In the project “Studies on integrated biohydrogen production process-Experimental and Modeling”,a co-culture (mixture of two microorganisms in a single reactor) study of a dark fermentative and photofermentative microorganism was done to assess its hydrogen production performance. For modeling purpose, Artificial Neural Network and Genetic Algorithm has been used as a stochastic technique. The optimized data from batch study was successfully used to run a photobioreactor in continuous mode. A mechanistic model was developed for a continuous co-culture setup using data from literature and solved using MATLAB.
STUDIES ON INTEGRATED BIO-HYDROGEN PRODUCTION PROCESS-EXPERIMENTAL AND MODELING
1. STUDIES ON INTEGRATED BIO-
HYDROGEN PRODUCTION PROCESS-
EXPERIMENTAL AND MODELING
Presented By
Arghya Dhar
Exam Roll No - M4CHE1712
Under the Guidance of
Prof. (Dr.) Ranjana Chowdhury
DEPARTMENT OF CHEMICAL ENGINEERING
JADAVPUR UNIVERSITY
2. INTRODUCTION
PRESENT ENERGY STATUS
Fossil fuels (non-renewable)
1. Primary energy resource,
2. Pollution prone,
3. Dwindling reserves.
Biomass (renewable)
1. Eco-friendly energy
resource,
2. Abundant.
1. Highest energy density(120
MJ/kg)
2. Less energy intensive.
3. Remediation of
waste/wastewater.
BIO-HYDROGEN( A
POTENTIAL SOLUTION)
4. AIMS AND OBJECTIVES
To select a compatible pair of dark fermentative and photofermentative microorganism for co-culture
study.
To formulate a nutrient medium for co-culture study.
To determine growth kinetic parameters of both microorganism through individual batch studies.
To perform single stage batch experiments using ratio of initial inoculum of microorganisms, initial
substrate concentrations and base dose concentration as parameters.
To optimize operating parameters using GA on ANN and validate ANN predicted data with experimental
data.
To set up and operate a flat plate continuous photobioreactor using optimized parameter values.
To develop a mathematical model for a continuous system using MATLAB.
9. ANALYTICAL METHOD
SAMPLE ANALYTICAL METHOD
BIOMASS •DETERMINATION OF O.D. VALUES OF BATCH
SAMPLES THROUGH SPECTROPHOTOMETRIC
ANALYSIS
•DETERMINATION OF CELL CONCENTRATION
USING DRY CELL WEIGHT METHOD AND
STANDARD CURVE.
HYDROGEN ANALYSIS OF A SAMPLE USING GAS
CHROMATOGRAPHY(GC)
ANALYSIS OF CO2 OF OTHER SAMPLES USING
ORSAT GAS ANALYZER
STOICHIOMETRIC ESTABLISHMENT OF
HYDROGEN PRODUCTION FROM CO2 ANALYZED
10. THEORETICAL ANALYSIS
GROWTH KINETICS:
The rate of growth of a bacteria growing in a batch reactor is expressed by,
In rearranged form after integration,
From linearized form of Monod’s equation,
1 X
X
dC
C dt
µ =
0
ln X
X
C
C
tµ=
1 1 1S
m S m
K
Cµ µ µ
= ∗ +
11. MATHEMATICAL MODELING(GOVERNING
EQUATIONS)
Biomass balance(X) equations,
Substrate(S) balance equation,
Volatile fatty acid (V) balance equation,
1
1 1
X
X
dC
C
dt
µ=
2
2 2
X
X
dC
C
dt
µ=
( )
1
1 1
1S
S in S
X
S
X
dC
D C C
dt Y
Cµ= − −
( )
2
1 21 2
1
V
In V
X
X XC
dV
D V V
d
CY
t Y
µ µ= − + −
12. Hydrogen balance equation,
Carbon-dioxide balance equation,
Here D < min (µmax 1
,µmax 2
) to prevent washout of the biomass.
2 2
1 2
1 2
2
1 2H
X X
XHXC
dH
Y Y
dt
Cµ µ= +
2 2
1 2
2
1 1 2 2CO O
X X
XCXC
dCO
Y Y
dt
Cµ µ= +
Argon balance equation,
2 2dH dCOdAr
dt dt dt
= − −
13. RESULTS AND DISCUSSION
Batch study of microorganism 1 (X1) – C. acetobutylicum
Growth profile of C.acetobutylicum displays a typical sigmoidal curve. Growth curve figure shows Lag phase starts from 0 to 4 hours,
exponential phase ranges from 4 to 16 hours and stationary phase starts after 16 hours. After 16 hours biomass concentration remains
constant, growth stabilizes and stationary phase occurs.
14. Growth profile of Rhodopseudomonas sp. displays a typical sigmoidal curve. Growth curve figure shows Lag phase starts from 0 to 10
hours, exponential phase ranges from 10 to 38 hours and stationary phase starts after 38 hours. After 38 hours biomass concentration
remains constant, growth stabilizes and stationary phase occurs.
