You are on page 1of 11

A COMPUTATIONAL FLUID DYNAMICS MODEL OF

ALGAL GROWTH: DEVELOPMENT AND VALIDATION


J. L. Drewry, C. Y. Choi, L. An, P. E. Gharagozloo

ABSTRACT. Biofuels derived from algae are becoming an increasingly viable alternative to petroleum-based fuels; how-
ever, research and development in the field must continue to advance the technology before biofuels can be produced in
an economical and environmentally friendly manner. Unlike with photobioreactors, there is no generally accepted model
for evaluating the growth of algae in open raceways because algal growth involves a large number of variables. For this
reason, computational fluid dynamics (CFD) could prove to be a valuable and effective tool for the design, optimization,
and operation of large-scale raceway ponds under local environmental conditions. CFD can elucidate and quantify the
complex sets of variables that govern heat, mass, and flow patterns within the pond and provide spatiotemporal data con-
cerning algal concentration and other water quality variables, such as light and temperature. The corresponding out-
comes will enable designers to create more efficient ponds and more accurately predict growth under a variety of scenar-
ios, as well as optimize the ponds operation in order to produce the maximum amount of biomass possible within a given
locality. The present study focuses on developing user-defined functions capable of capturing key parameters, verifying
the CFD outcomes against existing experimental data, providing computational solutions, and assessing the sensitivity of
the model.
Keywords. Algae, Biofuels, Computational fluid dynamics (CFD).

I
ncreasing attention is being focused on alternative and rate of algal biomass production in the most environmental-
renewable fuels that can reduce our nations depend- ly and economically friendly manner.
ence on fossil fuels. Biofuels derived from algae offer Optimal raceway design depends on several important
one alternative since they can be grown using non- factors that affect algal growth, including solar intensity
potable water and on non-arable land, and unlike corn, for and duration, nutrient concentrations, dissolved gas concen-
example, which is currently used to make ethanol but is trations, temperature, hydrodynamics, and algal species, as
also a major food crop, algae are not considered a major well as many others. Overall growth rate is further compli-
food source. Algae also have a high lipid density as com- cated by the fact that not all the algae in a given environ-
pared with other terrestrial crops (Clarens et al., 2010). ment will be exposed to the same levels of light and nutri-
The most efficient way to produce algal biomass on a ents at any given time. To determine the optimal design and
scale that would be needed for transportation fuel produc- accurately predict biomass production, these factors must
tion is in open raceway-type ponds, which, compared with be integrated into a comprehensive model that can accu-
closed systems, cost relatively little to construct and oper- rately describe algal growth as well as the other aspects of
ate (Lundquist et al., 2010). However, before any reliable transport phenomena. It is also important that the interac-
estimation of their efficiency can be made, further research tions between these parameters be modeled such that the
must be conducted to determine the amount of algal bio- effects of the interactions are captured.
mass (and consequently biofuel) that can be produced by To simulate algal growth, a variety of models have been
these open ponds under a variety of conditions. Further- developed; some models fit experimental data using several
more, the design and operation of algal culture systems parameters, while other, more complex models attempt to
should be optimized in ways that will achieve the highest predict growth by modeling the specific mechanisms driv-
ing growth. A considerable number of models focus their
efforts on the effect that light intensity has on the rate of
Submitted for review in August 2013 as manuscript number ITSC photosynthesis, and these models have obtained promising
10372; approved for publication by the Information, Technology, Sensors, results. Eilers and Peeters (1988), for example, developed a
& Control Systems Community of ASABE in January 2015. model that defined a relationship between light intensity
The authors are Jessica L. Drewry, Doctoral Student, and
Christopher Y. Choi, ASABE Member, Professor, Department of and the rate of photosynthesis based on physiological
Biological Systems Engineering, University of Wisconsin, Madison, mechanisms. Wu and Merchuk (2002) and Yoshimoto et al.
Wisconsin; Lingling An, Assistant Professor, Department of Agricultural (2005) further expanded on this work by adding particle
and Biosystems Engineering, University of Arizona, Tucson, Arizona;
Patricia E. Gharagozloo, Senior Member of Technical Staff,
tracking and a more sophisticated model based on the Ru-
Thermal/Fluid Sciences and Engineering, Sandia National Laboratories, BisCO enzyme, respectively.
Livermore, California. Corresponding author: Christopher Choi, 460 While both of these models are good at assessing growth
Henry Mall, University of Wisconsin, Madison, WI 53706; phone: 608- when light is the most important parameter (as could be the
262-0607; e-mail: cchoi22@wisc.edu.

