PubAg

Main content area

Implementation of machine learning based real time range estimation method without destination knowledge for BEVs

Author:
Yavasoglu, H.A., Tetik, Y.E., Gokce, K.
Source:
Energy 2019 v.172 pp. 1179-1186
ISSN:
0360-5442
Subject:
algorithms, artificial intelligence, data collection, decision support systems, energy, environmental factors, prediction
Abstract:
In this work, an advanced range estimation method based on experimental test data including environmental factors and dynamic vehicle parameters with driver and road type predictions is proposed for electric vehicles.The focus point of the given study is to predict remaining range in general, at first start-up, without knowing future driving profile, in terms of giving an idea of how much distance can be travelled with the remaining amount of energy. Road type and driver profile are estimated by utilizing decision tree method and periodogram of jerk trace, respectively. Based on the preliminary results, utilized decision tree algorithm classifies the road type with an accuracy over 98%.Vehicle range is estimated online by using machine learning algorithm based on experimental train data sets where chassis dynamometer tests were conducted by performing specific driving cycle in various conditions. For a real life verification, the vehicle is driven for a 50.4 km distance in a road having mostly urban driving characteristics. The results of real life measurements show that the proposed method predicts range with a low margin of error and estimates final remaining capacity 11.3% better than rated one.
Agid:
6346335