Anaerobic Digestion Model with Multi-Dimensional Architecture (ADM-MDA)
Ph.D. Candidacy exam2011-04-27
David L. F. Gaden
Outline
Introduction
Problem
Proposed solution
Literature review
Methodology
Code development
Results
Tasks remaining
Introduction
What is anaerobic digestion?◦ The breaking down of biomass in the absence of oxygen
◦ Treats waste products while simultaneously producing renewable energy (biogas)
Primary purpose:◦ Waste treatment
◦ Energy production
◦ Pollution reduction
◦ Odour mitigation
Introduction
Applications◦ Industrial wastewater treatment
◦ Municipal wastewater treatment
◦ Agricultural wastewater treatment
Introduction
Stages of anaerobic digestion
1. Disintegration
Introduction
Stages of anaerobic digestion
2. Hydrolysis
Introduction
Stages of anaerobic digestion
3. Acidogenesis
Introduction
Stages of anaerobic digestion
4. Acetogenesis
Introduction
Stages of anaerobic digestion
5. Methanogenesis
Introduction
Types of digesters
Influent Effluent
Biogas
Plug flow digester
Introduction
Types of digesters
Influent
Effluent
Biogas
STR digester
Introduction
Types of digesters
Influent
Effluent
Biogas
Upflow anaerobic sludge blanket digester
Introduction
Types of digesters
Influent
Effluent Biogas
Anaerobic clarigester
Problem
Reliability issues
Unreliable
Well established
Problem
Modelling anaerobic digesters◦ Current state of the art is ADM1:
Batstone, D. J.; Keller, J.; Angelidaki, I.; Kalyuzhni, S. V.; Pavlostathis, S. G.; Rozzi, A.; Sanders, W. T. M.; Siegrist, H.; Vavlin, V. A.. 2002. "Anaerobic Digester Model No.1 (ADM1)", Scientific and Technical Report No.13, IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. IWA Publishing, London, United Kingdom.
Rosen, C.; Jeppsson, U.. 2006. “Aspects on ADM1 implementation within the BSM2 framework,” Technical Report No. LUTEDX/(TEIE-7224)/1-35/(2006). Department of Industrial Electrical Engineering and Automation, Lund University, Lund, Sweden.
Problem
Blumensaat, F.; Keller, J.. 2005. "Modelling of two-stage anaerobic digestion using the IWA Anaerobic Digestion Model No. 1 (ADM1)," Water Research, v. 39, pp. 171-183.
Problem
Ozkan-Yucel, U. G.; Gökçay, C. F.. 2010. "Application of ADM1 model to a full-scale anaerobic digester under dynamic organic loading conditions," Environment Technology, v. 31, n. 6, pp. 633-640.
Problem
Modelling anaerobic digesters◦ Current state of the art is ADM1:
◦ A bulk model
0z
S
y
S
x
S
o No spatial variation
o Uniform properties
Proposed solution
Objective: a spatially-resolved ADM1
ADM-MDA – Anaerobic Digestion Model with Multi-Dimensional Architecture
Methodology – Theory – ADM1
Liquid volume
Gas volume
Sgas,var (3)
Svar (21)Xvar (12)
Qliq
Qliq
Qgas
Methodology – Theory – ADM1
reacvaroutvarinvarvar mmm
dt
dm,,,
varliqoutvarliqinvarliqvar
liq rVSQSQdt
dSV ,,
varvarinvar
liq
liqvar rSSQ
V
dt
dS,
Methodology – Theory – ADM1
Methodology – Theory – ADM1
lilihydlifachchhydS XkfXkrsu ,,, 1
Methodology – Theory – ADM1
Ssu
Saa
Sfa
Sch4
Sh2
SI
Scat
San
Xc
Xch
Xpr
Xli
Xsu
Xaa
Xfa
Xc4
Xpro
Xac
Xh2
XI
Sgas,h2
Sgas,ch4
Sgas,co2
Sbu Shbu
Sbu-
Sva Shva
Sva-
Spro Shpro
Spro-
Sac Shac
Sac-
SIC SCO2
SHCO3-
SIN SNH3
SNH4+
Sh+ Soh-
pgas,h2
pgas,ch4
pgas,co2
pgas,total Qgas
pH
ODE variables
Derived variables
Methodology – Theory – ADM1
Ssu
Saa
Sfa
Sch4
Sh2
SI
Scat
San
Xc
Xch
Xpr
Xli
Xsu
Xaa
Xfa
Xc4
Xpro
Xac
Xh2
XI
Sbu Shbu
Sbu-
Sva Shva
Sva-
Spro Shpro
Spro-
Sac Shac
Sac-
SIC SCO2
SHCO3-
SIN SNH3
SNH4+
Sh+ Soh-
pH
Sgas,h2
Sgas,ch4
Sgas,co2
pgas,h2
pgas,ch4
pgas,co2
pgas,total Qgas
ODE variables
Derived variables
Algebraic variables
Methodology – Theory - CFD
Governing equations:◦ Conservation of mass:
◦ Conservation of momentum:
◦ Conservation of energy:
