47
L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema I. Nikolic SPM 4530 25 March 2013 Evolving climate change resilient electricity infrastructures Modeling electricity network evolution

L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Embed Size (px)

Citation preview

Page 1: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

L. Andrew BollingerPhD studentSection Energy & IndustryFaculty of Technology, Policy & ManagementTU Delft

Supervisors:M.P.C. WeijnenG.P.J. DijkemaI. Nikolic

SPM 453025 March 2013

Evolving climate change resilient electricity infrastructuresModeling electricity network evolution

Page 2: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

PART 1 The Problem

Power outages as a result of Hurricane Sandy

Page 3: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Source: Renewables International

Minutes

Reliability of the Dutch electricity infrastructure

Average interruption time per customer per year (2007)

Page 4: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

man

ufac

ture

r

grid

des

ign

insta

llatio

n

oper

atio

n

exca

vatio

n

subs

iden

ce

moi

sture

obse

lesc

ence

/ wea

r

weath

er

secu

rity

othe

r0

10

20

30

40

50

60

70

2007

2006

2005

2004

2003

Causes of power failures in the Dutch high-voltage grid

Source: EnergieNed

Reliability of the Dutch electricity infrastructurePe

rcen

t

Page 5: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

(1) De Groot, 2006(2) Wilbanks, et al, 2008(3) Rothstein and Halbig, 2010(4) Bresser, et al, 2005

The (anticipated) impacts of climate change

Page 6: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Climate change and energy infrastructures

(1) De Groot, 2006(2) Wilbanks, et al, 2008(3) Rothstein and Halbig, 2010(4) Bresser, et al, 2005

The (anticipated) impacts of climate change on energy infrastructures

Page 7: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

• Thermal power plants: Reduced output due to cooling water shortages or restrictions

• Thermal power plants: Reduced generation efficiency

• Hydroelectric plants: Reduced resource availability

• Increased A/C and refrigeration demand• Increased market penetration of A/C• Power lines and cables: Increased resistance• Overhead power lines: Increased line sag and

increased risk of flashover• Underground cables: Increased risk of failure

due to soil movement

Reduced generation capacity

Immediate increased load demand

Long-term increase in peak summertime load demand

Reduced network capacity

Increased network losses

Increased potential for network disruption

COMPONENT IMPACTS NETWORK IMPACTS

Potential impacts of a heat wave on electricity systems

Page 8: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

The electricity infrastructure is a network

Page 9: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Research question & approach

Thesis: If we want "climate proof" infrastructures, we have to understand how changes in weather conditions may affect the performance of the electricity network as a whole, not just its individual components.

Modeling frameworkSimulation model 1

Infrastructure performance

Simulation model 2Infrastructure evolution

Extreme events

Component impacts

Network impacts

Power grid investments

Generation investments

Research question: How can we effectively support the resilience of the Dutch electricity infrastructure to climate change?

Agent-based model

Page 10: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

PART 2 Modeling electricity transmission network evolution

Image source: TenneT TSO

The Dutch electricity transmission network

Page 11: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Research (sub)question: What are the possible impacts of various climate change mitigation policies on the structure and properties of the Dutch electricity transmission network?

Research question and approach

Approach – 2 stages:

1. Exploratory model – How can we address this question using ABM?

2. Case model – More extensive model (calibrated with real data) used to directly address the research question.

Page 12: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

System identification and decomposition

What are the relevant components and how do they relate to one another?

Page 13: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

System identification and decomposition

Page 14: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Exploratory model – agents and infrastructure components

Transmission system operator (TSO)

Power producer

invests in

substations

power lines

transformers

invests in generators

distribution grids

large loads

manually determined by the user

AGENTS INFRASTRUCTURE COMPONENTS

Page 15: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Model setup - decision rules

A TSO agent must…

1. ENSURE CONNECTION: accept all applications for connections to the transmission grid, and construct connections to the respective component(s).

2. ENSURE FUNCTIONALITY: remove or replace grid components that have reached the end of their lifetime.

3. ENSURE SUFFICIENT CAPACITY: ensure that the capacity of lines is sufficient to satisfy demand under peak conditions.

4. ENSURE REDUNDANCY: ensure that a given fraction of components are embedded in loop structures.

5. ENSURE EFFICIENCY:

• implement all investments in the least cost manner.

• link substations exceeding a given supply/demand threshold to the EHV grid

6. LIMIT EXPENDITURES: maintain annual expenditures below a certain (user-set) level.

Page 16: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Model setup - decision rules

A power producer agent must…

1. ENSURE SUFFICIENT CAPACITY: invest in a new generator if his projections indicate a deficit of generation capacity within his planning horizon.

2. MINIMIZE VARIABLE COSTS: choose the technology with the least cost per MWh when investing in a new generator.

3. FIND SUITABLE LOCATIONS: locate a new generator on a parcel of land with suitable land-use characteristics.

Page 17: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Random landscape consisting of 100 unconnected distribution grids (green circles)

Load centers

Land values

Model setup - environmentDistribution grids

Keep in mind…

1. This is just a random starting point chosen for the sake of simplicity.

2. The quantity and configuration of distribution grids, load centers and land values can be changed to reflect different scenarios.

