IEEE Power & Energy Magazine - March/April 2018 - 74

Bandcenter (V)

126
125
124
123
122
121
120
119
118
117
116
16:00

Typical DSDR Implementation

16:34

17:07 17:41 18:15
Time (hh:mm)
Setpoint

18:48

19:22

Moving Average

figure 7. The voltage regulator bandcenter scatter plot.

necessary circuit improvements can potentially provide a
secondary benefit and should not be overlooked. DEP realized a reduction of summer peak active power losses totaling an estimated 6 Mw, which equates to approximately
30,000 Mwh of energy savings on an annual basis.
another essential prerequisite for being able to optimize
per-phase distribution voltage is a back-end control system
with advanced functionality capable of maintaining the
maximum reduction due to the dynamic nature of the grid.
this is where aDMS came into play-with a model of the
entire grid topology and corresponding equipment characteristics, typical customer load profiles, reoccurring LF and
SE processes with near-real-time SCaDa measurement
input, and VVO. aDMS was in its relative infancy at the
beginning of the project, and each of these functions was
available at different levels as part its base package. Some
required considerable design, evaluation, and modification
for the desired requirements of DSDr.
a key component that should be mentioned is the quality of the distribution grid data provided to the aDMS
model. the input from gIS-a common mapping tool for
most utilities-may be sufficient, depending on the accuracy. DEP had done a reasonably good job of documenting
its grid attributes within its gIS system but also found that
aDMS demanded an additional level of detail, resulting in
the previously discussed field and phasing verifications. an
elaboration of the data quality element is warranted, as it has
the potential to significantly impact the results of VVO and,
ultimately, led to a change for DSDr.
the VVO algorithm bases its control decisions on its
LF/SE values for the grid versus actual SCaDa measurements, and it is possible for model errors to cause a deviation
between the two. the same can be said for grid device measurements with old (due to lost communications) or erroneous values. aDMS does have the ability perform checks to
identify and exclude any such values from its SE process.
74

ieee power & energy magazine

this was not the case for model data, which were initially
incorrect, and DEP faced this issue with its LV portion of the
grid. the original intent was for DSDr to optimize based
on its LF/SE results for customer voltage because this is
the most limiting constraint when attempting peak demand
reduction. However, DEP's service size and length information from gIS was found to have the least amount of accuracy while also being the most difficult to verify because of
quantity and underground installations. the impact on the
quality of the LV results was evident and led to concerns
with regard to the effectiveness of DSDr.
an enhancement was consequently added to VVO that
provided an option of optimization based on the voltage
results for the MV network, which had better overall quality
and a greater quantity of measurements from the grid. a lesson learned from this experience is that a deviation between
LF/SE and SCaDa measurements is to be expected for multiple reasons (model quality, measurement accuracy, etc.) and
utilities should develop appropriate circuit evaluation metrics
based on their comfort level and specific use of VVO.
VVO command of voltage regulators during DSDr was
another important development. Multiple options were considered during its design, but DEP ultimately decided to change
the regulator control voltage bandcenter setting. a primary factor in the decision was the total number of commands required
for a maximum voltage reduction that included almost 4,000
voltage regulators-with individual tap lower commands approximately five times as many commands would be needed
during initial activation-as well as the corresponding demand
on SCaDa and the cost of cellular communications bandwidth. a secondary benefit of bandcenter command was that
the control could continue to regulate voltage in an automatic
fashion and therefore respond more quickly to significant variations, versus waiting for the next VVO retrigger.
while there was not necessarily a speed requirement,
there was an operational expectation for such a resource
to deliver its benefit in a reasonable and consistent amount
of time. the inclusion of VVO command validation-
advanced functionality that verifies successful change of the
setting within a specified time frame and performs retries
if selected-was deemed necessary for true optimization
but also caused apprehension due to the potential for added
implementation time.
these decisions-as with all aspects involved in the creation of DSDr-necessitated a substantial amount of testing
and fine-tuning of aDMS functionality, infrastructure, and
grid devices. the end result was the delivery of a state-ofthe-art system that successfully met the DSDr business case
by achieving an over 3.6% optimized voltage reduction during both summer and winter peak seasons. It is capable of
completing the initial mode transition in 30 min or fewer,
while averaging more than 40,000 total commands with a
success rate consistently over 98%. a scatter plot of typical DSDr bandcenter setting values over an entire activation
is shown in Figure 7. the voltage regulator command rate
march/april 2018

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - March/April 2018