IEEE Power & Energy Magazine - November/December 2017 - 42

In the literature, hundreds of ways to improve forecast
quality are described. One concept is based on the use of different NWPs as well as different weather-to-power transformation models, along with the subsequent combination of the
resulting power forecasts. At time horizons of up to approximately 5 h, the assimilation of power measurements into the
power forecast models leads to further significant reductions
in forecast errors. These strategies are well established within
the model chains of forecast developers.
Most extreme forecast errors result from erroneous interpretation and propagation of atmospheric processes within
the NWP models. To address such challenges, research
projects focus on improving NWP quality specific to the
RES forecasts. Within the publicly funded German project
EWeLiNE (http://www.projekt-eweline.de/en/index.html)
and the U.S. Department of Energy-sponsored Wind Fore cast
Improvement Project (WFIP)/Solar Forecast Improvement Project (https://www.esrl.noaa.gov/gsd/renewable/
wind.html), researchers have already successfully shown
the benefit of such focused research. Successful results
include improved predictions of nightly low-level jets
and low-stratus cloud situations, leading to better RES
power forecasts.
However, most of these improvements have emphasized
deterministic power forecasts tuned for the best average
performance in terms of an expectation value. Today, such
forecasts have undergone many improvements to achieve
high predictive quality, but it is important that forecasts additionally quantify their uncertainty or warn of alternative
weather conditions.

Interfacing Forecast Systems
Operators today are increasingly faced with managing rapidly
changing energy delivery conditions due to weather-dependent
RES having different uncertainty characteristics that depend
on the weather, time of day, and season of the year. When
system operators are asked what they optimally want from a
forecasting system, the answer invariably is advanced warning of pertinent events and operating guidance so that network
configuration and controllable generation resources can be
efficiently managed and dispatched.
Recent events, as shown in Figure 1, illustrate the types
of short-duration, difficult-to-predict weather conditions that
challenge even the most sophisticated forecasting tools. These
events represent the extreme conditions (or "tails") where forecasting is very difficult. But understanding limitations and coupling new visual capabilities that complement existing forecasts
can provide enhanced situational awareness and a bird's eye
view of changing local and regional solar and wind conditions
to support new operating response.
Working with state-of-the-art forecasts, energy management system (EMS) providers are coupling solar and wind
integrated forecast tools (SWIFT) and using a network of resource sensors to manage island systems. Hawaii operators,
for example, are routinely operating with 40% or greater
42

ieee power & energy magazine

penetration of wind and solar resources on the grid. Efforts
are under way to integrate system-, regional-, and circuit-level
probabilistic forecasts from SWIFT into EMS and distribution management algorithms. Load estimation and state estimation algorithms require more intelligence and data from
distributed energy resources along with load and renewable
forecasts to support dynamic dispatch needs, especially at
high penetrations. Targeted and tailored forecasts that hone in
on event-driven conditions and EMS playback tools will help
operators develop new strategies for handling uncertainty
in forecasts, planning reserves, and addressing contingencies that account for variability under related uncertainty
conditions. Utilities and their suppliers are actively working
together to refine and integrate these forecast capabilities
and establish a more proactive and complete picture of predictable events, potential impacts, and mitigating actions-all to
better manage the grid.
To relate weather and RES forecasts, a geographic graphical user interface has been developed within both the
EWeliNE and WFIP initiatives that enables the synchronized display of spatial and temporal weather and power
forecasts (Figure 2). Wind and solar power forecasts for
an entire system, an area, or individual transformer substations can be visualized on a map, together with selected
meteorological parameters such as wind speed, solar irradiance, resource variability, and temperature. For some
meteorological parameters, there are also probabilistic forecasts-such as the probability of exceeding wind speed and
cloudiness thresholds as well as low-stratus-cloud risk levels-that provide a first warning of weather situations that
could impact grid stability.

How Forecast Quality Impacts
Real-Time Operations
The prevailing practice in integrating RES is to procure a certain amount of reserve to compensate for the uncertainties of
net load, which is directly linked to RES forecast errors. Such
reserves are mainly affected by demand, wind, and solar
forecasting errors and unplanned outages of thermal groups. In
this sense, the quality of RES forecasting has a great influence on both reliability and operation efficiency. To procure
excessive reserve would lead to an inefficient operation,
while an inadequate amount of reserve could result in a
potential reliability issue. However, as RES forecast errors are
time varying, some judgment should be applied to determine
an appropriate tradeoff between economics and risk management. At the Electrical Reliability Council of Texas (ERCOT),
for example, the volume of the reserve is sized based on the
net-load forecast errors with a confidence level assigned as a
function of the net-load ramp magnitude. Figure 3 depicts a
3-h ahead load-forecast error versus a 3-h ahead wind-forecast
error in 2016 at ERCOT. A large amount of wind generation
greatly exaggerates the forecast error of net load at ERCOT.
Information about future forecast uncertainty already
plays a crucial role in safeguarding the security of supply.
november/december 2017

http://www.projekt-eweline.de/en/index.html https://www.esrl.noaa.gov/gsd/renewable/

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2017