Transcript Document

Covered Anaerobic Lagoon Simulation Using
Computational Fluid Dynamics
Jason G. Fleming and Richard R. Johnson
Objectives
This research project proposes to do the following: (1) adapt a conventional
anaerobic process model for use in a computational fluid dynamic simulation; (2)
create a CFD simulation of a covered lagoon; (3) investigate the effect of fluid
dynamics on the performance of a covered lagoon digester; and (4) provide
recommendations concerning the design and operation of covered lagoons.
Figure 1. Currently, waste material from
large scale swine operations in North
Carolina is typically handled using open
lagoons. Open lagoons provide some
waste treatment (depending on
temperature and loading rate), but they
also release methane and ammonia to the
atmosphere. Large land areas are
required for the application of lagoon
effluent.
Figure 2. Anaerobic digestion is an attractive
alternative treatment process because it typically
removes 90% of the Chemical Oxygen Demand
(COD) from the waste, it prevents ammonia from
escaping, it captures the valuable biogas, and it
uses naturally occurring microorganisms. The
disadvantages of anaerobic digestion are
expense and instability. For example, this
European anaerobic digestion plant is too
expensive to be practical in North Carolina.
Figure 5. The overall behavior of the covered anaerobic lagoon is
governed by several interdependent processes simultaneously. The fluid
velocities were calculated using the Semi-Implicit Method for Pressure
Linked Equations (Patankar, 1980). SIMPLE is a conservative finite
volume method that iterates line-by-line to find velocity in incompressible
flow. Once the velocities were known, the effect of bulk fluid motion on
species transport was modeled with an advection method for
incompressible flow (LeVeque, 1996). Gravity creates a downward
settling flux for non-dissolved species such as biomass and raw
substrate. This flux was set to 5% per day.
Figure 11. The transient response of
the conventional model shows no
benefit from a gradual start, and
actually predicts that steady state is
reached more quickly with an abrupt
start. The transient response of the
multidimensional simulation is much
more realistic, with more gradual starts
providing higher steady state
performance.
Figure 6. The multidimensional anaerobic digestion simulation
starts by solving for the fluid velocity field. The concentration
gradients are formed by iterating through the models that
describe the important processes inside the covered lagoon:
advection, settling, chemical reaction and biological reaction.
Post processing consists of comparison with experimental data
as well as 3D visualization. The overall model results will be
validated with experimental biogas production data from the
Barham Farm covered lagoon.
Changes to Reaction Model
Figure 3. Covered lagoons represent a “middle way,” combining the low cost of an open
lagoon with the controlled environment of an anaerobic tank reactor. However, conventional
anaerobic digestion models are insufficient for covered lagoon process design. The
Barham Farm covered lagoon (in Zebulon, NC) is pictured here.
Figure 10. In order to test the dynamic
response during startup, three cases of
unsteady inlet boundary conditions
were used. In the “slow startup” case,
the inlet concentration is ramped up
from zero to full load over a period of
150 days. In the “fast startup” case, the
ramp time is 60 days, and in the
“abrupt startup” case, the ramp time is
zero.
The microbial model from Hill et al. (1983) was modified by removing the terms
involving hydraulic retention time (HRT). The HRT—used to account for the
concentration difference between mass entering and leaving the reactor—becomes
redundant when fluid velocity and species transport are explicitly simulated. The kinetic
parameters from the model were used directly, without adjustment or calibration.
Figure 12. Contours of steady state methane production are shown for the
gradual and abrupt startup cases. Waste is flowing in from the right and out to
the left, while solids settle to the lagoon floor. A large unproductive area is
evident in the abrupt loading case.
Conclusions and Recommendations
The conventional model showed no real difference resulting from different
startup procedures. On the other hand, the multidimensional model predicts
better performance from a gradual startup, validating the recommendations
found in the literature.
Background
Furthermore, the 3D visualization revealed the mechanism behind the
performance enhancement: gradual startup avoids localized inhibition near the
lagoon inlet. Based on these results, it is recommended that covered lagoons
similar to the Barham lagoon ramp up the inlet concentrations over a period of
at least 120 days during initial startup.
Covered lagoon system design requires an accurate model of the biochemical and physical
phenomena underlying the process. The literature on the biochemistry of anaerobic
digestion is mature and well developed. However, physical modeling of anaerobic tank
reactors is crude; perfect mixing and constant temperature are always assumed. These
assumptions result in a process model consisting of ordinary differential equations (ODEs) which
are easy to solve but do not match well with empirical observations in real systems (Heinzle, et
al., 1993). The assumptions of perfect mixing and constant temperature are particularly
inappropriate for covered lagoons because covered lagoons are not mixed and have
seasonally varying temperatures.
dS

dt
dA

dt
S 0  S M

HRT
Y
A0  A  M 1  Y    c M c
HRT
Y
Yc
dM    k d  1 

M
dt  HRT 
d M c   c  k dc  1 

Mc
dt
 HRT 
Figure 4. A conventional model from Hill, et al (1983) represents conservation of mass for four
species: The S and A terms represent waste concentrations and the M and Mc terms represent
microbial concentrations. HRT is hydraulic retention time.
References
Heinzle, E., Dunn, I.J., and Ryhiner, G.B. 1993. Modeling and control for
anaerobic wastewater treatment. In Advances in Biochemical
Engineering/Biotechnology No. 48. Berlin: Springer-Verlag.
Figure 7. The fluid velocity in the lagoon
is visualized with white stream tubes (fluid
enters on the right and progresses toward
the left). Landscape graphics add scale
and context.
Figure 8. Biodegradable volatile solids
(BVS) are the primary component of raw
waste. The effect of the settling model on
BVS concentration is clearly visible here (the
contours of concentration slope downward
toward the exit).
Figure 9. Methane generation is highest
near the center of this sludge blanket
because the inlet side is overloaded while
the outlet side is underloaded.
Application to Startup Procedures
In order to measure the performance of this model against that of a conventional model, the transient startup
responses were compared. The physical characteristics of the models were selected to match that of the Barham
Farm covered lagoon. The startup transient was chosen as a “standard problem” because anaerobic digesters require
extra attention during startup; the natural initial concentration of methanogenic bacteria in the environment is small and
these bacteria tend to become inhibited before developing sufficient biomass.
One recommended startup procedure is to dilute the raw waste until the microbial community can establish itself (Dalla
Torre and Stephanopoulos, 1986). This recommendation was tested with three sets of inlet boundary conditions (see
Figure 10).
Hill, D.T., Tollner, E.W., and Holmberg, R.D. 1983. The kinetics of inhibition in
methane fermentation of swine manure. Agricultural Wastes 5, pp. 105-123.
LeVeque, R.J. 1996. High resolution conservative algorithms for advection in
incompressible flow. SIAM J. Numer. Anal. Vol. 33, No. 2, pp. 627—665.
Patankar, S.V. 1980. Numerical heat transfer and fluid flow. Washington, DC:
Taylor & Francis.
Torre, A.D., and Stephanopoulos, G. 1986. Mixed culture model of anaerobic
digestion: application to the evaluation of startup procedures. Biotechnology
and Bioengineering. Vol. 28, pp. 1106—1118.
Acknowledgements
We would like to acknowledge financial support (fellowship, internship, and
equipment grant) from SGI, Cray Research, and the North Carolina
Supercomputer Center. We would also like to thank Dr. Jay Cheng and Mr.
Julian Barham for their collaborative effort.