Loss of Mains Protection

Dr. Adam Dyko

University of Strathclyde
Glasgow, UK
e-mail: a.dysko@strath.ac.uk
Outline
Background

LOM methods and practices in different countries

Existing research work

LOM relay performance assessment ETR 139
Risk assessment study performed for NVD WG
ROCOF relay performance during 28 Sept. event
Dynamic modelling of the UK transmission system

Proposed changes of settings

Possible ROCOF alternatives

Future research work

Background
System inertia is lowering due to growing
penetration of non-synchronous generation

Recent experience in Ireland has demonstrated short time wind

penetration of 50% of the demand.

It is anticipated that up to 100% penetration level is achievable by 2020.

In UK n-1 contingency will increase from 1300MW to 1800MW in April

2014.

UK recommended ROCOF setting is 0.125Hz/s (G59/2), in Ireland the

recommended setting is 0.4..0.55Hz/s (G10) but the increase to 2Hz/s
is proposed.

Continually increasing the ROCOF setting my lead to poor ROCOF

dependability.
Loss-Of-Mains LOM
Loss of Mains (or islanding) occurs when part of the
public utility network (incorporating generation) loses
connection with the rest of the system.
OPEN
Utility network

If LOM is not detected the Interconnected Psys

generator could remain system
Qsys
connected, causing a safety Pload
hazard within the islanded Qload
part of the network.

Passive LOM detection is

difficult when the local load
closely matches the generator G
output.
Pgen
Stability under remote faults Qgen
Generator
and system wide events is network
also an issue.
Do we really need a dedicated LOM
protection?

Loss-Of-Grid AND AND

connection (e.g. fault

on the connecting Safety
circuit) AND hazard
Personnel in
the vicinity
Load/Generation in AND
close balance (both P Probability (?)
and Q)

Non-detection zone
duration longer than AND

the acceptable limit,

e.g. 5s
Safety hazard probability tree
Do we really need a dedicated LOM
protection?

Health and Safety at Work Act 1974

Individual risk from

inadvertent sustained
power island
Do we really need a dedicated LOM
protection?

Health and Safety at Work Act 1974

Individual risk from

inadvertent sustained
power island
Evidence of unintentional islanding

It was reported at CIGRE 2010 session that two

unintended
islanded operations were experienced in 2009 in
Brazil with small hydro generation.
Synchrophasor measurements for LOM applications was
proposed

Islanded operation with PV generation was

experienced in Spain during maintenance
disconnection (up to 40 min.)
Safety of maintenance staff was compromised
Introduce the requirement for the telecontrol systems to
manually disconnect all PV generation
 LOM performance requirements
sensitivity/dependability
LOM should be sensitive under all possible load and
generation scenarios.

The most challenging scenario is when the local load closely

follows the generator output both in terms of active and
reactive power.

P [MW] load profile

Interconnected 30
system generation
profile
feeder Pload
Qload 20
~ } non-detection zone
10

LOM

G t1 t2 t3 t4 24
Time [h]

Pgen
Qgen Risk of undetected island
LOM performance requirements
stability/security
LOM should be stable under remote faults cleared by
the utility system as well as under system dynamic events.

It is undesirable to issue a false trip as it leads to the

unnecessary disconnection of the generator.
Interconnected
system

feeder Pload
Qload
~

LOM

G
Pgen
Qgen
Current Practice in dedicated LOM protection

Rate of Change of Frequency (ROCOF)

Good sensitivity but prone to spurious tripping

Voltage Vector Shift (VS)

Fast but poor sensitivity to genuine LOM events and prone to spurious
tripping

Reverse VAR protection

Can fail if the load power factor is close to unity and/or the island
contains long cables

Intertripping
Best performance but cost is high and can become overcomplicated in
some parts of the system.

There is still a need for a reliable passive LOM method.

Rate-Of-Change-Of-Frequency (ROCOF) Method

The ROCOF method is based on the local measurement of the

generator voltage and estimation of the rate of change of frequency.

