Big data analytics in smart grids

Tayade Ashutosh Mahendra (213287)

1. Abstract:

The current power system can benefit greatly from the use of massive volumes of data from
the electricity network, meteorological information system, geographic information system,
etc. in the age of big data. Additionally, this will enhance societal welfare and consumer
service. To develop the application of big data analytics in actual smart grids, however,
additional challenges must be overcome. These challenges relate to methodology, awareness,
synergies, etc.

Keywords: Big data, smart grids

2. Introduction:
As digital technology and cloud computing rapidly evolve, an increasing amount of data is
generated by digital devices and sensors, including smart phones, computers, cutting-edge
measuring infrastructures, etc., as well as through human activity and communications. For
instance, exabytes (1018) and zettabytes (1021) are now used to describe the quantity of data
on the internet (Emani et al., 2015). The rational, effective, and efficient examination of these
data has a significant positive impact on both our personal and professional lives.
Yet, the amount of data being gathered is increasing exponentially, and their structure is also
getting much more intricate.
The potential data that may be gathered with smart grid infrastructure's improved metering is
first explored. The paper then quickly examines the fundamentals of data analytics as well as
some of the most common methods. The article concludes by demonstrating in-depth data
analytics applications in smart grid.

3. Big data in smart grid:

3.1. Concept of big data
Big data is currently not well defined or standardized. Nonetheless, there is agreement across
the many descriptions: this is a new technical issue caused by a dataset with a vast volume,
numerous categories, and complex structures that calls for a novel framework and
methodologies to effectively extract relevant information. As a result, the capabilities of data
mining algorithms and the related technology to handle huge volume datasets determines
what constitutes "big data" (Zikopoulos & Eaton, 2011). Instead of having an absolute
definition, it is an idea. According to Kaisler et al. (2012), big data refers to data volumes that
are too large for current technology to handle, store, and process effectively. As ICT
technologies advance, so do data volumes.
3.2. Concept of smart grid:
The smart grid is a power system that incorporates an information layer that enables two-way
communication between the logistical units and local actuators to swiftly and digitally react
to emergency circumstances involving physical elements. Zhang and others the fluctuating
demand for electricity, Energy Informatics (2018) 1:8 Page 2 of 24. The E.U. defines a smart
grid as an electrical network that can intelligently integrate the actions of generators,
consumers, and those who do both. To provide sustainable, economical, and secure electricity
supplies, this is necessary (SmartGrids European Tech, 2010).
3.3. Big data characteristics in smart grid:
1) Volume - This term describes the enormous volume of data produced, which results in data
sets that are too big to store and analyze using conventional database technologies.
Distributed systems, which store data in several locations and connect them via networks and
software, are one potential answer to this issue.
2) Variety describes the different forms of data that we can presently employ. In the past, we
have concentrated on structured data—information that neatly fits into tables or logical
databases—such as financial or meteorological information. With the use of big data
technologies, we can manage many unstructured data kinds, such as messages, social media
chats, digital photos, sensor data, video, and voice recordings, and combine them with more
conventional, structured data.
3.4. Data sources in smart grids:
The smart grid, an intelligent energy and information system, is a rich source of data that
includes information from the processes of power generation, transmission, distribution, and
consumption. These data, which are depicted in Fig. 1, include non-electrical information
such as marketing, meteorological, and regional economic data, as well as electrical
information from distribution stations, distribution switch stations, and electricity meters. The
gathering and analysis of these data is crucial for scheduling power plants, running
subsystems, maintaining important power equipment, and understanding consumer behavior
in marketing.

Fig.1
4. Data analysis techniques:
Data analysis is the most crucial stage of the big data processing system because it serves as
the foundation for finding vital information and assisting in decision-making (Fan et al.,
2018; Cheng et al., 2018).
From a broad perspective, data analytics, also known as data mining, is a computational
process that uses tools like database, statistics, pattern recognition, machine learning, etc. to
identify probable relationships between variables. However, because of the variety of
sources, the data sets gathered may vary in quality in terms of noise, redundancy, and
consistency.
4.1. Data preprocessing:
As Fig. 3 shows, the use of data pre-processing techniques is required to increase data
quality. The goal of data integration techniques is to effectively combine data from several
sources to create a single view (Roya et al., 2018). For instance, the attribute "date time"
would appear twice if the datasets for weather condition records and power system
interruption occurrences were combined. Yet, it appears that the following data analytics
procedure only requires one characteristic of "date time."

