Adaptive Partitioned Hierarchical Model Predictive Control for Virtual Power Plants: A Dynamic Clustering Approach to Air Conditioning Load Aggregation

This paper introduces an adaptive hierarchical model predictive control (MPC) framework tailored for virtual power plants (VPPs) to optimize air conditioning (AC) load aggregation through dynamic clustering and multi-granular control. The increasing integration of intermittent renewable energy sources highlights the critical need for flexible demand-side resources. However, conventional AC load aggregation methods often lack the responsiveness required to adapt to real-time grid conditions and thermal dynamics effectively. To address this limitation, the proposed framework employs a graph-based spectral clustering technique that dynamically partitions aggregated AC loads into hierarchical subsets, adapting to grid events and thermal coupling between units. Each cluster is managed by a distributed MPC controller, which optimizes grid stability, user comfort, and control effort through a multi-objective cost function with real-time weight adjustments based on electricity prices. Furthermore, the framework incorporates a thermal inertia model to harness the inherent energy storage potential of buildings, enabling predictive coordination of AC units while adhering to thermal comfort constraints. The system integrates seamlessly with VPP infrastructure by replacing traditional demand forecasts with real-time grid signals and dispatching optimized control actions via edge-computing devices and low-latency communication networks. Experimental validation demonstrates that the proposed approach significantly outperforms static aggregation strategies in terms of flexibility and responsiveness, providing a scalable and efficient solution for demand-side management in modern power systems. The key innovation lies in the synergistic integration of adaptive partitioning, hierarchical MPC, and thermal inertia exploitation, collectively enhancing the VPP’s ability to deliver grid services without compromising end-user requirements. This work advances the role of VPPs in achieving grid stability and sustainability through intelligent demand-side optimization.