Transactions of the ASABE


Vol. 58(2): 203-213 2015 American Society of Agricultural and Biological Engineers ISSN 2151-0032 DOI 10.13031/trans.58.10372 203
case in a controlled laboratory), they are not expected to be d a
able to predict growth reliably when dealing with less con- Salgae = K ag a K ar a K am a a
dz
trolled environments. Thus, a number of researchers have (1)
a a
developed other models that have been able to capture and Z zoo
aggregate the effects of such variables as light, temperature, a a + lpomlpom + zoo zoo
and nutrient concentrations to create more comprehensive
models. Two of these, one developed by Quinn et al. where
(2011) and the other by Packer et al. (2011), are capable of Kag = algal growth rate (s-1)
validating bulk growth and lipid accumulation on a scale Kar = algal respiration rate (s-1)
well suited to industrial production. In addition, Huese- Kam = algal mortality rate (s-1)
mann et al. (2013) developed a simple model targeted at Salgae = algal flux (kg m-3 s-1)
screening algal strains to predict growth in photobioractors z = control volume height (m)
and raceway ponds with few model parameters. These bulk Z = net growth rate of zooplankton (s-1)
models have also been shown to be effective at predicting = zooplankton grazing factors
algal growth in more controlled environments (laboratories = species concentration (kg m-3)
and photobioreactors). However, they are not capable of a = algal setting rate (m s-1)
resolving the fine gradients of mass, momentum, and ener- subscript a = algae
gy transport due to the less advanced numerical models subscript zoo = zooplankton
used, any and all of which could be crucial to modeling subscript lpom = liable particulate organic matter.
algal growth in a wide variety of scenarios. The respiration and mortality rates can be combined into
Several other models have been developed to capture the one value (Kbmr), the basal metabolic rate. The algal growth
fluid dynamics of the algal culture system as well as algal rate is dependent on the multiplicative effects of tempera-
growth. Sato et al. (2010), for example, added fluid dynam- ture, light, pH, and algae and nutrient concentrations (Cole
ics to the photobioreactor model developed by Yoshimoto and Wells, 2008):
et al. (2005) and their predecessors. James and Boriah
(2010) and James et al. (2013) used existing fluid dynamics Kag = Pm temp light nutrients pH (2)
and water quality codes (EDFC and CE-QUAL) to create a
where Pm is the maximum growth rate under ideal condi-
comprehensive model targeted at simulating algal growth
and fluid flow in open raceway ponds. tions (s-1), and stands for the combined effect of the forc-
All these models are capable of simulating some of the ing functions based on actual conditions. Due to differences
detailed transport phenomena associated with a variety of in the computational software used, the term for settling of
situations, and then predict growth, but none can handle all algal particles does not have to be directly coded. In addi-
of the key variables. In contrast, a computational fluid dy- tion, losses due to predators are neglected in the current
namics (CFD) model should be able to achieve all the model. All of the parameters affecting growth can be ad-
above-stated requirements for modeling algal growth be- justed to model a specific species of algae based on exper-
cause it can effectively model all transport phenomena oc- imental data, thus making this model adaptable to any sce-
curring within the open pond. Therefore, the objectives of nario. The final governing equation is:
the study are to (1) develop and validate a model of algal C A
growth in open ponds based on species, water quality, envi- = DAB 2C A + S (3)
t
ronmental conditions, and pond design using CFD; and
(2) evaluate the sensitivity of biomass production to model Temperature Effects
parameters. A CFD model should be able to elucidate the The forcing function for temperature models the varia-
transport phenomena associated with algal growth, thereby tion in algal growth in response to non-ideal temperatures,
creating an accurate model capable of predicting algal with growth increasingly inhibited as temperatures rise
growth under a variety of conditions. above or fall below the ideal temperature range. The form
of the equation is (Cole and Wells, 2008):

METHODS
( )
2
exp K1 Topt ,1 T for T < Topt ,1
GROWTH MODEL
Our model expanded on the CE-QUAL model, which temp = 1 for Topt ,1 T Topt ,2 (4)
can capture multiple aspects of algal growth and is flexible
exp K T T
( )
2
enough to be used under a variety environmental and water 2 opt ,2 for Topt ,2 < T

quality conditions and for most algal species. The spatio-
temporal flux of algae as a function of growth, respiration, where
excretion, mortality, settling, and losses that can be at- K = experimental constant (C-2)
tributed to grazing by zooplankton is described as follows T = water temperature (C)
(Cole and Wells, 2008): Topt = optimal water temperature (C).
The constant K and Topt (optimal water temperature) are
dependent on the specific species of algae being modeled

204 TRANSACTIONS OF THE ASABE


and should be taken from experimental data. This model is as a function of available nutrients. The forcing function,
especially effective since it is capable of capturing growth based on Monod kinetics, can be described by the follow-
limitations at different rates above and below the ideal ing equation (Cole and Wells 2008):
temperature range.
p c
n
Light Effects nutrients = min , , (9)
The Beer-Lambert law can be used to integrate the effect n + K n p + K p c + K c

