Smart Meter Data Analytics
Smart Meter Data Analytics
Smart Meter Data Analytics
Springer
To Our Alma Mater
–Tsinghua University
Foreword
Smart grid is a cyber-physical-social system where the power flow, data flow, and
business flow are deeply coupled. Enlightened consumers facilitated by smart me-
ters form the foundation of a smart grid. Countries around the world are in the midst
of massive smart meter installations for consumers on the pathway towards grid
digitalization and modernization. It enables the collection of extensive fine-grained
smart meter data, which could be processed by data analytical techniques, especially
now widely available machine learning techniques. Big data and machine learning
terms are widely used nowadays. People from different industries try to apply ad-
vanced machine learning techniques to solve their own practical issues. The power
and energy industry is no exception. Smart meter data analytics can be conducted
to fully explore the value behind these data to improve the understanding of con-
sumer behavior and enhance electric services such as demand response and energy
management.
This book explores and discusses the applications of data analytical techniques
to smart meter data. The contents of the book are divided into three parts. The first
part (Chapters 1-2) provides a comprehensive review of recent developments of
smart meter data analytics and proposes the concept of “electricity consumer behav-
ior model”. The second part (Chapters 3-5) studies the data analytical techniques
for smart meter data management, such as data compression, bad data detection,
data generation, etc. The third part (Chapters 6-12) conducts application-oriented
research to depict the electricity consumer behavior model. This work includes elec-
trical consumption pattern recognition, personalized tariff design for retailers, socio-
demographic information identification, consumer aggregation, electrical load fore-
casting, etc. The prospects of future smart meter data analytics (Chapter 13) are also
provided as the end of the book. The authors offer model formulations, novel al-
gorithms, in-depth discussions, and detailed case studies in various chapters of this
book.
One author of this book, Prof. Chongqing Kang, is a professional colleague. He
is a distinguished scholar and pioneer in the power and energy area. He has done
extensive work in the field of data analytics and load forecasting. This is a book
worth reading; one will see how much insight can be gained from smart meter data
vii
viii Foreword
alone. There are definitely broader qualitative understanding that can be gained from
massive data collected in the realm of generation, transmission, distribution, and
end-use of the smart grid.
ix
x Preface
eling and distribution system operations. The research framework of the smart meter
data analytics in this book can be summarized in the following figure.
This book consists of thirteen chapters. It begins with an overview of recent de-
velopments of smart meter data analytics and an introduction on the electricity con-
sumer behavior model (ECBM). Since data management is the basis of further smart
meter data analytics and its applications, three issues on data management, i.e., data
compression, anomaly detection, and data generation, are subsequently studied. The
main components of electricity consumer behavior model include the consumer him-
self, appliances, load profiles, and the corresponding utility function. The following
works try to model the relationships among these components and discover the inher-
ent law within the behavior. Specific works include pattern recognition, personalized
price design, socio-demographic information identification, and household behav-
ior coding. On this basis, this book extends consumer behavior in both spatial and
temporal scales. Works such as consumer aggregation, individual load forecasting,
and aggregated load forecasting are introduced. Finally, prospects of future research
issues on smart meter data analytics are provided.
To help readers have a better understanding of what we have done, we would like
to make a simple review of the thirteen chapters in the following.
Chapter 1 conducts an application-oriented review of smart meter data analytics.
Following the three stages of analytics, namely, descriptive, predictive and prescrip-
tive analytics, we identify the key application areas as load analysis, load forecasting,
Preface xi
and load management. We also review the techniques and methodologies adopted or
developed to address each application.
Chapter 2 proposes the concept of ECBM and decomposes consumer behavior
into five basic aspects from the sociological perspective: behavior subject, behavior
environment, behavior means, behavior result, and behavior utility. On this basis,
the research framework for ECBM is established.
