On June 28, 2015, a SpaceX Falcon 9 rocket bound for the International Space Station exploded almost 2½ minutes after launching from Cape Canaveral, Florida. In the process of trying to determine the root cause of the rocket failure, engineers had to develop a clear timeline of the events leading to the failure from roughly 3,000 sources of telemetry.
"The biggest thing that’s needed in the short term is the ability to just gather all the data, and create a very precise timeline, so that, by the millisecond, we know what each sensor was reading, and we can correlate that with ground video.”says Mr. Elon Musk, founder and chief executive of SpaceX. “One of the biggest challenges is matching things to the exact time. When you're dealing in milliseconds all that stuff actually makes quite a big difference. So that's the biggest effort we've been engaged in thus far, is just putting together a super-detailed timeline and then just making sure that we have the sequence of events down as precisely as possible.” *
The challenge of quickly and accurately identifying the order of events leading to the rocket failure cost SpaceX time, money, and negatively impacted their reputation. Ironically, the same need to align time and sequence data happens daily here on Earth when electrical events occur inside an energy consumer’s facility or out on the utility grid. The difference is that, unlike Musk, most do not grasp the importance of accurately time stamping metering data to analyze and resolve problems.
Time and Power Monitoring Systems
Although “time” is essentially a measure of change, it is typically defined by its measurement. The need to apply timestamps to events, establish the duration of events, and/or sequentially order the occurrence of multiple events is a fundamental human trait. It is also an indispensable scientific parameter used to describe our physical world, including electrical properties. In the energy domain, time is used to calculate consumption rates and demand costs, develop load profiles, quantify severity of power quality phenomena, facilitate event progression and reconstruction, operate, protect and control electrical systems and grids, and much more.
Power monitoring networks are spatially distributed systems, consisting of two or more digital metering devices that are communicatively interconnected to provide accurate and timely data from different locations within a facility’s electrical system. Digital meters can measure and store both periodic (logged) and asynchronous (event) data. When an event occurs, information typically captured includes the event’s time of occurrence, type, duration, origin, and possibly a sampled waveform. Each of these is an important factor when trying to identify the root cause of the event.
Digital meters typically incorporate “on-board” clocks to time stamp the acquisition of data points and event information. Each discrete meter’s on-board clock operates independently from the other meters in the power monitoring system, so the clocks are not inherently synchronized with each other. This offset of time between discrete metering devices is known as clock skew. The time associated with each device’s clock also deviates over a given period due to internal and external influences, which may exacerbate the extent of clock skew. This deviation in the rate time is measured by a device’s on-board clock is defined as clock drift.
Clock skew and clock drift become an issue when attempting to aggregate and analyze event data from a multitude of discrete metering devices with nonsynchronous time stamps. While it may be possible to quickly determine some aspects of an event (e.g., event type), discrepancies in time stamp data from multiple devices can lead to confusion regarding how the event propagated through an electrical system. If the clock skew is severe, it may even be concluded more than one event occurred. Consequently, this issue may protract the task of identifying an event’s root cause and mitigating the problem to avoid future reoccurrences.
Accurately and continuously determining each metering device’s on-board clock offset with respect to all other metering device on-board clocks has several benefits. By quantitatively resolving temporal (time) relationships across power monitoring system devices and the data they produce, it is easier to identify, measure, and order occurrences and events to better characterize the electrical system’s behavior. Changes in energy consumption and perturbations of source voltages are not exclusive; they often impact broad portions of the electrical system (and the loads served therein). Using synchronous data from multiple metering devices yields propagation rates and the sequence of these changes as they ripple throughout the electrical system. It also permits a better understanding of the electrical system’s response to these changes so that it is possible to identify weaknesses and improve the system’s response to these perturbations. Characterizing system device clocks in this way provides temporal context to monitoring system data, which is essential to understand, troubleshoot and resolve problems.
It is impossible to perfectly synchronize distributed systems (at least cost effectively), so alternative approaches are used to quantify and minimize the clock skew and clock drift within monitoring systems. Several approaches are used today to provide temporal context across multiple independent metering devices, each with its own constraints and accuracy.
On-board Clocks – On-board clocks are the most prevalent form of time keeping in metering devices. Because they are generally integrated into the metering hardware, they are essentially free with the cost of the meter. On-board clocks are often set by humans with wristwatches when a device is configured, so precision cannot be presumed. On-board clocks are also susceptible to many influences that affect the rate at which they keep time (i.e., clock drift). The vast majority of on-board meter clocks experience drift due to internal and external influences such as power source stability, quality of components, temperature variations, and so forth. Moreover, on-board clocks may not incorporate regional or national time changes, so clocks may lose or gain an hour due each year due to daylight savings time. The expected accuracies of on-board clocks are unpredictable, and error could easily range from seconds to days.
PC-based Clocks – This approach uses the head-end software to periodically update individual device clocks. At some regular interval (e.g., midnight), the software writes the onboard time from its clock in the time registers of each connected discrete device. There will typically be propagation delays (e.g., due to communications lag) between when the time data was sent from the software and when the device’s onboard clock is actually reset with the time data to match the software. Ethernet-connected devices will typically experience smaller lag times (milliseconds) than serially-connected devices (may be greater than a second).