3. We can also start with an existing transmission grid and explore how the system develops further under different scenarios.

Page 18: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Software implementation

Octaveconnectextension

(Power flow analysis software)

Page 19: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Simulation – what happens when we press “go”?

During the course of a simulation…

1. The demand of distribution grids grows/shrinks at user-defined rates.

2. Large loads are constructed/decommissioned at a user-defined rate.

3. Power producers and the grid operator act according to their defined decision rules.

Page 20: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Simulation – what happens when we press “go”?

1 2

3 4 75 years

0 years

blue lines 150kV (HV) lines

red lines380kV (EHV) lines

gray lines under construction

line widthline capacity

line intersectionssubstations/transformers

green circles distribution grids

blue circleslarge generators

brown circleslarge loads

Page 21: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Results for the default case

Summary of metric values over 100 runs at the default parameter settings

3 examples of an emergent network after 75 years

Page 22: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiments – Parameters and metrics

Metrics tracked during experimentation

Parameters varied during experimentation

Page 23: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 1 – Varying the TSO’s redundancy requirement (looped percentage)

Low redundancy requirement (looped percentage)

High redundancy requirement (looped percentage)

Page 24: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 1 – Varying the TSO’s redundancy requirement (looped percentage)

Page 25: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 2 – Varying the demand growth rate

Low demand growth High demand growth

Page 26: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 2 – Varying the demand growth rate

Page 27: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 3 – Varying the cost of distributed generation

High cost of distributed generation Low cost of distributed generation

Page 28: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 3 – Varying the cost of distributed generation

Page 29: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 4 – Varying the TSO’s annual expenditures cap

Low expenditures cap High expenditures cap

Page 30: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Experiment 4 – Varying the TSO’s annual expenditures cap

Page 31: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Case model

Page 32: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Power plants Power grid Electricity demand

Case model - Infrastructure data

Page 33: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Infrastructure evolution model

Infrastructure configuration• Locations and properties

of generators• Locations and properties

of grid components• Development of demand

Infrastructure data

TSO agent decision rules

Power producer agent decision rules

Case model

Page 34: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

2013 2023 2033

Model 2 – Preliminary results

Page 35: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Simulation model 1Infrastructure performance

Simulation model 2Infrastructure evolution

Policy scenariosCl

imat

e sc

enar

ios

Extreme events

Component impacts

Network impacts

Power grid investments

Generation investments

Future work

Test different policy and climate scenarios -> Identify robust policy options for supporting infrastructure resilience.

Page 36: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Contact:

L. Andrew BollingerDelft University of TechnologyFaculty of Technology, Policy and ManagementEmail: [email protected]

Page 37: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Simulation – preliminary results under different scenarios

Default case

• 126 substations• 146 lines• 21 loops• mean degree: 2.87

High demand case

• 177 substations• 199 lines• 24 loops• mean degree: 3.366

Page 38: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Simulation – preliminary results under different scenarios

Low cost of distr. gen.• 111 substations• 124 lines• 15 loops• mean degree: 2.48

Low expenditures case

• 93 substations• 92 lines• 0 loops• mean degree: 1.98

Page 39: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Simulation – growing a transmission infrastructureMetrics

Page 40: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Exploratory model – an initial attempt

Page 41: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Research (sub)question: How can various carbon taxation schemes and RES support mechanisms be expected to affect the structure and properties of the Dutch electricity transmission network?

Problem owner: The Dutch transmission system operator

Scope:• The Netherlands• The electricity transmission network

Problem formulation and actor identification

Page 42: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Exploratory model – initial attempts

The challenge:

• A set of electricity consumers and producers are distributed randomly in a landscape.

• Each piece of the landscape is characterized by a value representing the feasibility/efficiency of putting a transmission line across it.

The goal:

• Link consumers to producers in an efficient way.

Page 43: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

EACH TIME STEP:1. Calculate power flows through each line2. Remove the link with the least power flowing through it

REPEAT UNTIL removing the next link will disrupt supply to the consumer

Exploratory model – initial attempts

Page 44: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Exploratory model – initial attempts

Limitations:

1. Doesn’t capture growth & evolution

2. Only bottom-up

3. Computationally expensive

Page 45: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Approach – 3 cycles

Cycle 1 – Exploratory model:• What are the relevant components and relationships? • Who are my agents? How do they interact? • Which software platform should I use? • Get feedback from the problem owner.• Go back to the system decomposition.

Cycle 2 – Generic model• Elaborate the decision procedures.• Implement the model based on these decision procedures.• Get feedback from the problem owner.

Cycle 3 – Case model• Calibrate the model with real-world data.• Improve the decision procedures, as necessary.• Perform experiments and address the research question.

Page 46: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Results for the default case

Total path length of the Dutch transmission grid (km) Degree distribution Of the Dutch transmission grid

Page 47: L. Andrew Bollinger PhD student Section Energy & Industry Faculty of Technology, Policy & Management TU Delft Supervisors: M.P.C. Weijnen G.P.J. Dijkema

Results for the default case