The state-of-charge (SOC) estimation is a crucial parameter of a lithium-ion battery it depends on numerous incalculable factors such as battery chemistry, ambient environment, aging factor, etc. This paper develops a state of charge (SOC) estimation model for a lithium-ion battery using an improved deep neural network (DNN) approach in electric vehicle applications. The DNN is suitable for SOC estimation due its sufficient hidden layer which is capable of predicting the SOC of unseen drive cycle during training. A series of DNN models with varying number of hidden layers and its training algorithm is developed to investigate the training performance of different drive cycles. It is observed that adding hidden layers in DNN decreases the error rate and improves the SOC estimation. This study also shows that the 7-layer of DNN training on dynamic stress test (DST) drive cycle is capable of predicting SOC values unexpectedly well of other unseen drive cycles such as federal urban driving schedule (FUDS), Beijing dynamic stress test (BJDST), and supplemental federal test procedure (US06), respectively.