Multi-agent Reinforcement Learning-based Dynamic Pricing for EV Charing/Discharging

The number of electric vehicles (EVs) is projected to continue increasing over the coming decades. However, without access to clean and renewable energy, EVs may struggle to fulfill one of their primary objectives: replacing fossil fuels in the transportation sector. While technologies for renewable energy generation and EV charging have matured, synchronizing EV charging times with periods of high renewable energy output remains challenging. Consequently, a control mechanism is necessary to regulate EV charging patterns in alignment with the variable nature of renewable energy generation. Indirectly controlled charging can serve as an effective demand-side management tool by utilizing pricing signals to influence EV charging and discharging behaviors. However, there is currently a lack of a transparent and rational pricing model that accurately reflects real-time power system conditions and the pricing preferences of EV owners. To address this gap, this paper proposes a multi-agent reinforcement learning-based pricing model that makes pricing decisions based on real-time data regarding EV owners’ preferences and power system conditions. Preliminary results indicate that this proposed pricing model has the potential to balance power demand and supply while simultaneously reducing the overall charging costs for EV owners.