The measured rate of change of frequency is compared with a preset

threshold.

Additional time delay can also be applied.

The rate of change of frequency following an LOM event is directly

proportional to the amount of active power imbalance between local
load and the generator output.
P f
ROCOF
2 Sn H
ROCOF estimated rate of change of frequency [Hz/s]
P change in active output power during LOM event [MW]
Sn nominal generator rating [MVA]
f generator rated frequency [Hz]
H inertia constant of the generator [s]
Voltage Vector Shift (VS) LOM Method
The relay measures voltage phase changes in
consecutive cycles (or half cycles) and compares
the value with the preset threshold.

Zero crossing technique is often used as method of angle

measurement

VS is very fast in comparison to other methods such as

ROCOF

VS is sensitive to network faults (both resulting in islanding

situation and remote faults cleared by the utility)

VS is not sensitive to rate of change of frequency

Low sensitivity to genuine LOM events. The setting of 6

requires imbalance of more than 30%Sn to cause operation.
Voltage Vector Shift (VS) LOM Method

Before LOM After LOM

100

50
V [V]

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

-50

-100

t [s]

VT1
VT2
Example Vector Shift Relay Operation
Vector Shift is used to protect an Embedded Generator.
Calculate the voltage angle change if the generator output
increased from 15MW to 25MW as a result of an LOM event. Before
and islanding the generator operated at unity power factor pf1=1.
Assume that at the time of islanding the generator terminal voltage was
VT1 = 1/0pu.
Generator: Sn = 30 MVA Vn = 33 kV
Xd = 0.23 pu Ra = 0.05 pu
Xd
Xd
X d
IL R

Ed E d E d VT

The resulting angle shift would be = 6.23

This is way below the recommended setting of 12.
DG protection in different countries

Italy
New directive 84/2012/R/EEL (8 March 2012) was issued by the Italian
Regulatory Authority for Electricity and Gas (Aeeg), with the aim of
integrate the CEI 0-16 (Reference technical rules for the connection of
active and passive consumers to the HV and MV electrical networks of
distribution) with a technical document from TERNA (the Italian
TSO) defining the system requirements of the DG.

f f<47.5 Tripping after 4s

f>51.5 Tripping after 1s

f<49.7
OR
f<50.3

V V0>thr. AND Tripping after 0.1s

V2>thr. OR

V1<thr.
Systematic LOM relay testing
ETR 139
Systematic LOM relay testing ETR 139

Main objectives
To establish the behaviour, under various scenarios, of different Loss
of Mains (LOM) protection relays that are typically applied to the
protection of distributed generation installations
Produce a matrix of recommended settings for LOM
Propose standard test scenarios for LOM relays

Case studies
Scenario 1: Generator of 30MVA connected
to 33kV network
Scenario 2: Generator of less than 5MVA
connected to a section of 11kV network

It was assumed that maximum system

< 0.16Hz/s

Systematic LOM relay testing ETR 139

Type of studies
Genuine loss of mains events sensitivity
Cleared system faults (where there is no actual loss of mains) on
adjacent circuits stability

Generator types
Gas turbine with synchronous machine
Wind turbine driven DFIG generator
Induction machine driven generator
(11kV scenario only)
DC/AC inverter connected source
(11kV scenario only)
Systematic LOM relay testing ETR 139
Network model

Fault Position A (20% Fault Position B (50% Fault Position C (80%

Retained Gen Voltage) Retained Gen Voltage) Retained Gen Voltage)
Systematic LOM relay testing ETR 139

Typical results sensitivity (SM based generator)

Maximum settings to ensure sensitivity under genuine LOM event
1

0.9

0.8
ROCOF setting [Hz/s]

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
-10 -5 -2.5 0 2.5 5 10
Generation - local load balance [%Sn]
Systematic LOM relay testing ETR 139

Typical results stability (SM based generator)

