterhead leaf lettuce (Lactuca sativa L., cv. Ostinata)
was grown in a hydroponic system over a range of constant
Figure 2–Average daily outdoor solar PPFD integral (mol-m–2) for daily light integrals (Both, 1995; Both et al., 1998). A total
Ithaca, N.Y., 1983-1996. of 35 separate lighting treatments were investigated over an
eight-month period (October 1992-May 1993). A linear
However, control should ensure off-peak electricity hours relation was found between shoot dry mass at final harvest
are used to the maximum extent. (35 days after seeding) and total accumulated light between
Limited work is reported in the greenhouse literature seeding and final harvest (fig. 3):
pertaining to controlling light to a target integral. Artificial
lighting for day length control (Bakker et al., 1995) and DMf = –1.86 + 0.0131 PPFDaccumulated, R 2 = 0.87 (1)
time clock or photocell control (Aldrich and Bartok, 1994)
are suggested. Aldrich and Bartok suggest supplemental
lighting be activated when sunlight falls to twice the light Equation 1 is based on temperature control to 24°C during
intensity provided by the lights. Even when supplemental the day and 19°C during the night (conditions which were
lighting is used, it is often installed at minimum capacity found to lead to best growth for this lettuce cultivar for
(relative to typical solar irradiation). This is one reason, ambient carbon dioxide concentration).
perhaps, why control to a daily integral has not been a Shoot dry mass development observed during each
subject of serious previous study. When supplemental treatment was fitted with a general, second-order
lighting capacity is small, the lights can contribute only exponential polynomial of the form (fig. 4):
modestly to daily integrals.
Dynamic optimization of supplemental lighting has
DM = exp a + bT + c T2 (2)
been a subject of interest (e.g., Heuvelink and Challa,
1989; Ishii et al., 1993). Dynamic optimization combines
crop modeling with greenhouse dynamics and energy Analysis of the coefficients a, b, and c for all 35 treatments
considerations to determine optimum and time-variant revealed coefficients b and c could be treated as constants
intensities of greenhouse lighting. Such an objective is very but coefficient a was a linear function of the daily light
different from the one sought here, which is to achieve integral.
consistent daily target integrals of PPFD so the resulting
plant growth meets market demand where crop timing,
quality and day-to-day consistency are essential
(e.g., contracted greenhouse lettuce production for a food
service operation). Furthermore, even when dynamic
optimization leads to a projected PPFD integral optimized
for a day, a means to control to that integral is still
required.
A more sophisticated dynamic optimization approach to
supplemental lighting is suggested by Carrier et al. (1994).
Their work is based on an expert system approach and uses
heuristic rules as well as a plant growth model to activate
supplemental lights for tomatoes when economically
beneficial to do so, but not to exceed a specified daily light
integral (the authors used approximately 14 mol-m–2,
based on results from Bruggink, 1987). The rules consider
factors such as time of day, plant photoperiod
requirements, and a daily light forecast. Obtaining an
accurate daily light forecast is problematic at best.
The objective of the work reported here was to develop
an algorithm, or set of rules that could be implemented on a Figure 3–Relationship between total accumulated PPFD, mol-m–2,
and final shoot dry mass, g (35 days after seeding) for the butterhead
greenhouse control computer, by which supplemental lettuce cultivar Ostinata.

422 TRANSACTIONS OF THE ASAE


se 1672 ms 7/9/01 7:46 AM Page 423

A second goal was to avoid including adaptive control,


or program “learning”. Such alternative control strategies
could be equally valid but, for this study, were assumed
applicable only to a secondary level of control—to refine
values of parameters used within the primary algorithm.
A starting assumption in the algorithm is as follows. The
function of lighting control is to turn lights off, not on.
Control is active when lamps are off. The control algorithm
assumes (each hour) that lights are on and searches for
reasons to switch them off. This approach proved to be
more successful than the opposite, searching for reasons to
switch lights on, but is not suggested as essential.
A second fundamental assumption in developing the
algorithm was to idealize the form of daily solar PPFD.
The daily PPFD history was assumed to follow a sine curve
in a perfect situation (see fig. 5), starting at sunrise and
ending at sunset (ignoring contributions of twilight to
photosynthesis). Atmospheric transmittance varies as a
function of solar altitude and distorts the sine curve
representation but, for simplicity, constant transmittance is
Figure 4–Selected growth curves based on equations 2 and 3 and six
different daily integrated PPFD levels, mol-m–2, for the butterhead
assumed. Later discussion will return to this assumption. In
lettuce cultivar Ostinata. this idealized situation, instantaneous PPFD can be
calculated from:

