Chapter 11: Introduction to

Power System Stability

 On a good day for a power system
• Things change over time
• But things change as expected:
• Load follows the forecast reasonably well
• Generators come online as expected
• Changes in network topology occur in a controlled manner
• All variables remain within expected limits
• No unexpected outages of generators or network components
• No faults
• System is in quasi-steady state
• Study its evolution using the algebraic power flow equations
• As opposed to differential equations
© 2023 Daniel Kirschen 2
 What can go wrong?
• Fault on a line, transformer, or busbar

• Generator failure

• Unexpected rapid decrease in renewable production

• Unexpected rapid increase in load

© 2023 Daniel Kirschen 3

 What usually happens after such a contingency?
• System is designed and operated with a sufficient safety margin

• System state moves to another stable operating point

• Operators take action to restore the safety margin

• System ready to handle another contingency

• (That’s why we don’t have movies about engineers…)

© 2023 Daniel Kirschen 4

 What else can happen?
• Contingency is beyond the design and operation criteria
• Severe storm causes the disconnection of multiple lines at the same time
• Contingency is compounded by a human error
• Mistake in operational planning
• Mistake in operation
• Example: cruise ship incident in Europe
• The process is usually to blame, not the individual
• Contingency is compounded by a protection failure
• Fault is not cleared in a timely manner
• More components than necessary are disconnected

© 2023 Daniel Kirschen 5

 What then?
• System may go unstable
• Protection system disconnects more equipment
• Cascading outages
• e.g., cruise ship incident
• Can lead to a partial or complete system collapse: blackout
• Partial collapse:
• The opening of circuit breakers isolates the part of the system that collapses
• If the load/generation balance can be maintained, the rest of the system, does not collapse

© 2023 Daniel Kirschen 6

 Equilibrium
• Equilibrium state:
• The state of the system won’t change except due to an external perturbation
• Solving the power flow equations
 Equilibrium state of the power system for given loads and generations
• Change in loads or generation change
 The equilibrium state of the power system changes
• Is this equilibrium stable?

© 2023 Daniel Kirschen 7

 What makes a system stable?
Potential energy

System State

x
© 2023 Daniel Kirschen
(state variable describing the system)
8
 Stable System
Potential energy

Perturbation

Return to equilibrium state

© 2023 Daniel Kirschen 9

 Unstable System

Potential energy

Equilibrium state

x
© 2023 Daniel Kirschen 10
 Unstable System

Potential energy

This equilibrium is unstable!

Perturbation

x
© 2023 Daniel Kirschen 11
 Slow Evolution of a Stable System

Potential energy

New conditions
Original conditions

x
© 2023 Daniel Kirschen 12
 Steady State Stability

Potential energy

Instability!

The system has evolved to

a state where it does not
return to its equilibrium
point after a small
perturbation

x
© 2023 Daniel Kirschen 13
 Transient Stability
Potential energy

x
© 2023 Daniel Kirschen 14
 Transient Stability
Potential energy How large a perturbation would be
Needed to make the system unstable?

Is the stability margin large enough

to withstand a large perturbation?

Stability margin
Large
perturbation

x
© 2023 Daniel Kirschen 15
 Steady state stability vs. transient stability
• Steady state stability is a function only • Transient stability depends not only
of the system and its state on the system and its state but also on
the size of the perturbation
• Any system is unstable if the
perturbation is sufficiently large…
• Must limit the analysis to “credible”
perturbations

© 2023 Daniel Kirschen 16

 Stability limits
• Steady state stability: • Transient stability:
• Will the system collapse if a line or • Will a fault kick the system into
generator is disconnected? instability?

© 2023 Daniel Kirschen 17

 PV Curves
• constant
𝑉 𝐴 =𝑉 𝐴 ∠ 𝛿 𝑉 𝐵 =𝑉 𝐵 ∠ 0 ° • Line modeled as pure reactance
𝐼
𝑃 + 𝑗𝑄
•

© 2023 Daniel Kirschen 18

 PV Curves 𝑉 𝐴 =𝑉 𝐴 ∠ 𝛿 𝑉 𝐵 =𝑉 𝐵 ∠ 0 °
𝐼
𝑃 + 𝑗𝑄

© 2023 Daniel Kirschen 19

 PV Curves – Unity power factor load (
(a.k.a. “Nose Curves”)
1.2

𝑉𝐵
and
1

0.8
𝑉 𝐴 =𝑉 𝐴 ∠ 𝛿 𝑉 𝐵 =𝑉 𝐵 ∠ 0 °
0.6
𝐼
𝑃 + 𝑗𝑄
0.4

0.2
No solution for !
0
0 0.1 0.2 0.3 0.4 0.5 0.6
© 2023 Daniel Kirschen 𝑃 20
 PV Curves – Lagging power factor load (

1.2

𝑉𝐵
and
1
pf = 1.0

0.8
𝑉 𝐴 =𝑉 𝐴 ∠ 𝛿 𝑉 𝐵 =𝑉 𝐵 ∠ 0 °
pf = 0.9 lag
0.6
𝐼
𝑃 + 𝑗𝑄
0.4

0.2
Voltage drops faster as
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Maximum is smaller
© 2023 Daniel Kirschen 𝑃 21
 PV Curves – Leading power factor load (

1.2

𝑉𝐵 pf = 0.9 lead
and
1

pf = 1.0
0.8
𝑉 𝐴 =𝑉 𝐴 ∠ 𝛿 𝑉 𝐵 =𝑉 𝐵 ∠ 0 °
pf = 0.9 lag
0.6
𝐼
𝑃 + 𝑗𝑄
0.4

0.2 Voltage initially increases as

Maximum is larger and happens
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 for a larger
© 2023 Daniel Kirschen 𝑃 22
 Maximum active load
𝑉𝐵

The two solutions coincide

© 2023 Daniel Kirschen 23

 Maximum active load

• The two solutions coincide when

© 2023 Daniel Kirschen 24

 Example 11.1
𝐴 𝐵
𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

• Maximum load that this system can deliver for load power factors of
1.0, 0.9 lag, and 0.9 lead?

© 2023 Daniel Kirschen 25

 𝐴 𝐵
Example 11.1 𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

Power factor
Unity 0 1.6667 0.7071
0.9 lag 1.0446 0.5901
0.9 lead 2.6591 0.9415

Practical maximum loading Less than the 0.95 p.u. that is

is smaller than critical load usually considered acceptable
© 2023 Daniel Kirschen 26
 𝐴 𝐵
Example 11.2 𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

• Practical loading limit?