Minimum settings to ensure stability
10

8
ROCOF setting [Hz/s]

0
f1a f1ab f1abc f2a f2ab f2abc f3a f3ab f3abc

Fault scenario
Systematic LOM relay testing ETR 139
Main findings
The LOM performance is affected primarily by the generation
technology the most challenging is a synchronous generator
Stability causes more problems than sensitivity
Improvement in stability can be achieved by providing additional
time delay which does not significantly compromise the sensitivity
Significant difference in
performance between
different relay manufac-
turers was noted, mainly
in terms of stability

Example ROCOF Stability

results for as synchronous
generator
Risk assessment of NVD
protection requirement
Safety hazard probability tree

11kV line single AND AND

(i) phase-to- earth NVD operation required

fault during NDZ Safety
AND hazard
Personnel in
Local AND the vicinity
Load/Generation
(ii) P&Q Balance
(LOM non-
detection zone)

Non-detection
zone duration AND
(iii)
longer than the
acceptable limit,
e.g. 5s
Example NVD Requirement probability calculation
(ENW 11kV data Circuit 1)
0.7
Pload

Assumptions 0.6
Qload

o ENW 11kV data Circuit 1 0.5

P[MW], Q[MVAR]
0.4
o 5km of 95mm2 (5MVA) 0.3

cable is present 0.2

o Non-detection zone 8% (due to 0.1

presence of phase-to-earth fault 0

0 1 2 3 4 5 6 7 8
time [s] x 10
4

a value less than 10% is assumed)

o Generator at 0.98pf (lead)
o Generator sizes are spread evenly between
PDG,min=100kW and PDG,max=5MW
o Maximum acceptable non-detection zone duration
TNDZ,max = 5s
o Number of affected generators in 2020 is nDG=18,000x0.75
(75% LV connected).
Example NVD Requirement probability calculation
(ENW 11kV data Circuit 1)
Average time of non-detection zone:
TNDZ ,avr 54.972 s

Number of expected incidents in a single scheme:

n 6.76
N NVD,1DG P23 OHE l 0.000220 5 0.000074 incidents p/a
100 100
National number of expected incidents:

N NVD N NVD,1DG nDG 0.000074 18000 0.75 1.003 incidents p/a

Total annual time of expected incidents:

TNVD N NVD TNDZ ,avr TNDZ ,max 1.03 54.972 5 50.10 s

Annual probability of existence of hazardous islanded system

condition resulting from the relaxation of the NVD protection
requirement is: T 50.10
PNVD NVD 1.59 10 6
Ta 8760 60 60
 System event
28 September 2012
 PMU Frequency Record

50.2
50.15
50.1
50.05
Frequency (Hz)

50
49.95
49.9
Strathclyde
49.85
49.8 Manchester
49.75 Imperial
49.7
49.65
01:48:00.00 01:49:26.40 01:50:52.80 01:52:19.20 01:53:45.60 01:55:12.00
GPS Time
PMU Frequency Record
PMU Frequency
50.1
London
Manchester
50.05 Glasgow

50

49.95
f [Hz]

49.9

49.85

49.8

49.75

49.7
512 513 514 515 516 517 518 519 520
t [s]
PMU Phase Angle Difference

PMU Phase Difference

45
PMUPhaseGla - PMUPhaseLon
PMUPhaseGla - PMUPhaseMan

40

35
angle [deg]

30

25

20
512 513 514 515 516 517 518 519 520
t [s]
PMU df/dt Record

1.5 Strathclyde
Manchester
1 Imperial

0.5
df/dt (Hz/s)

0
01:48:31.97 01:48:32.83 01:48:33.70 01:48:34.56 01:48:35.42 01:48:36.29
-0.5

-1

-1.5

-2
GPS Time
df/dt calculated from PMU frequency
as 0.5s average
1.5 Strathclyde
Manchester
1 Imperial

0.5
df/dt (Hz/s)

