A customer recently called me about the information on his power quality monitor screen. He said about a dozen different events were showing up, but the lights had just blinked once. He wanted to know what all of the other nonsense was supposed to tell him. Why did his process stop when the lights blinked off for a second? It was certainly a valid concern, but my attempts to explain the validity of the different types of transients and other events that were listed along with the sag (aka, the blink, or “dip” in other parts of the world) fell on deaf ears.


By the commonly used power quality standards, such as IEEE 1159 or IEC 61000-4-30, “a sag (dip) is a decrease to between 0.1 and 0.9 pu in rms voltage or current at the power frequency for durations of 0.5 cycles to 1 minute.” Right away we have some confusion, as “pu” or “per unit” is not a common term in many users’ vocabularies. If you multiply the number by 100, it can be thought of as a percentage. So, a sag is a reduction to somewhere between 10 and 90 percent of the nominal or normal voltage, whether it is 120 volts (V) nominal (12 to 108V) or 480V (48 to 432V) or even 345 kilovolts (kV) (34,500 to 310,500V).


This variation in the rms voltage has a companion on the high side, called a swell, when the voltage goes above 110 percent, or 1.1 pu. An interruption (or outage in the old terminology) is when voltage is below 10 percent of nominal. In some other parts of the world, 5 percent is the line for an interruption, but there aren’t many pieces of equipment that care whether it is 10 percent, 5 percent or 1 percent. The same thing happens: the equipment fails to operate.


The IEEE Std 1159 gives further labels or characteristics to the sag by how long it lasts: instantaneous, 0.5–30 cycles; momentary, 0.5–3 seconds; and, temporary, 3 seconds to 1 minute. In a few words, these labels help describe the seemingly complex waveforms and rms voltage plots. Some power quality instruments add another label upstream or downstream to finding the problem origin. This is very effective in determining where to go to fix the problem. 


Another IEEE standard, 1564, is a guide to further reduce a lot of data into indices. More specifically, 1564 methods for quantifying the severity of individual rms variation events, for quantifying the performance at a specific location (i.e., single-site indices) and for quantifying the performance of the whole system (i.e., system indices).


Another common question about sags is, “How does my site’s data compare to others?” This can be misleading because there are a lot of variables that can make a comparison to other data less meaningful. For example, the results of typical sag magnitudes from the Electric Power Research Institute (EPRI) Distribution Power Quality (DPQ) project in the 1990s produced a graph of percent magnitudes of typical values (as in Figure 1) from data recorded on distribution feeders. Would this be the same as on the welding line of an automobile factory outside Detroit or for a data center on the 57th floor of an New York office tower? Probably not. As many studies have shown, a large percentage of power quality disturbances originate within a facility. These events aren’t likely to have shown up in the DPQ data.


How long the lights were off is another critical parameter in trying to answer the customer’s question. It could actually have been a disruptive sag that was so short in duration and minor in magnitude that the blink wouldn’t even register in the eye of the person.


Figure 2 shows a 3-D plot of the rms variations at an industrial facility in central New Jersey, a heavily populated area with a residential, commercial and light industrial mix. There are no events between 90 and 110 percent of nominal because they programmed the instrument not to record between that range of variation. This graph says that 20 percent of the events were interruptions (<10 percent) while 47 percent were between 80–90 percent of nominal. On the duration side, 86 percent are only 10 cycles or less, and most of those are less than five cycles. Note that all of the lower remaining voltage events took longer than five cycles. This is a result of the distribution system protection operating times or the fault-clearing time.


Another potential confusion between the two data sets is the use of “temporal aggregation” or the combining events that occur within a given time window into a single event, counting it only once in the statistics. From the utility’s point of view, a recloser sequence can result in three or more interruptions as it tries to clear the fault before it decides to “lock out.” None of the events in Figure 2 were sustained interruptions from lockouts. Otherwise, there would have been a count in the 120 seconds, 0–10 percent magnitude block. However, if a temporal aggregation of 1 minute (as used in the EPRI DPQ project) were applied, the 24 percent of the events being interruptions would likely have been a much lower number.


You could just count the blinks and compare to the process interruptions, but knowing your equipment susceptibilities and the voltage profile provided by the electric utility—as well as what really is happening to the power quality—is the best way to keep your lights on and the process running.