a = –8.596 + 0.0734(PPFDintegral), R2 = 0.83 Π t – SR


PPFD t = PPFD max sin (4)
b = 0.4822, standard deviation = 0.0861 SS – SR

c = –0.006225, standard deviation = 0.0016 (3) The idealized daily PPFD integral to time t is obtained by
integrating equation 4:
This direct relationship between dry mass accumulation
and daily light integral motivated development of light SS – SR
control in such a way that plants receive a constant daily PPFD integral = 0.0036 PPFD max
light integral, independent of daily solar irradiance. In Π
general, this approach requires (movable) shading in t – SR
summer and supplemental lighting in winter. In addition, × 1 – cos Π (5)
supplemental PAR from high pressure sodium (HPS) light SS – SR
sources was assumed to have approximately the same
effect on lettuce growth and development as PAR from where the factor of 0.0036 converts µmol-s–1 to mol-h–1.
sunlight (and later confirmed by observation). Equations 4 and 5 are graphed in figure 5. If light intensity
is modulated to reach the desired target precisely by sunset,
PPFDintegral = PPFDtarget for t = SS. Equation 5 can be
ALGORITHM DEVELOPMENT inverted to solve for PPFDmax as a function of PPFDintegral
SUPPLEMENTAL LIGHTING
A first goal in developing a light control algorithm was
to achieve suitable control considering neither historical
weather data nor daily weather forecasts. In a practical
sense, historical weather data are not easily obtained
outside the research community, i.e., by greenhouse
installation contractors or growers. Furthermore, averaged
historical conditions may poorly represent any given day of
weather. For example, in Ithaca, N.Y., the average daily
solar PPFD integral (outdoors) during July is
approximately 42 mol-m–2. Yet, there are occasional July
days with barely 4 mol-m–2 and others near 60 (see fig. 2).
Requiring a weather forecast for the coming day is
problematic for, although on-line weather forecasts may be
imported to control programs, access is not yet universal
and adds a level of complexity that is unnecessary if other
adequate, but simpler, control strategies can be identified.
Furthermore, it is not clear that weather forecasting is
sufficiently advanced that an accurate prediction of the
Figure 5–Idealized trace and integral of daily (normalized) solar
coming day’s light integral will be possible soon. PPFD.

VOL. 43(2): 421-431 423


se 1672 ms 7/9/01 7:46 AM Page 424

and, thereby, PPFDtarget. Equation 5, rewritten to express b. Calculate that day’s sunrise and sunset hours,
the idealized daily PPFD integral from sunrise to time t in based on the latitude and longitude of the
terms of the daily target, becomes: greenhouse and the day of the year.
A third preliminary calculation is made for each
PPFD integral = hour of the weather data set:
c. The total (potential) PPFD that could be
t – SR accumulated using only supplemental lighting if
0.5 PPFD target 1 – cos Π (6) lamps were to be on starting at the beginning of
SS – SR the next hour and remain on until the following
sunrise hour or end of the off-peak period (and
Equation 6 provides a means to establish a moving possibly de-activated for a dark period for
benchmark of accumulating PPFD during a day, sensitive photoperiod purposes), whichever comes first.
to the sunrise and sunset hours and the daily PPFD target 1. If time clock control is included and the current hour
integral. is during the period when lamps should be off,
Note that, implicit in the development of what follows, control is activated and the lamps are turned off.
each “day” for control purposes is assumed to begin at Time clock control was inactive for everything
sunrise and end at the following sunrise. Sunrise and sunset reported herein. Lettuce was the crop of interest and
times are calculated for each day from solar geometry lettuce needs no dark period to thrive.
equations and rounded down to the nearest hour for sunrise Justification: There may be certain hours of the day,
and up for sunset. This leads to occasional days having possibly changing monthly, when lamps must not
fewer or more than 24 h for PPFD accumulation purposes operate—perhaps to avoid excessive electricity
but was found not to create problems. Control decisions are demand or time-of-use charges, or for photoperiod
based on solar time (which, of course, must be converted to considerations.
local clock time for control actions). 2. Do not permit lights to operate early in the morning
The algorithm for lighting and shade control was during the brighter months of the year. Anticipate
developed and tested using an extension of program eventual adequate insolation until sufficient data has
FANDUTY, described by Albright (1994). The program been accumulated to show a high probability the day
acted as a greenhouse simulator in which the algorithm and will be dark. In the program, keyed to conditions of
associated lighting parameters could be embedded. the northeastern U.S., the specific tests were:
Weather data typically includes insolation for hourly time a. For months of greatest solar irradiation, keep
steps, which the program converted to PPFD units based on lamps off between sunrise and H1 hours after
data provided by Ting and Giacomelli (1987). sunrise. However, if the daily accumulated PPFD
The structure to search for reasons to turn off lights was is not equal to at least one-quarter of the daily
the IF/ELSE; IF/ELSE; IF... construct rather than a series target by solar noon, permit lights to remain on
of IF statements. Each rule was structured to switch off regardless of the value of H1.
lights when the corresponding Boolean test was true. The b. For late summer (when days are still sunny, but
sequence of rules follows below. The daily PPFD solar intensity has lessened), keep lamps off
accumulation was updated (a running sum) hourly. between sunrise and H2 hours after sunrise.
Admittedly, some of the tests listed below appear However, if the daily accumulated PPFD is not
occasionally to overlap, but repeated variations of the tests equal to at least one-quarter of the daily target by
suggested such overlap is useful to address the wide variety solar noon, permit lights to remain on regardless
of daily insolation histories that can occur during a year. of the value of H2.
During development of the algorithm, many rules were c. For spring and autumn months, keep lamps off
tested and discarded. The ones below are what remained between sunrise and H3 hours after sunrise.
after detailed examinations of how the rules responded to d. For the rest of the months of the year, keep
one year of available weather data, 1988 data for Ithaca, lamps off between sunrise and H4 hours after
N.Y. (Albright, 1998). sunrise.
The rules were applied hourly although a time step of Justification: During months with greater solar
this length is not required for a control program. On the irradiation, avoid adding supplemental light during
other hand, very short time steps are not likely to be needed cloudy morning hours; anticipate more insolation
or, perhaps, even desirable. Eight rules follow, preceded by will become available later in the day. This
preliminary calculations. expectation is viable for fewer hours during late
0. A preliminary calculation is made only at the first summer and early autumn, and even fewer during
hour of the weather data set: winter months (H1 > H2 > H3 > H4 and suitable
a. The integrated supplemental PPFD achievable by values are discussed below).
operating lights during the entire off-peak period 3. If solar PPFD accumulated to this hour meets or
(less a possible dark period for photoperiod exceeds the accumulation target (eq. 6) for the hour,
control). This assumes time-of-day electricity turn the lights off.
rates do not change during the year. Justification: To this hour, there is no PPFD integral
A second preliminary calculation is made at deficit.
1:00 A.M. each day: 4. If: (a) the hour is during the time of year with more
sunlight and between sunrise and sunset, (b) the
PPFD left to be accumulated can be achieved by

