IEEE Power & Energy Magazine - November/December 2021 - 78

zero fuel expenses. This change is increasing the commercial
pressure on existing conventional power plants, potentially
leading to their closure earlier than originally planned. As
the share of the supply mix where the fuel source is weather
dependent expands, the importance of accurate weather as
well as wind and solar generation forecasting similarly grows.
Furthermore, the proliferation of distributed energy resources,
such as behind-the-meter solar PVs, batteries, and electric
vehicles, is providing opportunities for new market participants
with innovative new business models, such as virtual
power plants.
This confluence of factors is leading to a renewed focus on
the optimal integration of forecasting and markets to ensure
✔ uncertainty and risk are prudently managed day to day
✔ essential system services are adequately procured and
compensated for
✔ appropriate market signals exist to incentivize and
achieve the necessary investment
✔ reliable energy is delivered cost-effectively.
The Evolution of Forecasting and Its
Increasing Importance
Load-forecasting techniques have developed since the 1950s,
and the field has a high level of maturity. It is already widely
integrated into the operational procedures of system operators
and market agents. However, existing load-forecasting
methods require adjustments to tackle two new challenges:
✔ accounting for the influence of distributed generation,
demand response programs, and extreme weather
events in load profiles
✔ exploiting smart metering data in a hierarchical loadforecasting
framework across multiple voltage/system
levels.
For long-term horizons (years ahead), load forecasting is
becoming highly dependent on technology adoption, such as
the proliferation of PVs, electric vehicles, and smart appliances.
This means research in technology forecasting should
be revisited to improve forecast skills.
Figure 1 depicts the model and value chain for renewable
energy forecasting. As a first step, numerical weather prediction
(NWP) models generate forecasts of weather variables
for horizons up to several days ahead, usually with updates
every 6 h. Spatial resolution can go down to 1 km and temporal
resolution to 1-3 h. For very short horizons (several minutes
to a few hours), forecasts of the weather variables can be
generated from satellite- or sky imager-based models.
These forecasts, together with measurements from renewable
energy source (RES) plants, are then used as inputs to
dedicated RES forecasting models to predict the power output
of these plants. If the horizon is very short (typically up
to 2 h), forecasts can be based only on onsite measurements
as well as satellite and sky imager data.
The last step is a decision-aid phase, which can be either
a human expert who visualizes the forecast products and
makes a choice or an automated software tool that involves
78
ieee power & energy magazine
some optimization functions. This value chain presents several
gaps and bottlenecks at different levels, which are also
shown in Figure 1.
The use of forecasts and market clearing (and, more generally,
decision making) are inextricably linked across the majority
of electricity markets around the globe. What may differ is
how they are integrated, who is responsible for providing them,
and what type of forecast is needed at which operation stage.
Accurate long-term load and renewable energy forecasts
are necessary, not just for prospective renewable generation
investors to know their potential for revenue to justify
builds but, increasingly, for all other prospective technologies.
These forecasts can help investors and system planners
understand the need for all technologies to support resource
adequacy. They also provide insights on prices that may
impact the profitability of each technology in different ways.
Regions in the United States and Europe typically have dayahead
markets, which allow for price certainty in determining
the efficient unit commitment of resources through either
centralized or decentralized unit commitment procedures.
Day-ahead load, wind, and solar forecasts are crucial for these
markets and have been the domain of specialized forecast providers
worldwide. These forecasts may be provided by the market
participants or system operator to ensure reliability through
subsequent processes, such as the reliability unit commitment
that follows the day-ahead market in the United States.
Despite their name, real-time markets are typically cleared
in advance of the operating time. Some markets may close
more than 60 min before the operating hour, and market clearing
may still occur 5-30 min ahead. While the accuracy of
forecasts in these horizons tends to be far better than those
used in the day-ahead market, they are still not easy to compute.
Precise forecasts during this timeframe are critical to
ensure the operating reliability and economic dispatch of all of
the technologies. A poor forecast in system operation can also
have severe economic, financial, and reliability consequences.
This may depend on the forecast horizon and market where it
occurs as well as the magnitude of the forecast error.
The reliance on accurate forecasting and its impact will
continue to increase in importance (see Figure 2) as variable
renewable energy installed capacity grows. Unit commitment
procedures may decline in importance due to decreases in
resources that require ample time to commit. However, the decision
to charge batteries and other storage is heavily impacted by
the forecasts of prices, which are driven by the predictions of
everything else. Price-responsive demand also requires accurate
forecasts to position a plant to react to price events while minimizing
the impact on its operations and resources.
Load, renewable energy, and fuel forecasts are used to derive
a series of other predictions. These will impact future prices,
thereby affecting offers of fuel-limited resources, such as natural
gas and energy storage. Contingency analysis is driven by
forecasts, as the possibilities that system operators are most
worried about may not simply be single events. Increasingly,
they could be extreme, weather-driven, multiple-contingency
november/december 2021
IEEE Power & Energy Magazine - November/December 2021

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2021