• Unity power factor ( ):

• of critical load

© 2023 Daniel Kirschen 27

 𝐴 𝐵
Example 11.3 𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

• System operates at maximum practical limit ()

• One of the lines is disconnected
 Impedance between source and load increases

© 2023 Daniel Kirschen 28

 𝐴 𝐵
Example 11.3 𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

𝑉𝐵

Two lines
To avoid a system collapse if
One line one of the lines were to be
disconnected, the loading
must be less than the critical
value with only one line in
service

© 2023 Daniel Kirschen

𝑃 29
 𝐴 𝐵
Example 11.4 𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

• Both lines are in service

• Unity power factor using reactive power compensation

• Well below critical loading
• Reactive power compensation fails  0.9 p.f. lag

© 2023 Daniel Kirschen 30

 𝐴 𝐵
Example 11.4 𝑋=0.6 𝑝. 𝑢.
𝑃 + 𝑗𝑄
𝑉 𝐴 =1.0 𝑝. 𝑢.
𝑋=0.6 𝑝. 𝑢.

𝑉𝐵
A
Unity pf

B Unacceptably low voltage

0.9 pf lag
Dangerously close to critical
load

© 2023 Daniel Kirschen

𝑃
31
 Cascading overloads

100 MW

100 MW
A 100 MW
B

Each line is rated at 140 MVA

System A is exporting 300 MW to system B

© 2023 Daniel Kirschen 32

 Cascading overloads

100 MW

100 MW
A 100 MW
B

Each line is rated at 140 MVA

© 2023 Daniel Kirschen 33

 Cascading overloads

150 MW

150 MW
A B

• Remaining lines are overloaded

• Unless generation in system B is rapidly increased, they will trip

© 2023 Daniel Kirschen 34

 Cascading overloads

A B

• What happens if the systems separate, and no action is taken?

• Excess generation in system A  frequency increases
 generators trip on overspeed
• Deficit of generation in system B  frequency decreases
 generators trip on underfrequency
© 2023 Daniel Kirschen 35
 Example 11.5
𝐴 1 𝐵 Line Rating Impedance
𝑃𝐴
1 50 MVA
2 50 MVA
3 𝑃 𝐿=100 𝑀𝑊 3 40 MVA

• Generator A is cheaper than generator B

 Want to generate as much as possible with generator A
 Must avoid the risk of cascading overload
 Can’t rely on fast response from the generators
 How much can generator A produce?

© 2023 Daniel Kirschen 36

 Example 11.5
𝐴 𝐵 Line Rating Impedance
𝑃𝐴
1 50 MVA
2 50 MVA
𝑃 𝐿=100 𝑀𝑊 3 40 MVA

• Will the disconnection of one of the lines overload the remaining ones?
• Need to perform a contingency analysis
• How does distributes itself among the remaining lines?

© 2023 Daniel Kirschen 37

 Parallel paths – Power divider
Path A
𝑥𝐴
P 𝑃𝐴? P
𝑃 𝐵?
𝑥𝐵
Path B

Analog to current divider

𝑥𝐵 𝑥𝐴
𝑃 𝐴= 𝑃 𝑃 𝐵= 𝑃
𝑥 𝐴+ 𝑥𝐵 𝑥 𝐴+ 𝑥𝐵
© 2023 Daniel Kirschen 38
 Example 11.5
𝐴 1 𝐵 Line Rating Impedance
𝑃𝐴
1 50 MVA
2 50 MVA
𝑃2
3 𝑃 𝐿=100 𝑀𝑊 3 40 MVA
𝑃3

• Contingency 1: disconnection of line 1

• Flow through line 2:
• Flow through line 3:
• We must have
• We must also have
 less constraining

© 2023 Daniel Kirschen 39

 Example 11.5
𝐴 𝑃
1
𝐵 Line Rating Impedance
𝑃𝐴 1
1 50 MVA
2 50 MVA
3 𝑃 𝐿=100 𝑀𝑊 3 40 MVA
𝑃3

• Contingency 2: disconnection of line 2

• Similar to contingency 1

© 2023 Daniel Kirschen 40

 Example 11.5
𝐴 𝑃
1
𝐵 Line Rating Impedance
𝑃𝐴 1
1 50 MVA
2 50 MVA
𝑃2
3 𝑃 𝐿=100 𝑀𝑊 3 40 MVA

• Contingency 3: disconnection of line 3

• Flow through line 1:
• Flow through line 2:
• We must have
 less constraining

© 2023 Daniel Kirschen 41

 Example 11.5
𝐴 1 𝐵 Line Rating Impedance
𝑃𝐴
1 50 MVA
2 50 MVA
3 𝑃 𝐿=100 𝑀𝑊 3 40 MVA

• The most severe contingency is thus the disconnection of line 1 or line

2
• To avoid overloading line 3 if this contingency were to occur, we must
have
• Hence

© 2023 Daniel Kirschen 42

 European incident of 4 November 2006

© Daniel Kirschen & University of Washington

 European incident of 4 November 2006
Line disconnected to let the
ship pass safely

Other lines disconnected

© 2023 Daniel Kirschen 44

 European incident of 4 November 2006

Source: UCTE
© Daniel Kirschen & University of Washington
 Frequency in the three areas following the separation

Source: UCTE
© 2023 Daniel Kirschen 46
 Electromechanical stability
• Commonly known as transient stability
• Interactions between the mechanical and electrical systems
• Mechanical system: rotating generators providing power
• Electrical system: export power from the generator to the loads
• These systems are coupled
• How does this electromechanical system respond to a fault?

© 2023 Daniel Kirschen 47

 Mechanical model of synchronous generators

𝑇𝑚 𝑇𝑒
Prime Mover Grid
𝜔𝑚
2
𝑑 𝜃𝑚
Newton’s law of motion : 𝐽 2
=𝑇 𝑚 − 𝑇 𝑒
𝑑𝑡
angular position of the rotor
moment of inertia of prime mover and generator

© 2023 Daniel Kirschen 48

 Synchronous generator in the steady state
~ ~
𝑃𝑚 𝑃𝑒
𝑇𝑚 𝑇𝑒
Prime Mover Grid
𝜔𝑚
~ ~
𝑃 𝑚 = 𝑃 𝑒 ⇒ 𝑇 𝑚 =𝑇 𝑒
2
𝑑 𝜃𝑚
𝐽 2
=𝑇 𝑚 − 𝑇 𝑒 = 0
𝑑𝑡
𝑑 𝜃𝑚
=𝜔 𝑚=constant=𝜔𝑚 ,𝑠𝑦𝑛 Synchronous speed
𝑑𝑡
𝜃𝑚 , 𝑠𝑦𝑛 ( 𝑡 )=𝜔𝑚 , 𝑠𝑦𝑛 𝑡 +𝜃 𝑚,0
© 2023 Daniel Kirschen 49
 Deviations from synchronous speed
~ ~
𝑃𝑚 𝑃𝑒
𝑇𝑚 𝑇𝑒
Prime Mover Grid
𝜔𝑚
~ ~
𝑃𝑚 ≠ 𝑃𝑒⇒ 𝑇𝑚 ≠ 𝑇𝑒
𝜔 𝑚 ≠ constant
𝛿𝑚 ( 𝑡 )=𝜃𝑚 ( 𝑡 ) − 𝜃𝑚 , 𝑠𝑦𝑛 (𝑡 ) Angular deviation from synchronous
reference

