Developing effective platforms for economic energy management is considered a pivotal issue in the field of Electric Vehicles (EVs). To implement a cost-effective Energy Management Platform (EMP), developers must overcome two major challenges. The first challenge lies in the environmental dynamic nature such as EV location, energy price fluctuations, storage levels, and parking availability at charging stations. This causes most traditional one-shot optimizations to fail. The second challenge pertains to the lack of regulation in EV energy exchanges. To address these challenges, we propose a cost-aware two-stage EMP based on blockchain and deep reinforcement learning (DRL), namely TEMP. Specifically, TEMP first develops a sharding-based blockchain energy management framework, which guarantees trust, security, privacy, traceability, and accountability without the need for intermediaries. Then, considering the complex and high-dimensional environment, TEMP devises a two-stage cooperative scheduling scheme by combining ant colony optimization (ACO) with proximal policy optimization (PPO) to enhance learning effectiveness. Evaluations show that TEMP outperforms the two state-of-the-art baselines by 12.3% and 4.4% in terms of long-term profits while reducing costs by 6.7% and 2.8%, respectively. Moreover, energy transaction efficiency can be ensured when the EV number of blockchain networks is gradually increased.

中文翻译：

---

TEMP：区块链支持的电动汽车成本感知两阶段能源管理


开发有效的经济能源管理平台被认为是电动汽车（EV）领域的一个关键问题。为了实施具有成本效益的能源管理平台（EMP），开发人员必须克服两大挑战。第一个挑战在于环境动态性质，例如电动汽车位置、能源价格波动、存储水平和充电站的停车位可用性。这会导致大多数传统的一次性优化失败。第二个挑战与电动汽车能源交易缺乏监管有关。为了应对这些挑战，我们提出了一种基于区块链和深度强化学习（DRL）的成本感知两阶段 EMP，即 TEMP。具体来说，TEMP首先开发了基于分片的区块链能源管理框架，无需中介机构即可保证信任、安全、隐私、可追溯和问责。然后，考虑到复杂和高维的环境，TEMP通过将蚁群优化（ACO）与近端策略优化（PPO）相结合，设计了一种两阶段协作调度方案，以提高学习效果。评估显示，TEMP 在长期利润方面比两个最先进的基准分别高出 12.3% 和 4.4%，同时成本分别降低了 6.7% 和 2.8%。而且，当区块链网络的EV数量逐渐增加时，可以保证能源交易效率。