Many drivers are attracted by electric cars that minimize emissions of greenhouse gases, although lack of faith in the charging facilities discourages others from buying. It is not easy to build a reliable charging station network since aggregating data from the independent operators of stations is a bit challenging. Researchers have come up with an AI for analyzing users’ reviews in the charging stations. This helps to identify areas where there are out-of-service or insufficient stations.
Omar Asensio, who works at the Georgia Institute of Technology, in the School of Public Policy as an assistant professor and a principal investigator said that they are spending billions of private and public dollars building electric vehicle infrastructure. However, they don’t have a better understanding of how these funds are serving the general public and its interest. To try to tackle the unclear infrastructure for charging, electric vehicle drivers have started creating communities on charge facility locator apps that leave reviews. Now, the researchers are collecting those reviews and analyze them to help identify the problems the users are facing.
With the help of the AI, Omar and his team could tell if a particular station was functional on a specific day. Most of these communities are located in Midwest and West states such as South Dakota, Utah, Nebraska, Oregon, and Hawaii. Omar said that when users engage and share information about their experience on a charging station, they engage in pro-environmental and prosocial behaviour, giving behavioural information for machine learning. The co-author of the Georgia Institute of Technology, Sameer Dharur, said that analyzing texts using a computer is more challenging since some have many words with various topics and misspellings. Sameer added that some users through emojis or smiley face into the texts.
For them to solve this problem, Asensio and his colleagues developed an algorithm for electric vehicle transportation lingo. Using 12,750 US charging stations, they classified reviews into eight categories: availability, location, functionality, service time, cost, user interaction, dealership, and range anxiety. The AI attained an accuracy of 91% and high learning efficiency in analyzing the reviews in minutes. Omar said that this is a significant step in the transition since the AI tools perform well than humans.
The AI could help reduce research costs and offer real-time standardized data compared to previous charging infrastructure evaluations that depend on periodic self-reported and costly surveys. By 2027, the charging market of the electric vehicle is anticipated to grow to $27.6 billion. The new methods that have been deployed can lead to rapid policy analysis and make it easy for government and private companies to manage charging infrastructures.https://cryptotodays.com/