Utilizing Data-Driven Techniques to Improve Predictive Modeling of Connected Electric Vehicle Energy Consumption

Electric vehicles (EVs) have emerged as a promising solution for environmental preservation. However, a major hurdle remains: range anxiety caused by limited driving range and inaccurate energy consumption estimates. This is because energy use in EVs varies significantly based on environmental factors (like temperature) and driving conditions (like traffic congestion). To address this challenge, this paper leverages the recent trend of connected electric vehicles (CEVs). These vehicles are equipped with sensors and onboard computers that gather real-time data on various aspects that affect energy consumption. The paper then introduces a Tri-component data-driven machine-learning model that utilizes this data from connected EVs. The model focuses on optimizing route planning for energy efficiency by predicting energy consumption along different road segments. It considers various external factors like temperature, traffic congestion, and road incline to predict three key elements: velocity (how fast the vehicle will go), traction power (energy needed to move the vehicle), and auxiliary power (energy used by features like A/C and onboard computer). Tested in real-world scenarios, the model demonstrates a significant reduction in energy consumption estimation errors, with a remarkable 2.38% error rate for battery state-of-charge (SoC) and 0.52 kWh for energy consumption.