Charger Monitoring Stream Analysis Using Artificial Neural Networks

This paper develops a trace and prediction model for the energy load imposed by electric vehicle charging in Jeju City, based on artificial neural networks as well as the massive amount of charger monitoring streams collected for about 1 year. The monitoring system generates a data archive, in which a single report essentially embraces charger ID, timestamp, and battery level reading. Our data analysis module, consisting of MySQL, R, and C program implementation across Linux and Windows PC domains, not only processes raw data files to insert into a database table but also finds the active charger set. The distributions of the number of reports and the power consumption for each charger are investigated. For the dayby-day power consumption stream for vehicle charging, a neural network model traces and predicts the time series having 124 sequential values. The experiment result shows that our model catches up with unpredictable spikes just with small time lag.