Hosseini, Seyedmohammad2019-09-232019-09-232019-09-232019-09-20http://hdl.handle.net/10012/15135In this study, an innovative energy management system (EMS) employing the promising reinforcement learning (RL) method is proposed. The EMS intelligently administrates the power flow between the main battery which is fed through the alternator and a solar-powered auxiliary battery which is used for the vehicle idle time reduction via providing energy for auxiliary loads which force the engine to be running, although the service vehicle is stopped. RL, which is an exquisite artificial intelligence technique, endeavors to offer a sub-optimal performance for this control problem compared to the really time consuming Dynamic Programming approach, which determines the optimal solution through exhaustive search. A service vehicle is modeled in the Matlab/Simulink environment. Different parts of the model are described in detail, and the dynamics of the considered vehicle are discussed. The simulation results express a better functionality compared to an existing rule-based controller and the idled engine case, turning the proposed RL-based EMS into an effective method for implementation in vehicular solar idle reduction (SIR) systems. Double DQN is also utilized to come up with the continuous observation space. The results are showing that Deep-RL can be a promising method in control tasks like the EMS of vehicular systems. Furthermore, a cost-effective and efficient data acquisition system is designed, tested, and implemented using the renowned Raspberry Pi board, and some sensors to collect voltage, current, and temperature data. The required electrical enclosures are also designed to keep the whole package safe. The validation of the system results is done and the process is discussed in detail. This data acquisition system can be employed to read the required information from vehicle and its loads, in order that the intelligent EMS system can wisely decide which action to take in a real-time manner.enenergy management systemreinforcement learningsolar idle reductiondata acquisitiondeep learningglobal warmingservice vehiclesauxiliary loadshybrid vehiclesMaster Thesis