An Operational Model for Optimal Non‐Dispatchable Demand Response 

for Continuous Power‐intensive Processes

oror

An Operational Model for Optimal Production Scheduling under Hourly varying electricity pricesunder Hourly‐varying electricity prices

for Continuous Power‐intensive Processes

Sumit Mitra, Ignacio E. Grossmann

EWO Meeting, 03/09/2011

1

Perspective of the Future Electric Power System

Renewable StFACTS,

Main challenge: balance supply and demand This is the motivation for “smart grids”!

Renewable StorageFACTS,

$

Demand Response

EnergyStorage

HVDC IndustrialCustomer

PEV

Renewable Energy

Source: U.S.-Canada Power System Outage Task Force

High Observabilityenabled by increased communication capabilities

2

enabled by increased communication capabilities

This slide is based on a presentation slide by Prof. Gabriela Hug, Department of Electrical Engineering, Carnegie Mellon University, ESI seminar, January 13 2011

Demand Response (DR): Overview and recent work for industrial processes

B i id T b l l d d d f tBasic idea: To balance supply and demand of a power system, one can manipulate both: supply and demand  demand response (DR).Utility provider perspective: distinguish dispatchable and non‐dispatchable DR

ABB, CMU (2009, 2010)(Cement)

F BOC R t (2002)

Al O k Rid N i l L b (2009)

Former BOC, Rutgers (2002)UBuffalo (2007)(Air separation)

Goal of our current research: Develop a generic model for

Air Products, Lehigh (2010)(Air separation)

Alcoa, Oak Ridge National Lab (2009)(Aluminum)

Develop a generic model for non‐dispatchable DR

3

( p )

Other power‐intensive processes:‐ Chlor‐alkali synthesis‐ Paper production

Air Separation Plant

Industrial customer’s perspective:           Problem Statement for an Air Separation Plant

Due to Air Separation Plant

Liquid Oxygen

Due to compressors!

Electricity costs

Liquid Nitrogen

Liquid Argon

Gaseous Oxygen

Gaseous Nitrogen

Strategic DecisionsOperational DecisionsPrice forecast

Demandforecast

InvestmentCosts

Additional equipment?

Additional tanks?

Equipment upgrade?

• When to produceand amounts?• Inventory levels?

4

q p pg

Overview: Research Methodology

Step 4a (projected)Model uncertainty in electricity prices

data

stochastic

St 1 St 2

stochastic

Step 1Obtain an efficientmodel for the short‐term scheduling

Step 2Model investment 

decisions:additional equipment,upgrade equipment,

Step 3Develop a suitable

algorithm to solve thelarge‐scale optimization

problem

deterministic

add. storage tanks;(retrofit, multi‐period)

problem

decisionsoperational strategicSingle plant

Focus of this talk

Step 4b (projected) Multiple plants

decisions

solution methodmodeling

operational strategicSingle plant

5

p p

Operational Model

Objective function

Feasible region Mass balances

Logic constraints for transitions

6

Operational Model

Feasible region

ibl i j i i d

Feasible region

Mass balances

Feasible region: projection in product spaceModes: different ways of operating a plantEnergy consumption: requires correlation with production levels Mass balanceswith production levelsMass balances: differences for products with and without inventory

7

Modeling of transitions

Off Trans OnImmediate [2] After some hrs

Immediate [1]

St t di f t iti

Immediate [1]: Minimum uptime Immediate [2]: Minimum downtime Link between state and transitional variables

State diagram for transitions

Enforce minimum stay in a modeLogic constraints for transitions

Forbidden transitions

Coupling between transitions

Modes: different ways of operating a plantTransitions: between modes to enforce e.g. min. uptime/downtime

8

Result: Comparison of production schedules

Electricity consumption profiles for test case B Inventory profiles for test case B

9

Case studies: Air separation plantsProblem sizesProblem sizes

For the same demand, P2 can produce cheaper due to higher flexibility

cyclic is theharder problem 

Savings compared toconstant operation

Demand Savings CPU time 

Demand Savings CPU time 

D i d d h d t l d t hi h ti l fl ibilitComputational results

10

Decreasing demand = harder to solve due to higher operational flexibilityComputational results

Synergies for integration of operational and strategic decisions

OperationalCosts

P1

P2

Regime of feasible operationOperational 

Improvements(Scheduling)

StrategicImprovements

(Investment decisions)

( g)

Operational Flexibility

Take‐home message: Operational flexibility is key to achieve economic savings with demand response.

11

Operational flexibility is key to achieve economic savings with demand response.