Chapter 3 provides a highly efficient data compression technique to reduce the
great burden on data transmission, storage, processing, application, etc. It applies
the generalized extreme value distribution characteristic for household load data and
then utilizes it to identify load features including load states and load events. Finally,
a highly efficient lossy data compression format is designed to store key information
of load features.
Chapter 4 applies two novel data mining techniques, the maximum information
coefficient (MIC) and the clustering technique by fast search and find of density
peaks (CFSFDP), to detect electricity abnormal consumption or thefts. On this basis,
a framework of combining the advantages of the two techniques is further proposed
to boost the detection accuracy.
Chapter 5 proposes a residential load profiles generation model based on the gen-
erative adversarial network (GAN). To consider the different typical load patterns of
consumers, an advanced GAN-based on the auxiliary classifier GAN (ACGAN) is
further to generate profiles under typical modes. The proposed model can generate
realistic load profiles under different load patterns without loss of diversity.
Chapter 6 proposes a K-SVD-based sparse representation technique to decom-
pose original load profiles into linear combinations of several partial usage patterns
(PUPs), which allows the smart meter data to be compressed and hidden electricity
consumption patterns to be extracted. Then, a linear support vector machine (SVM)-
based method is used to classify the load profiles into two groups, residential cus-
tomers and small and medium-sized enterprises (SMEs), based on the extracted pat-
terns.
Chapter 7 studies a data-driven approach for personalized time-of-use (ToU) price
design based on massive historical smart meter data. It can be formulated as a large-
scale mixed-integer nonlinear programming (MINLP) problem, Through load pro-
filing and linear transformation or approximation, the MINLP model is simplified
into a mixed-integer linear programming (MILP) problem. In this way, various tar-
iffs can be designed.
Chapter 8 investigates how much socio-demographic information can be inferred
or revealed from fine-grained smart meter data. A deep convolutional neural network
(CNN) first automatically extracts features from massive load profiles. Then SVM is
applied to identify the characteristics of the consumers. Different socio-demographic
characteristics show different identification accuracies.
Chapter 9 uses smart meter data to identify energy behavior indicators through
a cross-domain feature selection and coding approach. The idea is to extract and
connect customers’ features from the energy domain and demography domain. Smart
meter data are characterized by typical energy spectral patterns, whereas household
information is encoded as the energy behavior indicator. The proposed approach
xii Preface
This book made a summary of our research about smart meter data analytics achieved
in recent years. These works were carried out in the Energy Intelligence Labora-
tory (EILAB), Department of Electrical Engineering, Tsinghua University, Beijing,
China.
Many people contributed to this book in various ways. The authors are indebted
to Prof. Daniel Kirschen from the University of Washington, Prof. Furong Li and
Dr. Ran Li from the University of Bath, Dr. Tao Hong from the University of North
Carolina at Charlotte, Dr. Ning Zhang, Dr. Xing Tong, Mr. Kedi Zheng, Mr. Yux-
uan Gu, Mr. Dahua Gan, and Mr. Cheng Feng from Tsinghua University, who have
contributed materials to this book.
We also thank Mr. Yuxiao Liu, Mr. Qingchun Hou, Mr. Haiyang Jiang, Mr. Yinx-
iao Li, Mr. Pei Yong, Mr. Jiawei Zhang, Mr. Xichen Fang, and Mr. Tian Xia at Ts-
inghua University for their assistance in pointing out typos and checking the whole
book.
In addition, we acknowledge the innovative works contributed by others in this
increasingly important area especially through IEEE Power & Energy Society Work-
ing Group on Load Aggregator and Distribution Market, and appreciate the staff at
Springer for their assistance and help in the preparation of this book.
This book is supported in part by the National Key R&D Program of China
(2016YFB0900100), and in part by the Major Smart Grid Joint Project of National
Natural Science Foundation of China and State Grid (U1766212). The authors really
appreciate their supports.
Yi Wang
Qixin Chen
Chongqing Kang
xiii
Contents
xv
xvi Contents
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164