IEEE Power & Energy Magazine - November/December 2017 - 47

Information about future forecast uncertainty
already plays a crucial role in safeguarding
the security of supply.

The integration of information from geographically distributed time series or from a grid of NWP is known to
improve the uncertainty forecast accuracy and is the focus
of recent research. The disadvantage is that this process
does not produce a spatiotemporal representation of forecast
uncertainty, nor is there a physical dependency on the results,
because it is based on past climatology.
The second process can be considered as postprocessing
of a set of NWP ensemble members, i.e., a set of NWP forecasts produced by perturbing the initial or boundary conditions or the result of different parameterization schemes
of one NWP model (the so-called multischeme approach) is
converted in a subsequent phase into power with a curvefitting method. The NWP ensemble is configured to represent the physical uncertainty of the weather ahead of time
rather than uncertainty as a function of past experience. In
practice, this means that the NWP ensembles, especially the
multischeme approach, are event driven, produce outliers, and
also catch extremes, even with return periods of 50 years.
This is clearly distinct from statistical methods because even
long time series of historic data contain too few extreme
events to impact the learning algorithms.
Statistically based ensemble forecasts result from statistical approaches based on copula theory applied to generate
scenarios from the probability distributions produced by a
statistical model. These scenarios have some similarity to
the physically based ensembles.
The uncertainty representation of both the physical and
statistical ensembles are built to capture spatial and temporal variability such as ramps and detect, with probability
characterization, possible outliers such as extreme events
above wind turbine cut-out wind speeds. Statistical scenarios, however, require an extreme event analysis to determine
these outliers.

Using Forecast Uncertainty
in Business Practices
In the International Energy Agency's wind task 36 (the
"Wind Energy Forecasting" work package 3; www.ieaforecasting.dk), the first overview concerning the current use
of wind power uncertainty forecasts shows that there are
many different levels of knowledge about the application of
uncertainty forecasts in the power industry today. In some
countries, regulations lack transparency, and insecurity is
spread among the market players and the investors; in other
countries, the wind penetration is not yet high enough for
november/december 2017

uncertainty in production to become a bottleneck to efficient
integration of renewables. It is, therefore, interesting to note
that uncertainty forecasting has been established most fully
where the penetration has exceeded a certain level (>30%
of gross annual energy supply) and the cost of integrating
RES has increased rapidly, such that a change of operating
practice was required for unit commitment and reserve allocation decisions-and for decision making in general.
Although it has been found in University of Washington
laboratory experiments that decisions based on uncertainty are
generally better, there is still insecurity in the industry regarding how to make use of uncertainty forecasts. Although risk
assessment and reserve allocation are based on probabilities
of exceeding a predefined limit, there are many organizations
that have difficulties establishing the tools and mechanisms to
deal with uncertainties and employ uncertainty forecasts for
more efficient use of reserves.
One aspect that is gaining more attention recently is the
use of uncertainty forecasts for situational awareness in
operations. Due to the variable output nature of RES, where
small differences in wind speed can have a large impact on
the power grid, operators in small island grids and those with
high penetration levels request information about the probability that strong ramps could lead to congestion or a shortage of power on the grid. This is especially important for
high-speed cut-out events or at the peak-load ramps in the
morning and evening.
To summarize, uncertainty forecasts are today mostly
used in industry for
✔ setting operating reserve requirements
✔ unit commitment and economic dispatch
✔ market bid optimization, e.g., minimizing imbalance
costs
✔ virtual power plant operation
✔ predictive grid management and flexibility allocation
✔ maintenance scheduling
✔ long-term commitment and portfolio planning.
As the previous discussion suggests, TSOs are initiating
integration of or already using uncertainty forecasts for certain decision-making problems. Conversely, with the advent
of smart-grid technology, DSOs are beginning to explore
renewable energy forecasts in the following use cases:
✔ forecast grid operating conditions for the next several hours
✔ improve scheduling and technical assessment of transformer maintenance plans
ieee power & energy magazine

47
http://www.ieafore http://www.casting.dk
Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2017
