Electric Autonomous Vehicles (EAVs) are becoming increasingly popular because of its lower pollutants, fewer emissions, energy savings and smart technology. However, they face a significant obstacle called as the range anxiety which is one of the most significant impediments to the widespread adoption of EAVs. It is a fact that estimating the Driving Range (DR) accurately can help to alleviate drivers' range anxiety. The DR expresses the distance which can be driven by the EVs with the remaining energy in the battery at any instant. The DR is one of the significant reasons for the range anxiety which is a common term used. In order to improve the DR of the EVs, the first choice is to utilize a battery with larger capacity. However, a large battery is not preferred due to its high costs and weight. In this paper, a back-propagation-based multi-layer neural network machine learning algorithm has been constructed to perform the DR estimation by using the various vehicle data such as the vehicle speed, motor angular speed, motor voltage, and state of charge. These data have been obtained from the electric vehicle simulation in MATLAB/Simulink. The results of the DR estimation have been compared according to the root mean square error values for each driving cycle. As a result of the comparison, it has been determined which of the DR prediction values obtained from the inputs used gave less state error. (This study was produced from the thesis of the first-ranked author.)
Anahtar Kelimeler: Back-propagation-based multi-layer neural networks, Driving range estimation, Electric autonomous vehicle, Range anxiety
|
