This paper presents an optimal state of energy (SOE) estimation strategy of a lithium-ion battery using the back-propagation neural network (BPNN). Two heuristic optmization techniques named backtracketing search algorithm (BSA) and particle swarm optimization (PSO) algorithm are applied to improve the accuracy of BPNN model. Optimization algorithms are developed to determine the optimal value of hidden layer neurons and learning rate of BPNN model. Three most influencing factors including current, voltage and temperature are considered as the inputs to the optimal BPNN model. Federal Urban Driving Schedule (FUDS) is used to check the model robustness at 0°C, 25°C and 45°C. The model performance is evaluated based on the root mean square error (RMSE) and mean absolute error (MAE). The results show that the proposed model obtains good accuracy with an absolute error of ±5%. The BPNN based BSA model improves the SOE estimation accuracy by reducing RMSE and MAE by 2.8% and 4.4% compared to BPNN based PSO model at 25°C.