II.

ANALYSIS OF THE BIOETHANOL PROCESS CONTROLLER

A. Process Description

The figure below illustrates ethanol production using a yeast fermenter with

constant substrate inflow and outflow. The reaction volume is maintained at constant

levels.

Initially a biomass solution, which is a suspension of yeast, is fed to the

fermenter. Since the process is carried out in the presence of oxygen, a baker’s yeast

(saccharomyces cerevisiae) is used. There is a continuous supply of substrate, a

solution of glucose, which feeds the microorganisms, to the fermenter. For the

formation of coenzymes, inorganic salts are added together with the biomass. The

equilibrium concentration of oxygen in the liquid phase is influenced by inorganic

salts because of the salting-out effect. Yeast hydrolyses glucose into ethanol and

carbon dioxide. The contents of the reactor are biomass, substrate, and product. The

product is continuously withdrawn along with biomass and substrate.

Figure 2.1 Fermenter Process

 B. Mathematical Model

The mathematical model used is the Monod equation derived from Michaelis-

Menten kinetics for the growth of microorganism which has been proposed and

described by Aiba et al. and Z.K. Nagy, respectively.

Temperature of fermentation should be considered when modeling the

behavior of the yeast and the development of ethanol. At higher temperatures, yeast

tends to grow faster than low temperatures and generate ethanol by fermentation. The

same concentration of alcohol promotes cell death, which at higher temperatures will

also be faster, resulting in a faster decay of living cells. The optimal method is

therefore executed to achieve an optimum temperature point for high production of

ethanol and steady growth rate of cell mass.

For a temperature driven system:

E μi
μi= μ i0 exp (− )
RT , i = G, M, N (1)

E Ki
K i= K i 0 exp (− )
RT , i = G, M, N (2)

EK ' i
K 'i= K ' i0 exp(− )
RT , i = G, M, N (3)

The dynamic equations describing the cell growth, product formation and

substrate utilization were developed by applying the principle of conservation of mass,

resulting in the systems of first-order ordinary differential equations. The change in

substrate concentration depends on four terms: substrate assimilation into biomass

 μX qX
( ) ( )
YX , substrate assimilation into extracellular product Y P , substrate

utilization for cell growth ( GsX ) and substrate utilization for maintenance energy

( MsX ).

dX
=μX - k d X
dt (4a)

dP
=qX + M p X
dt (4b)

dS μX qX
=− - -G X -M S X
dt YX YP S (4c)

Modeling Specific Growth Rate, Cell Growth Rate, Sugar Consumption, Ethanol

Production, Temperature and Fusel Alcohol are needed for the development of the

dynamic model on MATLAB Simulink.

2.2.1 Modeling Specific Growth Rate

μG G
μ1 =
K G+ G (5)

μM M K 'G
μ2 =
K M +M K ' G +G (6)

μN N K 'G K'M
μ3 =
K N +N K ' G+G K ' M +M (7)

Where μi is the maximum specific consumption rate of the sugar i, Ki is

the saturation constant for the sugar i, and K’I, is the constant of inhibition

caused by the sugar i.

 In this batch fermentation, more than one fermentable sugar was

utilized, glucose(G), maltose(M), and maltotriose(N).

2.2.2 Sugar Consumption Model

a. Glucose

dG
=−μ1 X
dt (8)

b. Maltose

dM
=−μ 2 X
dt (9)

c. Maltotriose

dN
=−μ3 X
dt (10)

The basic consumption levels of these sugars have been represented by

equations showing a kinetic pattern of preferential use of these substrates, i.e.

the preferred sugar (G) is first used until its complete exhaustion; the second

sugar (M) of intermediate preference is consumed; and finally, the third sugar

(N), the least preferred, is consumed (Ramirez and Macciejowsli, 2007).

2.2.3 Modeling Cell Growth Rate

Consider an unstructured model for cell growth,

dX
=μ X X
dt

Where

μ X = (Y XG μ1 +Y XG μ1 +Y XM μ 2 +Y XN μ3 ) (11)
 Based on biological data µ(S), the specific growth rate is often defined by

a classical function. The specific growth rate µ(S) is related to the concentration

of a single inhibitory and growth limiting substrate represented by Haldane law

(Muthamilselvi & Karunanithi, 2010).

The availability of unsaturated fatty acids and lipids in the wort limits

the growth of yeast. The cell division cycle can only continue as long as the

requisite structural components can be produced from the media or

synthesized in the cells. The cell cannot synthesize unsaturated fatty acids in

the membrane structure in the absence of oxygen. Therefore, in the anaerobic

condition of the fermenter, the available unsaturated fatty acids in the media

are depleted (Ramirez and Macciejowsli, 2007).