of light variation at the surface and the attenuation of light
based on depth and biomass concentration on algal growth, where is the concentration of a nutrient (kg m-3), and
when culture density is low (Chapra 1997): Kn,p,c is the half-saturation constant (kg m-3). This model, by
assuming that only the scarcest nutrient controls the limita-
I = (1 ) I0 exp ( z ) (5) tion of the rate of growth, allows the user to expand the
function to include any number of nutrients (such as nitro-
where gen, phosphorous, and carbon dioxide, for example).
I = shortwave solar irradiation at the control volume (W The flux of nutrients in and out of the system must also
m-2) be considered. The model assumes that nutrient flux is pro-
Io = shortwave solar irradiation at the surface of the portional to the flux of algae based on the Redfield ratio
computational cell (W m-2) (Cole and Wells, 2008):
z =depth of the computational cell (m)
= attenuation coefficient (m-1) Snutrients = CRedfield ( Salgae + S DOM ) (10)
= fraction of solar radiation absorbed at the water sur-
face. where
The attenuation factor is based on Rileys empiricism CRedfield = Redfield ratio
(Di Toro et al., 1971): Snutrients = flux of nutrients (kg m-3 s-1)
SDOM = flux of dissolved organic matter (kg m-3 s-1).
(
= kb + 0.0088 a + 0.054 2a / 3 ) (6) This relationship assumes that the ratios of nutrients
within a cell remain constant, and the nutrients that result
where kb is the attenuation due to sources other than algae from the death of algae due to metabolic functions can be
(m-1), and a is the chlorophyll-a concentration (mg m-3). recycled through dissolved organic matter. The ratio of
Equations 5 and 6 are coupled with the light forcing func- nutrients for a specific strain can also be measured experi-
tion, based on Steeles equation, which considers inhibition mentally.
at high light levels (Di Toro et al., 1971): Dissolved Organic Matter
I As each individual alga dies, the nutrients contained
I
light = exp + 1 (7) within it are converted into dissolved organic matter over
I I
s s time. The concentration of dissolved organic matter (DOM)
depends on the basal metabolic rate as well as the rate of
where Is is the saturating shortwave solar irradiation (W m-2). mineralization of organic matter to inorganic nutrients:
This series of equations allows for the effects of both low
and high light intensities on algae to be considered. The val- S DOM = K bmr a K om DOM (11)
ue of Is is calculated by the average light intensity over the
previous three days in order to account for the algaes ability where KOM is the rate of mineralization of a nutrient. These
to adapt to the light intensity of the environment over time effects are more important on long time scales but could be
(Cole and Wells 2008): an important consideration in algal growth for biofuel pro-
duction.
(
I s = max I s,min , I o,avg exp ( d opt ) ) (8) pH Effects
The pH of the pond system affects the availability of nu-
where trients by changing the ratios of carbon and nitrogen spe-
dopt = optimal depth of photosynthesis (m) cies available for assimilation and also by affecting the
Is,min = minimum saturating solar irradiation (W m-2) enzymes required for metabolism and other cell functions.
Io,avg = weighted average of solar irradiation at the sur- It is therefore crucial to model the effect of non-ideal pH on
face over past three days (W m-2). growth rate. The forcing functions for pH and the values of
The value of Is,min should be taken from experimental da- the hydration constants are as follows (James et al., 2013):
ta specific to each algal species. This series of equations
allows for the quantification of light as a function of depth, [H + ]
algae concentration, and algae biological response. The pH = (12)
[H + ] + kOH + [H + ]2 / kH
equations assume that the photosynthetically active radia-
tion is proportional to the shortwave solar irradiation (Brit-
ton and Dodd, 1976). kOH = 1013 ( 8T 2 500T + 8000 ) (13)

Nutrient Effects kH = 107 ( 5T 3 + 300T 2 5000T + 30000 ) (14)


The Monod relationship is used to quantify algal growth

58(2): 203-213 205


where 1 1
[H+] = hydrogen ion concentration (g m-3) k Henry = ( 0.034 ) exp 2400 (17)
T 298.15
kOH = rate constant for the deprotonation of critical en-
zymes The aeration coefficient was calculated using relation-
kH = rate constant for the protonation of critical en- ships for oxygen transport in streams (Owens and Gibbs,
zymes. 1964):
The values of kOH and kH were taken from experiments
with Chlorella vulgaris and heterotrophic bacteria in a 5.32 U 0.67
K aer ,O2 = (18)
chemostat reactor (Mayo, 1997) and will not necessarily be H 1.85
applicable to other species or for open pond conditions.
However, since C. vulgaris has an ideal pH range similar to where U is the local water velocity (m s-1), and H is the
that of many common species used for biofuel production, depth of the pond (m). This can be converted to the coeffi-
it is assumed that this relationship will provide accurate cient for carbon dioxide using the relationship (32/44)0.25
results. In addition, James et al. (2013) incorporated this (Chapra, 1997). It is important to consider the effects of
relationship into an existing algal growth model with posi- aeration on the carbon dioxide balance of the pond system,
tive results. since adding carbon dioxide affects growth as well as the
economics.
Carbon Dioxide Exchange at Pond Surface
The exchange of carbon dioxide (CO2) between the at- ENERGY MODEL
mosphere and the pond plays an important role in algal The exchange of thermal energy between the open pond
growth and pond optimization. Carbon dioxide is essential to and its surroundings plays a significant role in algal growth,
algal growth and the regulation of pH within the pond; there- with respect to temperature, as seen in equation 4. An ener-
fore, CO2 levels must be monitored and optimized so that the gy balance was constructed at the ponds surface in order to
levels present are optimal for growth and, hence, the eco- estimate the water temperatures of the pond based on data
nomics of algal production. The losses to the atmosphere obtained from a weather station as well as other experimen-
during addition, and the influx when levels in the pond are tally obtained data. Equations for radiation, conduction,
low, must be accounted for in the model. The flux of carbon convection, and evaporation are implemented by modifying
dioxide in the surface layer of the pond can be described as: the source term of the energy equation at the boundary (fig.
SCO2 = K aer ( c,sat c ) (15) 1). Environmental models commonly use relationships
based on weather station data to derive surface heat balanc-
where Kaer is the rate of carbon dioxide exchange (s-1) and es (Cole and Wells 2008). However, with respect to this
c,sat is the saturation concentration of carbon dioxide in the model, the aim was to develop a surface heat balance using
pond (kg m-3). Additional relationships can be added to a mechanistic model based on heat and mass transfer equa-
account for the effects of wind and water velocity. The sat- tions. The governing equation is as follows:
uration concentration can be found from Henrys law: T 1 2
= T +S (19)
c,sat = pCO2 kHenry (16) t

