What does time synchronisation mean for Sampled Values?

Time synchronisation has been discussed and used in substations since the mid 1980's as SCADA systems started to provide time stamped data such as Sequence of Events recording.  Clearly the same event recorded by two different devices should be reported with the same time stamp, or two different events occurring say 30 milliseconds apart should have their time stamps also show the 30 millisecond difference in the time stamp.  Hence in SCADA systems we talk about time accuracy and time stamp resolution in the order of milliseconds, hence IRIG-B or even NTP (Network Time Protocol) was sufficient to achieve that coherency of time stamping between the devices.

However the protection functions and algorithms as say the overcurrent, distance or differential relays do not rely on time synchronisation to work - you could remove the time synchronisation connection and the relay would still work properly ... just as electromechanical relays have for over 100 years!!.

The adoption of IEC 61850-8-1 MMS and GOOSE messages starting in 2002 does require each device to coherently know the real time so that time stamps contained within the messages have the same meaning.

But still these don't require any better than the time stamp accuracy provided by IRIG-B and NTP across the Station Bus or the Process Bus, i.e. one millisecond accuracy and resolution is quite sufficient.

Implementing IEC 61850-9-2 Sampled Values however introduces a new paradigm and standard of performance for coherency of the time synchronisation of the Merging Units

Coherency effectively boils down to accuracy and resolution of the time synchronisation being received and applied consistently to each Merging Unit.

Merging Units need much better than 1 millisecond .. in fact they need better than 1 microsecond accuracy.


Consider a substation with a main bus VT with its MU, and then a series of 20 feeder located CTs and MUs

click to enlarge

These MUs are physically quite some distance apart from each other and at different physical distances from the Grand Master clock.  Hence the latency of the time synch from the Grand master to each MU can be noticeably different

Back in the control room, the distance relay algorithm in Relay 1 will subscribe to MU-V and MU-I1
Relay 2 will subscribe to MU-V and MU-I2
......
Relay 20 will subscribe to MU-V and MU-I20

Or the current differential will just subscribe to all the current Merging Units MU-I1 to MU-I20

The relays only receive the Sampled Value streams which contain the field “SeqNbr” which effectively tells the relay “this is sample #2345 within this 1-second sampling window”

click to enlarge 

Therefore all the MUs must coherently start their 1-second sampling window at precisely the same instant.

When Relay1 subscribes to MU-V and MUI-1 it knows that sample count #2345 from both of them was in fact taken at exactly the same instant

Relays with direct CT and VT analogue inputs do their own internal sampling of I and V so it knows the relative time alignment of the samples.

But with separate independent MUs, the relay algorithms can easily get confused if it uses voltage sample from MU-V count #2345 within the 1-second window and current sample from MU-I1 count #2345 but they were taken say 1 millisecond different in real time

The need to use IEEE 1588 v2 Precision Time Protocol (2008)

Distance relays only need samples (as per IEC 61869-9) at 4800 samples per second (sps) [96 samples per cycle (spc) at 50 Hz or 80 samples per cycle (spc) at 60 Hz]  i.e. one sample taken every 208 microseconds.

So all the MUs to be “aligned” start each 1-second window and reset the sample count “SeqNbr” to zero within at worst perhaps within 1 microsecond of each other. Note it is not the absolute time itself, but the coherency of the start of the 1-second window between the different MUs that must be accurately synchronised.

Hence the MUs need to be synchronised using IEEE 1588 (2008) Precision Time Protocol (PTP) as IRIG-B just won’t achieve the required coherency across all the MUs
(N.B. specifically the so called v2 of IEEE 1588).

This then relies on how the IEEE 1588 message is distributed through the LAN network and its switches.  The switch-clocks must therefore be compatible IEEE 1588 clocks with appropriate use of the :

  • Grand Master
  • Boundary Clock(s)
  • Transparent Clocks
  • Ordinary Clocks

.... and the appropriate operation of those in Slave or Master port modes and use of end-to-end or peer-to-peer mechanisms defined in IEEE 1588 (2008)

The use of these clocks is clarified in the IEEE C37.238 PTP Profile.

Looking ahead, IEC 61869-9 has Sampled Values as fast as 96 kHz - i.e. one sample every 10 microsec so coherency will probably need to be better than say 0.1 or even 0.01 microsec