IEEE Power Electronics Magazine - December 2014 - 18

need for higher reliability is driving fault-location identification and system restoration, allowing autonomous
decisions on changing the network topology following a
fault, thereby putting volt-var control assets in nonoptimal locations. Overall, this has created a very complex
operating environment, triggering massive investments
in integrated volt-var control, where network topology,
predicted load and sunshine, and grid assets, along with
their operating constraints, are all integrated into a massive optimization program, providing actionable commands to LTCs and capacitor banks. This has still been
treated as a centralized command and control problem
where AMI or other data drive complex nonlinear optimization computation, resulting in commands to switch
assets to achieve an overall cost function.
It is interesting to note that almost without exception,
this has been treated as a primary-side problem. However, the real goal is to manage the secondary-side voltage for thousands of customers to be within ANSI limits.
The AMI data are revealing an unexpected surprise. The
voltage drop DV across distribution transformers, always
assumed to be minimal, is unexpectedly high (2-13 V on
a 240-V base-more than half the ANSI band), mainly

245
2V

Volts

13 V
240

235

230

Nodes
(a)
245

Volts

10 V

240

235

230

Nodes

fig 2 (a) The minimum voltage occurs at node 13 at 10 p.m.
(b) The minimum voltage occurs at node 29 at 11 p.m. The
green line shows an estimate of the primary-side voltage, and
the dark blue line shows the measured secondary-voltage profile along an actual feeder with 33 nodes.

IEEE PowEr ElEctronIcs MagazInE

DV . I L X L 1 - PF 2 + I L R L PF,
where DV = line and leakage drops, I L = line and load current, X L = line and leakage inductance, PF = power factor,
and R L = line and transformer resistance. For real transformers, as the PF decreases below 0.98-0.99, the voltage
drop across X L becomes dominant. For a 25-kVA transformer at 1-pu loading and PF = 0.8, the voltage drop
across the transformer is >10 V. Figure 2(a) and (b) show
the real measured secondary-side voltage data for a typical
feeder at 10 p.m. and again at 11 p.m., showing that the lowest-voltage node is not at the end of the feeder and that it
moves continuously with local load. Further, it can be seen
that the drop across the distribution transformer keeps
changing over time and has a large range of 2-13 V, depending on transformer loading and instantaneous power factor.
Figure 3(a) shows the actual measured voltage profile
along a feeder over an 8-h period, when the substation voltage is set at 1.035 pu, a normal practice that allows utilities
to keep low-voltage points within the ANSI band. The node
voltages are generally within-band, only infrequently going
below the ANSI band. Figure 3(b) shows the same feeder
on a similar day, where the substation voltage is reduced
by 3% for energy conservation. More than 50% of the measured nodes now see persistent low voltages, well below
the ANSI band.
The data show high variability and completely uncorrelated behavior from node to node. If centralized command and control were to be used for ANSI compliance,
the substation voltage would have to be raised to a1.04
pu, completely defeating the energy-conservation objective and raising the possibility of overvoltage for a lightly
loaded feeder with high PV penetration. This is a highly
distributed problem that centralized command and
control based on electromechanical controls with slow
response time and limited actuation life cannot fully
solve. What is needed is an additional layer of control
at the grid edge that acts in a distributed, decentralized
manner to decouple local secondary-side voltage issues,
allowing the overall centralized command and control
assets to achieve the targeted system-level objectives, as
shown in Figure 4.

Grid-Edge Control Using Power Electronics

(b)

because of the drop across the transformer leakage reactance. Further, this voltage varies continuously and rapidly
with load current ^ I Lh and power factor (PF), as per the
equation below, and is not aligned with the average loading
or PF measured at the feeder:

z December 2014

A near-term need for utilities is a cost-effective distributed
voltage control at the grid edge. The dynamics of subcycle
voltage volatility and continuously varying correction suggest that an electromechanical solution may not work and
a power electronics solution is needed. Distribution transformers in North America are rated 15-100 kVA for

Table of Contents for the Digital Edition of IEEE Power Electronics Magazine - December 2014