where pCO2 is the partial pressure of carbon dioxide in the Radiation


atmosphere, and kHenry is Henrys constant (mol kg-1 bar-1). Longwave and shortwave radiation, as well as the reflec-
An empirical relationship can be used to calculate Henrys tion of these components, was considered. Shortwave ra-
constant as a function of temperature in fresh water (Carroll diation data can be easily acquired from a weather station.
et al., 1991): However, longwave radiation must be calculated because
weather station sensors do not typically detect this part of

qconv qevap qrad

Salgae

qcond

Figure 1. Two-dimensional domain of CFD model, 0.2 m deep 1 m long, showing mesh and locations where source terms were applied.

206 TRANSACTIONS OF THE ASABE


the spectrum. The formula used to find longwave radiation where
is: qevap = evaporative heat flux (W m-2)
hfg = latent heat of vaporization (J kg-1)
(4
q"rad = Tsky Ts4 ) (20) hm = mass transfer coefficient (m s-1)
sat = concentration of water vapor at saturation (kg m-3)
where = concentration of water vapor in the surroundings
qrad = longwave radiation flux (W m-3) (kg m-3).
= Stefan-Boltzmann constant (J m-2 s-1 K-4) The mass transfer coefficient can be found through rela-
= emissivity of water. tionships based on the heat and mass transfer analogy:
Tsky = sky temperature (K)
h
Ts = temperature at the air and water interface (K). = C p Le1 n (25)
Sky temperatures can be estimated from empirical rela- hm
tionships (Martin and Berdahl, 1984).
where
Convection Cp = specific heat (J kg-1 K-1)
Heat transfer between the pond and the environment will = density (kg m-3)
also occur because of convection. This flux can be calculat- Le = Lewis number
ed using the following equation: n = constant.
This series of equations can easily account for evapora-
q"conv = h (T Ts ) (21) tive heat losses under a variety of environmental condi-
tions.
where
qconv = convective heat flux (W m-2) Conduction
h = average convective heat transfer coefficient (W m-2 Conductive heat transfer will occur between the soil and
K-1) the pond and can be described by the following equation:
T = temperature of the surrounding air (K).
The convective heat transfer coefficient can be obtained (
q"cond = k / L T pond Tsoil ) (26)
from formulas based on the geometry of the system, the
type of flow, and the thermal conductivity. The following where
equations are used and assume geometry approximated as a qcond = convective heat flux (W m-2)
flat plate (Incropera et al., 2007). k = thermal conductivity (W m-1 K-1)
For laminar flow: Tpond = temperature at the water soil interface (K)
Tsoil = temperature of the soil (K).
h = k / L ( 0.664Re1/ 2 Pr1/ 3 ) (22) The soil temperature can be taken from weather station
data. The thermal conductivity can be determined from
For mixed laminar and turbulent flow: experimental data and will be a function of soil type and
moisture content (a value of 0.6 W m-1 K-1 was used for the
( )
h = k / L 0.037Re4L/ 5 871 Pr1/ 3 (23) study).
In order to calculate the energy balance of the system,
where equations 18, 20, and 23 and the shortwave irradiation can
k = thermal conductivity of the air (W m-1 K-1) be combined to give the surface heat flux, while equation
L = length of pond surface (m) 25 can be used calculate the heat flux between the bottom
Pr = Prandtl number surface of the pond and the soil.
Re = Reynolds number.
These equations can be solved if air temperature, wind CFD MODEL
speed, and the thermal conductivity and kinematic viscosity A representative two-dimensional geometry of the cross-
of the air are all known, all of which can be readily ac- section of a typical algae raceway was created (fig. 1). This
quired from weather station data or commonly available geometry would elucidate the transport phenomena occur-
experimental data. ring within the entire raceway and facilitate validation of
Evaporation the code. The number of computation cells was set in order
Evaporation is also an important component of heat to ensure resolution of transport at the boundaries and to
transfer. Assuming that all the energy from phase change is ensure that the ratio of length to width of each cell re-
removed from the pond, the flux can be found using the mained small enough to create a stable domain, with a spe-
following equation, which is based on the amount of water cific mesh developed for each model (1075 cells with a
evaporating at the surface and the latent heat of vaporiza- maximum face size of 5e-2 m). In addition, the time step
tion: was chosen to maintain stability during transient simula-
tions; a convergence criterion of 10e-6 was used. ANSYS
q"evap = h fg hm ( sat ) (24) Fluent, version 14.0, was used for all simulations (Fluent,
2012). The species transport model was used to simulate
algae and nutrient species. The concentration of algae and