© 2023 Daniel Kirschen 50

 Deviations from synchronous speed
~ ~
𝑃𝑚 𝑃𝑒
𝑇𝑚 𝑇𝑒
Prime Mover Grid
𝜔𝑚

𝛿𝑚 ( 𝑡 )=𝜃𝑚 ( 𝑡 ) − 𝜃𝑚 , 𝑠𝑦𝑛 (𝑡 )
𝛿𝑚 ( 𝑡 )=𝜃𝑚 ( 𝑡 ) − 𝜔𝑚 , 𝑠𝑦𝑛 𝑡 − 𝜃 𝑚 ,0
𝑑 𝛿𝑚 𝑑 𝜃𝑚
= − 𝜔𝑚 , 𝑠𝑦𝑛
𝑑𝑡 𝑑𝑡
2 2
𝑑 𝛿𝑚 𝑑 𝜃𝑚 
2
= 2
© 2023 Daniel Kirschen
𝑑𝑡 𝑑𝑡 51
 Swing equation
~ ~
𝑃𝑚 𝑃𝑒
𝑇𝑚 𝑇𝑒
Prime Mover Grid
𝜔𝑚
2
𝑑 𝛿𝑚
𝐽 2
=𝑇𝑚 −𝑇 𝑒
𝑑𝑡
~
𝑃 𝑚 =𝜔 𝑚 𝑇 𝑚
~
𝑃 𝑒 = 𝜔𝑚 𝑇 𝑒
2
𝑑 𝛿𝑚 ~ ~
𝐽 𝜔𝑚 ( 𝑡 ) 2
=𝑃𝑚 − 𝑃𝑒
𝑑𝑡
© 2023 Daniel Kirschen 52
 Swing equation
~ ~
𝑃𝑚 𝑃𝑒
𝑇𝑚 𝑇𝑒
Prime Mover Grid
𝜔𝑚
2
𝑑 𝛿𝑚 ~ ~
𝐽 𝜔𝑚 ( 𝑡 ) 2
=𝑃𝑚 − 𝑃𝑒
𝑑𝑡
2
𝐽 𝜔 𝑚 (𝑡 ) 𝑑 𝛿𝑚 : power in per unit
=𝑃 𝑚 − 𝑃 𝑒
𝑆𝑟𝑎𝑡𝑒𝑑 𝑑𝑡 2
: MVA rating of the generator

© 2023 Daniel Kirschen 53

 Introducing the inertia constant
2
𝐽 𝜔 𝑚 (𝑡 ) 𝑑 𝛿𝑚
=𝑃 𝑚 − 𝑃 𝑒
𝑆𝑟𝑎𝑡𝑒𝑑 𝑑𝑡 2

The moments of inertia of generators vary over a wide range  inconvenient

1 2 Ratio of the kinetic energy at

𝐽 𝜔𝑚 ,𝑠𝑦𝑛
Inertia constant of a generator: 𝐻=
2 synchronous speed to the MVA
𝑆𝑟𝑎𝑡𝑒𝑑 rating of the generator:
2
2 𝐻 𝜔𝑚 ( 𝑡 ) 𝑑 𝛿𝑚
2 2
= 𝑃𝑚− 𝑃 𝑒
𝜔 𝑚 , 𝑠𝑦𝑛 𝑑𝑡

© 2023 Daniel Kirschen 54

 Electrical vs. mechanical angles
One mechanical rotation of a generator with poles produces periods of the
electrical output

𝑃
𝜔 𝑠𝑦𝑛 = 𝜔𝑚 , 𝑠𝑦𝑛
2

𝑃
𝜔
Electrical ( 𝑡 )=
angles ( 𝑡 ) angles
𝜔 Mechanical
2 𝑚

𝑃
𝛿 ( 𝑡 )= 𝛿 (𝑡 )
2 𝑚

© 2023 Daniel Kirschen 55

 Swing equation in electrical angles
2
2 𝐻 𝜔𝑚 ( 𝑡 ) 𝑑 𝛿𝑚
2 2
= 𝑃𝑚− 𝑃 𝑒
𝜔 𝑚 , 𝑠𝑦𝑛 𝑑𝑡
𝑃
𝜔 𝑠𝑦𝑛 = 𝜔𝑚 , 𝑠𝑦𝑛
2 2 𝐻 𝜔 ( 𝑡 ) 𝑑2 𝛿
2 2
=𝑃 𝑚 − 𝑃 𝑒
𝑃 𝜔 𝑠𝑦𝑛 𝑑𝑡
𝜔 ( 𝑡 )= 𝜔 (𝑡)
2 𝑚

𝑃
𝛿 ( 𝑡 )= 𝛿 (𝑡 )
2 𝑚

© 2023 Daniel Kirschen 56

 Final form of the swing equation

2 𝐻 𝜔 ( 𝑡 ) 𝑑2 𝛿
2 2
=𝑃 𝑚 − 𝑃 𝑒
𝜔 𝑠𝑦𝑛 𝑑𝑡 2𝐻
2
𝑑 𝛿
𝜔 𝑝𝑢 ( 𝑡 ) 2 =𝑃 𝑚 − 𝑃 𝑒
𝜔 (𝑡 ) 𝜔 𝑠𝑦𝑛 𝑑𝑡
𝜔 𝑝𝑢 ( 𝑡 )=
𝜔 𝑠𝑦𝑛

Swing equation as a second-degree differential equation

© 2023 Daniel Kirschen 57

 Swing equation as first order differential equations

𝑑𝛿
=𝜔 ( 𝑡 ) −𝜔 𝑠𝑦𝑛
2 𝑑𝑡
2𝐻 ( ) 𝑑 𝛿
𝜔 𝑝𝑢 𝑡 =𝑃 𝑚 − 𝑃 𝑒
𝜔 𝑠𝑦𝑛 𝑑𝑡
2
𝜔𝑠𝑦𝑛
𝑑𝜔
= ( 𝑃 𝑚− 𝑃 𝑒 )
𝑑𝑡 2 𝐻 𝜔 𝑝𝑢 ( 𝑡 )

More convenient for simulation using numerical integration

© 2023 Daniel Kirschen 58

 Example 11.6
𝑃𝑚 𝑃𝑒

Prime Mover Grid

• Generator in the steady state and operating at synchronous speed

• Bolted three-phase faults at its terminals at
 for
• remains unchanged for the time interval considered
• How does the angle change over time?

