Loading content...
Electric vehicle production, charging
infrastructure, and Electric Vehicle Supply Equipment have
all grown in response to the fast adoption of EVs spurred by
government subsidies and clean energy requirements. It is
now crucial to monitor and detect abnormalities in EV
charging behaviour to maintain secure and efficient
operations, especially as EVSEs interact with cloud platforms,
payment systems, and battery management units. To improve
anomaly detection capabilities, this work uses StandardScaler
to normalise charging session data and an MTM-based
approach to extract critical behavioural aspects. An RL-based
model is utilised to detect anomalous relational patterns in EV
charging procedures; the challenge is presented as a reasoning
assignment across huge KG. The system is able to effectively
distinguish between conventional and abnormal charging
behaviours thanks to a bespoke incentive algorithm that
maximises accuracy, detection efficiency, and path diversity.
By achieving a high detection accuracy of 96.38%, the
suggested KG-RL model seems to be successful in finding
anomalies in EV charging networks, according to
experimental evaluation. Future improvements in predictive
maintenance, fraud prevention, and safe EV charging
ecosystems can be facilitated by this method, which showcases
the potential of intelligent KG-RL frameworks in assuring
dependable EVSE management.
The growth of the EV market has garnered
significant interest from governments, automakers, and
energy companies. EV are thought to be a practical
solution to the growing pollution of the environment and
the depletion of fossil fuels. EV charging facilities have
proliferated in tandem with the development of EVs. On
the other hand, power grid safety and optimal dispatching
are threatened by the unpredictable, intermittent, and
volatile nature of the load. In order to increase the
popularity of EV, rationalize the building of charging
stations, and improve prediction precision for optimal
dispatching, it is necessary to establish a scientific and
realistic short-term load forecasting model for EV charging
stations. Focusing studies on load predictions for EV
charging stations is thus highly significant[1]. advancement of distributed energy resources, such as EVs,
distributed photovoltaics and others, has recently attracted
social attention as a means to lessen the impact of pollution,
energy crises, and climate change. As the number of
distributed energy resources on the load side continues to
rise, more and more people are shifting from being
traditional power consumers to prosumers, or people that
can generate their own electricity[2]. As mentioned, the
worldwide trend towards a society with sustainable
renewable energy depends on prosumers' active
involvement at the end user energy side. Typical
prosumers on the grid include commercial buildings with
solar PV panels and EV charging stations. The building
energy management system can optimize the energy use of
these structures under the incentives of electricity pricing.
Electricity cost savings, load levelling, and consumption of
distributed power are thus promising areas for commercial
building energy management. The importance of predicting EV charging
demands in power system management has grown in
tandem with the booming EV market. While centralized
EV charging has many benefits, it also has the potential to
significantly increase grid load and strain the power
system. An imbalance between the supply and demand for
power can cause instability in the system or power outages
if these load changes are not correctly predicted.
Conversely, efficient energy scheduling and management
are two outcomes of reliable demand forecasting that
contribute to the grid's capacity to incorporate renewable
energy sources[3]. Nevertheless, there are still a lot of
obstacles to overcome, even though EV charging load
prediction technology are constantly improving[4]. There
are notable spatiotemporal features to EV charging loads.
While daily, seasonal, and holiday variations affect
charging loads in the time dimension, charging demand
varies by region in the spatial dimension, with notable
distinctions between urban and suburban areas as well as
between commercial and non-commercial areas. Large
scale EV deployment poses significant hurdles for
regulators, power providers, and urban planners despite the
fact that it promises a great deal of sustainability[5]. The
hardest of them is predicting charging demand accurately
given varied charging patterns and user behavior.