424 TRANSACTIONS OF THE ASAE


se 1672 ms 7/9/01 7:46 AM Page 425

delaying supplemental lighting until the next hour conjunction with the supplemental light control algorithm.
even if solar PPFD drops suddenly to insignificance, The following two rules were then imposed, hourly.
and the PPFD deficit to this point could be made up 1. Starting two hours after sunrise and continuing until
by a scaled portion of the off-peak PPFD potential, one hour before sunset, use the measured PPFD
turn off the lights. The off-peak PPFD potential is averaged over only the previous hour to calculate
scaled so decisions during the early hours of the the corresponding value of PPFDmax, based on
solar day are not based on an expectation of using equation 4. (Measure the hourly PPFD within the
all the off-peak PPFD potential to make up for the greenhouse, but not under the shade, if the shade is
current deficit, saving none to make up for deficits already deployed.) Use the calculated value of
during later hours. The scaling function used in the PPFDmax, and equation 5, to project the idealized
program was a multiplying sine function that rose integrated PPFD for the remainder of the day. If
from a zero value at sunrise to unity at sunset. A integrated PPFD (from all sources) since sunrise,
linear rise (ramp) may have done as well. plus the idealized integrated solar PPFD projected
Justification: This pushes as much supplemental for the remainder of the day, exceeds the PPFD daily
lighting as possible into the off-peak electric rate target, deploy the shade immediately. Otherwise,
hours and is particularly sensitive to this need during leave it retracted (or retract it if already deployed).
the months of the year having higher solar PPFD 2. Starting two hours after sunrise and continuing until
expectations. one hour before sunset, use the integral of
5. Turn off the lights if the hour is between sunrise and (unshaded) solar PPFD since sunrise. Using the
sunset and the PPFD left to be accumulated could be integral and equation 5, calculate the value of
accumulated by turning on the lights at the next hour PPFDmax that would have produced the same
even if the solar PPFD drops immediately to integral to the current hour. Then, using equation 5
insignificance and remains there for the rest of the (for t = SS), project the idealized integrated PPFD
day. expected for the remainder of the day. If integrated
Justification: Somewhat an echo of test 4 above, but PPFD (from all sources) since sunrise, plus the
applies to all months. idealized integrated solar PPFD projected for the
6. If: (a) the hour is at sunset or between sunset and an remainder of the day, exceeds the PPFD daily target,
hour before the start of off-peak electric rates and deploy the shade immediately. Otherwise, leave it
(b) the accumulated PPFD deficiency to this hour retracted (or retract it if already deployed).
could be achieved during off-peak hours alone, turn Each of the above rules is, essentially, a means to
off the lights. extrapolate the integrated PPFD received during daylight
Justification: Use off-peak electric rates as much as hours to an eventual value at sunset. Application of each
possible. rule is delayed until two hours after sunrise to ensure at
7. If the hour is before off-peak electric rates start, but least some solar PPFD has been accumulated so
any remaining PPFD to be added by supplemental extrapolation can be made with some degree of confidence.
lighting will be achieved before the off-peak period The two hour delay was determined to work adequately
ends, turn off the lights. after examining the effects of delays of various lengths,
Justification: Push as much of the supplemental from one to four hours. Obviously, if the morning hours are
lighting as possible into off-peak hours. dark, but the day becomes brighter, the second rule will
As a final test, after the above conditions are considered, respond slowly while the first will respond relatively
a safety check is imposed based on a separate IF statement quickly.
not part of the IF/ELSE statement described above.
8. Be sure lights are not turned off if the hour is during
the dark part of the year and there remains more TEST CONDITIONS FOR THE CONTROL
integrated PPFD to be added than can be met by the
lamps, alone, operating from the next hour until the ALGORITHM
following sunrise. GENERAL
Justification: Avoid a combination of conditions The proposed algorithm was tested in detail by
whereby the daily PPFD integral is not met, there is determining its response to one year of hourly weather data
little expectation of substantial solar PPFD (Ithaca, N.Y., for 1988) in the greenhouse climate
accumulation for the rest of the day, and the lights simulation program described above. The focus of this
remain off during the next hour. project was on climates similar to those in the northeastern
U.S., which are characterized by relatively cloudy winters
MOVABLE SHADE but relatively sunny summers. Ithaca, N.Y., is located in
Many of the same concepts embedded in the central New York State, leeward (winter prevailing winds)
supplemental lighting control rules were used to develop an of two of the Great Lakes, and experiences some of the
algorithm for movable shade control. However, it was cloudiest winter weather in the United States. After testing
found preferable to assume that, each hour, the movable the algorithm in detail against a single year of weather data
shade is retracted. The algorithm searches for reasons to (and selecting parameters suitable for the one year) it was
deploy the shade. tested for generality against seven additional years of
The calculation of sunrise and sunset hours for each day hourly weather data for Ithaca, and one year of
preceded application of shade control rules, done in representative weather data from southern New Jersey.