58(2): 203-213 207


nutrient species within the pond was considered to be small EXPERIMENTAL AND NUMERICAL VALIDATION
enough as to not affect the properties of water. Starting Greenhouse Pond
from the top surface and moving clockwise, the boundary In order to validate the CFD model of algal growth in an
conditions were set as wall, symmetry, wall, and symmetry. open pond located within a greenhouse, a Runge-Kutta
This model was used as the basis for the development and method and experimental data were used. A standard
validation of the code in all scenarios. fourth-order Runge-Kutta method was used to solve the
species conservation equations (eq. 3). Data on Nannochlo-
USER-DEFINED FUNCTIONS ropsis salina grown in Albuquerque, New Mexico, during
User-defined functions (UDF) were implemented to mod- November 2011 (DOY 314-328) were used for validation.
el transport phenomena not built into the commercial CFD The culture pond had a 0.9 m radius and 0.211 m depth and
software. UDF allow for the modeling of (1) algal growth was located within a greenhouse. Temperature, irradiance,
and nutrient depletion by modification of the mass source pH, nutrient addition, and biomass concentration were rec-
terms within the cell zone and (2) pond temperature by mod- orded throughout the experiment (Timlin et al. 2012).
ification of thermal boundary conditions. In addition, UDF
are used to define boundary and initial conditions, weather Open Raceway Pond
conditions as a function of time, and species thermal proper- In order to validate the CFD model of algal growth and
ties. All functions are written in the C programming lan- pond temperature in an open raceway pond, a finite differ-
ence (FD) method and experimental data were used. An
guage and are interpreted by the CFD software. A flowchart
of the interaction between the CFD software and the UDF is implicit finite difference model, second order in space and
shown in figure 2. The UDF pass variables back and forth to first order in time, was used solve the governing equation
(eq. 19). In order to ensure accurate and stable results in the
the CFD software at each iteration to calculate values needed
to solve the transport and continuity equations. The UDF can finite difference model, the Fourier number was kept below
include the growth model, temperature model, or an integrat- 0.25, and discretization was increased until the error in pre-
dicted average temperature was below 1%. Data on Nanno-
ed model of growth and temperature depending on the simu-
lation requirements. User-defined functions allow the user to chloropsis granulata grown in an open raceway pond at
model more advanced transport phenomena than is available Arizona State Universitys Polytechnic Campus in Mesa,
through the user interface. Arizona, during July 2012 (DOY 194 to 202) were used for

Figure 2. Flowchart of interaction between computational fluid dynamics (CFD) software and user-defined functions (UDF).

208 TRANSACTIONS OF THE ASABE


validation. The pond had an area of 59.3 m2 and a depth of Table 1. Initial and boundary for the Chapra problem.
0.2 m. Temperature was measured and recorded at 15 min Parameter Unit Value[a]
Cp[b] - 1.5
intervals at a single point approximately 0.1 m from the Io W m-2 186
surface using a multiparameter water quality monitoring Is W m-2 116
unit (model 5200A-DC, YSI, Inc., Yellow Springs, Ohio). kb m-1 0.1
Biomass was measured based on ash-free dry weight of Kbmr 1/s 1.57e-7
Kp kg m-3 2.00e-6
pond samples. A detailed explanation of all experimental Photoperiod - 0.5
data can be found in Gharagozloo et al. (2014). Air and soil Pm s-1 1.57e-5
temperature, wind speed, relative humidity, and shortwave z m 10
solar irradiation data for the experimental period were tak- ao kg m-3 5.00e-7
po kg m-3 9.70e-6
en from the Mesa, Arizona, station of the Arizona Meteoro- [a]
Values taken from Chapra (1997, pp. 614-615).
logical Network (AZMET, 2013). [b]
Cp = ratio of phosphorous to chlorophyll-a.

STATISTICAL ANALYSIS
A statistical analysis was conducted to determine the
models sensitivity to parameters using the statistical soft-
ware JMP 9 (JMP, 2011). Since these parameters must be
taken from experimental data, it was important to assess the
degree of error that could be allowed in these parameters
without causing a significant change in biomass output.
Consequently, a preliminary ANOVA was conducted to
find the most significant parameters for further analysis.
The models parameters were increased and then decreased
by 5%, 10%, and 20%, changing only one parameter at a
time while the rest were kept at the baseline (i.e., no
change), and the biomass at 120 h was recorded. A half-
fractional factorial design was selected to analyze the signifi-
cant parameters selected from the preliminary ANOVA. The
error was estimated by aliasing the three-way interactions.

Figure 3. Chlorophyll-a (Chla) and phosphorous (P) concentration as


RESULTS a function of time for the Chapra problem using computational fluid
VALIDATION OF ALGAL GROWTH MODEL dynamics (CFD) and fourth-order Runge-Kutta (RK4) methods.
AGAINST NUMERICAL METHOD
Table 2. Greenhouse model parameters.
The numerical solution, using a fourth-order Runge-
Parameter Unit Value[a]
Kutta method, was compared with the numerical results B:C[b] - 45:01:00
obtained from the CFD code that had been modified using C:N:P[c] - 358:38:01
UDF. This validation is necessary to ensure that the UDF CO2 flow rate kg m-3 s-1 4.4e-7 or 1.7e-6
were properly coded and implemented into the commercial Is,min W m-2 17
CFD software. A slightly modified version of example 33.2 K1 C-2 0.69
-3
K2 C 0.007
from Chapra (1997, pp. 614-615) was solved. Because the kb m-1 0.45
system is assumed to be a fully mixed, batch system, the Kbmr s-1 1.57e-7
concentration of chlorophyll-a, a function of algae concen- Kc kg m-3 2.80e-5
tration, and phosphorous can be found by establishing the kchla m m3 (g B)-1 0.314
Kn kg m-3 1.00e-5
mass balance of the system. The boundary and initial con- KOM,n s-1 1.70e-7
ditions are listed in table 1, and the results of the simulation KOM,p s-1 1.20e-6
are shown in figure 3. The CFD model was found to be in Kp kg m-3 2.00e-6
good agreement with the fourth-order Runge-Kutta solution Pm s-1 1.20e-5
T1,opt C 18
of the Chapra (1997) problem, indicating that the UDF T2,opt C 22
were functioning correctly. ao kg m-3 1.54e-2
co kg m-3 9.00e-4
VALIDATION OF ALGAL GROWTH MODEL no kg m-3 5.47e-2
AGAINST EXPERIMENTAL DATA po kg m-3 3.10e-3
[a]
Values taken from James et al. (2013).
The CFD model was used to simulate data obtained from a [b]
B:C = ratio of algal biomass to carbon.
greenhouse experiment, and its outcomes were compared to a [c]
C:N:P = ratio of carbon to nitrogen to phosphorous in algal biomass.
fourth-order Runge-Kutta model and an Environmental Fluid
Dynamics Code (EFDC) model used to model the same ex- trations from the CFD, fourth-order Runge-Kutta, and EFDC
periment. The initial and boundary conditions, as well as the models, along with experimental data obtained from the
models constants, are listed in table 2. Average algal concen- greenhouse experiment, are shown in figure 4. Results of the

