Emergency Computation Offloading for Virtual Power Plants Under Edge Server DDoS Attacks

Cloud-edge collaboration enables virtual power plants (VPPs) to efficiently manage demand-side resources (DSRs). However, it also exposes VPPs to the threat of edge server distributed denial of service (DDoS) attacks, which can undermine the dispatchable power capability of a VPP by exhausting its computational resources. This article formulates an emergency computation offloading model to guide the collaboration between cloud servers and edge servers to maximize the dispatchable power of VPP. The model identifies analytically the range of attack intensities where a VPP can maintain full dispatchable power and where the VPP cannot. It also uncovers the convexity property of VPP’s dispatchable power. The proposed model is submodular so that an efficient submodular optimization algorithm can be developed to alleviate the curse of dimensionality inherent in the computation offloading problem. Numerical studies were carried out on the VPP with 9,951 DSRs to validate the effectiveness of the proposed method. A small-scale VPP was also selected to show that the proposed method offers comparable performance, but much less computational cost, to deep reinforcement learning methods.