Batch study of microorganism 2 (X2) – Rhodopseudomonas sp.
15. Calculation of µmax and KS of C. acetobutylicum and Rhodopseudomonas sp.
KINETIC PARAMETERS C. acetobutylicum Rhodopseudomonas sp.
µmax µm1 = 0.519 h-1
µm2 = 0.07 h-1
KS KS1 = 1.76 g/l KS2 = 1.86 g/l
Plot of 1/µ vs 1/ (C. acetobutylicum ) Plot of 1/µ vs 1/ (Rhodopseudomonas sp. )sC sC
16. RANGE FOR INPUT AND OUTPUT PARAMETERS USED IN BACK PROPAGATION NEURAL
NETWORK MODEL (BPNN) MODEL
PARAMETERS MINIMUM MAXIMUM UNIT
Photo:dark bacteria 2 6
Base dose concentration 50 135 mM
Inlet substrate
concentration
10 20 g/L
Hydrogen yield 6.7 11.8 mL/g glucose
ANN diagram(3-6-1-1)
17. Optimization of Input Parameter value for batch co-culture studies ( using Artificial
Neural Network and Genetic Algorithm)
Run Initial ratio
photo:dark
bacteria
Base dose
concentration(mM)
Inlet substrate
concentration(g/L)
Hydrogen yield(mL/g glucose)
Experimental Predicted by ANN
1 2 85 10 7.1 7.2
2 2 85 20 6.8 6.9
3 2 130 15 6.9 6.912
4 2 50 15 6.7 6.98
5 6 50 10 8.5 8.552
6 6 50 20 8.7 8.89
7 6 85 15 11.8 11.1
8 6 85 20 11 10.0904
9 6 130 10 9.1 9.9871
10 6 130 20 10 10.0736
11 10 50 15 6.8 6.9
12 10 85 20 9.8 10.34
13 10 85 10 9.4 9.45
Root Mean Square Error (RMSE) = 0.44
Standard error of prediction (SEP) = 5.11%
18. Run Initial ratio
photo:dark
bacteria
Base dose
concentration(mM)
Inlet substrate
concentration(g/L)
Hydrogen yield( mL/g
glucose)
Experimental Predicted
by ANN
1 6 120 8 8.8 8.98
2 6 115 12 9.9 10.16
Experimental design for testing the ANN model
Root Mean Square Error (RMSE) = 0.223
Standard error of prediction (SEP) = 2.39%
19. Weight matrix 1 Bias matrix 1 Weight matrix 2 Bias matrix 2
[2.0372 3.163 -3.4253;
-2.9285 1.7327 -3.7827;
2.1602 3.9422 2.3833;
-3.6083 3.5871 0.00020587;
3.4457 -3.7164 -0.44995;
-0.065737 2.2425 4.5666]
[-5.0879;
3.0528;
-1.0176;
-1.0176;
3.0528;
-5.0879]
[0.51641 0.55307 1.6895
1.4359 0.36062 -1.4904]
[-1.5326]
WEIGHT AND BIAS MATRICES
ANN details
Data division Random(dividerand)
Train trainscg
Performance rmse
Number of neurons 6
Transfer function logsig
20. Validation of experimental with ANN predicted data
Correlation coefficient between the experimental data and data predicted by ANN was found to be 0.965
21. Ratio photo:dark bacteria
Base dose
concentration(mM)
Inlet substrate
concentration(g/L)
Hydrogen yield( mL/g
glucose)
Experimental Predicted
by GA
6 117 10 10.1 10.2402
Optimization of ANN data using Genetic Algorithm(GA)
Solver Genetic Algorithm
Population type Double vector
Creation function Constraint dependent
Fitness scaling(scaling function) Rank
Selection function Stochastic uniform
Reproduction(Crossover fraction) 0.8
Mutation function Constraint dependent
Crossover function Scattered
Stopping criteria(Generations) 1000
22. Experimental Result : Continuous flat plate photobioreactor(PBR)
Dilution rate(hr-1
) Ratio
photo: dark
bacteria
Base dose
concentration(mM)
Inlet substrate
concentration(g/L)
Hydrogen yield
( mL/g glucose)
0.06 6 117 10 9.5
The dilution rate was set such that it was less than maximum specific growth rate of both the microorganisms to
prevent the washout condition.
25. CONCLUSION
• Both the dark and photo bacteria followed Monod growth model when grown separately in common nutrient
medium.
• C. acetobutylicum and Rhodopseudomonas sp bacteria can produce hydrogen in a co-culture system under
strict anaerobic conditions and under illumination.
• In the small scale batch co-culture system, it was found that maintenance of neutral pH of the medium is an
important factor as there is reduction in pH after a span of time.
• The optimum base dose concentration was found to be 117 mM, ratio of photofermentative to dark fermentative
bacteria 6:1 and base dose concentration to be 10 g/L.
• ANN along with GA can be used to predict and optimize hydrogen yield.
• Flat plate photobioreactor was successfully setup and operated in continuous mode for hydrogen gas
production.