© 2023 Daniel Kirschen 59

 Example 11.6
𝑑𝜔 𝜔𝑠𝑦𝑛
= ( 𝑃 𝑚− 𝑃 𝑒 )
𝑑𝑡 2 𝐻 𝜔 𝑝𝑢 ( 𝑡 )

For simplicity, we will assume that

𝑑𝜔 𝜔 𝑠𝑦𝑛
= 𝑃𝑚 with the initial condition
𝑑𝑡 2𝐻
𝜔 𝑠𝑦𝑛
𝜔 ( 𝑡 )= 𝑃𝑚 𝑡 + 𝜔 𝑠𝑦𝑛
2 𝐻 𝑑𝛿 𝜔 𝑠𝑦𝑛 𝜔 𝑠𝑦𝑛 2
=𝑃 𝑚 𝑡 𝛿 ( 𝑡 )= 𝑃 𝑚 𝑡 + 𝛿 ( 0)
𝑑𝛿 𝑑𝑡 2𝐻 4𝐻
=𝜔 ( 𝑡 ) −𝜔 𝑠𝑦𝑛
𝑑𝑡
is determined by the pre-fault
© 2023 Daniel Kirschen
loading of the generator 60
 Example 11.6
How much would it require for a 60 Hz generator with an inertia constant operating at
rated power and a rotor angle to reach following such a fault?

© 2023 Daniel Kirschen 61

 Example 11.6
𝜔 𝑠𝑦𝑛 2
𝛿 ( 𝑡 )= 𝑃 𝑚 𝑡 + 𝛿 ( 0)
4𝐻
𝜔 𝑠𝑦𝑛
𝜔 ( 𝑡 )= 𝑃𝑚 𝑡 + 𝜔 𝑠𝑦𝑛
2𝐻

60 Hz generator 

Rated power 
2
𝛿 ( 𝑡 )=31.42 𝑡 + 0.262𝑟𝑎𝑑  for

 A small change in speed results in a

significant change in the angle
© 2023 Daniel Kirschen 62
 Electrical model of synchronous generators
• Electrical dynamics of generators are very complex
• A detailed analysis of these dynamics is outside the scope of this
course
• We will limit our analysis to the simplified, “classical” model

© 2023 Daniel Kirschen 63

 Subtransient, transient and synchronous reactances

Subtransient state Transient state Steady state

Subtransient reactance Transient reactance Synchronous reactance

′ ′ ′
𝑋 < 𝑋 < 𝑋 𝑆
© 2023 Daniel Kirschen 64
 Subtransient, transient and steady periods

Subtransient state Transient state Steady state

Valid for a few 10’s ms Valid for a few 100’s ms

Compatible with mechanical transients

© 2023 Daniel Kirschen 65
 The classical dynamic model of generators
• Same structure as the steady-state
model
𝐼
′
𝑋

• Synchronous reactance replaced

by transient reactance
𝐸′ =𝐸′ ∠𝛿 𝑉𝑡
• is a valid approximation during
the electromechanical transients

© 2023 Daniel Kirschen 66

 The classical dynamic model of generators
• Generator unloaded:

𝐼
′
𝑋 • Generator loaded:

• , : voltage induced in the stator by
𝐸′ =𝐸′ ∠𝛿 𝑉𝑡 rotor magnetic field
• thus represents the position of
the rotor
• is thus the same angle as in the
mechanical model

© 2023 Daniel Kirschen 67

 Connecting the mechanical and electrical models

Electrical and mechanical angle

2
2𝐻 𝑑 𝛿
𝜔 𝑝𝑢 ( 𝑡 ) =𝑃 𝑚 − 𝑃 𝑒
𝜔 𝑠𝑦𝑛 𝑑𝑡
2

Electrical power

© 2023 Daniel Kirschen 68

 Modeling the rest of the system
• Stability is a characteristic of the entire system
• Must connect the electromechanical model of the generator to a model
of the rest of the system
• Simplest model of the rest of the system: infinite bus beyond a reactance
• Infinite bus = ideal voltage source
• Constant voltage magnitude
• Constant frequency
• Constant phase angle
• Simplistic but useful to understand what affects stability
• We’ll consider more realistic models later

© 2023 Daniel Kirschen 69

 One-machine vs. infinite bus model

𝑋
′
𝐼 𝑋

′
𝐸 =𝐸 ∠ 𝛿
′
𝑉𝑡 𝑉 ∞ =𝑉 ∞ ∠0°

Generator under study Transmission network Rest of the system

©©2023
Daniel Kirschen
Daniel & University of Washington
Kirschen 70
 Electric power in the one-machine vs. infinite bus model

𝑋
′
𝐼 𝑋

′
𝐸 =𝐸 ∠ 𝛿
′
𝑉𝑡 𝑉 ∞ =𝑉 ∞ ∠0°

′
𝐸 𝑉 ∞
𝑃𝑒 (𝛿)= ′
sin 𝛿
𝑋 +𝑋
©©2023
Daniel Kirschen
Daniel & University of Washington
Kirschen 71
 Stable and unstable operating points
𝑃
𝑃 𝑒 (𝛿)
• Steady state:
S U
𝑃𝑚
• Operating point S at

• What about operating

point U at ?

0 𝛿0 90 𝛿𝑈 180

© 2023 Daniel Kirschen

(degrees)
72
 Small disturbances around the operating points

𝑃
𝑃 𝑒 (𝛿)

S1 U1
S 𝑃 𝑚 − 𝑃 𝑒 <0 𝑃 𝑚 − 𝑃 𝑒 <0 U
𝑃𝑚
𝑃 𝑚 − 𝑃 𝑒 >0 𝑃 𝑚 − 𝑃 𝑒 >0
S2 U2

0 𝛿0 90 𝛿𝑈 180

© 2023 Daniel Kirschen

(degrees) 73
 Stable and unstable operating points
𝑃 • Any point S with is a
𝑃 𝑒 (𝛿)
stable operating point

• Any point U with is an

𝑃𝑚
S U unstable operating point

• Valid only for small

disturbances

0 𝛿0 90 𝛿𝑈 180

© 2023 Daniel Kirschen

(degrees)
74
 Steady state stability limit
𝑃𝑚 Steady state stability limit
𝑃 𝑒 (𝛿)

Generators are normally operated

with because a large
disturbance (i.e., a fault ) could
cause an instability
Normal operating zone

0 90 180

© 2023 Daniel Kirschen

(degrees)
75
 Example 11.7
A synchronous generator with a transient reactance delivers apparent power
at 0.95 pf lagging and nominal voltage.
Calculate the rotor angles corresponding to the stable and unstable operating points.

𝑃
𝑃 𝑒 (𝛿)

𝑃𝑚 S U

0 𝛿0 90 𝛿𝑈 180
© 2023 Daniel Kirschen (degrees) 76
 Example 11.7
A synchronous generator with a transient reactance delivers apparent power
at 0.95 pf lagging and nominal voltage.
Calculate the rotor angles corresponding to the stable and unstable operating points.