Fig. 2
A logarithm can be used to "correct" the distribution form of data with high skewness
because some data analytics methods are sensitive to imbalanced data. If the initial dataset
only contains the highest and minimum temperature values, other attributes like the
temperature difference can be computed in the pre-processing step. The newly created
features often aid in increasing the accuracy of data analytics findings.
4.2. Data analytics techniques:
According to whether each item in datasets has a label, the most popular data mining or
machine learning methods are typically characterized as supervised or unsupervised learning,
as illustrated in Table 5. The data analytics model can be trained using the provided data for
the supervised learning algorithms to determine the relationship between data attributes and
the related categories or values. The data analytics model is typically created to identify
potential categories among all the items, while for those without labels (Di Zhua & Zhang,
2018).

5. Big data analytics in smart grid:

5.1. Fault detection:
The driving force and building goals of the smart grid, which is envisioned as a decentralized
system, are the reduction of carbon emissions and environmental sustainability. An efficient
way to use common renewable energy sources like wind and solar energy is currently
provided through the use of distributed generator units in modern power distribution systems.
The expectation of a low-carbon world depends on these new microgrids.
5.2. Power quality monitoring:
Electric power quality (PQ) is a global issue that concerns the amount, frequency, and
waveform of voltage and current in a power system. It is closely tied to both the secure
operation of the power grid and consumer happiness. Harmonic distortions and unstable
conditions commonly develop in the power grid due to the growing use of nonlinear and
power electronics-based loads and generators. According to Balouji and Salor (2017), deep
learning is successfully used to classify PQ occurrences in the electricity networks. Deep
learning techniques are used to process the picture files of the three-phase PQ events for
categorization rather than sampling the voltage data of the PQ event data as was done by the
earlier analysis methods.
5.3. Open issues for the application of big data analytics in smart grids:
Electric power quality (PQ) is a global issue that concerns the amount, frequency, and
waveform of voltage and current in a power system. It is closely tied to both the secure
operation of the power grid and consumer happiness. Harmonic distortions and unstable
conditions commonly develop in the power grid due to the growing use of nonlinear and
power electronics-based loads and generators. According to Balouji and Salor (2017), deep
learning is successfully used to classify PQ occurrences in the electricity networks. Deep
learning techniques are used to categorize the images of the three-phase PQ events rather than
sampling the voltage data as in previous analysis approaches.
In smart grids, big data analytics is a broad and challenging topic that includes more than just
mathematical techniques or algorithms, but also how systems operate, how autonomous users
behave, how ICT technologies are used, and how knowledgeable the field is. Hence, if we
want to witness the benefits of it in the smart grids, there needs to be a synergy between
professionals from many sectors.

6. Conclusion:
Big data in smart grids and the accompanying cutting-edge analysis techniques have been
studied and addressed in this article. Smart meters deployed in the power system, the
electrical market, GIS, meteorological information systems, social media, and other sources
are used to gather data that may contain important information. Advanced ICT technology in
the power system links the traditional physical characteristics of the power system to external
variables in order to discover upcoming regulations and scientific difficulties. Smart grids can
benefit from the eleven data analytics applications described in the paper, including
operations, maintenance, load forecasting, protection, fault detection, and fault location.
Given that the use of data analytics in smart grids is a broad and complex topic that involves
ICT technologies, electrical engineering, computer science, and other disciplines, it requires
collaboration among specialists in various domains as well as strategic visions for the best
designs.

7. References:
1. Ghosh D, Ghose T, Mohanta DK (Aug. 2013) Communication feasibility analysis for
smart grid with phasor measurement units. IEEE Trans. Ind. Informat. 9(3):1486–
1496
2. Gillis JM, Alshareef SM, Morsi WG (2016) Nonintrusive load monitoring using
wavelet design and machine learning. IEEE Transactions on Smart Grid 7(1):320–328
3. Ak R, Fink O, Zio E (2016) Two machine learning approaches for short-term wind
speed time-series prediction. IEEE Transactions on Neural Networks and Learning
Systems 27(8):1734–1747
4. Bauman K, Tuzhilin A, Zaczynski R (2017) Using social sensors for detecting
emergency events: a case of power outages in the electrical utility industry. ACM
Transactions on Management Information Systems 8(2–3)
5. Granell R, Axon CJ, Wallom DCH (Nov. 2015) Impact of raw data temporal
resolution using selected clustering methods on residential electricity load profiles.
IEEE Trans Power Syst 30(6):3217–3224
6. Shvachko, K., Kuang, H., Radia, S., et al.: ‘The Hadoop distributed file system’. Proc.
IEEE 26th Symp. on Mass Storage Systems and Technologies (MSST), 2010, pp. 1–
10
7. Fang, B., Yin, X., Tan, Y., et al.: ‘The contributions of cloud technologies to smart
grid’, Renew. Sustain. Energy Rev., 2016, 59, pp. 1326–1331