0
01:48:31.97 01:48:32.83 01:48:33.70 01:48:34.56 01:48:35.42 01:48:36.29
-0.5

-1

-1.5

-2
GPS Time
df/dt calculated from PMU frequency
as 0.5s average
0.05 Strathclyde
Manchester
Imperial
0
01:48:31.97 01:48:32.83 01:48:33.70 01:48:34.56 01:48:35.42 01:48:36.29

-0.05
df/dt (Hz/s)

-0.1

-0.15

-0.2

-0.25
GPS Time
ROCOF relay testing
Three COMTRADE waveform records have been synthesised on a
cycle by cycle basis from the available PMU frequency profiles
and subsequently injected into the MiCOM P341 relay.

Delay Trip setting Tripping time No trip setting

Record
(s) (Hz/s)* (s)** (Hz/s)***
0 0.16 0.96 0.18
Glasgow 0.25 0.15 1.15 0.17
0.50 0.12 1.32 0.14
0 0.11 0.80 0.12
Manchester 0.25 0.10 1.14 0.12
0.50 0.10 (2/5 times) 1.27 0.11
0 0.16 0.31 0.18
London 0.25 0.10 0.50 0.13
0.50 0.10 (2/5 times) 1.72 0.11

* 5/5 injections result in a trip unless otherwise stated

** This is the average of 5 (or less) injections' tripping times recorded at the trip setting
*** 0/5 injections result in a trip
Observations

df/dt seen by the DG depends on the position in the

network and position in relation to the initiating event
df/df seen by the relay depends on the frequency and
ROCOF calculation method (i.e. mainly the averaging
period)
Applying additional time delay increases relay stability
Applying additional frequency dead band (e.g. 49.5Hz to
50.5Hz) will block ROCOF operation during the majority
of system wide events.
Equivalent dynamic model of
the UK transmission system
UK Transmission System model

21 bus equivalent
transmission network model
1 is proposed
2

4
3
7YS was used to obtain the
load and generation data in
each zone.

Available PSS/E full load

flow model was used to
obtain the circuit equivalent
impedances.
France

PMU data was used to verify

the dynamic performance.
 Dynamic Validation: PMU Data

PMU data captured during a major event in the UK transmission system 28 September 2012

70 50.05
50 GL
60
49.95
MN
50 49.9
LN
49.85
40 49.8
30 49.75
49.7
20 GL-MN 49.65
GL-LN 49.6
10
49.55
0 49.5
14:03:22 14:03:23 14:03:25 14:03:26 14:03:28 14:03:29 14:03:31 03:22.003:22.603:23.203:23.903:24.503:25.203:25.803:26.4

Phase difference Frequency

The analysis of the voltage angle shift from different parts of the system,
prior to, and after the event can assist the load flow solution validation.
The frequency response can be used to validate the inertia and system
dynamics, potentially revealing power flow oscillations. 65
Simulation of the 28 September event

No PSSs included
50.05
70
50
GL
60
49.95 MN
50 49.9 LN
49.85
40
49.8
30 49.75

20 49.7
GL-MN
49.65
10 GL-LN
49.6
0 49.55
0.0 1.4 2.9 4.3 5.8 7.2 8.6 10.111.512.914.415.817.318.7 0.0 0.7 1.4 2.1 2.9 3.6 4.3 5.0 5.8 6.5 7.2 7.9 8.6 9.4

Phase difference Frequency

66
Adding stabilisers

STAB1 provided by PSS/E

is used in following buses:
1,2,3,8,9,10,13,19,20 STAB1 transfer function

0.15
4
Eigenvectors and 0.1
3

participation factors of 0.05

20
7
6
5

19
each lightly damped 0 9 1
10

oscillation mode indicate 14

1816
17
15
12
13118

-0.05

the need for a stabiliser.