𝑃
𝑃 𝑒 (𝛿)

𝑃𝑚 S U Hence:
.
By symmetry:

0 𝛿0 90 𝛿𝑈 180
© 2023 Daniel Kirschen (degrees) 77
 Simplest example of large disturbances
𝐺 ∞

𝐶𝐵 1 𝐶𝐵 2

• Generator injects power into the infinite bus through a single line of
reactance
• Bolted three-phase fault immediately to the right of
• Fault cleared by opening of
• Effect on the electromechanical dynamics of the generator?
• assumed constant

© 2023 Daniel Kirschen 78

 Simplest example of large disturbances
𝐺 ∞

𝐶𝐵 1 𝐶𝐵 2

• Pre-fault conditions:
•
• Gives the initial value of for the differential equations
• Fault conditions:

• Post-fault conditions

© 2023 Daniel Kirschen 79

 Simplest example of large disturbance
Momentum

𝑃 C Momentum (kinetic energy) is exhausted

D Back towards the stable operating point

𝑃 𝑒 (𝛿)
S U
𝑃𝑚

A 𝑃 𝑒= 0 B
0 𝛿0 𝛿𝑐𝑙𝑒𝑎𝑟 90 𝛿𝑙𝑖𝑚 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 80

 Simplest example of large disturbance
𝑃 C Momentum

𝑃 𝑒 (𝛿)
U Kinetic energy not exhausted before reaching
S
𝑃𝑚 the unstable operating point
Angle continues to increase  instability

A 𝑃 𝑒= 0 B
0 𝛿0 𝛿𝑐𝑙𝑒𝑎𝑟 90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 81

 Marble Analogy

• Marble is released
• Gravity causes the marble to accelerate

© 2023 Daniel Kirschen 82

 Marble Analogy

• Marble reaches the bottom

• Acceleration becomes negative because it goes uphill
• Marble does not stop because it has acquired momentum

© 2023 Daniel Kirschen 83

 Marble Analogy – Stable case

• Marble has exhausted its momentum before reaching top of the next hill
• Speed is zero
• Marble starts moving downwards because of gravity

© 2023 Daniel Kirschen 84

 Marble Analogy – Stable case

• Marble settles at the bottom when friction (damping) has consumed

all the kinetic energy acquired during the initial downhill
• Note that our simple power system model does not include damping

© 2023 Daniel Kirschen 85

 Marble Analogy – Unstable case

• Marble still has some momentum when it reaches the top of the next hill
• Marble starts accelerating again
• Instability

© 2023 Daniel Kirschen 86

 Simplest example of large disturbance
𝑃 C
D
𝑃 𝑒 (𝛿)
S U
𝑃𝑚

System remains stable if

A B
0 𝛿0 𝛿𝑐𝑙𝑒𝑎𝑟 90 𝛿𝑙𝑖𝑚 𝛿𝑈 180
(degrees)
© 2023 Daniel Kirschen 87
 Equal area criterion
• Generator speeds up during the fault
 It acquires kinetic energy during the fault
• Generator slows down after the fault is cleared
 It injects this kinetic energy back into the system
• Can it return enough kinetic energy (i.e., slow down) before it reaches
the unstable equilibrium point?
• Develop a simple criteria
• Visualize the factors that influence stability

© 2023 Daniel Kirschen 88

 Equal area criterion
• Swing equation:
• Assume
• Justification: a small increase in speed result in a large change in the
angle
• Simplified version:
• Integrate between two arbitrary angles :

© 2023 Daniel Kirschen 89

 Measuring kinetic energy

• Swap and and integrate over rather than :

•
• , : speeds of the generator when it is at angles and
• Observe that
•
•

© 2023 Daniel Kirschen 90

 Measuring kinetic energy

Change in kinetic energy Change in kinetic energy

represents a change in kinetic energy

© 2023 Daniel Kirschen 91

 represents a change in kinetic energy
Kinetic energy returned
to the grid after the fault
𝑃 Is cleared ()
𝑃 𝑒 (𝛿)
𝐴2
S U
𝑃𝑚

Kinetic energy
acquired during
the fault () 𝐴1

0 𝛿0 𝛿𝑐𝑙𝑒𝑎𝑟 90 𝛿𝑈 180
(degrees)
© 2023 Daniel Kirschen 92
 System will remain stable if the generator
Stability criterion can return this kinetic energy to the grid
before reaching the unstable equilibrium
point U.

𝑃 Mathematically: .
𝑃 𝑒 (𝛿)
𝐴2
S U
𝑃𝑚

𝐴1

0 𝛿0 𝛿𝑐𝑙𝑒𝑎𝑟 90 𝛿𝑈 180
(degrees)
© 2023 Daniel Kirschen 93
 Critical clearing angle
such that ?
𝑃
𝑃 𝑒 (𝛿)
𝐴2
S U
𝑃𝑚

𝐴1

0 𝛿0 𝛿 𝑐𝑟𝑖𝑡 90 𝛿𝑈 180
(degrees)
© 2023 Daniel Kirschen 94
 Calculating the critical clearing angle
is such that:
𝑃
𝑃 𝑒 (𝛿)
𝐴2
S U
𝑃𝑚 Equivalent to:

𝐴1 𝐴3

0 𝛿0 𝛿 𝑐𝑟𝑖𝑡 90 𝛿𝑈
© 2023 Daniel Kirschen
(degrees) 180
95
 Calculating the critical clearing angle

© 2023 Daniel Kirschen 96

 Example 11.8
𝐺 ∞

𝐶𝐵 1 𝐶𝐵 2

•
•
• for each line
•


© 2023 Daniel Kirschen 97

 This case is informative but not realistic…
𝐺 ∞

𝐶𝐵 1 𝐶𝐵 2

• Before the fault, the generator injects power through one line of reactance
• During the fault, the generator injects no power at all
• After the fault, the generator injects power through the same line of reactance
• In other words, the system is back to its original state
• This is usually not the case
• Let us develop some more sophisticated examples

© 2023 Daniel Kirschen 98

 A somewhat more realistic scenario
𝐺 𝑋 ∞ ′
𝑝𝑟𝑒 𝐸𝑉∞
Pre-fault 𝑃 ( 𝛿 )= sin 𝛿
𝑋 𝑒
′ 𝑋
𝐶𝐵 1 𝐶𝐵 2
𝑋+
2
𝐺 ∞
𝑓𝑎𝑢𝑙𝑡
Fault 𝑃 𝑒 ( 𝛿) =0
𝐶𝐵 1 𝐶𝐵 2

𝐺 ∞ ′
𝑝𝑜𝑠𝑡 𝐸 𝑉∞
Post-fault
𝑃 𝑒 ( 𝛿 )= ′
sin 𝛿
𝑋 +𝑋
𝐶𝐵 1 𝐶𝐵 2

© 2023 Daniel Kirschen 99

 Power angle curves for this scenario
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿)

𝑝𝑜𝑠𝑡
𝑃𝑒 ( 𝛿)
D E
𝐴2
A F U
𝑃𝑚

𝐴1
B C
0 𝛿0 𝛿1 𝛿𝑐𝑙𝑒𝑎𝑟 90 𝛿𝑙𝑖𝑚 𝛿𝑈 180
© 2023 Daniel Kirschen
(degrees) 100
 Example 11.9
𝐺 ∞

𝐶𝐵 1 𝐶𝐵 2

•
•
• for each line
and
•

© 2023 Daniel Kirschen 101

 Example 11.10
𝐺 ∞

𝐶𝐵 1 𝐶𝐵 2
𝑋 𝑋
2 2
•
•
• for each line
•

© 2023 Daniel Kirschen 102

 Example 11.10
𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

𝐺 Pre-fault ∞ ′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝑗 0.4 𝑝. 𝑢.

