Edge computing resource optimization allocation method for virtual power plants

Abstract

Virtual power plants (VPPs) aggregate photovoltaic units, battery storage, electric-vehicle chargers, flexible loads, and optical/electrical sensing devices through cloud-edge-terminal platforms. Centralized cloud scheduling is often too slow for dispatch verification, emergency frequency-control, and burst telemetry processing. This paper proposes a priority- aware edge computing resource allocation method for VPPs. A cloud-edge-terminal model is established, heterogeneous VPP computing tasks are described by data size, CPU cycles, deadline, and operational priority, and a rolling-window allocator jointly optimizes offloading decision, CPU frequency, bandwidth quota, utilization balance, and emergency reserve. The method combines a deadline-priority score, feasibility repair, and dynamic reserve guidance so that routine forecasting and settlement tasks do not occupy capacity required by high-priority control tasks. A simulated IEEE 33-bus VPP with 280 distributed energy-resource endpoints is used for validation. Compared with cloud-only scheduling, the proposed method reduces average latency from 168 ms to 64 ms and decreases deadline violations from 18.6% to 4.9%. Ablation tests further show that priority scoring, feasibility repair, and reserve protection are all necessary for stable real- time operation. The results indicate that edge-side resource allocation can improve computational reliability and operational scalability for automated VPP dispatch.