Contemporary power solutions are revolutionizing the existing trend in residential and transportation services. Mobile realizations, in particular, are advancing based on innovative mobile power supply solutions. Batteries are essential to several key applications, such as electric vehicles (EVs), satellites, and mobile phones. Additionally, batteries are elementary parts of the energy storage for distributed renewable microgrids. The are just a few applications based on the batteries-based power supply. Li-ion batteries are frequently used because of their exceptional qualities, which include endurance, high power density, and compact size. Besides these notable benefits the Li-Ion batteries are expensive. To effectively compensate for this high initial cost, the "battery management systems" (BMSs) are used. These solutions are effective in maximizing the Li-Ion battery life and also guarantee a safer operation. Based on the advantages of modular design, the contemporary Li-Ion batteries can be comprised of thousands of individual cells. It results in complex and sophisticated BMSs and can render a notable overhead in terms of power consumption. The focus of this research is on enhancing the existing Li-Ion BMSs by incorporating novel event-based feature mining and machine/ensemble learning techniques. The event-based feature extractor mines pertinent information and promises significant real-time compression. In order to predict the capacity of Li-Ion cells, the produced feature set is then processed utilizing robust machine/ensemble learning algorithms, and the performance of considered regressors is compared. A real Li-Ion battery dataset is used to demonstrate the applicability. For the case of a bagging-based ensemble regressor, the method obtained a correlation coefficient of 0.9996.