Energy Consumption Estimation of Electric Vehicles using different Navigational Parameters

Abstract

The demand of Battery Electric Vehicles (BEVs) is increasing at a very fast rate and the auto- mobile industry is considering them as the future of transport. They have number of advantages such as pollution free environment, lower cost of ownership and comfortable ride etc. but even with all these advantages customers are hesitant to buy BEVs due to driver’s range anxiety, scarcity of charging infrastructure and long charging time. The driver’s range anxiety is emerg- ing as the major concern for customers. Hence, a system has been developed to accurately predict the energy consumption for a BEV based on the different influencing factors such as traffic, road elevation, auxiliary loads, environmental temperature and wind speed etc. The proposed system contains two main modules, one for energy consumption estimation and the other for traffic speed prediction. For estimating the energy consumption based on different parameters two solutions have been proposed namely, Basic Energy Estimation (BEE) model and Improved Energy Estimation (IEE) model. The BEE model uses a Convolutional Neural Network (CNN) to predict en- ergy consumption based on three parameters namely, vehicle speed, tractive effort and road elevation. Multiple experiments with different variations are performed to explore the impact of number of layers and input feature descriptors. The IEE model uses a hybrid approach by combining a multi-channel CNN with Bagged Decision Tree for energy estimation. The IEE model provides better performance as it considers a number of other factors (such as auxiliary loads, battery’s initial state of charge, wind speed and environmental temperature etc.) also for prediction. Unlike existing techniques, the proposed approaches do not require internal vehi- cle parameters from manufacturer and can easily learn complex patterns even from noisy data. Comparison of results with existing techniques shows that the developed IEE model provides better estimates with least mean absolute energy deviation of 0.08 ± 0.069. To predict the traffic speed a deep learning based approach namely, Multistep Traffic Speed Pre- diction (MTSP), has been developed using both the spatial and temporal dependencies. To con- sider spatio-temporal dependencies, nearby road sensors at particular instant are selected based on the attributes such as traffic similarity and distance. Two pre-trained deep auto-encoders were cross connected using the concept of latent space mapping and the resultant model was trained using the traffic data from selected nearby sensors as input. The MTSP model was trained using the real world traffic data collected from loop detector sensors installed on differ- ent highways in Los Angeles. The traffic data is freely available from web portal of California Department of Transportation Performance Measurement System (PeMS). The effectiveness of the MTSP model was verified by comparing it with number of machine/deep learning ap- proaches. The proposed IEE model and MTSP model are integrated to form a system which can provide real-time accurate energy consumption prediction for different routes to a destination based on different traffic, road and environmental conditions. Hence, the system can be used to guide the driver for selecting the best possible route with minimum energy consumption so that he/she can reach the destination with the remaining charge present in the battery. This will greatly improve the confidence of the driver and hence will reduce driver’s range anxiety.