Time Protocols – There are several time protocols for Ethernet-connected devices that may be used to improve time keeping. A few examples include Network Time Protocol (NTP), Simple Network Time Protocol (SNTP), and Precision Time Protocol (PTP) among others. Each of these protocols can be used to provide varying degrees of time accuracy for a discrete device, and the cost and constraints will vary with each. The time accuracy of these may range from milliseconds to microseconds for Ethernet-connected devices, with some protocols (e.g., PTP) requiring additional hardware in each capable metering device. Because these are internet protocols, gateways with serially-connected devices will still be limited in their ability to precisely correct clock skew. If a device is Ethernet-connected, it is possible to achieve accuracies in the microsecond range.
Global Positioning Satellites (GPS) – This method involves introducing a special time signal into the meters from an external time source (e.g., a GPS receiver). In this case, satellites broadcast an accurate and stable time signal from their on-board atomic clocks to earth. The GPS antenna on earth acquires the signal and passes it to the metering device through the GPS receiver. Each metering device receives the GPS-sourced time signal at approximately the same time, and updates its respective on-board clock accordingly. When an event occurs, the time stamped event data will be properly sequenced because the on-board metering clocks are nearly synchronous. This method is very accurate, but requires special firmware and hardware (GPS antenna, receiver, cabling, I/O cards) to acquire, disseminate, and incorporate the external time source signal. Synchrophasor technologies generally use GPS time data to time stamp measurements in near real-time. The accuracy is in the microsecond to millisecond range, depending on the hardware being used in the devices.
General Packet Radio Service (GPRS) – GPRS provides the ability to transfer data over cellular networks via terrestrial cell towers. By using a single cell tower with a satisfactory signal, time information may be transferred to connected devices efficiently and accurately. Again, special hardware is required to assimilate time data using this method, which will increase the overall system cost.
The Cost of Time
As alluded to above, accurate time keeping across a monitoring system is subject to several considerations. The metering hardware must be capable of receiving and incorporating time reference signals (e.g., SNTP, PTP, IRIG-B, DCF77, etc.). This may be impossible with some devices and easy to configure in others. If accurate time keeping is critical for your operation, it is important to review the hardware specifications to ensure this feature may be facilitated.
Devices without direct Ethernet connectivity (i.e., serially-connected devices) will experience greater propagation delays if the time data is provided through Ethernet gateways. In some cases, this delay may be overcome if the device is able to circumvent the normal communications network and integrate a GPS time synchronization signal directly into the device.
Cost and availability are two important factors when considering implementation of accurate time keeping in metering systems. Bringing in external time signals can be expensive, requiring receivers, cabling, special hardware, capable meters, and system configuration. If Ethernet time protocols are leveraged, it may require legacy meters and/or serial communications systems be replaced. Ultimately, the cost of aligning data from distinctive devices will correspond with the level of synchronicity needed.
As the number of data collecting devices increases, the availability, accuracy and alignment of system clocks becomes more essential. Access to synchronous real-time parameters by utilities and consumers alike will become more critical for the safe and reliable operation of the electrical grid.
As the electrical grid becomes “smarter” (e.g., smart grids, meters and loads), applications will require redundant time keeping techniques to ensure data synchronicity, facilitating optimal control of the electrical system. Resolving systematic issues on electrical systems also generally require accurate sequencing of events based on time of occurrence. “Time” is the fundamental thread weaving electrical systems together from the generator to the outlet. Just as extensive acquisition of data has become necessary, so too the need for improved data synchronicity across metered systems.
Jon Bickel is a fellow engineer, Power Applications at Schneider Electric.
IS YOUR CLOCK KEEPING THE CORRECT TIME?
An important technique in maintaining a stable energy grid is balancing the energy produced by utilities (generation) with the energy consumed by their customers (load). This allows better stability and control of the system frequency and voltage, and thus, the electric grid. Because loads are always changing and the generators are constantly seeking to meet those changes, the grid frequency is always varying slightly over time (between 59.95-60.05 Hertz).
Some electric clocks connected to the grid (e.g., ovens, coffeemakers, etc.) use the grid’s frequency to control their time (i.e., one second per 60 voltage cycles). Traditionally, utilities and regulatory agencies have sought to help support the time accuracy of these electric clocks by adjusting the system frequency to maintain their correct time over a given period (e.g., 1-5 days). This precedent has prompted valid concerns that adjusting the system frequency over quicker intervals merely to help maintain the accuracy of electric clocks may be unwarranted and put the grid at greater risk of instability issues in certain situations.
In late 2016, the Federal Energy Regulatory Commission (FERC) filed a petition to eliminate regulations governing time error correction, trading off the accuracy of grid-connected clocks with the ability to better manage utility grids (with the goal of improved stability). Since the retirement of these regulations earlier this year, grid-connected clocks may be running a bit faster or a bit slower. The impact over the course of a day may not be noticeable, but it may be prudent to start checking the accuracy of your grid-connected clocks a few times to be on the safe side.