IEEE 519-2022 does not compare each reading with the limit. From the 3 s values accumulated over a day it calculates the daily 99th percentile, and from the 10 min values accumulated over a week it calculates the weekly 95th and 99th percentiles, always by linear interpolation. Each one is compared with the table multiplied by 1.0, 1.5 or 2.0.
Short answer
IEEE 519-2022 does not require that no reading exceed the limit. It requires calculating percentiles of the records and comparing them with the limits. The 3-second values are accumulated over a day and their 99th percentile is calculated. The 10-minute values are accumulated over a week and their 95th and 99th percentiles are calculated. Each percentile is compared with the table multiplied by a different factor.
What a percentile is
The 99th percentile is the value exceeded only during 1 % of the period. The 95th percentile is the value exceeded during 5 %. To obtain them, the measured values are sorted from lowest to highest and the point that leaves 99 % or 95 % below it is located.
It is a way to ignore the highest peaks. A load that starts a couple of times a day generates very high values for a few seconds; the 99th percentile of the 3-second values leaves out the highest 1 % of each day and measures the level the installation maintains almost all the time.
Clause 4.4 indicates that percentiles are calculated with a linear interpolation algorithm: if the percentile position falls between two ordered values, it is interpolated linearly between them, instead of taking the nearest one.
What is accumulated and what is calculated
Clause 4.4 defines the periods and the percentiles:
| Starting values | Accumulated over | Percentile calculated |
|---|---|---|
| Very short time (3 s) | One day | 99th percentile, for each 24-hour period |
| Short time (10 min) | One week | 95th and 99th percentiles, for each 7-day period |
These statistics are used for both voltage and current harmonics, with one exception: the weekly 99th percentile of short-time values is not recommended for voltage harmonics.
How the 3-second and 10-minute values are built is in Measuring Harmonics for IEEE 519: the 12-Cycle Window. Here we start from the assumption that they are already available.
Which factor each one is compared with
Clauses 5.1, 5.3, 5.4 and 5.5 give the factor that multiplies the table value in each criterion:
| What is compared | Voltage (Table 1) | Current (Tables 2, 3 and 4) |
|---|---|---|
| Daily 99th percentile, 3 s values | Less than 1.5 times the table | Less than 2.0 times the table |
| Weekly 99th percentile, 10 min values | Not recommended | Less than 1.5 times the table |
| Weekly 95th percentile, 10 min values | Less than the table | Less than the table |
The pattern is that the shorter the averaged value, the larger the factor: 2.0 for the 3-second values and 1.5 for the 10-minute values. The weekly 95th percentile is the only one compared directly with the table value.
Illustrative example
The numbers below are made up to show the comparison. They do not correspond to a real installation.
A plant at 13.2 kV with an Isc/IL ratio of 30 is in the “20 to less than 50” row of Table 2. The 5th current harmonic is in the range 2 to less than 11, with a table value of 7.0 % of IL. The limits are as follows:
| Criterion | Limit | Record value (made up) | Result |
|---|---|---|---|
| Weekly 95th percentile of 10 min | less than 7.0 % | 6.2 % | Meets |
| Weekly 99th percentile of 10 min | less than 10.5 % (1.5 × 7.0) | 9.1 % | Meets |
| Daily 99th percentile of 3 s | less than 14.0 % (2.0 × 7.0) | 12.8 % | Meets |
If the weekly 95th percentile had come out at 7.4 %, the criterion would not be met even though the other two are, and the installation exceeds the limit for that harmonic. Each criterion is evaluated separately, for each harmonic.
For voltage, with a PCC at 13.2 kV and the 5th harmonic, the Table 1 value is 3.0 %: the weekly 95th percentile of 10 min must stay below 3.0 % and the daily 99th percentile of 3 s below 4.5 %.
How much data is behind it
A full day of 3-second values has 28,800 values, and the 99th percentile leaves about 288 above it. A full week of 10-minute values has 1,008 values: the 95th percentile leaves about 50 above it and the 99th, about 10. These are arithmetic counts over an uninterrupted record; a record with gaps has less data.
That is why the record should be continuous: a percentile calculated over a few hours is worth little compared with a full week. The standard defines the evaluation over periods of one day and of seven days.
Possible reading errors
- Comparing the maximum with the table. The maximum of the record is not the standard’s criterion; the percentile is.
- Applying the same factor to everything. There are three different factors depending on the criterion and the type of quantity.
- Using the weekly 99th percentile for voltage. The standard does not recommend it for voltage harmonics.
- Mixing periods. The daily percentile is calculated from 3 s values; the weekly ones, from 10 min values.
The values they are compared against are in IEEE 519-2022 Voltage Harmonic Limits at the PCC and Harmonic Current Limits and TDD by Isc/IL; the point where they apply, in PCC in IEEE 519: Where Harmonic Limits Apply. Our free harmonic distortion THD and TDD tool compares a spot spectrum with the table; the percentile evaluation is done over the complete record, as explained here.
We help with the evaluation
At Tiza Group we carry out power quality studies, in Colombia, or elsewhere on request: we process the record of your PCC with the IEEE 519-2022 percentiles, compare it with the limits that apply to you and deliver a report that shows which criterion is met and which is not. If you need to measure, see also Power Quality Analyzer.
Related guides
- Measuring Harmonics for IEEE 519: the 12-Cycle Window
- IEEE 519-2022 Voltage Harmonic Limits at the PCC
- Harmonic Current Limits and TDD by Isc/IL
- Power Quality Analyzer
- IEEE 519: Harmonic Limits and Key Concepts
References
- IEEE Std 519-2022, IEEE Standard for Harmonic Control in Electric Power Systems: clause 4.4 (periods of one day and one week, 95th and 99th percentiles, linear interpolation); clause 5.1 and Table 1 (factor 1.5 for voltage); clauses 5.3 to 5.5 and Tables 2 to 4 (factors 2.0 and 1.5 for current).
Frequently asked questions
What does it mean that the 99th percentile is below the limit?
That the value exceeded only during 1 % of the period is below the limit. Put the other way: up to 1 % of the values may exceed it without non-compliance under that criterion.
Why is the weekly 99th percentile not used for voltage?
Clause 4.4 clarifies that the weekly 99th percentile of short-time values is not recommended for voltage harmonics. For voltage, the daily 99th percentile of 3 s and the weekly 95th percentile of 10 min are evaluated.
How are percentiles calculated if the value falls between two data points?
The standard indicates that percentiles are calculated with a linear interpolation algorithm: when the percentile position falls between two ordered values, it is interpolated linearly between them.