Grid Modernization – Meters, Sensors and Time
As discussed in a previous blog, the use of meters as sensors for Grid Control will be an important aspect for the evolution of the grid and Grid Modernization. Having a subset of the meter population (bellwether meters) sending in Voltage interval data and alarms on a much higher frequency basis than the typical daily read, will be valuable to operators insight into grid operations. An area that must be considered is the way this information is time-stamped within the system.
When we consider a meter on a mesh network, there are effectively two logical components 1) Meter Register 2) Mesh Network Interface Card (NIC). In typical operation, these components are working on different times. The NIC gets network time from NTP at some given frequency. It then provides this time to the Register in order to keep these clocks, somewhat synchronized. But, if the time is “close enough” between the two, the Register does not update its clock.
It is very common to have the Register maintain its clock if the NIC and Register are within 30s or so. If they drift apart more than this, the Register is updated. If the Register is much farther out of synchronization, say 3 mins, the Register may not update and only send an alarm indicating the times have become very different between the two. Fundamentally, this means there are two clocks maintained on the meter/communications combined hardware.
The question then arises “why not just sync the Register and NIC very often?”, effectively keeping them in constant synchronization. The issue is that whenever time is changed on the meter, the interval being recorded must be marked as LONG or SHORT, dependent upon which way the clock is changed. This in turn causes an issue with the Meter Data Management (MDM) system when performing validations. Many validation scenarios require that the validation fail, and manual validation be performed when there is a long/short interval. Imagine now this happening for many (all bellwether) meters. The staff required to complete validation in a timely way would increase dramatically.
From a systems perspective, the fact that two clocks are running has an impact on outage. When a meter experiences an outage the system will typically attempt to send a Positive Outage Notification (PON), so that a meter that is still powered may be able to forward this on to the head-end. Since most meters can only stay up for a limited number of seconds after an outage, the PON is time-stamped from the NIC, to ensure that an alarm was sent out. On the other hand, the Register records the time stamp of the event using meter time. It means that outage PONs and events often do not reflect the same time. Also, if the meter is being used as a Voltage monitor, the Voltage readings will likely not be time aligned for a given set of meters, since their clocks are all different.
This can be corrected by using the NIC clock for sensor information and the meter clock for billing data, but not many vendors have taken this into account yet. This must be considered when planning to expand smart metering solutions to grid sensors. Time is a critical aspect of sensor operation in all Industrial Internet of Things (IIoT) systems and care must be taken when analyzing and operating on time-sensitive data.