predictive modeling and rheological estimation of mixed...

Predictive Modeling and

Rheological Estimation of

Mixed Feedstocks

Narani A, Coffman P, Stettler A, Tachea F, Li

C, Pray T, and Tanjore D*

Process Engineer

Advanced Biofuels (and Bio-products) PDU (AB-PDU)

Lawrence Berkeley National Laboratory

Berkeley, CA

1

Acknowledgements

DOE EERE (Energy Efficiency and Renewable Energy) –

BETO (BioEnergy Technologies Office) Division and

ARRA (American Recovery and Reinvestment Act)

Idaho National Laboratory (INL) – Kevin Kenney, Vicki Thompson, Garold Greesham, Allison Ray and David Thompson

Sandia National Laboratory (SNL) – Blake Simmons and Murthy Konda

Joint BioEnergy Institute (JBEI) – Noppadon Sathitsuksanoh, Gabriella Papa and Seema Singh

Staff of Advanced Biofuels Process Demonstration Unit

2

ABPDU Process Execution & Development

Feedstock

storage

&

Size

reduction Homogenous

and/or

Heterogeneous

Pretreatment

& Catalysis

Solvent/

Catalyst

Recovery

S/L Separation

Fermentation: 0.2L to 300L

Seed Train

Process Development / Material Production

10L 50L2L

producing enzymes, biofuels, and chemicals

Sugars and

Precursors

.2L

producing long chain carbon, Sugars and Furans, Lignin, and Biofuels

Recovery (Sugars, Aldehydes,

Acids, etc.)

Recovery(Enzymes, Terpenes,

Lipids, etc.)

Centrifugation 300L

Analytics: separation sciences, calorimetry, compositional analysis, online tools (FT-NIR,

MGA, etc.), techno-economics, design of experiments

Catalysis: 10ml to 100L

Recovery and Rheology: bench-scale liquid extraction, filtration & preparative

chromatography, oscillatory stress controlled rheometer

10L 50L2L.2L S/L Separation300L

3

Providing tools to de-risk investments for sugar (bio-based) production

• What are the feedstocks?

• How much will it cost?

• What treatment and process parameters should I use?

Integrate

• Least cost modeling of the available biomass (INL)

• Predicting Deconstruction Process Parameters (LBNL)

o Identify and Optimize Traditional Pretreatment Methods

o Biomass Mixture Compositions

to Maximize Sugar Yield and Minimize Furfural Production

LBNL – INL Collaboration, Objective:

4

Predictive modeling defined

• “Predictive modeling is a mathematical algorithm that predicts target variable

from a number of factor variables.” - 56th Annual Canadian Reinsurance Conference

• Day to day example of Predictive modeling

• Has been applied in other renewable sectors: wind and solar

5

BETO funded Mixed feedstock

Collaborative Project: FY 2015-17

15

LBNL

Development of an

Applicable

Predictive Model

SNL

TEA model comparing

IL with DA

INL

Least Cost Modeling,

Feedstock blends supply and

characterization

FY 2015

• Feedstocks – Several ratio combinations

o Corn Stover (CS)

o Switchgrass (SG)

o Energy Cane (EC)

• Pretreatments – Categorical variable

o Acid (Ac) – 1% w/w H2SO4

o Alkali (Al) – 1% w/w NaOH

o Ionic Liquid (IL) - [C2 mim] [OAc]

• Temperature – Scaled variable (1 – 100%)

o Acid – 140 to 180°C

o Alkali – 55 to 120°C

o Ionic Liquid – 120 to 160°C

• Time – Scaled variable (1 – 100%)