-0.1

-0.2 0 0.2 0.4 0.6 0.8 1

Eigenvectors of one lightly

damped oscillation mode
 Simulation of the 28 September event

Dynamic performance after adding PSSs into the system

70
50.05
GL
60 50
MN
49.95
50 LN
49.9
40 49.85
49.8
30
49.75
GL-MN 49.7
20
GL-LN 49.65
10 49.6
49.55
0
0.00.20.50.70.91.11.41.61.82.12.32.52.73.03.23.43.73.94.14.44.64.8
0.0 0.6 1.2 1.9 2.5 3.1 3.8 4.4 5.0 5.7 6.3 6.9 7.6 8.2 8.8 9.4

Phase difference Frequency

68
 PMU Data - 28 September

Original PMU record

70 50.05
50 GL
60
49.95
MN
50 49.9
LN
49.85
40 49.8
30 49.75
49.7
20 GL-MN 49.65
GL-LN 49.6
10
49.55
0 49.5
14:03:22 14:03:23 14:03:25 14:03:26 14:03:28 14:03:29 14:03:31 03:22.003:22.603:23.203:23.903:24.503:25.203:25.803:26.4

Phase difference Frequency

ROCOF settings adjustment
Initial Thoughts on Minimising System Risks
New plant connecting in and beyond Apr 2014
must not have LoM protection sensitive to RoCoF; or
the RoCoF setting must be at or above 1Hz/s and with a
minimum timer setting of 0.5 s
New plant connecting after Apr 2013 but before Apr 2014
must not have LoM protection sensitive to RoCoF; or
the RoCoF setting must be at or above 0.5Hz/s with a minimum
timer setting of 0.5 s with a provision of changing to 1 Hz by Apr
2014
Existing plant
If LoM protection is sensitive to RoCoF, its setting is required to
be changed to 0.5Hz/s or above and with a minimum timer setting
of 0.5s as soon as practically possible
Provision should be made to change to 1 Hz by Apr 2014
this may be an ideal solution from system security perspective, but
Issues to consider

Can the setting of 1Hz/s and 0.5s delay provide

acceptable LOM sensitivity?
Are there any time coordination issues with the
delayed LOM operation (in excess of 500ms)?
Can the ROCOF setting recommendation be
made dependent on the generating technology?
Should ROCOF be removed from the acceptable
LOM methods?
What LOM methods (if any) should be adopted in
the future?
Other LOM methods
 Phase Angle Drift (PAD) LOM Protection
f[n] fnest
fn-D
fn-D-W
fn

TW TD n
Local frequency is measured by the relay

Grid frequency is estimated using linear extrapolation of recorded

historical data

The PAD algorithm is based on a threshold comparison of an

accumulated voltage phase angle derived from the difference
between the current measured frequency and estimated frequency
H.T. Yip, G. Lloyd, A. Dysko, G.M. Burt, R.M. Tumilty, Islanding Detection Using an Accumulated Phase Angle Drift Measurement, 10th
International Conference on Developments in Power System Protection, Manchester UK, 2010.
 Satellite based LOM protection

The LOM protection algorithm

uses the difference between Transmission
the received signal and the System

locally measured frequency to ~

estimate the voltage angle
shift according to the following trip
formula:

n n 1 2 ( f n f nref )t DG
LOM
GPS stamped
frequency signal

Protection

n calculated angle
fn frequency measured locally
fnref reference frequency received from the satellite
t angle calculation time step
A tripping signal is sent when n exceeds the preset threshold.
Satellite channel latency can be easily compensated using GPS
A. Dyko, G.M. Burt, P.J. Moore, I.A. Glover, J.R. McDonald, Satellite Communication Based Loss-of-Mains Protection, 9th International
Conference on Developments in Power System Protection, Glasgow UK, vol. 1, pp. 687-692, March 2008.
 Satellite based LOM protection

Relay response to 28 September event

Phase Angle Difference calculated from frequency difference
7
Manchester and
PAD from (fGla - fLon) London PMU record
6
as a reference
PAD from (f -f ) frequency
Gla Man

5
Glasgow PMU
4 record as a DG local
angle [deg]

frequency
3
Accumulated phase
2 angle difference
contained within
1 7deg.

