Context Aware-Resource Optimality in Electric Vehicle Smart2Charge Application: A Deep Reinforcement Learning-Based Approach
Muddsair Sharif, Gero Lückemeyer, Hüseyin Şeker
Abstract
Electric vehicle (EV) adoption is expanding, posing new issues for grid operators, fleet operators, charging station operators, and EV owners. The challenge is to devise an efficient and cost-effective strategy for managing EV charging that takes into account the demands and objectives of all parties. This study offers a context-aware EV smart charging system based on deep reinforcement learning (DRL) that takes into account all participants’ requirements and objectives. The DRL-based system adjusts to changing contexts such as time of day, location, and weather to optimize charging decisions within an instantaneous fashion by balancing the trade-offs among charging cost, grid strain reduction, fleet operator preferences, and energy efficiency of charging station maintainer while providing EV owners with a convenient and cost-effective charging experience for its ability to handle sequential decision-making, capture complex patterns in data, and adapt to changing contexts. The proposed system’s performance has been evaluated using simulations and compared with existing solutions. The results demonstrate that the proposed system is capable of balancing the trade-offs between different objectives and providing an energy-efficient solution which is approximately 15% better than traditional approach, and about 10% more cost-effective charging experience for EV owners while reducing grid strain by 20% and CO2 emissions by 10% as a result of using a natural energy source. The proposed system has then resulted in achieving the needs for efficient and optimised resource scheduling of fleet operators and charging station maintainers.