A model predictive control framework with customer-priority tiers for virtual power plant resilience during extreme weather: A UK heatwave case study

TL;DR

This work develops a model predictive control framework for a technical Virtual Power Plant coordinating PV, batteries, and loads to enhance resilience during extreme weather. By formulating a multi-objective MIQCP with customer-priority tiers and a delta-LinDistFlow power-flow approximation, it enables receding-horizon optimization that adapts before, during, and after outages. The approach is validated on the IEEE 33-bus network using UK heatwave data, showing MPC achieves 11-20% higher resilience than traditional full-horizon optimization under forecast and model uncertainties, while maintaining voltage quality and enabling aggressive pre-charging for islanded operation. The results demonstrate the practical viability of MPC-based, priority-aware VPP coordination for resilient, low-carbon grid operation in extreme conditions, with a clear path for future uncertainty-aware extensions and broader DER integration.

Abstract

Due to changes in frequency and intensity of extreme weather events, such as heatwaves and storms, power systems around the globe are having to deal with increased imbalance between demand and supply and additional risk of loss of supply, calling for advanced control strategies that strengthen system resilience. This paper develops a Model Predictive Control (MPC) framework for coordination of Virtual Power Plants (VPPs) that manages photovoltaic (PV) systems, batteries, and loads before, during, and after extreme weather events. A multi-objective mixed-integer quadratically constrained program is solved to enforce customer-priority tiers, serving critical loads first, while minimizing operating cost and PV curtailment under network and device constraints. Simulations on the IEEE 33-bus distribution network with real UK heatwave data show that, under realistic forecast errors and modeling uncertainties, MPC improves resilience by 11-20% relative to traditional full-horizon optimization. These results indicate the practical viability of receding-horizon coordination for resilient, low-carbon VPP operation during extreme weather.

Figures (5)

• Figure 1: Overall VPP architecture showing physical assets (PV, ESS, loads, distribution network), exogenous inputs (weather and price signals), forecasting/error model, MPC-based control, and optimization core across pre-, during-, and post-event phases.
• Figure 2: VPP control strategies. (a) Full-Horizon optimization: a single plan is computed for the entire simulation period, with feasibility checks and corrective re-optimization when violations occur. (b) Model predictive control (MPC): a rolling-horizon optimization is solved at each time step, and only the current control actions are implemented before forecasts and states are updated.
• Figure 3: MPC vs Full-Horizon under ideal conditions.
• Figure 4: MPC vs Full-Horizon under forecasting errors and systematic bias.
• Figure 5: Total battery energy under time-varying battery efficiency.