VOL. 43(2): 421-431 425


se 1672 ms 7/9/01 7:46 AM Page 426

Table 1. Greenhouse dimensions and other data, Table 2. Luminaire layouts to achieve a range of
base case for comparison and discussion (averaged) supplemental PPFD values
Factor Value 24-h
Ave. No. Installed PPFD
Latitude (Ithaca, N.Y.) 42.2 N Design PPFD Luminaires kW Integral
Longitude (Ithaca, N.Y.) 76.5 W
Floor area (m2) 1713 1 200 mol-m–2-s–1 1177 559 17.3 mol-m–2
Maximum solar transmittance of glazing 0.7 2 175 1030 489 15.1
3 150 883 419 13.0
4 125 736 350 10.8
Southern New Jersey is sunnier during winter and subject 5 100 589 280 8.64
to less extreme swings of insolation during summer.
A representative greenhouse was assumed for the
program. The greenhouse was based on the Work Sheet avoid creating “hot” spots of light but is one of the more
No. 4 example presented by Aldrich and Bartok (1994), (energy) efficient units. Luminaire layout designs were
which is a four-section, gutter-connected glass structure completed using the lighting design program LumenMicro,
29.3 × 58.5 m (96 × 192 ft), with 2.4-m (8-ft) high side version 6 (Lighting Technologies, Inc., Boulder, CO).
walls. Pertinent greenhouse data are listed in table 1. Designs were completed to provide supplemental PPFD
A maximum transmittance (solar noon at the summer values of 100, 125, 150, 175 and 200 µmol-m–2s–1, a range
solstice) of 0.7 was chosen to represent a typical that includes levels appropriate for greenhouse lighting to a
glasshouse. (See, for example, Pearson et al. (1995) who daily integral without being so high that plants will be
show PAR quantum transmission through various glazing excessively heated by infrared emissions from the lamps
materials approximates 90%. Additional shading by (Dietzer et al., 1994). A summary of designs is in table 2.
framing members is expected.) Solar transmittance was
assumed to vary over the day and year (lower at more ATMOSPHERIC TRANSMITTANCE FACTOR AND MOVABLE
oblique sun angles). A daily transmittance factor was SHADE RULES
assumed to vary as one-half a sine wave, rising from 0.6 at During development of the movable shade rules, it was
sunrise to 1.0 at solar noon, and returning to 0.6 at sunset. observed that the rules were often triggered somewhat too
The yearly variation was assumed to vary as a full sine late, with the result being a consistent overshoot of the
wave, varying from 0.8 on the winter solstice to 1.0 on the daily PPFD target integral during summer. This was
summer solstice. These variations were assumed for attributed to omitting consideration of atmospheric
illustrative purposes only. Daily and yearly transmittance transmittance in predicting future solar gains. The effect of
values modified the assumed maximum transmittance by solar altitude on atmospheric transmittance is well
multiplication. established and, moreover, early morning haze is common
The rules stated above are based on months and include in humid climates. In a real greenhouse, early morning dew
four time delay intervals, H1, H2, H3, and H4. The values and condensation on the greenhouse glazing may also
and associated months were H1 = 12 h (May, June, and contribute to this effect. To accommodate the effects, an
July), H2 = 9 h (August), H3 = 7 h (March, April, and atmospheric “haze factor” was hypothesized and used to
September), and H4 = 2 h (all other months). Additionally, divide the PPFD integral extrapolated from the current
lighting rule 4 was applied for March through September, hour to sunset, as follows:
months with days longer than nights. These selections of
months, and delay times, are not suggested as universal and 1
their importance is discussed later. PPFD rest of day = PPFD SR–SS – PPFD SR–t (7)
fhaze
Lettuce is known to suffer from the physiological
disorder, tip burn, when growth is too rapid. Both (1995)
has shown it is possible to grow the lettuce cultivar where equation 5 is used to calculate PPFDSR-SS and
Ostinata without tip burn at ambient carbon dioxide PPFDSR–t, based on an estimate of PPFDmax which was
concentrations, 24/19°C day and night air temperatures, also determined using equation 4 or 5, depending on the
and 17 mol-m–2 of daily integrated PPFD if supplemental rule under consideration. A linear haze factor was assumed,
air mixing fans are used, directed downward onto the plant rising from 0.5 at sunrise to 1.0 at solar noon and
canopy to enhance transpiration. This daily target of remaining at 1.0 until sunset. Functions for atmospheric
17 mol-m–2 was assumed for the simulations to test the transmittance are usually expressed as exponential
light and shade control algorithms. Other daily targets functions of an atmospheric extinction coefficient and are
(not reported here) were also tested and showed results symmetric around solar noon, but this simple linear form
comparable to those for 17. was found to be adequate for the movable shade rules.

