
including multi-algorithm combination strategy,
reinforcement learning strategy, optimized
scheduling algorithm, and prediction and dynamic
adjustment, providing a reference for achieving
more intelligent and efficient load balancing,
improving the operational efficiency and stability of
power systems, [14].
Online anomaly detection for smart meter data
was done a few years ago. The solution includes (1)
Data preprocessing: including data cleaning,
scaling, and transformation to adapt to model inputs;
(2) Anomaly detection model construction: The use
of prediction models to learn normal behavior and
perform anomaly detection based on prediction
errors; (3) Abnormal score calculation: Calculate the
abnormal score for each data point based on
prediction error and data history; (4) Online
learning: When new data arrives, update the model
to adapt to the new data and continue learning, [15].
Some research in relevant domains on still image
recognition was done, [16]. Some effective online
anomaly detection algorithms like the Gaussian
Mixture Model were used for vectored area
navigation and detecting spectrum access violations
etc., [17], [18].
As part of smart grid upgrades, traditional
electricity meters are being replaced with smart
meters that can improve accuracy, efficiency, and
visibility in electrical energy consumption patterns
and measurements. However, in most of the
deployments, smart meters are only used to digitally
measure the energy usage of consumer premises and
transmit that data to the utility providers. Despite
this, smart meter data can be leveraged into
numerous potential applications such as demand
side management and energy savings via consumer
load identification and abnormality detection.
Anyhow, these features are not enabled in most
deployments due to high sampling rate
requirements, lack of affordable communication
bandwidth and resource constraints in analyzing a
huge amount of data. The suitability of the
embedded edge computing paradigm which not only
enriches the functionalities but also overcomes the
limitations of smart meters is demonstrated through
the relevant study. It achieves significant
improvements in accuracy, latency, and bandwidth,
[19].
Machine learning can be used in many kinds of
industry domains for prediction research. A
structural graph-coupled advanced machine learning
ensemble model for disease risk prediction is
utilized in a tele-healthcare environment, [20].
Some key applications using data analytics,
machine learning, and deep learning in health
sciences and biomedical data are explored in data
analytics in biomedical engineering and healthcare.
The areas cover such as predictive health analysis,
electronic health records, medical image analysis,
computational drug discovery, and genome structure
prediction using predictive modeling. Case studies
demonstrate big data applications in healthcare
using the MapReduce and Hadoop frameworks,
[21], [22].
The relevant research method is also used in
construction and related industries. For example,
heating load and cooling load forecasting are crucial
for estimating energy consumption and
improvement of energy performance during the
design phase of buildings. Since the capacity of
cooling ventilation and air-conditioning system of
the building contributes to the operation cost, it is
ideal to develop accurate models for heating and
cooling load forecasting of buildings. A machine-
learning technique for the prediction of the heating
load and cooling load of residential buildings is
proposed. The proposed model is a deep neural
network (DNN), which presents a category of
learning algorithms that adopt nonlinear extraction
of information in several steps within a hierarchical
framework, primarily applied for learning and
pattern classification. The output of DNN has been
compared with other proposed methods such as
gradient boosted machine (GBM), Gaussian process
regression (GPR) and mini max probability machine
regression (MPMR). To develop the DNN model,
the energy data set has been divided into training
(70%) and testing (30%) sets. The performance of
the proposed model was benchmarked by statistical
performance metrics such as variance accounted for
(VAF), relative average absolute error (RAAE), root
means absolute error (RMAE), coefficient of
determination (R2), standard deviation ratio (RSR),
mean absolute percentage error (MAPE), Nash–
Sutcliffe coefficient (NS), root means squared error
(RMSE), weighted mean absolute percent error
(WMAPE) and mean absolute percentage Error
(MAPE). DNN and GPR have produced the best-
predicted VAF for cooling load and heating load of
99.76% and 99.84% respectively, [23].
These methods have room to improve though
they are valuable. This is because none of them
provided good accuracy for large-scale usage. No
study has applied any cutting-edge deep learning
methods for smart meter malfunction detection,
even though deep learning methods have been
successfully used for several other malfunction
detection problems in recent years.
To avoid huge waste on direct smart meter
physical testing or replacement, the new abnormal
WSEAS TRANSACTIONS on INFORMATION SCIENCE and APPLICATIONS
DOI: 10.37394/23209.2024.21.33
Jingxuan Fang, Fei Liu,
Lingtao Su, Xiang Fang