𝐶𝐵 1 𝐶𝐵 2

Fault 𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

𝐺 ∞
′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝑗 0.2 𝑝 .𝑢. 𝑗 0.2 𝑝 .𝑢.
𝐶𝐵 1 𝐶𝐵 2

𝐺 Post-fault ∞ 𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝐶𝐵 1 𝐶𝐵 2

© 2023 Daniel Kirschen 103

 Example 11.10 – Fault conditions
𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑗 0.2 𝑝 .𝑢. 𝑗 0.2 𝑝 .𝑢.
𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.

Replace by Thevenin equivalent

𝑋 𝑇𝐻 =0.2 ∥0.4=0.133 𝑝. 𝑢. 𝑗 0.3 𝑝 .𝑢. 𝑗𝑋 𝑇𝐻 = 𝑗 0.133 𝑝 .𝑢 .

′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 𝑇𝐻 =0.333 ∠ 0°𝑝. 𝑢.

© 2023 Daniel Kirschen 104

 Example 11.10
𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

𝐺 Pre-fault ∞ ′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝑗 0.4 𝑝. 𝑢.

𝐶𝐵 1 𝐶𝐵 2
𝑗 0.3 𝑝 .𝑢. 𝑗𝑋 𝑇𝐻 = 𝑗 0.133 𝑝 .𝑢 .
𝐺 Fault ∞ ′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 𝑇𝐻 =0.333 ∠ 0°𝑝. 𝑢.

𝐶𝐵 1 𝐶𝐵 2

𝐺 Post-fault ∞ 𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝐶𝐵 1 𝐶𝐵 2

© 2023 Daniel Kirschen 105

 Example 11.10
𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.
′ ′
𝑝𝑟𝑒 𝐸𝑉∞ 𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝑃 𝑒 ( 𝛿 )= sin 𝛿 𝑗 0.4 𝑝. 𝑢.
′ 𝑋
𝑋+
2
𝑗 0.3 𝑝 .𝑢. 𝑗𝑋 𝑇𝐻 = 𝑗 0.133 𝑝 .𝑢 .
′
𝑓𝑎𝑢𝑙𝑡 𝐸 𝑉 𝑇𝐻
𝑃 𝑒 ( 𝛿) = ′ sin 𝛿 ′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 𝑇𝐻 =0.333 ∠ 0°𝑝. 𝑢.
𝑋 + 𝑋 𝑇𝐻

′ 𝑗 0.3 𝑝 .𝑢. 𝑗 0.4 𝑝. 𝑢.

𝑝𝑜𝑠𝑡 𝐸 𝑉∞
𝑃 𝑒 ( 𝛿 )= ′
sin 𝛿 ′
𝐸 =1.2∠ 𝛿 𝑝. 𝑢. 𝑉 ∞ =1.0 ∠0° 𝑝. 𝑢.
𝑋 +𝑋

© 2023 Daniel Kirschen 106

 Example 11.10
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿)

𝑝𝑜𝑠𝑡
E 𝑃𝑒 ( 𝛿)
D
𝐴2
A F U
𝑃𝑚
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿)

0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 107
 Example 11.10
for :
𝑃 𝑃
𝑝𝑟𝑒
𝑒 ( 𝛿)

𝑝𝑜𝑠𝑡
E 𝑃𝑒 ( 𝛿)
D
𝐴2
A F U
𝑃𝑚
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿)

0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 108
 What factors affect stability?
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿) • What decreases ?
• What increases ?
𝑝𝑜𝑠𝑡
E 𝑃𝑒 ( 𝛿)
D
𝐴2
A F U
𝑃𝑚
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿)

0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 109
 What factors affect stability?
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿) • Faster relaying and
circuit breakers
𝑝𝑜𝑠𝑡 Reduces
E 𝑃𝑒 ( 𝛿)
Increases
D
𝐴2 • Limitation?
A F U
𝑃𝑚 • Extinguishing an arc is
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿)
not easy

0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 110
 What factors affect stability?
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿) • Increase inertia
Reduces
E
𝑝𝑜𝑠𝑡
𝑃𝑒 ( 𝛿) Increases
D • Limitation?
𝐴2 • Adding unnecessary
A F U
𝑃𝑚 steel is expensive
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿) • Inverter-based
generation (wind,
B solar) don’t contribute
to inertia
0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180
© 2023 Daniel Kirschen (degrees) 111
 What factors affect stability?
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿) • Reduce and
Reduces
E
𝑝𝑜𝑠𝑡
𝑃𝑒 ( 𝛿) Increases
D • Limitations?
𝐴2 • Decreases production
A F U
𝑃𝑚 from cheaper
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿)
generating units
• Increases production
from more expensive
B
generating units

0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 112
 What factors affect stability?
𝑃 𝑝𝑟𝑒
𝑃 𝑒 ( 𝛿) • Decrease pre- and
post-fault
𝑝𝑜𝑠𝑡
E 𝑃𝑒 ( 𝛿)
Reduces
D
𝐴2 Increases
A F U
𝑃𝑚
𝐴1 C
𝑓𝑎𝑢𝑙𝑡
𝑃𝑒 ( 𝛿)

0 𝛿0 𝛿 1 𝛿𝑐𝑙𝑒𝑎𝑟 𝛿𝑙𝑖𝑚90 𝛿𝑈 180

© 2023 Daniel Kirschen (degrees) 113
 What can be done to improve stability?
• Limit pre-fault power injected by the generators
• Limits depends on the operational conditions
• Lines that are in service
• Other generation
• Affects the cost of producing power
• System should be designed to allow generators to produce their rated
power under normal conditions

© 2023 Daniel Kirschen 114

 What can be done to improve stability?
• Reduce reactance between generation and load
• Locate generation closer to the load
• Build parallel lines or parallel paths between generation and load
• Build higher voltage lines
•  per unit impedance decreases with the square of the voltage
• Series compensation

© 2023 Daniel Kirschen 115

 Series compensated line
Intrinsic resistance and inductance of the line

Series compensation

Intrinsic susceptance of the line

© 2023 Daniel Kirschen 116
 Series compensation equipment

© 2023 Daniel Kirschen 117

 Problems with series compensation
• Protection of series compensated lines is difficult
• Risk of subsynchronous resonance
• Resonance frequency of an LC circuit:
• Can match the torsional resonance frequency of turbogenerators
• Creates torsional oscillations in these turbogenerators
• Leads to metal fatigue and failure of the shaft

© 2023 Daniel Kirschen 118

 High Voltage DC Transmission
• DC does not introduce stability issues
• Power transfer controlled using power electronics
• Ideal for transmitting large amounts of power over long distances

© 2023 Daniel Kirschen 119

 Existing and planned HVDC lines

Blue: Existing links

Red: Proposals

© 2023 Daniel Kirschen

Source: Iowa State University
120
 Transient stability analysis using time domain
simulation
• Equal area criterion
• Provides useful insights on what affects stability
• Applicable only to simple cases
• Inaccurate
• Need a more rigorous method for assessing stability
• More accurate results
• Applicable to multi-machine systems
• Numerical integration of the differential equations

© 2023 Daniel Kirschen 121

 What is a differential equation?
• Given: 𝑥

𝑥 (𝑡 )?

