CVR/VO subcommittee follow-up: Performance Thresholds May

CVR/VO subcommittee follow-up: Performance Thresholds
May, 2015
Josh Rushton
RTF Contract Analyst
The purpose of this note is to frame and organize technical questions related to performance thresholds
used in the November 2012 draft of RTF’s Simplified Voltage Optimization Standard Protocol (VO
Protocol). The central theme of these questions is this: What performance thresholds are needed to
“reliably” estimate the average ΔV for a feeder using the formula in the protocol?
Here, reliably should be read in the sense of “right-on-average.” If the formula gets ΔV too high for
some feeders, and too low for others, it may still be sufficiently reliable for claiming conservation credit
provided it is not systematically biased.
Also, note that the RTF Staff/CAT plan is to recast necessary requirements as measure eligibility
requirements, rather than performance thresholds.1 This means that only those systems that meet all
of the requirements should use the VO Protocol. Other systems would default to custom evaluation
protocol. In some cases, these custom protocols may incorporate elements of the VO Protocol (such as
the VO factor tables); as described in Section 4.3.3 of the RTF Savings Guidelines, the practitioner is
responsible for determining reliable analysis methods for custom protocols on a case-by-case basis.
Simplified Voltage Calculations in the VO Protocol
The VO Protocol estimates a feeder’s average annual voltage based on one week of one-hour interval
metering at the feeder’s source and at EOL.2 The voltage formula for feeders using fixed voltage
reduction3 is as follows:
𝑉 = 𝑉𝑆𝑒𝑑 − ∗ 𝐴𝑣𝑒 (𝑉𝑂𝑒𝑑, 𝑖 − 𝑉𝐸𝑂𝐿, 𝑖 ) ∗
= Regulator set point voltage setting
= Hour-i metered regulator output voltage on 120 V base
To avoid losing valuable information, we will also try to find a home for the system improvement steps (cast as
recommendations, maybe in an appendix).
The protocol defines the “EOL” location to be point on the feeder that typically experiences he lowest voltages,
not necessarily the literal endpoint of the line.
For systems using line drop compensation or automated voltage feedback control, the protocol’s voltage formula
includes a correction for volt rise.
= Hour-i metered EOL primary voltage on 120 V base
π·π‘Žπ‘›π‘›π‘’π‘Žπ‘™ = Average annual kW demand (based on measured historical data)
π·π‘šπ‘’π‘‘π‘’π‘Ÿ = Average kW demand during the one-week metering period (metered at source)
The protocol applies this formula separately pre and post (with a full week of metering for each case)
and takes the difference to estimate the average annual ΔV for customers on the feeder.
The formula’s central assumption is that voltage drop is approximately linear (so the average voltage
drop experienced along the feeder is half of the full voltage drop from source to EOL). It also assumes
the average voltage drop changes in proportion to demand.
Performance thresholds
The simplified voltage calculation assumes a linear voltage drop along the length of the feeder, and it
assumes the average voltage drop is proportional to demand. These assumptions do not hold for all
systems. For example, in the subcommittee call, Bob Fletcher noted that they found negative savings
for some feeders which were discovered to have poor power factors (around 80%). The question we
need to answer is what threshold requirements are essential to getting unbiased estimates of ΔV.
Note that poor system performance can threaten the ΔV either by directly invalidating the linearity
assumption, or by causing low-voltage complaints that may result in a utility reverting to the highervoltage settings that prevailed prior to the VO measure. The “right” set of performance thresholds
should mitigate both of these risks.
Also, while we would like to determine the minimum essential requirement for each performance
metric, we will probably have to guess at some of the boundaries. In such cases, it would be interesting
to know something about the distribution of values found in the region. For instance, if systems with
phase imbalances over 20% are extremely rare, we would not want to waste a lot of time trying to
determine whether the protocol will work with phase imbalances of 30%. Insights of this sort are
welcome too.
In summary, the three specific questions we should ask of each threshold are:
1. Can the requirement be relaxed without threatening assumptions in the ΔV calculation?
2. Can the requirement be relaxed without threatening power-quality complaints when the
measure is activated?
3. How common are systems that narrowly miss the requirement? (How much might be gained by
fiddling with the boundary?)
Below are the threshold requirements currently stated in the VO Protocol, along with comments
collected in the May subcommittee meeting. Please provide additional comments, especially if you have
insights related to the three questions just identified.
Power factor (3-phase total, at source):
– Minimum (hourly) greater than 0.96
Average (for week) greater than 0.98
Subcommittee unclear about extent to which these can be stretched. The cut-offs used
in the protocol may have been based on the mean power factor of the “acceptable”
portion of the sample used in the NEEA study? Should the lower quartile be referenced
instead, since the cut-offs set a minimum threshold?
– Need to work with NEEA on this (but all commenters are welcome to weigh in).
Phase load balance (3-phase lines, at source)
– Per-unit unbalance < 0.15
– Neutral < 40 amps
– Subcommittee okay with allowing imbalances up to at least 0.20, and dropping the
neutral lead requirement (it’s probably redundant anyway). Is the requirement that
remains essential to getting a reliable ΔV estimate?
Max-adjusted voltage drop (3-phase mean). (Max-adjusted drop is mean meter-period drop,
times annual peak kW, divided by mean meter-period kW)
– Primary max-adjusted drop < 3.3%
– Secondary max-adjusted drop < 4.0%
– Power factor comments probably apply here too.
Variation between feeder max voltage drops
– Compare feeders within substation control zone
– Must not differ by more than 2 Volts (on 120 V base)
– Subcommittee didn’t discuss.
Primary line minimum hourly voltage (measured near feeder’s expected low voltage point)
– At least 114 V + (1/2) Voltage regulation bandwidth + secondary max allowed voltage
– Subcommittee didn’t discuss. Josh’s comments: This is probably more of a measure
definition thing (be sure the voltage setting is in this range); this should probably be a
requirement for efficient-case system. This is part of ANSI C84.1 standard. Our measure
needs it because of question 2 above.
Primary line maximum hourly voltage (measured near expected high voltage point)
– Less than 126 V - (1/2) Voltage regulation bandwidth
– Subcommittee didn’t discuss. Josh’s comments: This is probably more of a measure
definition thing (be sure the voltage setting is in this range); this should probably be a
requirement for efficient-case system. This is part of ANSI C84.1 standard. Our measure
needs it because of question 2 above.
Conductor loading
– Source hourly loading (amps) less than design normal spec
– Subcommittee didn’t discuss. Josh’s comment: This is important for safety and system
reliability, but those concerns aren’t necessarily relevant to measure savings. However,
we will need to define what makes a system is obsolete (improvements that would
impact the measure are inevitable in the near term). This will be relevant to measure
life and/or eligibility. These questions are important, but for now we need to focus on
conditions needed for reliable savings estimates.