IEEE Power & Energy Magazine - May/June 2020 - 63

of intervals are all important first steps in facilitating data
integrity. The ability to compare the total of the interval load
data to the corresponding billing month data will always
serve as the first line of defense. However, additional issues
such as zeros, valleys, gaps, and spikes still need to be monitored and corrected or eliminated appropriately. There are
a host of alternative strategies that the load researcher can
employ, but the automated VEE processes of meter data
management systems (MDMSs) may not be enough. Load
researchers would be well served to check and potentially
re-estimate data cleaned by an MDMS using more robust
strategies such as regression-based analysis.
Once the sample data are of the highest quality possible,
then the analysis, often referred to as the expansion of the
sample to the population, can begin. It has always been the
purview of the load researcher to ensure that proper weights
are applied to the sample data to protect the integrity of the
results. Of course, load researchers should always calculate and report variances, standard errors, and relative and
absolute precisions along with their estimates. These form
the very foundation of the "results" that can be summarized
tabularly and graphically in a wide array of forms during the
reporting cycle.
The last step of our load research lifecycle is "model,"
where the parameters from the current study are retained
for future use. This includes the coefficients of variation
and/or error ratios for the various demand variables of interest and provides insight and guidance for use in planning
future studies.
Historically, the load research lifecycle took 18-24
months to complete, assuming that a full year of data was
required, when sampling was used to develop an efficient
representation of the populations of interest. With the use
of AMI, assuming an ongoing retention of interval data collected, the lifecycle may shrink to 18-24 days or, perhaps,
18-24 h, and sampling will be unbounded. This does not
mean that, as an industry, we should abandon the principles
of the past, namely, the use of proper population identification and statistically valid sampling approaches; however,
today, our samples can be bigger and cover more domains,
allowing us to explore and use the data in a much richer form.

Interacting With Utility
Business Functions
In a recent survey conducted by DNV GL in 2018 covering 17
U.S.-based load research departments, the respondents were
asked to rank the importance that load research plays in servicing several utility business functions. Table 1 summarizes the
responses. Not surprisingly, the cost of service and rate design
ranked the highest. These were followed by fulfilling internal
and external data requests, load forecasting, special studies,
and demand-response planning and evaluation. Fulfilling internal and external data requests supports the notion of the load
research department as the internal load consultant to the rest
of the utility. The following sections briefly describe many of
may/june 2020

the current roles played by the load research department in
each of the functions listed in Table 1.

Cost of Service
The load research department has been responsible for developing the hourly or subhourly (i.e., 15-min) load profiles for
each rate (or tariff) class of interest. The load profiles are
used to develop the required factors that distribute cost allocations to the various customer classes. The allocation factors
can be as simple as the class or system peak demand of the
class or can be a mixture of variables. In the past, utilities
have reported using various cost-allocation factors, such as
the 12 coincident peak (12-CP) demands, the four-summer
CP (4-SCP) demands, the on- and off-peak energy use (utility
defined on-peak periods), the class contribution to the top 50,
100, or 200 h of system load, and something as elaborate as a
28-h allocator, which uses the 4-SCP peak periods, including
the three hours before and after each system peak.
table 1. The summary statistics for a recent load
research survey on the importance of load research
(1 = "not at all important" to
5 = "critically important").
Business
Functions

Average

Median

Number Number
of 5s
of 4s

Cost of service

4.8

Rate design

4.7

Fulfilling internal
data requests

3.9

Fulfilling external
data requests

3.5

Load forecasting

3.4

Special studies

3.2

Demand-response
planning and
evaluation

3.2

Key account
management

2.9

Energy efficiency
planning and
evaluation

2.9

Marketing studies

2.8

T&D planning

2.8

DER planning

2.6

Renewable
integration

2.5

Smart grid
initiatives

2.4

IRP

2.2

T&D: transmission and distribution; DER: distributed energy
resource; IRP: integrated resource planning.
ieee power & energy magazine

IEEE Power & Energy Magazine - May/June 2020

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - May/June 2020