• We want to calculate
𝑥0

0 𝑡

© 2023 Daniel Kirschen 122

 Numerical integration of differential equations
𝑥

• Given a value of , we can

calculate the derivative of
• We know that
𝑥0
• We can calculate
• This gives us the tangent to at
0 𝑡

© 2023 Daniel Kirschen 123

 Numerical integration of differential equations
• For a sufficiently small timestep , 𝑥
the tangent is a good
approximation of the function 𝑥 (𝑡 )?
•

𝑥0

0 ∆𝑡 𝑡

© 2023 Daniel Kirschen 124

 Step-by-step integration using Euler’s
formula
• Replace the function by a 𝑥
piecewise linear approximation
𝑥3

𝑥2
𝑥1
𝑥0
•…

0 ∆ 𝑡 2∆𝑡 3 ∆𝑡 𝑡

© 2023 Daniel Kirschen 125

 Example 11.11

0 0 2.0 1.0 1.0

1 0.01 2.04 1.02 1.02020134
• Analytical solution: 2 0.02 2.0808 1.0404 1.04081077
3 0.03 2.122416 1.061208 1.06183655
• Error increases progressively
4 0.04 2.16486432 1.08243216 1.08328707
5 0.05 2.20816161 1.10408080 1.10517092
6 0.06 2.25232484 1.12616242 1.12749685
7 0.07 2.29737134 1.14868567 1.15027380
8 0.08 2.34331876 1.17165938 1.17351087
9 0.09 2.39018514 1.19509257 1.19721736
10 0.1 2.43798884 1.21899442 1.22140276

© 2023 Daniel Kirschen 126

 Limitations of Euler’s method
• The time step must be very small for the numerical solution
to remain close to the actual solution
• Accuracy: small difference at each step
• Stability: numerical solution deviates completely from the
actual solution
• The smaller the time step, the longer it takes to calculate the
solution
• Euler’s method is simple but inaccurate and not very stable

© 2023 Daniel Kirschen 127

 Other numerical integration methods
• Higher order methods
• Use data from more than one previous step
• Predictor-corrector methods
• Refines the estimate of the derivative
• Example: modified Euler’s method
• Variable step methods
• Adjust the step size as a function of the dynamics
• Rapid changes  smaller step size
• Slower changes  longer step size
© 2023 Daniel Kirschen 128
 Numerical integration of the swing equation
Euler’s formula:
First order form of the swing equation:
𝑑𝛿
=𝜔(𝑡 ) − 𝜔 𝑠𝑦𝑛 𝛿𝑘+1=𝛿𝑘 + ( 𝜔 𝑘 − 𝜔 𝑠𝑦𝑛 ) × ∆ 𝑡
𝑑𝑡
𝑑𝜔 𝜔𝑠𝑦𝑛 𝜔
2
=
𝑑𝑡 2 𝐻 𝜔 𝑝𝑢 ( 𝑡 )
[ 𝑃𝑚 − 𝑃 𝑒 ( 𝛿 ) ] 𝜔 𝑘+1 =𝜔 𝑘 +
2 𝐻 𝜔𝑘
[
𝑠𝑦𝑛
𝑃 𝑚 − 𝑃 𝑒 ( 𝛿𝑘 ) ] × ∆

𝑝𝑟𝑒 𝐸𝑉
Initial conditions: 𝜔 0 =𝜔 𝑠𝑦𝑛 𝑃 𝑒 = sin 𝛿0 =𝑃 𝑚
𝑋𝑆
© 2023 Daniel Kirschen 129
 Numerical integration: pre-fault
Pre-fault conditions:
2
𝜔
𝜔 𝑘+1 =𝜔 𝑘 + [
𝑠𝑦𝑛
𝑃 𝑚 − 𝑃 𝑒 ( 𝛿𝑘 ) ] × ∆ 𝑡 𝜔 𝑘+1 =𝜔 𝑘=𝜔0 =𝜔 𝑠𝑦𝑛
2 𝐻 𝜔𝑘
¿0
𝛿𝑘+1=𝛿𝑘 + ( 𝜔 𝑘 − 𝜔 𝑠𝑦𝑛 ) × ∆ 𝑡 𝛿𝑘+1=𝛿𝑘=𝛿0

© 2023 Daniel Kirschen 130

 Numerical integration: fault
Fault conditions:
2
𝜔
𝜔 𝑘+1 =𝜔 𝑘 +
2 𝐻 𝜔𝑘
[
𝑠𝑦𝑛
𝑃 𝑚 − 𝑃 𝑒 ( 𝛿𝑘 ) ] × ∆ 𝑡 Speed increases

¿0
𝛿𝑘+1=𝛿𝑘 + ( 𝜔 𝑘 − 𝜔 𝑠𝑦𝑛 ) × ∆ 𝑡 Angle increases

© 2023 Daniel Kirschen 131

 Numerical integration: post-fault
Post-fault conditions:
2
𝜔
𝜔 𝑘+1 =𝜔 𝑘 +
2 𝐻 𝜔𝑘
[
𝑠𝑦𝑛
𝑃 𝑚 − 𝑃 𝑒 ( 𝛿𝑘 ) ] × ∆ 𝑡 Speed decreases

¿0
𝛿𝑘+1=𝛿𝑘 + ( 𝜔 𝑘 − 𝜔 𝑠𝑦𝑛 ) × ∆ 𝑡 Angle increases until
and then decreases unless
exceeds

© 2023 Daniel Kirschen 132

 Example 11.12: Simulation of Example 11.10
• Time-domain simulations of the system of Example 11.10
• Euler’s method with a time step
• Fault onset at
• Three different fault clearing times:
•  Stable
•  gets very close to the value of
•  exceeds at about
• Stable case oscillates because our model does not
include damping (i.e., losses)
© 2023 Daniel Kirschen 133
 Example 11.12: Simulation of Example 11.10
180

160
(degrees) Unstable
140 Re-acceleration beyond the
unstable operating point
120
Marginally stable
100 Fault cleared for the
marginally stable and
80
unstable cases
60