LAYOUT OF LUMINAIRES WEATHER DATA


Although this report focuses on lighting and shade Actual data was preferred to test the algorithm rather
control, luminaire layout must be considered in any than typical or averaged data in order to impose the
lighting design. Minimally, the number and type of frequent wide and rapid swings of insolation to which a
luminaires should be adequate to achieve, on average, the greenhouse can be subjected to. One year of Ithaca weather
design PPFD. For this discussion, 400W HPS luminaires as data (1988) was considered as rules for supplemental
reported by Albright and Both (1994) were assumed. The lighting control were developed. Daily PPFD integral
luminaire chosen for the sample simulations in this report values expected inside the example greenhouse for the
provided an intermediate degree of light pattern focus to weather data set and other parameters described above, but

426 TRANSACTIONS OF THE ASAE


se 1672 ms 7/9/01 7:46 AM Page 427

ACHIEVING DAILY PPFD INTEGRALS


Table 3 contains averaged monthly values of integrated
daily PPFD and the accompanying solar contribution as
predicted by the computer program operating under the
lighting control algorithm and assumed parameter values
described above. For a greenhouse transmittance as
assumed and Ithaca-like weather data, PPFD intensity
greater than 150 µmol-m–2-s–1 is required to approach the
daily target of 17 mol-m–2 closely for most months of the
year and an intensity nearer 200 is required if the target
must be obtained every day of the year. The degree of
control when both supplemental lighting (designed for
200 µmol-m–2-s–1) and movable shade control are imposed
is shown by the graph in figure 6 (1988 Ithaca data).
Several points can be deduced from figure 6. During the
dark time of the year (October through February), control
Figure 6–Daily greenhouse PPFD integrals, mol-m–2, for the Ithaca, to achieve the target PPFD integral is excellent at the
1988, data set, with control to 17 mol-m –2 imposed by the highest design PPFD. The target is achieved on even the
supplemental light and movable shade algorithms. darkest days because supplemental lighting is enabled two
hours after sunrise and no dark period is imposed for
no supplemental lighting or shade, are in figure 6. Rules photoperiod reasons. During the rest of the year, however,
and parameters developed from this one year of data were occasional days occur (even during the middle of summer)
then applied to seven additional years of Ithaca weather when the PPFD target integral is not achieved although
data and one year of data for southern New Jersey. These lights are enabled at solar noon on days when the
eight additional years of weather data provided an accumulated PPFD is less than one-quarter of the daily
important test of whether the rules were generally effective target at that time. There is insufficient capacity in any of
and not tailored to only the single year for which they were the lighting systems considered here to achieve the target
developed. PPFD integral on a very dark day if lighting can not begin
until solar noon. However, this restriction was imposed
because earlier enabling times wasted lighting on dark
RESULTS OF SIMULATIONS mornings followed by bright afternoons, but the lights had
EFFECT OF SUPPLEMENTAL LIGHTING DESIGN already been activated. The practical effect of slightly
As can be inferred from data in table 3, supplemental missing the target integral on these occasional days was
lighting that provides low PPFD intensities cannot be considered insignificant.
expected to provide consistent daily PPFD integrals as high Another view of the control is shown in the graphs in
as 17 mol-m–2, regardless of the control imposed, during figure 7. On the graph, points falling along the line of
the darkest days. The 24 h integral value determines the constant daily PPFD integral (the diagonal line of data
limits. points ending at 17 mol-m–2 on the x-axis) represent days
of successful control; the integral of solar and supplemental
PPFD equals the target. Points to the left of the constant
PPFD integral line represent days during the nominally
brighter time of year that were very cloudy and control did
Table 3. Simulated daily integrated PPFD with various supplemental
not achieve the PPFD target integral. Using H1, H2, and
PPFD intensities, Ithaca weather data, maximum
greenhouse transmittance of 0.7*
Average Supplemental PPFD (mol-m–2-s–1)
Month 100 125 150 175 200 Solar
Jan 14.6 15.8 16.6 16.9 17.0 6.3
Feb 15.6 16.4 16.9 17.1 17.1 9.6
Mar 16.2 16.5 16.7 16.8 16.8 15.7
Apr 15.6 16.0 16.3 16.6 16.8 16.4
May 17.0 17.1 17.2 17.4 17.3 24.7
Jun 17.9 18.0 17.9 17.9 17.9 30.2
Jul 17.3 17.5 17.6 17.6 17.6 27.5
Aug 17.4 17.4 17.5 17.5 17.6 24.1
Sep 16.1 16.4 16.6 16.7 16.8 17.0
Oct 15.4 16.2 16.8 17.0 17.0 9.4
Nov 13.6 15.0 16.1 16.7 16.9 5.7
Dec 13.3 15.0 16.2 16.8 17.0 4.9
Average 15.8 16.4 16.9 17.1 17.1 16.0
Lights 3.87 4.53 4.90 5.11 5.16 –
* The last column contains results for operation with no supplemental Figure 7–Supplemental and solar contributions to the total PPFD
lights or summer shading. The final row contains the average daily integral, mol-m–2, combinations for each day of the year, inside the
mol-m–2 added by supplemental lighting. greenhouse, for Ithaca, 1988, data set, with control to 17 mol-m–2.

