Privacy-Aware and Energy-Efficient Task Offloading for the Internet of Energy

With the continued evolution of intelligent energy systems, terminal devices in the Internet of Energy (IoE) generate large volumes of high-frequency, privacy-sensitive data related to user energy consumption behaviors, device operational states, and regional load characteristics. These developments impose increasingly stringent demands on data privacy, computational performance, and response latency. To address these challenges, this paper proposes a task offloading optimization framework based on multi-agent collaborative deep reinforcement learning, tailored for typical IoE scenarios. The ACTOS algorithm enhances privacy protection during the offloading process by explicitly incorporating a data privacy leakage cost as a constraint in the multi-objective optimization framework. In addition, it leverages a shared experience replay mechanism among multiple agents to achieve policy alignment and collaborative optimization across edge nodes, thereby improving the system’s adaptability and stability under dynamic network conditions. Experimental results show that ACTOS significantly outperforms representative baseline algorithms in terms of response latency and energy consumption, while maintaining excellent performance under privacy constraints. The results show that ACTOS delivers real-time performance and energy efficiency without compromising privacy, and is well-suited for future deployment in representative IoE engineering scenarios such as distributed renewable coordination and microgrid scheduling.