In the near future, microgrids will become more prevalent as they play a critical role in integrating distributed renewable energy resources into the main grid. Nevertheless, renewable energy sources, such as solar and wind energy can be extremely volatile as they are weather dependent. These resources coupled with demand can lead to random variations on both the generation and load sides, thus complicating optimal energy management. In this article, a reinforcement learning approach has been proposed to deal with this non-stationary scenario, in which the energy management system (EMS) is modelled as a Markov decision process (MDP). A novel modification of the control problem has been presented that improves the use of energy stored in the battery such that the dynamic demand is not subjected to future high grid tariffs. A comprehensive reward function has also been developed which decreases infeasible action explorations thus improving the performance of the data-driven technique. A Q-learning algorithm is then proposed to minimize the operational cost of the microgrid under unknown future information. To assess the performance of the proposed EMS, a comparison study between a trading EMS model and a non-trading case is performed using a typical commercial load curve and PV profile over a 24-h horizon. Numerical simulation results indicate that the agent learns to select an optimized energy schedule that minimizes energy cost (cost of power purchased from the utility and battery wear cost) in all the studied cases. However, comparing the non-trading EMS to the trading EMS model operational costs, the latter one was found to decrease costs by 4.033% in summer season and 2.199% in winter season.