IEEE Power & Energy Magazine - May/June 2017 - 60

18
:0
0
21
:0
0

00
15
:

00
12
:

00
09
:

00
06
:

00
:

1.9
1.8
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1.0

03
:

Load Model-np

customers) are dominant. At peak time (6:00-7:00 p.m.),
lighting appliances are the largest component. This information, in turn, enables the component-based approach to
estimate the aggregated time-varying load model [shown in
Figure 6(c)] without the need for actual measurements, thus
making it possible to assess any substation where the number of residential customers and peak demand are known.
The measurement-based load model was developed for
each of the 60 primary substations using data from field trials
carried out within CLASS throughout a year. In these trials,
both parallel transformers were tapped (on specific days and
at specific times) at one to three positions so as to reduce voltages by approximately 1.5-5%. The new tap position was held
for 15 min to capture the corresponding voltage and demand
changes. The noise and error in the measurements were then
removed using a filtering process. Finally, using a load model
formulation (chosen based on its practicality and the resolution of the available data), a curve-fitting technique determined the parameters that best matched the measurements.
Given that different load compositions have different
responses to voltage changes, the primary substations were
also categorized into 1) mainly residential, 2) nonresidential, and 3) mixed based on their expected demand share at
peak time. The resulting time-varying load models of primary substations within the same category were found to
be, in general, similar. In addition, due to the averaging process adopted to obtain these results, the intraday variability within a season (e.g., winter in Figure 7) is smaller than
the one found for a single primary substation [e.g., mainly
residential in Figure 6(c)]. Nonetheless, throughout the year
there are significant variations: np values that fluctuate from
0.67 (autumn) to 2.11 (summer) for mainly residential; from
0.86 (winter) to 1.98 (autumn) for mainly nonresidential; and
from 0.7 (winter/summer/autumn) to 1.91 (winter) for mixed.
More details and discussions of the results are available in
the CLASS final report.

Time (hh:mm), 24 h
Nonresidential
Mainly Residential
Mixed

figure 7. Field-trial results: average time-varying load
model per substation type (winter).
60

ieee power & energy magazine

The values provided by the component- and measurement-based load models were relatively close when considering mainly residential primary substations. Given that
the component-based approach considered only residential
demand (for the sake of simplicity, due to the challenging
task of modeling commercial and industrial demand), this
provided a degree of validation. However, the larger discrepancies found with the other two categories of primary
substations highlighted the contribution of nonresidential
demand. To account for these discrepancies, the componentbased load model was enhanced by embedding the measurement-based model for nonresidential substations, thus creating a more flexible and reliable load model applicable to any
substation without the need for extra measurements.

Voltage Capability
The extent to which the voltage at a primary substation can
be reduced (i.e., voltage capability) is limited by downstream
LV customers, whose voltages cannot be provided below
statutory limits, and also affected by upstream networks. For
this purpose, a three-stage approach quantifies the voltage
capability, taking into account these influences by considering, simultaneously, EHV and LV networks, as summarized
in Figure 5 and described in the following.
✔ EHV influence. Voltages in EHV networks affect the
tap position of primary substations, which, in turn,
dictates the number of available tap positions. For this
purpose, a real EHV network model was used to estimate voltages at the primary side of primary substation transformers (the blue square in Figure 8) and,
ultimately, the corresponding OLTC tap headroom.
✔ LV influence. Ensuring that voltages of LV customers are above the lower statutory limit determines the
extent to which voltages at primary substations can be
reduced. For this purpose, a set of 57 real LV feeders were analyzed using a Monte Carlo approach, in
which the lowest LV busbar voltage (the green square
in Figure 8) that does not affect customers is statistically quantified considering the variability in the
residential demand profiles. This allows DNOs to
realistically quantify the impact different LV busbar
voltages can have on customers, which is essential for
the voltage-capability quantification.
✔ HV study. Lastly, considering the EHV and LV influences as well as the time-varying load models and
demand profiles developed previously, a power-flow
analysis is carried out on the HV network associated
with the primary substation as a way to ultimately
quantify the corresponding voltage capability and potential demand reduction.
The volume of demand reduction that can be achieved
at a given primary substation is variable on a daily and
seasonal basis. Relevant factors include demand composition and interactions across voltage levels. For instance, a
mainly residential primary substation with a peak demand
may/june 2017

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - May/June 2017