Abstract: Accurate estimation of state of the charge (SOC) is vital for electric vehicle batteries. This paper presents a novel method to estimate the SOC based on a neural network which can be programed into a low cost microcontroller. The microcontroller monitors the battery voltage and takes four samples immediately after the battery is disconnected from the load and monitors the steady state terminal voltage when the vehicle is parked for more than 30 minutes to train a neural network. Each time the battery is disconnected from the load while driving, the microcontroller takes four voltage samples to estimate the SOC using the previously trained neural network. In order to increase the accuracy, the battery temperature is also taken as an input.