Fault40cleared for the

stable
20 case
Stable
0

Onset of the fault

© 2023 Daniel Kirschen Time (s) 134

 Example 11.13 : Effect of generator loading
• Time-domain simulations of the system of Example 11.10
• Euler’s method with a time step
• Fault onset at and cleared at
• Three cases:
•  Stable
•  Marginally stable
•  Unstable

© 2023 Daniel Kirschen 135

 Example 11.13: Effect of generator loading
180

160
(degrees)
140 Unstable
120

100
Marginally stable
80
Fault
60 cleared
40
Onset of
20
the fault Stable
0

© 2023 Daniel Kirschen Time (s) 136

 Dynamics of multi-machine systems
• Assumption so far: 𝐺 ∞
• One machine vs. infinite bus
• Fault affects only one generator 𝐶𝐵 1 𝐶𝐵 2
• Rest of the system in unaffected
• In fact, faults affect all generators in the system
• Faults change the amount of electric power they inject in the network
• This changes the flows in the transmission lines
• Transients of each generator affect the other generators
• Must simulate these interactions

© 2023 Daniel Kirschen 137

 Dynamics of multi-machine systems
′

1
𝑗 𝑋1
• System with generators
′
𝐼1 • Each generator represented
𝐸 ∠ 𝛿1
1
by the classical model:
𝑗 𝑋2
′
• Constant voltage behind
2 transient reactance
′ 𝐼2 Transmission Network
𝐸 ∠ 𝛿2
2 and Loads

⋯
′
𝑗 𝑋𝑛
𝑛
′
𝐼𝑛
𝐸 ∠ 𝛿𝑛
𝑛

© 2023 Daniel Kirschen 138

 Dynamics of multi-machine systems
′

1
𝑗 𝑋1
• For each generator :
′
𝐼1 •
𝐸 ∠ 𝛿1
1

′
𝑗 𝑋2
2  first order differential
′ 𝐼2 Transmission Network equations
𝐸 ∠ 𝛿2
2 and Loads
• is a function not only of but of
⋯ the state of the entire system
′
𝑗 𝑋𝑛
𝑛
′
𝐼𝑛
𝐸 ∠ 𝛿𝑛
𝑛

© 2023 Daniel Kirschen 139

 Initialization: Generators’ internal emf
• Perform a power flow
• Generator is connected to bus injects

′
𝑗 𝑋𝑖 𝑘
𝑖
′ 𝐼𝑖
𝐸 ∠ 𝛿𝑖
𝑖

© 2023 Daniel Kirschen 140

 Initialization: Model loads as admittances
• Load at bus
•
• Voltage at bus is

© 2023 Daniel Kirschen 141

 Initialization: Build admittance matrix
′

1
𝑗 𝑋1
• Internal buses of generators
𝐼1 numbered to
′
𝐸 ∠ 𝛿1
1
• Other buses numbered to
2
𝑗 𝑋2
′
• Injected current only at internal
generator buses
′ 𝐼2 Transmission Network
𝐸 ∠ 𝛿2
2 and Loads

⋯
′
𝑗 𝑋𝑛
𝑛
′
𝐼𝑛
𝐸 ∠ 𝛿𝑛
𝑛

© 2023 Daniel Kirschen 142

 Initialization: Build admittance matrix

• : vector of internal emf of the generators

• Magnitudes assumed constant, angles variable

• : vector of voltages at other buses

• : vector of currents injected by the generators

© 2023 Daniel Kirschen 143

 Reduced system
′

1
𝑗 𝑋1
Only interested in angles
′
𝐼1
𝐸 ∠ 𝛿1
1

′
𝑗 𝑋2
2
′ 𝐼2 Transmission Network
𝐸 ∠ 𝛿2
2 and Loads

⋯
′
𝑗 𝑋𝑛
𝑛
′
𝐼𝑛
𝐸 ∠ 𝛿𝑛
𝑛

© 2023 Daniel Kirschen 144

 ~
Reduced system
′

1
𝑗 𝑋1
Only interested in angles

𝒀
′
𝐼1
𝐸 ∠ 𝛿1
1

′
𝑗 𝑋2
2
′ 𝐼2
𝐸 ∠ 𝛿2
2

⋯
′
𝑗 𝑋𝑛
𝑛
′
𝐼𝑛
𝐸 ∠ 𝛿𝑛
𝑛

© 2023 Daniel Kirschen 145

 Admittance matrices
• Study solid fault on linenext to bus
• Pre-fault admittance matrix
• Build matrix with all lines in service
• Faulted admittance matrix
• Voltage at bus is zero
• Bus is merged with ground
• Remove row and column from the matrix
• Admittances of other lines connected to node are now admittances to
ground
• Post-fault admittance matrix
• Remove linefrom pre-fault admittance matrix
© 2023 Daniel Kirschen 146
 Reduced system

© 2023 Daniel Kirschen 147

 Multi-machine system equations
• For all generator :
•

• With:
•

© 2023 Daniel Kirschen 148

 Integrating the differential equations
1. Set
2. For all generators, set
3. For all generators, calculate initial
4. Set
5. If , exit
6. Calculate for each generator using the latest values of
• If ,
• If , calculate using the faulted admittances
• If , calculate using the post-fault admittances
7. calculate and using Euler’s method
8. Go to step 4

© 2023 Daniel Kirschen 149

 Example 11.14
1
′
𝑗 𝑋1 4

𝐸′1 ∠ 𝛿1 𝐺1

′
𝑗 𝑋2 5
2 7
𝐸′2 ∠ 𝛿2 𝐺2 8 𝐿7

𝐿8

′
𝑗 𝑋3 6
3
𝐸′3 ∠ 𝛿3 𝐺3

© 2023 Daniel Kirschen 150

 Example 11.14: fault at and cleared at .
120
(degrees)
100

80
𝛿3 − 𝛿1
60

40
𝛿2 − 𝛿1
20

-20

-40
0 0.5 1 1.5 2 2.5

© 2023 Daniel Kirschen

Time (s) 151
 Example 11.14: fault at and cleared at .
200
(degrees)
160

𝛿3 − 𝛿1
120

80

40
𝛿2 − 𝛿1

-40
0 0.5 1 1.5 2 2.5

© 2023 Daniel Kirschen Time (s) 152

 Effect of damping
• Deviation from synchronous speed  flux linking damper windings is
not constant
• Currents induced in damper windings create a torque that opposes
the speed deviations
• This torque dampens the oscillations

© 2023 Daniel Kirschen 153

 Example 11.15: fault cleared at with
100

(degrees)
80

𝛿3 − 𝛿1
60

40

20

𝛿2 − 𝛿1
0

-20

-40
0 0.5 1 1.5 2 2.5
© 2023 Daniel Kirschen Time (s) 154
 Detailed dynamic models
• We have used the classical generator model
• Constant emf behind transient reactance
• This model is simple but does not accurately reflect the dynamic
electrical behavior of generators
• Include differential equations to reflect the electrical dynamics of
generators
• Very large system of coupled differential equations

© 2023 Daniel Kirschen 155