The balance for the biomass growth rate yields the following equation:

KX
μ X = ( Y XG μ1 + Y XM μ2 +Y XN μ 3)
K X +( X -X 0 )2 (12)

2.2.4 Ethanol Production

E= E0 +Y EG (G0 -G )+Y EM ( M 0 -M)+Y EN ( E 0 -E ) (13)

2.2.5 Temperature Model

a. For the reactor

dT x UA (T -T f )
=− ( ΔH 1 μ1 +ΔH 2 μ2 + ΔH 3 μ3 )−
dt ρ Cp ρ CpV (14)
 b. For the Cooling Jacket

dT j FC UA(T -T j )
= (T C -T j )+
dt V j ρC CpC V j (15)

2.2.6 Fusel Alcohol Model

dIB
= Y IB μV X
dt (16)

dIA
= Y IA μ L X
dt (17)

dMB
= Y MB μ I X
dt (18)

dP
= Y PE ( μV +μ I ) X
dt (19)

Fusel Alcohols are by-products of ethanol fermentation and are involved

in the biochemical pathways leading to the formation of esters. In this paper,

Isobutyl alcohol (IB), Isoamyl alcohol (IA), 2-methyl-1-butanol (MB), and n-

Propanol are considered.

Wherein,

1 dV
μV = −
X dt (20)

1 dL
μ L= −
X dt (21)

1 dI
μI = −
X dt (22)
 C. Controller Implementation

To follow the pre-defined temperature profile in the industry, temperature

controller is needed to be design. Feedback controller is used for the controlling of

temperature in the fermentation tank. This is efficient for fermentation tank which are

cooled using a cooling jacket or cooing coil.

+ e(t) u(t) T(t)

Set Controller Plant
Point
-

Figure 2.2 Control of the Bioethanol Fermenter.

The set of nonlinear differential-algebraic equations derived in the previous

section simulates the actual behavior of the industrial yeast fermentation continuous

fermenter. The mathematical model of batch fermenter was implemented as a

MATLAB Simulink. Simulink block diagram is shown in Figure 2.3 followed by its

corresponding response.
 Figure 2.3 Simulink model of a yeast fermenter.

Figure 2.4a Yeast Concentration (g/l) Figure 2.4b Ethanol Concentration (g/l)

Figure 2.4c Glucose Concentration (g/l) Figure 2.4d Oxygen Concentration (g/l)
 Figure 2.4e Temperature of the reactor (⁰C)

Figure 2.4f Temperature of the cooling jacket (⁰C)

Figures 2.4 Response of a fermenter to step decrease in substrate at an inlet temperature

of 25–23°C.

Figures 2.4 present the response of the batch fermenter to step decrease in the

substrate inlet temperature (from 25 to 23°C). When the substrate inlet temperature

suddenly decreased from its nominal value to 23°C and settled at this new steady state,

the temperature of the reactor and the jacket initially decreased, and finally achieved

new steady states lower than their respective nominal values. The decrease in the

substrate inlet temperature caused the product (bioethanol) concentration to decrease

and reach the new steady state.

Figure 2.5a Yeast Concentration (g/l) Figure 2.5b Ethanol Concentration (g/l)

Figure 2.5c Glucose Concentration (g/l) Figure 2.5d Oxygen Concentration (g/l)

Figure 2.5e Temperature of the reactor (⁰C)

 Figure 2.5f Temperature of the cooling jacket (⁰C)

Figures 2.5 Response of a fermenter to step decrease in substrate at an inlet

concentration of 60-40 g/L.

The substrate inlet concentration (Figures 2.5) has almost negligible impact on

the reactor temperature, the jacket temperature, and the concentration of oxygen but

can significantly change the concentration of yeast, glucose (i.e., substrate) and

ethanol.

Figure 2.6a Yeast Concentration (g/l) Figure 2.6b Ethanol Concentration (g/l)
Figure 2.6c Glucose Concentration (g/l) Figure 2.6d Oxygen Concentration (g/l)

Figure 2.6e Temperature of the reactor (⁰C)

Figure 2.6f Temperature of the cooling jacket (⁰C)

 Figures 2.6 Response of fermenter to unit step increase in substrate with an inlet flow

rate of 51–52 l/h.

A unit step increase in the substrate inlet flow rate (Figures 2.6) greatly

amplified the yeast concentration, the oxygen concentration, and the product

concentration but induced the glucose concentration to decay. For the increased flow

rate of substrate, the mathematical model accurately predicts the increased product,

hence, the model is a perfect representative of an industrial yeast fermentation

bioreactor.

Figure 2.7a Yeast Concentration (g/l) Figure 2.7b Ethanol Concentration (g/l)
Figure 2.7c Glucose Concentration (g/l) Figure 2.7d Oxygen Concentration (g/l)

Figure 2.7e Temperature of the reactor (⁰C)

Figure 2.7f Temperature of the cooling jacket (⁰C)

 Figures 2.7 Response of fermenter to step increase in coolant with an inlet flow rate of

18–25 L/h.