58(2): 203-213 209


Table 3. Minimum and maximum average weather data input values
for July 2012 from an AZMET weather station in Mesa, Arizona
(AZMET, 2013).
Input Parameter Unit Minimum Maximum
Irradiance W m-2 0 1069.40
Tair C 22.80 40.90
Tsoil C 32.70 36.40
Relative humidity % 10.30 91.00
Wind speed m s-1 0.20 7.70

EFDC code and experimental data come from James et al.


(2013). The results from the CFD and Runge-Kutta models
were so similar that it is nearly impossible to distinguish their
individual values on the graph, indicating that the commercial
CFD software correctly integrated the UDF code and no pro-
gramming errors were present. In addition, good agreement
was achieved between the CFD and EFDC models and the
experimental data. There was a slight variation between the
Figure 4. Average algae concentration as a function of time for com- outcomes of the EFDC model and the CFD model between
putational fluid dynamics (CFD), fourth-order Runge-Kutta (RK4), days 3 and 7. This was most likely because light-averaging
and Environmental Fluid Dynamics Code (EFDC) models along with
experimental data from a greenhouse experiment. Experimental data
equations are applied by the EFDC model but not by the CFD
were taken from James et al. (2013). model. Overall, the CFD model was able to predict algal
growth in a closed environment based on species, environ-
310 mental, and water quality parameters.
Tavg (CFD)
Tavg (FD)
Data
VALIDATION OF ENERGY BALANCE AGAINST THE
NUMERICAL METHOD AND EXPERIMENTAL DATA
The CFD model of pond temperature was compared
with experimental data (Gharagozloo et al., 2014) and the
Temperature(K)

305
FD numerical method. The minimum and maximum values
of weather data over the experimental period are listed in
table 3 (AZMET, 2013). A good agreement exists between
the CFD and FD models, indicating that there are no errors
300 in the UDF code of the CFD model (fig. 5). Agreement
between the model and experimental data was not exact,
although the trends are well matched (fig. 5). The absolute
difference between the experimental data and CFD model
did not exceed 5C, and the relative difference was below
295 2% (fig. 6). This was expected, since average values pre-
0 1 2 3 4 5 6 7
Time (days) dicted by the model were compared to a point measurement
Figure 5. Results of experimental validation of computational fluid within a large pond. To more thoroughly validate this mod-
dynamics (CFD) code against finite difference (FD) model and exper- el, an experiment with temperature measurements at multi-
imental data. Experimental data were taken from Gharagozloo et al. ple points within the pond would be required. With this
(2014).
validation, the code for the prediction of pond temperature
could be implemented into the full model of algal growth.

VALIDATION OF COMBINED GROWTH


AND ENERGY MODEL
The combined model of algal growth and pond tempera-
ture was validated against experimental data (Gharagozloo
et al., 2014). The parameters of algal growth for a similar
species of algae were taken from Gharagozloo et al. (2014).
The results of average algal biomass for the simulation and
the experimental data are shown in figure 7. The R2 value
for the model was 0.88. These results show that an increase
in biomass occurred during hours when light was present
(to fuel photosynthesis), and a decrease in biomass oc-
curred during dark hours (due to respiration). This model
effectively predicted growth in an open pond.

Figure 6. Absolute and percent difference between experimental data


and computational fluid dynamics results.

210 TRANSACTIONS OF THE ASABE


550

500

Algae Concentration (g C / m3)


Summer
450
Winter
400

350

300

250

200

150

100
0 1 2 3 4 5 6 7
Time (days)

Figure 8a. Average algal growth of N. Salina under summer (July)


and winter (January) conditions in Tucson, Arizona, during 2012.
Figure 7. Comparison of predicated values of biomass (CFD) and Solid line represents summer conditions, and dash-dotted line repre-
experimental data. Experimental data were taken from Gharagozloo sents winter conditions.
et al. (2014).