VOL. 43(2): 421-431 427


se 1672 ms 7/9/01 7:46 AM Page 428

H3 values of 12, 9 and 5 h prevented the lighting system calculated but, because hours were of integer data type,
from achieving its target during those days. The same days the result was truncated. This was done rather than
are shown on figure 6 as those falling below the PPFD rounding, or truncating and adding one hour to the result
integral target from March through September. Days where to overestimate the amount of lighting remaining. Failures
movable shade control was inadequate cluster in the region to achieve the full target occurred during the time of year
where the x-axis intersects the line of constant PPFD when movable shade control was necessary. Shade control
integral. Finally, the few points to the right of the line of failures resulted in slight errors above the target integral.
constant PPFD integral represent days where lighting Thus, slight lighting failures below the target
control was not completely successful and a small amount compensated, in part, for slight failures of shade control
of supplemental light was wasted. above the target. Modest electricity and money savings
Several observations arise from the simulation data that were a second benefit.
may not be obvious from the graphs in figures 6 and 7. The algorithm was tested by writing an output file to
First, designs with higher PPFD capacity provided not only show the lighting system state for every hour of the test
better control but also wasted fewer lighting hours (and, year. Fortunately, the file showed no instances where
thereby, less electricity). Higher PPFD designs permit two supplemental lighting cycled repeatedly during a day. If
desirable actions: (1) more lighting PPFD can be delayed lights remained on for part of the morning and then were
to the off-peak period; and (2) decisions can be delayed turned off, they remained off until late afternoon or
until later in the day when more data will have been evening and then were turned on until the PPFD integral
accumulated and the complete solar day can be more was obtained.
accurately predicted. The second action permits more The several parameters of the control algorithm
informed decisions regarding whether to use the lights. (e.g., H1, H2, H3, and H4) are fixed in the current
Because values of parameters and time delays were algorithm. Adding adaptive control to the algorithm might
chosen to provide best control in response to the 1988 be useful. Additional simulations showed an accurate
Ithaca weather data set and assumed greenhouse measure of greenhouse transmittance at solar noon on the
conditions, the next step was to apply the same rules and summer solstice (assumed to be the maximum potential for
parameter values to other sets of weather data and explore light transmission) is important for achieving good
their wider applicability. supplemental lighting control. For example, when the
maximum transmittance was assumed to be 0.6 instead of
EXTENSION OF SUPPLEMENTAL LIGHTING RULES TO 0.7, and all other factors remained constant, best control
ADDITIONAL YEARS OF WEATHER DATA was achieved when H3 applied only to September but not
In figures 8 and 9 are graphs showing the degree of to March or April. The magnitude of improvement was
control achieved for another seven years of Ithaca data only 2 to 3%, but the energy savings were estimated to be
(design instantaneous PPFD = 200 µmol-m–2-s–1) and one more than $12,500/ha-year of lighting to the daily target of
year of New Jersey weather data (design instantaneous 17 mol-m–2 in Ithaca. This suggests careful and continual
PPFD = 175 µmol-m–2-s–1). The difference in PPFD was monitoring of the control performance would be useful and
chosen due to the brighter winter solar climate of southern a careful assessment of maximum greenhouse
New Jersey. Other parameters are unchanged. transmittance should be accomplished prior to
implementing supplemental lighting control.