For the abrupt increase in the coolant inlet flow rate (Figures 2.7), the jacket

temperature dropped immediately, but within a short period of time, the jacket

started gaining enough heat so that the jacket temperature was increasing slowly until

reaching the new steady state. Because of increased reactor and jacket temperature,

the product concentration increased. The substrate inlet flow rate can serve as a

manipulated variable to regulate the product concentration, and the remaining input

variables can be treated as disturbances. Since the product concentration is a

secondary measurement, the ethanol (product) concentration can be controlled

through controlling reactor temperature by manipulating the inlet flow rate of

substrate.

D. Transfer Functions and Block Diagram of the Fermenter

The different classes of unstable second order time delay (USOPDT)

processes are available in the chemical and biochemical industries and which

can be written by the model equation:

k p e−Θs
G p=
(τ 1 s+1 )(τ 2 s+1 ) (23)
 Gp indicates the process transfer function, IMC controller is denoted by

Q, C(s) denotes to the feedback PI/PID controller which proceeds to the

equation:

Q (a1 s 2 + a2 s +1)( β 2 s2 β 1 s +1)

C( s)= =
(1−QG p ) K p [( λs+1 )4 −(1− ps )( β 2 s 2 + β 1 s+1 )e−θs ] (24)

Controller C(s) obtained in the Eq. (24) is not in the standard form of the PID

controller. Hence, Eq. (24) is simplified to the standard form of the PID with a lag

filter in series. Using the Taylor series, the time delay term approximated as

e−θs =1−θs in Eq. (24) and the controller can be written in the form of

(a1 s2 +a2 s+1 )( β2 s 2 β1 s+1 )

C( s )= 4 2
K p [( λs+1) −(1− ps )( β 2 s +β 1 s+1 )(1−θs )] (25)

Further, C(s) is written into the following form

(a 1 s 2 +a 2 s+1)( β 2 s 2 β 1 s+1)
C( s )=
K p (4 λ−β 1 +θ+ p )s [ χ 1 s3 + χ 2 s2 + χ 3 s+1 ]

2 2
(a 1 s +a 2 s+ 1)( β 2 s β 1 s+ 1)
=
K p hs[ χ 1 s3 + χ 2 s 2 + χ 3 s+1 ] (26)

Where,

h=(4 λ−β 1 + θ+ p )

(27)

χ 1 =( λ 4 − p θβ1 ) (28)
3
χ 2 =( 4 λ − p θβ 1 +θβ 2 + pβ 2 ) (29)

χ 3 =(6 λ 2− β2 −θp+ pβ 1 +θβ 1 ) (30)

The denominator term of the Eq. (26), can be factorized as

 χ 1 s3 + χ 2 s 2+ χ 3 s+1=( γs+1)(a1 s 2 +a2 s+1 )

(31)

Upon equating the corresponding coefficients on both sides of Eq. (31), we get

χ 1 =γa 1 (32)

χ 2 =γa 2 +a 1

(33)

χ 3 =γ +a2 (34)

By solving Eqs. (30) and (32), the coefficients β1, β2 and γ are calculated as:

y 4 z1 − y 2 z 2
β 1=
y 1 z4− y 2 z3 (35)

β 1 y 1−z 1
β 2=
y2

(36)

χ1
γ=
a1

(37)

Where,

y 1 =a1 pθ−a
12 (38)

y 2 =a2 pθ+a 1 ( p+θ )

(39)

y 3 =a1 ( p+θ )+a 1 a2

(40)

y 4 =a1− pθ (41)

z 1=4 a1 λ3 −a2 λ 4 −a 2 ( 4 λ+θ+ p )

1 (42)
 4 2
z 2= λ +a1 a2 ( 4 λ+θ+ p )−a 1 (6 λ −6 pθ ) (50)

The final form of PID controller Gc(s) is obtained as

1 1
C( s )=K c 1+
( τi s )
+τ D s
αs+1
(51)

Where,

λ = tuning parameter

β1
Kc=
K ph (52)

τ i =β 1

(53)

β2
τ D=
β1 (54)

χ1
α =γ =
a1 (55)

Finally, the last transfer function model to describe the plant is a second-order

process with time delay and lead equation (51) to:

−Θ s
Y (s ) Ke d (τ 3 s+1)
=
U (s ) (τ 1 s +1)( τ 2 s+1) (56)

By completing the analysis of each component of the control system and

obtaining a transfer function for each. These transfer functions can now be

combined so that the overall system is represented by the block diagram in the

figure below.

Fag + e(t) u(t) T(t) γ= Tf

SF PID
Ti -

Figure 2.8 Block Diagram for a Fermenter Control System.