Optimum biorefinery pathways

selection using MILP with

Integer-Cuts constraint method

Student:

Stefano Maronese Supervisors:

Andrea Lazzaretto

François Maréchal

Adriano Viana Ensinas

Project outline

1. Study of Biorefinery technologies

2. Create a superstructure of different pathways to convert

biomass into useful products

3. Define a methodology to select and rank different technologies

for converting biomass according to the objective function.

A rank enlarge the view and help in the decision process

Method robust and fast → MILP.

4. Apply the method to a case study: evaluate the best

technologies to convert wooden biomass (Switzerland)

Objective: systematic generation of a rank of pathways creating a

light and quick tool.

Tool: software developed @LENI and applied to the OSMOSE

Wood2CHem Platform, a joint project between SNSF, EPF Lausanne

and ETH Zurich.

The biorefinery concept

Biorefinery is the sustainable processing of biomass into a

spectrum of marketable products (food, feed, materials,

chemicals) and energy (fuels, power, heat)“ (IEA 2009)

Biorefineries are designed to be sustainable in the whole chain,

thus maximizing the efficiency and minimizing the waste.

The OSMOSE Wood2CHem Platform is a tool for joining numerous

models to create complex energy systems.

Elements of the structure:

Units

Represent the conversion processes

Layers

Mass and energy balance nodes

Streams

Connect units and layers

Type of unit inside the platform:

Process unit (technologies)

Resource unit (input)

Service unit (output)

The tool: OSMOSE Wood2CHem Platform

The tool: OSMOSE Wood2CHem PlatformCreating the superstructure

Steps to build a process unit for the Wood2CHem Platform:

1. Create flow-sheet model of the process (or take it from literature)

2. Set the boundary of the system and analyse the cross-boundary flows

3. Extract input/output data (mass, energy, economics)

4. Create the black box model to be plugged into the platform

ElectricityInvestment

Maintenance

Labour

Acknowledgement: FICFB gasifier and SNG synthesis (PhD thesis Gassner 2009)

The tool: Wood2CHem Platform models

Wood SNG

MILP problem statement

Objective function:

Subject to:

With:

Parameters:

• Efficiency (supposed constant)

• Investment and operating cost (supposed linear)

• Prices of resources and products

• Stream reference value es+ (scaled by fs)

Variables:

• multiplication factor

(real)

• unit use

(integer)

Existence of subsystem

Mass/energy balance at layers

positive (cost in

resource unit) or

negative

(revenue in

service unit)

Cost function linearized in the range [fmin, fmax]

Generate multiple pathways: Integer-Cuts Constraint

Integer Cut (original):

Single condition

‼ Non linear function

Integer Cut (Fazlollahi et al. 2012)

Linear

‼ Number of constraints increases at each run

Generate multiple pathways: computational issues

Definition of solution used to analyze the output of the Platform:

Simple solution: only one process utility;

Complex solution: two or more processes of different kind in parallel

Meaningless solution: two or more process of the same kind in parallel

(same kind: same input and output)

Fake solution: size of the unit is too small (or at least zero). Most

outstanding example of fake solution: combination of unit use=1 and unit

mult=0

Because of fake solutions, the Platform is slow. The solver analyzes all the

combinations of integer that gives the optimal solution before moving to

the sub-optimal one.

Unit 1

Unit 2

Unit 3

Unit 1 Unit 2 Unit 3

Solution f y f y f y1 S1 1 S2 1 0 0

2 S1 1 S2 1 0 1

3 S1 1 0 0 S3 1

Fake solution:

unit 3 is active

but size is zero

Generate multiple pathways:

Adding constraints to accelerate the Platform

Method to speed up needed → Additional constraints

Set a minimum size of each process:

Fake solution avoided

‼ Do not reduce the number of runs

Epsilon constraint:

Reduce number of runs

‼ Not reliable (loss of solutions!)

Sum of unit use:

Reduce number of runs

‼ Np must be chosen accordingly

Tests proved that ICC + fmin>0 + Σy≤Np are the best set of

constraints → rank of solutions in the shortest time

Case Study

Technologies and

superstructure for

converting wood into

energy products:

SNG

Methanol

DME

FT

Several configuration

for each technology.

Case Study: input dataThe superstructure is made of three sub-models:

1. Wood model → assess wood cost (harvesting + transport) and availability (Resource unit);

2. Techno-economic models of the processes → assessing efficiency and cost of each technology (Process Units)

3. Market condition → assessing prices for the products (Service units)

Parameter MeOH-a MeOH-b MeOH-c MeOH-d FT-EF_ind FT-EF_dir FT-a FT-b FT-c DME SNG-a SNG-b

Technology FICFB FICFB CFB CFB EF EF CFB CFB FICFB FICFB FICFB CFB

Output [kW/MW] 570 570 318 318 637 458 303 352 601 561 693 750

Electricity [kW/MW] -85 -59 -18 35 -14 55 155 126 -4 -48 37 26

20 MW biomass plant

Investment [M€] 27 28 15 15 7 10 11 12 19 23 24 16.6

400 MW biomass plant 100 MW biomass plant

Investment [M€] 365 363 156 136 88 101 115 133 295 311 103 51

Case Study: 20 MW input biomass

Top 10 solutions

FT-a + Eout

FT-b + Eout

FT-EFdir + Eout

SNG-a + Eout

Ein + FT-EFind

MeOH-d + Eout

FT-EFind + MeOH-d

FT-EFind + SNG-a

FT-a + MeOH-b

FT-EFind + SNG-b

Case Study: 200 MW input biomass

Top 10 solutions

FT-a + Eout

FT-b + Eout

SNG-a + Eout

MeOH-d + Eout

FT-EFdir + Eout

FT-a + MeOH-b

MeOH-b + SNG-a

FT-b + MeOH-b

MeOH-a + SNG-a

Ein + FT-EFind

Conclusions I

Methodology to generate pathways to convert

wooden biomass according to the objective function:

Rank of pathways obtained;

Optimal set of constraint developed (form -40% to -45%

computational time);

No loss of solutions;

Light and robust tool to analyse different concept and

technologies;

First analysis performed according to simple models.

Conclusions II

Conclusions from the case study

Trend in technologies is defined:

For Switzerland Fischer-Tropsch is the best conversion

process (but FT fuel needs further refinement)

Then comes SNG and MeOH

DME is the worst technology (any size)

Break even price of wood is from 30% (20 MW) 10% (200

MW) lower than wood cost

Break even price for the fuel produced is from 50% (20

MW) to 20% (200 MW) higher than the fossil counterparts.

Conclusions III – Perspectives & future research

The Wood2CHem Platform, as it was developed, opens a broad

variety of future developments:

Short term: expand the platform adding more processes in

series → compare different products and add recirculating

flows

Mid term: More detailed approach → Add energy integration

Long term: uncertainty in cost estimation → Depict scenarios

to assess most promising technologies that worth developing