Home > Research > Publications & Outputs > Renewable prediction-driven service offloading ...


Text available via DOI:

View graph of relations

Renewable prediction-driven service offloading for IoT-enabled energy systems with edge computing

Research output: Contribution to Journal/MagazineJournal articlepeer-review

<mark>Journal publication date</mark>4/08/2021
<mark>Journal</mark>Wireless Networks
Publication StatusPublished
<mark>Original language</mark>English


The emerging of the Internet of Things (IoT) enables the interconnection among everything. With edge computing serving low-latency services, IoT makes intelligent energy management become a possibility, thereby enhancing the energy sustainability for energy systems. Currently, renewable energy is widely applied in energy systems to alleviate the carbon footprint. However, the instability and discontinuity of renewable generation decrease the quality of service (QoS) of edge servers. To address the challenge, a renewable prediction-driven service offloading method, named ReSome, is proposed. Technically, a deep-learning-based approach is designed for renewable energy prediction firstly. Next, the service offloading process is abstracted to a Markov decision process (MDP). With the predicted renewable energy amount, asynchronous advantage actor-critic (A3C) is leveraged to determine the optimal service offloading strategy. Finally, by utilizing a real-world solar power generation dataset, the experimental evaluation validates the capability and effectiveness of ReSome.