Wide area blackouts can occur due to natural causes, technical failures, human error, or terrorism. Major blackouts in recent history have affected millions of people across multiple countries. Key factors that increase blackout risk include the liberalization of electricity markets and increased use of renewable energy sources. An optimization model is presented that can divide a power network into sections to isolate potential faults and maximize power delivered during a blackout situation. The model has been tested on power grid test cases of varying sizes.

1. Wide Area Blackouts: Causes and
Prevention
Waqquas A. Bukhsh

2. Outline
• Introduction to Power Systems
• Why Blackouts happen?
• Case Study of two major blackouts
• Optimization model to avoid large area
blackouts
• Conclusion

3. Introduction to Power Systems
• Power Systems
– A system of high voltage cables by which electricity is
distributed in a region.
– The flow of power in the cables is governed by Kirchhoff's
laws.
– Electricity can not be stored, so demand has to be
matched in real time.
– Frequency is global indicator of demand-generation
balance

4. Flow of Power from one point to other
• Power does not just flow on the
path of least resistance
• Power flows allocate themselves
along all possible paths in inverse
proportion to the resistance
N-1 Criterion
• It is basic principle in power
systems operation
• It says that any probable single
event leading to a loss of power
systems elements should not
endanger the security of whole
system
GB Transmission Network. Photo credit National Grid

5. Effect of football match on Electricity Demand
Source: National Grid

6. Effect of football match on Electricity Demand
Source: National Grid
640,000 Kettles
1,120,000 Kettles

7. Source: National Grid
Effect of football match on Electricity Demand

8. Causes of Blackouts
• Natural causes
– Lightening, rain, snow, wind storm
• Technical Failures
– Transformer faults, Short circuits
• Human error
– Error of judgment, communication errors between
operators
• Terrorism
– Bombing, cyber attacks, HEMP and IEMI

9. Blackouts Around the world

10. Risk of Blackouts
• Liberalization and privatization
– These reforms have reduced security margin.
• Renewable Energy
– A downside of renewable energy is volatile supply of
power
Source: IEEE Power & Energy Magazine

11. Worlds Worst Blackouts
Source: National Geographic
Brazil and Paraguay, 2009
Northeastern US and Canada, 1965
India, 2012
New York, 1965 Europe, 2006New York, 1977
Hunan Province, China, 2008 US and Canada, 2003

12. Blackout People Affected
(Millions)
Countries Affected Date
Indian Blackout 670 India 30-31 July, 2012
Indonesian Blackout 100 Indonesia 18 August, 2005
Brazilian Blackout 97 Brazil 11 March, 1999
Brazil-Prague
Blackout
87 Brazil, Prague 10-11 November,
2009
Northeastern
Blackout
55 US, Canada 14-15 August, 2003
Italian Blackout 55 Italy, Switzerland,
Austria, Slovenia,
Croatia
28 September, 2003
Northeast Blackout 30 US, Canada 09 November, 1965
Worlds Worst Blackouts

13. European Blackout, 2006
• November 04, 2006
• 15 million households lost
power

14. European Blackout, 2006
Chain of Events
• Event 1: Shut down of Conneforde-Diele Line
– On September 18, 2006, the shipyard Meyerwerft requested E-On to shutdown a high voltage cable over
river Ems on November 05, 2006 at 01:00 am
• Event 2: Wrong Calculations
– Meyerwerft requested E-On on November 03, 2006 to move forward the switch-off by three hours at 10:00
pm. At 09:29 pm on November 04, 2006 E-On performed a simulation for scheduled switching off of the
line, based on current state. A separate N-1 simulation was not performed.
• Event 3: Switching off Conneforde-Diele Line
– At 09:38 pm the line was opened. The power flows redistributed on other lines. E-On noticed loading of
other lines at 10:01 pm. At 10:07 pm the safety limit of Landesbergen-Wehrendorf exceeds.
• Event 4: Tripping of Landesbergen-Wehrendorf Line
– At 10:10:11 pm corrected measures were deployed. Landesbergen-Wehrendorf line tripped two seconds
later on 10:10:13
• Event 5: Blackout
– The cascading tripping of lines followed and this resulted in wide area blackout.

