IEEE Power & Energy Magazine - November/December 2015 - 54

PG&E Bay Area
PG&E Non-Bay Area
SCE Coastal
SCE Inland
SDG&E

San Francisco

* 4.49 kW-ac
* SunPower Inverter
(SPR-5000X, 240 V)
* 27 Modules (SunPower
210 W, SPR-210-WHT)
* 37.76281° N,
122.44313° W
* Commissioned
April 2008

figure 3. Detailed site information and weather data, aggregated to fleet level, for the three major investor-owned utilities
in California. (Courtesy of Clean Power Research.)

Challenges, however, do exist for this approach. While most
hardware, location and orientation information is typically well
documented, dynamic elements about each PV installation such
as shading and hardware system availability will also impact the
accuracy of the prediction. Suitable approaches are being developed to account for these impacts.
In applications where the objective is forecasting aggregated PV production or net load across a large system and/
or where detailed data do not exist, other less detailed prediction methods can be deployed based on combinations of
existing data, such as substation loads and regional meteorological data. This "top-down" approach has the advantage of
limited data requirements. Whereas the bottom-up approach
uses validated models applied to each system, the less detailed
top-down approach requires comparisons to metered data on
a statistically significant number of sites for benchmarking its
overall accuracy. Challenges with this approach center on the
representation of the statistical sample to adequately depict the
total solar contribution. Ongoing research is focused on determining the detail of system data required based on the given
PV penetration. It is important to recognize that different end
users may have different requirements in terms of the amount
of detail needed.
54

ieee power & energy magazine

As an example, the forecasts for transmission system
operators and distribution system operators in Germany
are calibrated and evaluated against estimations of the
solar production on a zip code level (Figure 4). Forecasts of
aggregate BTM PV production are developed and validated
based on the up-scaling of output from a select subset of
representative PV sites. The real-time solar production is
directly obtained from the solar inverters of these installations through companies that install PV modules and
monitor their performance. This approach has two main
advantages: first, the grid operator can predict solar output
in real time, and, second, the forecasts can be optimized on
a regional level.
Comparisons between such estimations and metered data
show a very strong correlation. Essential to this approach
is consideration in the selection of representative sites, for
which sufficiently granular and reliable production data must
be available. Further, real-time telemetry is required if these
sites serve as the basis for nowcasting (i.e., within horizons
of one hour or less). The approach offers a good example of
the data and telemetry requirements that will be important
for system operators, distribution utilities, and others to consider in their planning and interconnection processes.
november/december 2015

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2015