o Acid – 5 to 60 minutes

o Alkali – 1 to 24 hours

o Ionic Liquid – 1 to 3 hours

Predictive Model Input

7

Experimental Design generated by SAS JMP®

8

Whole plots

PT Temp% 0C Time %

Min CS SG EC

1 IL 1 120 39 106.8 0 1 0

1 Ac 1 140 100 60 0.3 0.4 0.3

1 Al 1 55 100 1440 0 0 1

1 Al 1 55 39 589 0 0.5 0.5

1 Al 1 55 100 1440 0 1 0

1 IL 1 120 100 180 0 0 1

2 Ac 100 180 1 5 0 0.6 0.4

2 Ac 100 180 60 38 1 0 0

2 Al 100 120 1 60 0 1 0

2 Al 100 120 1 60 1 0 0

2 Al 100 120 1 60 0 0 1

2 IL 100 160 1 60 0 0 1

3 IL 39 135 100 180 0.5 0.5 0

3 Ac 39 155 1 5 0 0 1

3 IL 39 135 1 60 1 0 0

3 Al 39 80 1 60 0.3 0.4 0.3

3 Ac 39 155 1 5 0.4 0.6 0

3 Al 39 80 100 1440 1 0 0

4 Ac 80 172 80 49 0 1 0

4 IL 80 152 80 156 1 0 0

4 IL 80 152 80 156 0 1 0

4 IL 80 152 1 60 0.5 0 0.5

4 Al 80 107 80 1159 0.2 0.4 0.4

4 Ac 80 172 80 48.8 0 0 1

• Compositional Analysis of Feedstocks

– NREL Protocols

• Solids loading during Pretreatments for Mixed Feedstock

– 10% w/w i.e. 1 g in 10 ml total volume in 25 ml (or 21 ml) SS316 tube (or Glass) reactors

• Enzymatic Hydrolysis on unwashed solids with CTec2 and HTec2

– Variable solids loading: < 4% w/w i.e.10ml pretreated slurry was brought to 25 ml total volume

– Variable pH adjustments for each test

– 2 enzyme loadings: 40 and 10 mg protein/g glucan in untreated biomass

• Sugar and furfural analysis on Dionex HPLC

– Aminex HPX 87 H-column

Methods and Materials

9

Pretreatment and Enzymatic Hydrolysis of

Mixed Feedstocks

10

Data Interpretation through Excel

can be Limiting

11

Understanding Predictive Model through

Ternary Plots

12

Mixture profile for predetermined yields from DA treated biomass

40 mg protein/ g glucan

Predictive Modeling to define Feedstock Mixtures

13

Application of the Model

Glucose Yield did not drop drastically

with Enzyme loading

14

Mixture profile for predetermined yields from DA treated biomass

40 mg protein/ g glucan 10 mg protein/ g glucan

Building an Applicable model…

Building the model

• A spectrum of data

• Several data points

Applying the model: Questions that need to be addressed,

• Does rheology impact scale-up sugar yields at high solids loading?

• What about commercialization?

14

BETO funded Mixed feedstock

Collaborative Project: FY 2015-17

15

LBNL

Development of an

Applicable

Predictive Model

SNL

TEA model comparing

IL with DA

LBNL

Scale up (10 and 100L)

deconstruction tests

LBNL

Scale up (300 L)

hydrolysate

fermentation

INL

Least Cost Modeling,

Feedstock blends supply and

characterization

LBNL

Rheological studies to

determine behavior of

mixed feedstocks at high

solids loading

UCR/ LBNL

High throughput

deconstruction for a

robust predictive model

LBNL

Validate the predictive

model

Why Rheology?Data from a 2011 study…

17

Rheology – Science of Deformation and Flow

16

Rotational Rheometer Parallel Plate Geometry

Oscillatory Measurements

Complex Modulus (G*) =Shear Stress

Shear Strain

Elastic Modulus G′ =Stress

Strain× Cos (phase angle) Viscous Modulus G" =

Stress

Strain× Sin (phase angle)

Variation in Rheology with Increasing

Solid loading

17

Energy Cane

10% 15% 30%

Energy Cane Blended with Switchgrass (50% each w/w)

Determination of Rheological parameters

for Alkali Pretreated Biomass at 5Hz

18

Oscillatory stress sweep of alkali pretreated energy cane at various solid loadings

Determination of Rheological parameters

for Alkali Pretreated Biomass blends at 5Hz

19

Oscillatory stress sweep of alkali pretreated energy cane blended with

switchgrass (50% w/w each) at various solid loadings

Comparing the Rheology of Single and Mixed

Feedstocks at 10% (w/w)

20

Comparing the Rheology of Single and Mixed

Feedstocks at 15% (w/w)

21

Comparing the Rheology of Single and Mixed

Feedstocks at 30% (w/w)

22

Rheological Estimation of IL Pretreated

Biomass at various solids loading

23

IL Pretreatment in Globe reactor

Corn Stover solid loading (w/w)

i) 10% ii) 15% iii) 30%

• Predictive Model could identify the

• Pretreatment catalyst and treatment conditions for an “optimal”

Biomass Mixture

• Optimal Biomass Mixture for a particular Pretreatment System

• In the future

• More deconstruction tests for a robust model

• Scale-up of deconstruction test to 10 and 100L

• Continue to better understand rheological implications on

pretreatment and enzymatic hydrolysis of mixed feedstocks

• Ferment-ability tests, i.e. sugar conversion to ethanol and other

molecules

Summary

24

Suggestions

25