Intelligent Energy Management Strategies for Hybrid Electric Transportation

Shortage of fossil fuels around the world, oil prices, and environmental concerns like climate change and air pollution are big challenges for the transport sector. Hybrid electric vehicles (HEVs) are an alternative solution to overcome the above challenge. HEVs require an efficient energy management strategy (EMS) to promote consumption efficiency in the different test cycles. The learning-based algorithms are frequently used to control how efficiently HEVs use energy. The tremendous processing intensity, the extensive data training, and the stringent necessity of accurately predicting the future state of operation all work against the maximum use of these systems. In this study, the model-free reinforcement control mechanism is used for the energy management of HEVs. A deep
$\mathbf{Q}$
network (DQN) and deep deterministic policy gradient (DDPG) are utilized to the enhance the learning process and reliability of the EMS framework. Since it is a memoryless random process, the Markovian decision process has been used to formulate the issue. Both learning algorithms are compared with different test driving cycles. The simulation results present that the deep DDPG performance is more reliable and fast converges than the deep
$\mathbf{Q}$
learning algorithm.