SCENARIOS AND PARAMETRIC STUDY


0.5 Summer
In order to determine the effectiveness of the model and Winter
its parameters in assessing algal growth, scenarios of algal

Light Forcing Function


growth, as well as a parametric study, were conducted. 0.4
Scenarios of N. salina growth under summer (July) and
winter (January) conditions in Tucson, Arizona, were run in 0.3
the CFD model. Average weather conditions for the months
of January and June were taken from AZMET on an hourly 0.2
basis for the year 2012 (AZMET, 2013). Model parameters
were taken from Gharagozloo et al. (2014). The time step
0.1
and mesh were selected such that the computational load
could be minimized without affecting the results.
Growth during summer months was much higher due to 0
0 1 2 3 4 5 6 7
the optimal temperature range and light intensity and availa- Time (days)
bility (fig. 8a). The dramatic increase in productivity during Figure 8b. Light forcing function for average algal growth of N. Sa-
the simulated summer months (539 g C m-3 vs. 270 g C m-3) lina under summer (July) and winter (January) conditions in Tucson,
could have implications for commercial production. The Arizona, during 2012. Solid line represents summer conditions, and
following scenarios demonstrate the interaction between the dash-dotted line represents winter conditions.
environment and algal growth. For example, when light and 1
temperature are more favorable, the biomass increases rapid-
0.9
ly, causing light, in turn, to become limiting (fig. 8b). In ad-
Temperature Forcing Function

dition, the maxima and minima of the temperature forcing 0.8


function coincide with the winter and summer conditions, 0.7
respectively (fig. 8c). This is because the heat of the day un-
0.6
der summer conditions is not favorable for growth, while the
heat of the day under winter conditions is the most favorable 0.5
for growth. Therefore, in order for commercial production to 0.4
be profitable, algae must be directly suited to the environ-
mental conditions, or significant wastes of energy could oc- 0.3

cur in the production of algal biomass. 0.2


The mean square values of each parameter are listed in 0.1 Summer
table 4: the greater the mean square value, the greater the Winter
effect of the parameter on biomass yield. The parameters 0
0 1 2 3 4 5 6 7
with a mean square greater than 1 were selected for a further Time (days)
analysis (i.e., factorial design) in order to look more closely Figure 8c. Temperature forcing function of algal growth under aver-
at the main effect and interactions and to assess their statisti- age summer and winter conditions in Tucson, Arizona. Solid line
cal significance. Mean square values were used because represents summer conditions, and dash-dotted line represents winter
conditions.
there were not sufficient degrees of freedom with which to
estimate the error term in the statistical ANOVA model. It is

58(2): 203-213 211


Table 4. Results of ANOVA for factor screening. Parameters with time. The statistical analysis indicated that great care
mean square values >1 were selected for further screening.
should be taken when selecting the significant parameters,
Parameter Mean Square
T2,opt 363.12
and especially when selecting Kag, T2, and K2, since their
Kag 127.03 interactions are also important. In addition, the interpreta-
K2 15.16 tion can be made that if significant increases in biomass are
Kbmr 14.48 to be achieved, then algae that are robust under high-
Is 12.93
kchla 2.89
temperature and high-irradiance conditions should be se-
T1,opt 0.44 lected. Statistical analysis allowed for the determination of
Kaer 0.4 model parameters that could significantly affect the model
Cc 0.063 results.
Pc 0.017
Cp 0.015
Cn 0.011
Pp
Pn
0.007
0.005
CONCLUSION
K1 0.002 Once thoroughly validated, this model showed promise
kOH <0.0001 as a tool for predicting algal biomass production based on
kH <0.0001 pond design, algae species, and weather conditions. The
ease with which this model can be adapted to suit any sce-
important to note that, due to the high temperatures during nario will allow users to apply it to commercial or research
the summer months, only those temperature parameters con- projects. However, it should be noted that this model is
trolling the upper boundary were significant. In addition, limited in that all the model parameters must be calibrated
nutrients and pH were not especially limiting, so the parame- for each specific strain of algae, which requires laboratory
ters controlling those interactions were not selected. experimentation. The sensitivity analysis indicated that the
An efficient, half-fractional factorial design was per- main effects of those parameters that control growth rate
formed for the chosen parameters (Kag, Kbmr, T2, K2, kchla, (basal metabolic rate, temperature and light effects, and a
and Is), each at two levels (10%). All six parameters were few of their interactions) are significant at 10%. Statistical
found to be significant. In addition, the interactions be- analysis can also serve to identify characteristics in algae
tween T2 and K2 and between Kag and T2 were also found to that would lead to significant increases in algal biomass.
be significant (fig. 9). A simultaneous p-value, rather than The model could be expanded into three dimensions to de-
the individual p-value, was chosen in order to control for termine if spatial effects of mixing have a significant affect
the family-wise type I error, as multiple terms (i.e., main on the accuracy of the model in a production setting. The
effects and interaction effects) need to be tested at the same current results obtained using this model demonstrate that

Figure 9. Results of fractional factorial screening with main effects and interaction terms. All main effects were found to be significant, and the
Kag T2 and T2 K2 interactions were also found to be significant as well.

