Machinelearning, RNN, LSTM, GRU, Automotive
The main factor influencing an electric vehicle’s range is its battery. Battery electric vehicles experience driving range reduction in low temperatures. This range reduction results from the heating demand for the cabin and recuperation limits by the braking system. Due to the lack of an internal combustion engine-style heat source, electric vehicles' heating system demands a significant amount of energy. This energy is supplied by the battery and results in driving range reduction. Moreover, Due to the battery's low temperature in cold weather, the charging process through recuperation is limited. This limitation of recuperation is caused by the low reaction rate in low temperatures. Technology developments for battery electric vehicles are mostly focused on maintaining the vehicle battery package temperature and state of charge. For battery management systems, state of charge and battery temperature estimations are important since they prevent over charge, over discharge, and thermal runaway. Estimation and controlling battery temperature and the state of charge guarantees safety, it will also increase the vehicle's life cycle. This study analyzes the effects of ambient and battery temperature on heating system energy demand and regenerative braking parameters. Moreover, different machine learning methods for estimating the battery temperature and its state of charge are compared and presented. The analysis is based on the BMW i3 winter trips dataset which includes data for 38 different drive cycles. Results show that every 3 degrees of ambient temperature drop results in a 1% increase in the heating energy share. Furthermore, the ability of machine learning methods such as LSTM and GRU has been demonstrated to successfully forecast battery temperature and state of charge.
Master of Applied Science
Mechanical, Automotive and Materials Engineering
Major Research Paper