Using Machine Learning to Gauge Consumer Perspectives of the Existing EV Charging Network
A new study from a research team in Georgia reports the use of machine learning techniques to provide the “best insight yet” into the attitudes of electric vehicle drivers about the existing charging network.
Although early efforts focused on increasing the quantity of electric vehicle (EV) charging stations and improving the EV charging network, something that will grow in importance as the number of mainstream EVs grows, a recent study by researchers from the Georgia Institute of Technology has found that the quality of the charging experience is just as important to EV users.
In a paper published in the June 2020 issue of the journal Nature Sustainability, the Georgia team, led by assistant professor Omar Isaac Asensio, looked at consumer perspectives of the existing EV charging network across the United States by using a machine learning algorithm.
In addition to providing valuable insight into consumer perspectives, the study demonstrates how machine learning tools can be used to quickly analyse data for real-time policy evaluation. This could have a profound impact on any number of key industries beyond the EEE space.
Using Machine Learning to Analyze Consumer Data
The study, which used the machine learning algorithm to analyze unstructured consumer data from 12,270 electric vehicle charging stations, found that workplace and mixed-use residential stations tend to get lower ratings from users.
Fee-based charging stations attracted the poorest reviews compared to free-to-use charging stations. Meanwhile, the highest-rated charging stations are usually found at hotels, restaurants, and convenience stores with other well-rated stations located at public parks, RV parks, and visitor centres.
A row of electric vehicles charging stations designed by and for Tesla vehicles
Asensio’s team used deep learning text classification algorithms to analyse data from popular EV users’ smartphone app. A task that would have taken up the best part of a year by using conventional methods was trimmed down to a matter of minutes by using the algorithms with accuracy on par with human experts.
Among consumers’ biggest gripes are frequent complaints about the lack of accessibility and prominent signage with stations in dense urban centres attracting the highest volume of complaints, around 12-15% more in contrast to stations in non-urban locations. Interestingly, the study found no statistically significant difference in user preference when it comes to public or private chargers, contrary to many early theories.
"Based on evidence from consumer data, we argue that it is not enough to just invest money into increasing the quantity of stations, it is also important to invest in the quality of the charging experience," assistant professor Omar Isaac Asensio says.
A Barrier to Adoption?
By now EVs are considered a crucial element of the solution to climate change. According to the study, however, a major barrier to adopting EVs is the perceived lack of charging stations and the so-called “range anxiety”—that is, how far an EV can travel on a single charge and the possibility of running out of charge in the middle of nowhere—that makes many consumers nervous about buying an EV. And although infrastructure has grown considerably in recent years, not enough work has gone into accounting for what consumers want, Asensio claims.
"In the early years of EV infrastructure development, most policies were geared to using incentives to increase the quantity of charging stations," Asensio said. "We haven't had enough focus on building out reliable infrastructure that can give confidence to users."
By offering evidence-based analysis of consumer perceptions, he claims that this study helps rectify that shortcoming and that overall, it points to the need to prioritize consumer data when considering how to scale infrastructure, particularly requirements for EV charging stations in new developments.
But it is not just EV policy that the study’s deep learning techniques could be applied to. They could also be adapted to a broad range of energy and transportation issues, enabling researchers to carry out rapid analysis with just a few minutes’ worth of computation.
"The follow-on potential for energy policy is to move toward automated forms of infrastructure management powered by machine learning, particularly for critical linkages between energy and transportation systems and smart cities," Asensio said.