0
Accumulated angle
resets if df<0.002Hz
-1
500 550 600 650 for more than 2s.
t [s]
 Satellite based LOM protection

Relay response to 28 September event

Phase Angle Difference calculated from frequency difference
16 Artificial time delay
PAD from (fGla - fLon) of 100ms has been
14 introduced.
PAD from (f -f )
Gla Man
12

10 Angle difference still

contained within
8
angle [deg]

reasonable margin
6 (16deg), i.e. GPS
accuracy in time
4 synchronisation is
2 not essential.
0

-2

-4
500 550 600 650
t [s]
Internet based LOM protection
General WAM system for LOM

Instantaneous phase difference

less hourly steady-state average

oscillation of phase angle

difference rarely exceeded
5

Laverty, D.M.; Morrow, D.J.; Best, R.J.; Crossley, P.A.; , "Differential ROCOF relay for Loss-of-Mains protection of Renewable Generation
using phasor measurement over Internet Protocol," Integration of Wide-Scale Renewable Resources Into the Power Delivery System, 2009
Reverse VAR Method
Reverse VAR relay measures the generator reactive
power flow Qgen and operates when it exceeds a fixed
threshold.
Trip if Qgen > Qs

Interconnected Psys Interconnected Psys=0

system system
Qsys Pload Qsys=0 Pload
Qload Qload

LOM LOM

G G
Pgen Pgen
Qgen0
Connected Qgen Islanded
 Centralised LOM protection using IEC-61850
The technique is based on communication of CB status
to the central controller.

Can be combined with conventional passive LOM method with

adaptive settings as a backup.
A B

CB1 CB2 CB3 CB1 CB2 CB3

D
CB1 CB3

C CB4
CB2 CB2
CB1 CB2
CB1 CB5
CB3

E
MAINS
CB3
CENTRAL
CONTROLLER

CB1 CB2 CB3

CB1 CB2

F G

CB4 CB5
CB3 CB4

Coffele, F.; Moore, P.; Booth, C.; Dysko, A.; Burt, G.; Spearing, T.; Dolan, P., Centralised Loss of Mains protection using IEC-61850,
Developments in Power System Protection (DPSP 2010).
Reverse VAR Method

The amount of reactive power from the generator Qgen may

become insufficient to activate the relay if the total capacitance of the
connecting cables is high delivering reactive power to the loads.

Interconnected Psys=0
system

Islanded Qsys=0 Pload

Qload

LOM Qc

G Qc

Pgen Qc

Qgen
Intertripping of Open Ring Feeders

Intertripping for open ring feeders can be facilitated using

Power Line Carrier (PLC) technique.
Intertripping signal is always delivered to the correct
generators regardless of the position of the open point.

Signal send

Signal block
G G
Signal receive Open point
Signal receive
Signal block

Signal send

G
Signal receive
Intertripping of Open Ring Feeders

Intertripping for open ring feeders can be facilitated using

Power Line Carrier (PLC) technique.
Intertripping signal is always delivered to the correct
generators regardless of the position of the open point.

Signal send

Signal block
G G
Signal receive Open point Signal receive

Signal block

Signal send

G
Signal receive
Active methods

Active methods are more reliable but may be

slower and loose performance in larger groups of
generators

The use of active frequency shift method was suggested

such as Sandia Frequency Shift.

This could also be combined with ROCOF to achieve the

best performance (combined use of active and passive
methods).

The Grid Connection Code in Japan mandates the

combined use of the passive and the active methods.
Future research

One Year Satellite Applications Catapult funded

LOM demonstrator project has been awarded and will
commence in March 2013.

A research team has been set up at Strathclyde to look

into equivalent modelling of the UK transmission system.

Hardware LOM relay testing under recorded and simulated

system wide disturbances (?).

Risk assessment of the LOM protection under new

proposed setting guidelines (?).