15. Northeastern Blackout, 2003
• August 14, 2003
– Temperature 88 degrees
– High demand was only an issue not a cause

16. • Event 1: Failed Contingency Analysis
– Software runs at automatic 5-minutes update. On that day an
analyst switched off the automatic update feature to trouble
shoot an unusual output. After trouble shooting analyst forgot
to restore the feature.
– The oversight went unnoticed for two hours and full automatic
operation was not restored until four hours later.
– Consequence: Loss of situational awareness
• Event 2: Tripping of Eastlake 5
– High demand of power cause the generator Eastlake 5 to trip.
Loss of Eastlake 5 caused more power to flow into the
Cleveland area, and loaded the transmission lines. However,
the grid remained in stable state.
– Consequence: The transmission lines get overloaded because of
the imports from other areas.
Northeastern Blackout, 2003

17. • Event 3: Tripping of lines from tree contact
– Roughly 90 minutes later after event 2, contact with
trees tripped 3 lines in less then 45 minutes.
– Consequence: The power carried by these lines had to
be carried by other lines near capacity.
• Event 4: Tripping of Sammis-Star Line
– Less than 30 minutes later after event 3, the Sammis-Star line
tripped. This line was carrying a load over 120 percent of its
rating.
– Consequence: The outage of this line resulted in cascading
outages of hundreds of lines in only minutes
Northeastern Blackout, 2003
Source: US/Canada Power System Outage Force

18. Some impacts of Northeastern
Blackout
• Daimler Chrysler
– Lost production at 14 of its 31 plants. 6 of those were assembly plants with paint shops. The
company reported that, in total, 10,000 vehicles were moving through the paint shop at the
time of outage had to be scrapped
• Ford Motor Company
– At Ford’s casting plant, the outage caused molten metal to cool and solidify inside ones of
plant’s furnace. The company reported a week would be required to clean the furnace
• Airports
– Airports were closed in Toronto, Newark, New York, Montreal, Islip, Cleveland, Erie and
Hamilton. Together they cancelled over 1,000 flights
• New York City
– New York City’s mayor estimated that the city would pay almost USD 10 m in overtime related
to the outage
• Marathon Oil Corporation’s
– The blackout was responsible for triggering emergency shutdown procedures at Marathon
refinery. During those procedures a carbon monoxide boiler failed to shut down properly,
causing a small explosion. As a precautionary measure police evacuated one-mile strip around
the complex.

19. Optimization Model to avoid blackouts
• Motivation and Assumptions
– We assume that there are parts of the network which
are suspected of having fault.
– Our aim is to split the network into disconnected
sections so that possible faults are all in one section
• Objective
– Maximize the amount of load delivered to customers

20. • Sections
– 𝑆1: Section 1 (Healthy section) and 𝑆0:section 2 (Sick section)
• Load Shedding
– 𝑝 𝑑
𝐷
= α 𝑑 𝑃𝑑
𝐷
, where 𝑃𝑑
𝐷
is the real demand, 0 ≤ α 𝑑 ≤ 1
• Sectioning Constraints
– 𝜌𝑙 ≤ 1 + 𝛾 𝐹 𝑙
− 𝛾 𝑇 𝑙
∀ 𝑙 ∈ 𝐿𝐿0
– 𝜌𝑙 ≤ 1 − 𝛾 𝐹 𝑙
+ 𝛾 𝑇 𝑙
∀ 𝑙 ∈ 𝐿𝐿0
– 𝜌𝑙 ≤ 1 − 𝛾 𝐹 𝑙
∀ 𝑙 ∈ 𝐿0
– 𝜌𝑙 ≤ 1 − 𝛾 𝑇 𝑙
∀ 𝑙 ∈ 𝐿0
– 𝛾 𝑏 = 0 ∀ 𝑏 ∈ 𝐵0
– 𝛾 𝑏 = 1 ∀ 𝑏 ∈ 𝐵1
• DC Line flow equations
– We use small angle approximation of Kirchhoff’s voltage laws (KVL). The implementation of KVL is given by following equations:
– 𝑝𝑙
𝐿
= −𝐵𝑙
𝐿
𝛿 𝐹𝑙
− 𝛿 𝑇 𝑙
– − 1 − 𝜌 𝑃𝑙
𝐿−
≤ 𝑝 − 𝑝𝑙
𝐿
≤ 1 − 𝜌 𝑃𝑙
𝐿+
– 𝜌𝑃𝑙
𝐿−
≤ 𝑝 ≤ 𝜌𝑃𝑙
𝐿+

21. Numerical Result
• This model has been tested on test cases ranging from 14 nodes to
2746 nodes.
• Here an islanding solution for IEEE 14 bus test case. In this system
we assume that bus 2 is uncertain. Red dotted line shows the
boundary of section 0 determined by our islanding model.

22. Conclusions
• Blackouts are rare events, but they happen
with considerable damage
• Understanding causes and consequences of
blackouts can provide insights into mitigation
• Mathematical modelling can help to suggest
techniques which can prevent paralyzing
cascading failures in future blackouts

23. Ken Andreas Paul Naiyuan
Waqquas Tim Ian
Edinburgh Energy and Optimization Group
http://www.maths.ed.ac.uk/optenergy/