Growing penetration of electric vehicles requires evolution of energy systems to enable sustainable and efficient charging. It is needSd to have systems that can smartly control the grid load and the energy consumption. Uncontrolled demand for charging causes system instability, higher operating costs and wasteful use of energy, especially during peak hours. Traditional charging management strategies such as rule-based scheduling and model predictive control are unable to adapt to real-time grid fluctuations. Such constraints lead to energy waste, higher prices for customers and a less effective use of renewable energy sources. To solve these issues, this study proposes a Deep Q-Learning Network based optimization framework for dynamic energy management in electric vehicle charging infrastructure. The proposed model can dynamically learn and adapt charging schedules based on real-time electricity pricing, grid load and the availability of renewable energy. This study uses reinforcement learning to schedule EV charging during off-peak hours. It also improves the price of charging and increases the use of renewable energy. Results show that the peak load on the grid declines by 22%. This results in a more consistent distribution of electricity within the network and minimises the risk of grid overload. The framework reduces EV charging costs by 15% which makes charging more affordable for users. Also, it promotes the renewable energy usage by 22%. This reduces dependence on conventional grid electricity and supports cleaner energy use.
Unlike prior works, this study combines real-time adaptive decision making with practical constraints such as dynamic pricing structures and grid limitations. The suggested architecture adapts continuously to real-time grid conditions, unlike typical charging systems. “Such a dynamic learning-based model offers a new scalable and practical solution to combine renewable energy with affordable EV charging.