D ³ PGSA: A Dual-Value Deep Deterministic Policy Gradient Based on Sample Augmentation for Dynamic Pricing of Electric Vehicle Charging Station

With the rapid growth of electric vehicles (EV), the demand for charging infrastructure has surged. Effective pricing strategies for electric vehicle charging station (EVCS) are critical for ensuring sustainable service delivery, promoting EV adoption, and facilitating a shift in energy usage. However, existing dynamic pricing approaches suffer from several limitations, such as the inefficiency of traditional algorithms at scale and the high data requirements and instability of reinforcement learning (RL) methods. In this paper, we propose a Dual-value Deep Deterministic Policy Gradient based on Sample Augmentation (D<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup>PGSA) algorithm to address the dynamic pricing problem for a single EVCS. The proposed method provides operators with a flexible pricing model tailored to dynamic market conditions, thereby enhancing both revenue and operational efficiency. Specifically, we first introduce an experience generation model to alleviate the issue of limited training samples, and to enhance the model’s ability to generalize to unseen states. Next, we introduce a feature generation model and a cluster-balanced (CB) experience replay mechanism to improve sample utilization and enhance decision quality. Finally, a dual-value network is integrated into the traditional framework to reduce error accumulation and improve learning stability. Extensive experiments conducted on two real-world datasets and four simulated datasets demonstrate that D<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup>PGSA can dynamically adjust prices in response to market demand and significantly outperforms other methods.