Dynamic adaptive model predictive control for prosumers-based energy communities

Abstract

Energy communities (ECs) offer a significant opportunity for decentralised energy production. However, realising their full potential is hindered by the significant challenge of managing the high volatility of renewable energy technologies (RETs) and dynamic electricity markets. To address this, the present work introduces a novel dynamic adaptive model predictive control (AMPC) framework designed to simultaneously reduce costs, minimise losses, and enhance RET integration in prosumer-based ECs. The methodology is built upon a high-fidelity dynamic model of the EC, operating with a 50 μs time step to accurately capture the switching dynamics of power electronics and ensure a realistic representation of system behaviour. The key innovation lies in the dynamic adaptation of AMPC weights and power constraints, enabling seamless transitions between a self-sufficiency mode during high-price periods and an economically optimised grid-interactive mode during favourable market conditions. The performance of the AMPC is rigorously benchmarked against fixed MPC strategies and the particle swarm optimisation (PSO) algorithm. The results demonstrate the profound superiority of the adaptive approach, showing reductions in operational costs and power losses of 6.13% to 44.92%, without compromising sustainability. The AMPC's average RET utilisation of 79.31% was superior to that of the fixed-MPC strategies, with improvements ranging from 0.45% to 13.34%. Furthermore, it demonstrated a highly efficient balance against the metaheuristic approach, where a minor 2.53% difference in utilisation was exchanged for significant gains in cost and efficiency. Finally, compared with an adaptive PSO strategy, it reduces 120% power losses and increases 28.33% the capacity utilisation. These results demonstrate a superior framework for achieving a costeffective, efficient, and sustainable operation.