Decentralised Model Predictive Control of Electric Vehicles Charging

Di Giorgio, Alessandro; Giuseppi, Alessandro; Germana, Roberto; Liberati, Francesco · 2019 · Crossref

DOI: 10.1109/smc.2019.8914040

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of managing simultaneous electric vehicle (EV) charging sessions to prevent overloads on electricity distribution grids while satisfying user preferences. The authors propose a decentralized Model Predictive Control (MPC) strategy that optimizes the trade-off between driver satisfaction (utility) and operator costs (aggregated power withdrawal). The motivation stems from the need to handle the flexibility of EV charging without requiring centralized control, which often necessitates sharing sensitive proprietary data between manufacturers and grid operators. By using a decentralized approach, the system preserves user privacy and limits information exchange to power levels and energy price signals. The methodology employs a discrete-time MPC framework where the control problem is formulated as a constrained optimization of social welfare, defined as the difference between total user utility and operational cost. User utility is modeled using a logarithmic function of charging power, while operator cost is modeled using a quadratic function penalizing high power withdrawal and rapid ramping. The core innovation lies in solving this coupled optimization problem via Lagrangian dual decomposition. This technique splits the global problem into independent subproblems for each EV and the grid operator, coordinated by Lagrangian multipliers that act as dynamic energy prices. Each agent solves its convex subproblem locally, and the multipliers are updated iteratively based on the imbalance between aggregated demand and supply until convergence is reached. Simulation results validate the approach using MATLAB. In a simplified scenario with four EVs, the controller successfully smoothed the aggregated power profile compared to an uncontrolled case where vehicles charged at rated power upon arrival. In a more complex scenario involving 20 EVs with overlapping charging windows, the uncontrolled case resulted in power peaks of approximately 90 kW. The proposed decentralized control significantly mitigated these peaks and variability, ensuring all vehicles reached their desired state of charge by their departure times. The iterative optimization process demonstrated convergence, achieving the exit condition in roughly 20 iterations for the simple case and 120 iterations for the congested scenario. The authors note that despite the higher iteration count in complex cases, the low computational complexity of individual subproblems (solving in under a second) makes the method suitable for real-time application. The significance of this work lies in demonstrating that decentralized MPC can effectively manage EV charging loads without centralized coordination. The approach ensures grid stability by shaving peak power withdrawal while guaranteeing user satisfaction. Furthermore, the use of Lagrangian decomposition allows for scalable implementation where agents only exchange minimal information, addressing privacy concerns and reducing communication overhead. This provides a viable framework for integrating large numbers of EVs into distribution networks while maintaining operational efficiency and user autonomy.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• teleoperation remote driving