0U
refref TTpDt
DU
U 2
0TTt
TU
Methodology – Theory – ADM-MDA
Two options:1. Start with ADM1 and write CFD into it; or
2. Start with CFD and write ADM1 into it.
ODE:
PDE:
varvarinvar
liq
liqvar rSSQ
V
dt
dS,
varvarvarvar rSSt
SU
Methodology – Theory – ADM-MDA
Three biochemistry strategies:1. Source term solver
2. ODE solver
3. Coupled solver
Methodology – Theory – ADM-MDA
Three biochemistry strategies:1. Source term solver
varvarvarvar rSSt
SU
varvarinvar
liq
liqvar rSSQ
V
dt
dS,
Methodology – Theory – ADM-MDA
Three biochemistry strategies:2. ODE solver
Transport variables (CFD)
Solve biochemistry (ODE)
Methodology – Theory – ADM-MDA
Three biochemistry strategies:3. Coupled solver
Methodology – Theory – ADM-MDA
Three biochemistry strategies:3. Coupled solver
Methodology – Theory – ADM-MDA
Time scale issue◦ ADM1:
◦ CFD: s 1dt
min 15dt
Transient solver (flow)
Steady state solver (biochemistry)
Boundary
conditio
ns c
hange
“multiSolver”
Methodology – Theory – ADM-MDA
Length scale issue
ADM1 CFD
Prolongation
Restriction
“dualGrid”
Methodology – Theory – ADM-MDA
Accessibility◦ “equationReader”
◦ The ability to read equations from a text file
Code development
Flow
Biochemistry
Transient solver
Steady state solver
Scalar transport
Non-Newtonian
Buoyancy
Particle model
PISO
Temperature
RANS turbulence
Steady state detection
Scalar transport
Particle model
Gas model
ODE solver
Implicit solver (Sh2)
Implicit solver (ions)
multiSolver
dualGrid
equationReader
Framework
Control
Code development
Scalar transport
Non-Newtonian
Buoyancy
Particle model
PISO
Temperature
RANS turbulence
Transient solver
Steady state detection
Scalar transport
Particle model
Gas model
ODE solver
Implicit solver (Sh2)
Implicit solver (ions)
Flow
Biochemistry
Steady state solver
multiSolver
dualGrid
equationReader
Control
?
?
Framework
Code development
Scalar transport
Non-Newtonian
Buoyancy
Particle model
PISO
Temperature
RANS turbulence
Transient solver
Steady state detection
Scalar transport
Particle model
Gas model
ODE solver
Implicit solver (Sh2)
Implicit solver (ions)
Flow
Biochemistry
Steady state solver
multiSolver
dualGrid
equationReader
Control
?
?
Framework
Code development
Bulk model◦ Written in Excel + Visual Basic macros
Post processing routines◦ Instantly produce comparisons between bulk model & ADM-MDA
◦ Written in Python
Results
Bulk model verification
Results
Boussinesqbuoyancy
model validation
Results
Two problems with the model:◦ Efficiency
◦ Transport deviation
Results
ChangeSimulation
TimeReal Time
Benchmark Estimate
Initial 27 [s] 6 [hr] >100,000 [yr]
(1000 cells, 1000 days)
Results
transS00 r
transr
trans SS
= +
Results
ChangeSimulation
TimeReal Time
Benchmark Estimate
Initial 27 [s] 6 [hr] >100,000 [yr]
Segregated flow
1 [dy] 18 [hr] 128 [yr]
(1000 cells, 1000 days)
Results
Semi-implicit Bulirsch Stoer
ddt
Solve derived
ODE ddt
Implicit Sh2 solver
Implicit ion solver
“innerLoops”
Results
ChangeSimulation
TimeReal Time
Benchmark Estimate
Initial 27 [s] 6 [hr] >100,000 [yr]
Segregated flow
1 [dy] 18 [hr] 128 [yr]
Inner loops1 [dy]
100 [dy]12 [min]
5:32[hr:min]144 [dy]
(1000 cells, 1000 days)
Results
Results
ChangeSimulation
TimeReal Time
Benchmark Estimate
Initial 27 [s] 6 [hr] >100,000 [yr]
Segregated flow
1 [dy] 18 [hr] 128 [yr]
Inner loops1 [dy]
100 [dy]12 [min]
5:32[hr:min]144 [dy]
Static yields 1 [dy] 62 [s] 12.5 [dy]
(1000 cells, 1000 days)
Results
Results
ChangeSimulation
TimeReal Time
Benchmark Estimate
Initial 27 [s] 6 [hr] >100,000 [yr]
Segregated flow
1 [dy] 18 [hr] 128 [yr]
Inner loops1 [dy]
100 [dy]12 [min]
5:32[hr:min]144 [dy]
Static yields 1 [dy] 62 [s] 12.5 [dy]
Function Pointers
1[dy]100 [dy]
58 [s]7 [min]
3.17 [dy]
(1000 cells, 1000 days)
The solver is now parallelized!
Results
1 day (with flow)
Results
100 days (with flow)
Results
100 days (no flow)
Transport Error
Quad precision
Pseudo-coupled solver
Mini timesteps for transport
Improve ADM1 numerical stability
Averaged noisy transport
Agglomeration
Timeline
Timeline
Judgement day ... (May 21st)
Solve transport error ... (June 1st)
Gas model ... (June 22nd)
Particle model ... (July 15th)
Dual-grid ... (August 1st)
Validate model ... (Octobruary 15th)
Write thesis ... (Novembgust 30th)