212 TRANSACTIONS OF THE ASABE


CFD, when combined with UDF, can accurately predict Huesemann, M. H., Van Wagenen, J., Miller, T., Chavis, A., Hobbs,
biomass and thereby serve as a cost and time effective S., & Crowe, B. (2013). A screening model to predict
method for designing and optimizing the open ponds used microalgae biomass and growth in photobioreactors and
for biomass production. raceway ponds. Biotech. Bioeng., 110(6), 1583-1594.
http://dx.doi.org/10.1002/bit.24814.
Incropera, F. P., Dewitt, D. P., Bergman, T. L., & Lavine, A. S.
ACKNOWLEDGEMENTS (2007). Fundamentals of Heat and Mass Transfer (6th ed.). New
Sandia National Laboratories is a multi-program labora- York, N.Y.: John Wiley and Sons.
tory managed and operated by Sandia Corporation, a whol- James, S. C., & Boriah, V. (2010). Modeling algae growth in an
ly owned subsidiary of Lockheed Martin Corporation, for open-channel raceway. J. Computer Biol., 17(7), 895-906.
the U.S. Department of Energys National Nuclear Security http://dx.doi.org/10.1089/cmb.2009.0078.
Administration under Contract No. DE-AC04-94AL85000. James, S. C., Janardhanam, V., & Hanson, D. T. (2013). Simulating
pH effects in an algae-growth hydrodynamics mode. J. Phycol.,
49(3), 608-615.
JMP. (2011). Using JMP. Ver. 9. Cary, N.C.: SAS Institute, Inc.
REFERENCES Lundquist, T. J., Woertz, I. C., Quinn, N., & Benemann, J. (2010).
AZMET. (2013). Tucson station data and reports. Tucson, Ariz.: A realistic technology and engineering assessment of algae
Arizona Meteorological Network. Retrieved from biofuel production. Berkeley, Cal.: University of California,
http://ag.arizona.edu/azmet. Energy Biosciences Institute. Retrieved from www.energy
Britton, C. M., & Dodd, J. D. (1976). Relationships of biosciencesinstitute.org/media/AlgaeReportFINAL.pdf.
photosynthetically active radiation and shortwave irradiance. Martin, M., & Berdahl, P. (1984). Characteristics of infrared sky
Agric. Meteorol., 17(1), 1-7. http://dx.doi.org/10.1016/0002- radiation in the United States. Solar Energy, 33(3-4), 321-336.
1571(76)90080-7. http://dx.doi.org/10.1016/0038-092X(84)90162-2.
Chapra, S. C. (1997). Surface Water-Quality Modeling. New York, Mayo, A. W. (1997). Effects of temperature and pH on the kinetic
N.Y.: McGraw Hill. growth of unialga Chlorella vulgaris cultures containing
Clarens, A. F., Resurreccion, E. P., White, M. A., & Colosi, L. A. bacteria. Water Environ. Res., 69(1), 64-72.
(2010). Environmental life cycle comparison of algae to other http://dx.doi.org/10.2175/106143097X125191.
bioenergy feedstocks. Environ. Sci. Tech., 44(5), 1813-1819. Owens, M., Edwards, R. W., & Gibbs, J. W. (1964). Some
http://dx.doi.org/10.1021/es902838n. reaeration studies in streams. Air Water Pollution, 8, 469-486.
Cole, T. M., & Wells, S. A. (2008). CE-QUAL-W2: A two- Packer, A., Li, Y., Andersen, T., Hu, Q., Kuang, Y., & Sommerfeld,
dimensional, laterally averaged, hydrodynamic and water quality M. (2011). Growth and neutral lipid synthesis in green
model, Version 3.6. Vicksburg, Miss.: U.S. Army Corps of microalgae: A mathematical model. Bioresource Tech., 102(1),
Engineers, Waterways Experiment Station. 111-117. http://dx.doi.org/10.1016/j.biortech.2010.06.029.
Di Toro, D. M., OConnor, D. J., & Thomann, R. V. (1971). A Quinn, J., de Winter, L., & Bradley, T. (2011). Microalgae bulk
dynamic model of phytoplankton population in the Sacramento- growth model with application to industrial-scale systems.
San Joaquin delta. In Advances in Chemistry 106: Bioresource Tech., 102(8), 5083-5092.
Nonequilibrium Systems in Natural Water Chemistry (pp. 131- http://dx.doi.org/10.1016/j.biortech.2011.01.019.
180). Washington, D.C.: American Chemical Society. Sato, T., Yamada, D., & Hirabayashi, S. (2010). Development of
Eilers, P. H., & Peeters, J. C. (1988). A model for the relationship virtual photobioreactor for microalgae culture considering
between light intensity and the rate of photosynthesis in turbulent flow and flashing light effect. Energy Conv. Mgmt.,
phytoplankton. Ecol. Model., 42(3-4), 199-215. 51(6), 1196-1201.
http://dx.doi.org/10.1016/0304-3800(88)90057-9. http://dx.doi.org/10.1016/j.enconman.2009.12.030.
Fluent. (2012). Fluent Users Guide. Ver. 13.1. Evanston, Ill.: Wu, X., & Merchuk, J. C. (2002). Simulation of algae growth in a
ANSYS, Inc. bench-scale bubble column reactor. Biotech. Bioeng., 80(2),
Gharagozloo, P. E., Drewry, J. L., Collins, A. M., Dempster, T. A., 156-168. http://dx.doi.org/10.1002/bit.10350.
Choi, C. Y., & James, S. C. (2014). Analysis and model of Yoshimoto, N., Sato, T., & Kondo, Y. (2005). Dynamic discrete
Nannochloropsis growth in lab, pond, and raceway experiments. model of flashing light effect in photosynthesis of microalgae. J.
J. Appl. Phycol., 26(6), 2303-2314. Appl. Phycol., 17(3), 207-214. http://dx.doi.org/10.1007/s10811-
http://dx.doi.org/10.1007/s10811-014-0257-y. 005-7908-y.

58(2): 203-213 213

You might also like