DISCUSSION MOVABLE SHADE CONTROL


SUPPLEMENTAL LIGHTING CONTROL Considering the simplicity of the proposed algorithm,
As a general comment, it was surprising how relatively control effectiveness was good. Simplicity was a goal of
well the candidate algorithm controlled supplemental this work and movable shade control based on only the
lighting and movable shades to reach a daily target, first of the two simple rules was found to be comparable to
considering that the algorithm drew from no historical control based on both rules. Combined with supplemental
weather data base and received no advance notice of the light control, the averaged yearly deviation from the PPFD
weather expected for the day. Additionally, although target is estimated to be 0.1 mol-m–2-day–1, an error of
differences were found between responses to the weather less than 1%. Most of this deviation occurred during
data of Ithaca and southern New Jersey, the differences summer, suggesting slight refinements of the algorithm
were not great. Because these two weather data sets could perhaps reduce the errors further. Whether this is
represent somewhat different weather types (solar data), important will depend on the penalty imposed by the
one can hope that a simple algorithm, such as presented summer PPFD excess, which was estimated to be not more
here, may apply to a fairly wide geographic area. The than 25% above the target for the least controlled single
ability to control the daily PPFD integral during the darkest day of the summer and, on average, not more than 5%
months of the year suggests the algorithm is adequate for above the target for the sunniest months (based on the
regions of the world with winters even darker than Ithaca Ithaca, 1988, weather data). One option to achieve a more
(northern Europe, for example). precise running average of the daily integral would be to
The graphs of figure 8 show some failures to achieve adjust the following day’s target based on the excess or
the full PPFD target (indicated by data points falling deficit of the current day. Simulations using this approach
slightly below the constant PPFD integral target diagonal were successful in achieving a better three day (for
line. This arose from a calculation in the simulation example) running average, although day-to-day control
program related to rule number 7. As part of the rule, the was not improved. Unpublished work at Cornell has
number of hours required to reach the target were shown lettuce is able to integrate light over three days

428 TRANSACTIONS OF THE ASAE


se 1672 ms 7/9/01 7:46 AM Page 429

Figure 8–Supplemental lighting and movable shade control adequacy (for control to mol-m–2) for seven years of insolation data from Ithaca
(1983-1987, 1989-1990) and one year of insolation data from southern New Jersey. Design supplemental PPFD values are 200 and 175 µmol-
m–2-s–1 for Ithaca and New Jersey data, respectively, with control to 17 mol-m–2.

without impairing growth as long as individual daily During the test year, 17 days were less than 1 mol-m–2
PPFD integrals do not differ more than 2 or 3 mol-m–2 removed from the target PPFD integral; 60 days were
from the desired long-term average. between 1 and 2 mol-m–2 removed from the target; 7 days
were more than 2 mol-m–2 in error; and 281 days received

VOL. 43(2): 421-431 429


se 1672 ms 7/9/01 7:46 AM Page 430

Figure 9–Supplemental and solar PPFD integrals, mol-m–2, for seven years of insolation data from Ithaca (1983-1987, 1989-1990) and one year
of insolation data from southern New Jersey. Design supplemental PPFD values are 200 and 175 µmol-m–2-s–1 for Ithaca and New Jersey data,
respectively, with control to 17 mol-m–2.

17.00 mol-m–2 of integrated PPFD, achieving the target SUMMARY AND CONCLUSIONS
exactly. For this test year, 316 (7 out of 8) days were A set of eight rules is proposed for greenhouse lighting
controlled to within 10% of the daily PPFD target integral. control and two for movable shade control. The rules are
shown to provide a consistent daily PPFD integral for a

