We examine the various machine learning methods to be adopted by Q-PLUS, a data-driven smart building energy management system under development by Qbots Energy. Q-PLUS aims to help commercial buildings to save on electricity cost, as well as to generate extra income via the provision of ancillary services, by shifting their electricity demand from the grid away from the peak hours. It is a battery storage, heating, ventilation and air-conditioning (HVAC) control system based on model predictive control (MPC), where machine learning tools are used to predict i), the half-hourly electricity demand of the building and ii), the response of the indoor environment to the HVAC controls. In this thesis, we test and compare different machine learning algorithms to find the most suitable set of tools for the development of Q-PLUS. We also design a battery control algorithm under the MPC framework and fully develop it in Python.