Abstarct: In this thesis, we seek a method to improve battery charge estimation. One of the commonly used methods for charge estimation in many studies is the Extended Kalman Filter (EKF), due to the non-linear nature of the battery model. However, this method often suffers from significant estimation errors caused by improper tuning of the filter parameters. To address this, the Adaptive Extended Kalman Filter (AEKF) was developed, which can achieve lower estimation errors compared to the standard method. Nevertheless, these adaptive methods also require the tuning of adaptation parameters. In this thesis, we aim to intelligently develop and improve the AEKF by determining the tuning parameters in an intelligent manner. Since the battery model and its charge estimation are highly dependent on environmental and operational conditions, model-free learning methods are recommended for finding the adaptation parameters. In this work, a reinforcement learning approach is used to obtain the optimal adaptation parameters. After reviewing a common adaptive charge estimation method, its tuning parameters are extracted and treated as discrete actions. The goal of improving the filter is considered as the reward function. After training, the optimal agent is selected, and the root mean square error of the charge estimation evaluated, showing improvement of 1.61%.