IEEE Power & Energy Magazine - May/June 2022 - 12

Tao Hong and Fangxing Li
witnessing a new wave
artificial intelligence and machine learning
P
12
POWER ENGINEERING PROFESsionals
are seeing another wave of artificial
intelligence (AI), which is arguably
the second wave in the power
industry and the third wave in the AI
community since the 1940s. The technological
advancement in AI has been
making it a universal solution to just
about everything. Compared to their
human counterparts, in some applications,
the AI-based solutions are more
accurate, faster in execution, and have
a higher tolerance level for tedious jobs
and extreme environments.
As people are getting used to AI in
their daily life, readers of this magazine
may wonder how AI and machine
learning (ML) affect the present and
future operation of the electric power
grid. Putting all bluffs and marketing
buzzwords aside, AI and ML techniques
have already been adopted in
real-world power industry applications,
though not as many as many people
have thought or claimed. Some readers
may be familiar with AI-based commercial
solutions for load forecasting.
Many readers may be interested to understand
what other power systems applications
AI can handle, whether we
can trust AI to run our power grid, and
what future grids may look like as AI
continues to evolve.
The maturity of AI and ML differs
across various power engineering
applications. This issue of IEEE
Power & Energy Magazine offers a
Digital Object Identifier 10.1109/MPE.2022.3156688
Date of current version: 19 April 2022
ieee power & energy magazine
collection of six articles and an " In
My View " column covering various
aspects of AI and ML for several
applications, such as load and price
forecasting, power system operations,
building energy management,
customer segmentation, and dynamic
modeling. The issue covers a historical
perspective and lessons learned as
well as the latest thinking.
Load forecasting is one of the first
power system applications of AI, dating
back to the late 1980s. During the
past three decades, people have seen a
significant gap between the academic
literature and industry practice, which
leads to various misunderstandings
and confusion about what AI can do
to load forecasting models. The first
article, " Artificial Intelligence for
Load Forecasting: History, Illusions,
and Opportunities, " by T. Hong and P.
Wang, examines five illusions associated
with developing AI-based models.
They also present clarifications to each
illusion to help improve the efficiency
of AI for load forecasting.
Restructuring of the energy sector
in the 1990s resulted in deregulated
energy markets, which made electricity
price forecasting a must-solve problem.
Electricity price forecasting models
have been evolving during the past 25
years from relatively simple models,
such as linear regression models, to
sophisticated ML models that rely on
more data and computing power. In
the second article, " Electricity Price
Forecasting: The Dawn of Machine
Learning, " A. Je
'
drzejewski, J. Lago,
G. Marcjasz, and R. Weron provide an
overview of the main trends and electricity
price forecasting models. They
also point out the areas to improve and
offer a list of best practices for electricity
price forecasters.
Many ML techniques have been
considered as black-box techniques.
Such characteristics prevent them from
being widely adopted for many power
systems applications in practice. In
the third article, " Machine Learning
in Power Systems: Is It Time to Trust
It? " S. Chatzivasileiadis, A. Venzke,
J. Stiasny, and G. Misyris attempt
to encourage trust
in artificial neural
networks by proposing two directions:
neural network verification and
physics-informed neural networks. The
former aims at providing guarantees
about the performance of a trained
neural network, while the latter focuses
on how to take advantage of physical
power system models.
Building energy management has been
a challenging problem due to possible
measurement and prediction errors and
the lack of generalization. ML-based
approaches have gained great interest in
this area in the smart grid community.
In the fourth article, " Demonstration of
Intelligent HVAC Load Management
With Deep Reinforcement Learning, "
Y. Du, F. Li, K. Kurte, J. Munk, and H.
Zandi discuss building load management
for heating, ventilation, and air
conditioning (HVAC) systems using deep
reinforcement learning (RL). The authors
describe and evaluate two deep RL
approaches for HVAC control. They
may/june 2022
guest editorial
IEEE Power & Energy Magazine - May/June 2022

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - May/June 2022
