Using publicly available data and battery models for energy consumption estimation in electric vehicles

Abstract

Improving the driving range of a battery-powered electric vehicle (EV) has been a signi cant concern in the automotive industry. The driving range is the actual distance an EV travels on a single charge. It strongly relies on the battery capacity and several other factors, such as road topology, tra c density, and weather conditions. These factors also in uence the velocity pro le or driving cycle of an EV which countries or organisations use as the standard procedure to assess the performance of vehicles. In contrast, less work was conducted to improve estimating the potential velocity pro- le for the vehicle based on the real-time tra c situation before the journey started. Estimating the driving pro le or driving cycle on a particular route will enhance the state-of-charge (SOC) estimation accuracy. Hence, it will improve studying the in uential factors towards battery behaviours and discharge processes. This thesis used public data from di erent resources related to the range estimation to predict the battery's remaining charge in a single trip. We further conducted several experiments to understand the battery behaviour of di erent models and used industrial open-source data to analyse the battery performance. The state-of-charge (SOC) is crucial in predicting battery life, de ned as the charge level relative to the battery's capacity. In this thesis, the determining factors of SOC are examined using tra c data obtained from Google Maps, HERE Maps and Tom- Tom routing data providers. Some data were collected using the API for these three map information providers based on two di erent routes on the map, including time, distance and road segments. The data were collected at di erent times to better understand the route tra c situation. The route segmentation was rst performed manually by specifying the waypoints on the map to separate the road parts where the speed is likely to be reduced due to possible stops. Furthermore, the waypoints were speci ed by the API dynamically to provide more route segments and to avoid API restrictions. This approach helps us construct realistic driving pro les to a certain extent despite lower accuracy due to insu cient data. This step neglected the tra c light's waypoints due to insu cient data, which has been applied as a random process. We added some noise to the velocity pro le to emulate the driving behaviour since the data returned from the API is for the average velocity for each segment. The nal driving pro les were used to discharge the batteries, and the results were investigated. Moreover, another approach was conducted to construct reliable driving pro les based on the route information. In this approach, we collected real-time data from di erent APIs, including route information, weather data, tra c light coordinates, and electric vehicle model. We incorporated these datasets into the SOC estimation algorithm. MATLAB and Simulink code was implemented, using the di erent datasets from di erent sources to estimate the real-time remaining range calculation. Throughout this thesis, we have investigated di erent battery models used in electric vehicle applications and analysed the battery behaviours under various conditions. In addition, we explored the public data in di erent scenarios, integrated di erent APIs to predict driving behaviours in di erent routes, and analysed and compared the results of the data sources. As a result, we generated di erent power demand pro les based on di erent data sources to estimate the energy consumption of electric vehicles. Some of the representative driving cycles were further analysed and validated using an actual battery in the lab. The validation results showed that our battery's estimation is within the range of the actual battery in terms of power demand and energy consumption. It also showed that the battery model dynamics are similar to the real one, which gives the model more validity to conduct further experiments and rely on its results.