Abstract

Due to urban pollution, transport electrification is being currently promoted in different countries. Electric Vehicles (EVs) sales are growing all over the world, but there are still some challenges to be solved before a mass adoption of this type of vehicles occurs. One of the main drawbacks of EVs are their limited range, for that reason an accurate estimation of the state-of-charge (SOC) is required. The main contribution of this work is the design of a Nonlinear Autoregressive with External Input (NARX) artificial neural network to estimate the SOC of an EV using real data extracted from the car during its daily trips. The network is trained using voltage, current and four different battery pack temperatures as input and SOC as output. This network has been tested using 54 different real driving cycles, obtaining highly accurate results, with a mean squared error lower than 1e-6 in all situations This work has been partially financed by the Spanish Ministry of Economy and Competitiveness within the framework of the project DEMS: “Sistema distribuido de gestión de energía en redes eléctricas inteligentes (TEC2015-66126-R)".