430 TRANSACTIONS OF THE ASAE


se 1672 ms 7/9/01 7:46 AM Page 431

climate having a highly variable day-to-day solar climate. Both, A. J., L. D. Albright, R. W. Langhans, R. A. Reiser, and B.
Testing by computer simulation showed the daily PPFD G. Vinzant. 1998. Hydroponic lettuce production influenced by
integral was within 10% of the target for seven of every integrated supplemental light levels in a controlled
eight days over a year. The success of the proposed environment agriculture facility: Experimental results. Acta
algorithm shows daily PPFD integral control is achievable Hort. 418: 45-52.
Bruggink, G. T. 1987. Influence of light on the growth of young
without needing historical weather data or a prediction of
tomato, cucumber and sweet pepper plants in the greenhouse:
each day’s solar conditions at the beginning of the day. Calculating the effect of difference in light integral. Scientia
However, parameters used within several of the rules Hort. 31: 175-183.
should be tailored to specific conditions of greenhouse Carrier, M., A. Gosselin, and L. Gauthier. 1994. Description of a
transmittance and, perhaps, other factors. The results also crop growth model for the management of supplemental
show the importance of appropriate lighting system design lighting in greenhouses. HortTechnol. 4(4): 383-389.
to achieve a PPFD that will permit the daily PPFD integral Dietzer, G., R. W. Langhans, A. Spomer, and T. Tibbits. 1994.
to be achieved on the darkest days. Furthermore, Guidelines for lighting of plants in controlled environments. In
supplemental lighting designs providing greater PPFD will Proc., Int. Lighting in Controlled Environments Workshop,
reduce operating cost by concentrating more of the lighting 391-393, Madison, Wisc. NASA Conference Publication CP-
3309. NASA Kennedy Space Center, Fla.: NASA.
to be done during off-peak electricity rate hours.
Heuvelink, E., and H. Challa. 1989. Dynamic optimization of
artificial lighting in greenhouses. Acta Hort. 260: 401-412.
ACKNOWLEDGMENTS. The authors would like to thank all Ishii, Y. M. Yokoyama, T. Murata, and I. Tanaka. 1993. Studies on
past and current members of the Cornell University the development of a supplemental lighting control system for
Controlled Environment Agriculture Program involved in sunlight in greenhouses. Plant Farm Soc. J. 5(1): 1-10.
this research effort and are indebted to Mr. A. R. Leed and Kozai, T., J. Goudriaan, and M. Kimura. 1978. Light Transmission
Mr. B. G. Vinzant for frequent discussions concerning and Photosynthesis in Greenhouses. Wageningen, The
strategies of greenhouse lighting control, and to Dr. D. R. Netherlands: PUDOC, the Centre for Agricultural Publishing
Mears of the Department of Bioresource Engineering of and Documentation.
Rutgers University for sharing hourly weather data Pearson, S., A. E. Wheldon, and P. Hadley. 1995. Radiation
transmission and fluorescence of nine greenhouse cladding
representative of southern New Jersey. Funding for the
materials. J. Agric. Eng. Res. 62(1): 61-70.
research described in this report was provided by the New Ting, K. C., and G. A. Giacomelli. 1987. Availability of solar
York State Energy Research and Development Authority photosynthetically active radiation. Transactions of the ASAE
(NYSERDA), the Empire State Electric Energy Research 30(5): 1453-1457.
Corporation (ESEERCO), and the New York State Electric
and Gas Corporation (NYSEG). Lettuce seed was
generously donated by Rijk Zwaan Export B.V., De Lier, SYMBOLS AND ACRONYMS
the Netherlands. a, b, c coefficients
DM shoot dry mass (g)
DMf shoot dry mass at final harvest (35 days
REFERENCES after seeding) (g)
Albright, L. D. 1994. Predicting greenhouse ventilating fan duty H1. . H4 lighting delay times after sunrise, keyed to
factors and operating costs. ASAE Paper No. 94-4576. St. month (h)
Joseph, Mich.: ASAE. HPS high pressure sodium
_____. 1998. Method for controlling greenhouse light. U.S. Patent PAR photosynthetically active radiation (400 to
No. 5,818,734. 6 October 1998.
700 nm)
Albright, L. D., and A. J. Both. 1994. Comparisons of luminaires:
Efficacies and system design. In Proc., Int. Lighting in PPFD photosynthetic photon flux density (400 to
Controlled Environments Workshop, 281-297, Madison, Wisc. 700 nm) (µmol-m–2-s–1)
NASA Conference Publication CP-3309. NASA Kennedy PPFDaccumulated accumulated PPFD integral between
Space Center, Fla.: NASA. seeding and final harvest (mol-m–2)
Aldrich, R. A., and J. W. Bartok. 1994. Greenhouse Engineering. PPFDintegral daily PPFD integral up to hour t (mol-m–2)
Ithaca, N.Y.: Northeast Regional Agricultural Engineering PPFDtarget daily PPFD integral target (mol-m–2)
Service, Riley-Robb Hall, Cornell University. PPFDmax maximum daily value of PPFD (µmol-m–2-
ASAE Standards, 41st Ed. 1994. EP411.2. Guidelines for measuring s–1)
and reporting environmental parameters for plant experiments in SR hour of sunrise
growth chambers, 569-572. St. Joseph, Mich.: ASAE.
SS hour of sunset
Bakker, J. C., G. P. A. Bot, H. Challa, and N. J. Van de Braak.
1995. Greenhouse Climate Control, An Integrated Approach. t time of day
Wageningen, The Netherlands: Wageningen Pers. T days after seeding (d)
Both, A. J. 1995. Dynamic simulation of supplemental lighting for
greenhouse hydroponic lettuce production. Ph.D. thesis. Ithaca,
N.Y.: Cornell University Libraries.

VOL. 43(2): 421-431 431


se 1672 ms 7/9/01 7:46 AM Page 432

You might also like