【Advances in Applied Energy】电力交通耦合环境中的充电服务商广义均衡定价策略

(1) 针对电力交通耦合环境下的充电市场均衡定价问题，提出了一种基于增强多智能体近端策略优化的多智能体强化学习算法，通过融合注意力机制减轻环境非静态性、维度和信息隐私对决策的影响；

(2) 提出一种顺序更新训练算法，保证不确定性环境中的智能体单调策略改进；

(3) 通过算例对比证明了所述方法在提升智能体策略均衡度、训练效率上的有效性。

(4) 基于所述方法，分析了竞争性充电市场对于缩减用户用能成本、提升新能源消纳率、降低电网运行成本的效果。

在完全竞争市场条件下，电力-交通交通网络的充电商定价为非合作博弈问题，牵涉电力、交通网络运行状态、充电商资源条件、用户出行特征等多种不确定性因素。已有市场均衡分析方法通常基于Karush-Kuhn-Tucker条件转化市场双层模型进行求解，难以处理非凸运行约束、隐私保护以及复杂不确定性，难以应用到复杂的竞争性充电市场均衡分析及决策中。

Transportation electrification, involving large-scale integration of electric vehicles (EV) and fast charging stations (FCS), plays a critical role for global energy transition and decarbonization. In this context, coordination of EV routing and charging activities through suitably designed price signals constitutes an imperative step in secure and economic operation of the coupled power-transportation networks (CPTN). This work examines the non-cooperative pricing competition between self-interested EV charging service providers (CSP), taken into account the complex interactions between CSPs' pricing strategies, EV users' decisions and the operation of CPTN. The modeling of CPTN environment captures the prominent type of uncertainties stemming from the gasoline vehicle and EV origin-destination travel demands and their cost elasticity, EV initial state-of-charge and renewable energy sources (RES). An enhanced multi-agent proximal policy optimization method is developed to solve the pricing game, which incorporates an attention mechanism to selectively incorporate agents' representative information to mitigate the environmental non-stationarity without raising dimensionality challenge, while safeguarding the commercial confidentiality of CSP agents. To foster more efficient learning coordination in the highly uncertain CPTN environment, a sequential update scheme is also developed to achieve monotonic policy improvement for CSP agents. Case studies on an illustrative and a large-scale test system reveal that the proposed method facilitates sufficient competition among CSP agents and corroborates the core benefits in terms of reduced charging costs for EV users, enhancement of RES absorption and cost efficiency of the power distribution network. Results also validate the excellent generalization capability of the proposed method in coping with CPTN uncertainties. Finally, the rationale of the proposed attention mechanism is validated and the superior computational performance is highlighted against the state-of-the-art methods.

Keywords（英文与中文）：

Coupled power-transportation network 电力交通耦合网络

Electric vehicles 电动汽车

Multi-agent reinforcement learning 多智能体强化学习

Nash equilibrium 纳什均衡

Pricing game analysis 定价博弈

图1. 所研究充电定价问题的形成因素、设定及其对能源转型和交通电气化的价值

图2. CSP定价博弈问题示意

图3. 所提Att-MAPPO算法的训练和执行过程

图4. 同轮次下所提算法与其它算法所得策略的纳什均衡度对比

图5. 同轮次下所提算法与其它算法的收益对比

图6. 所提策略在提升新能源消纳率和降低电力运行成本方面的效果

图7. 基于所述方法，分析不同用户弹性系数下竞争性市场和非竞争性市场的效益变化

图8. 基于所述方法，分析不同电动汽车渗透率及初始电量下车流空间分布变化

关于Applied Energy

点击“阅读原文”

提交文章吧

喜欢我们的内容？

点个“赞”或者“再看”支持下吧！