A Predictive Flexibility Aggregation Method for Low Voltage Distribution System Control

TL;DR

The paper addresses voltage management in low voltage distribution networks with high DER penetration by aggregating residential flexibility into a three dimensional chart of $P$, $Q$, and the associated cost. It introduces a three stage workflow comprising offline multiparametric optimization to build explicit region based mappings, an explicit model predictive control that accounts for forecasts, and a real time control layer that ties measurements to nodal exchanges and asset level disaggregation. The approach is privacy preserving and scalable since most computations are performed offline at the residential unit level and online computation is distributed and reduced to a central projection and disaggregation step. Validation on a 5 bus LV network demonstrates near optimal performance with improved privacy and potential cost savings, highlighting the practical potential for scalable, low cost distribution system management.

Abstract

This paper presents a predictive control strategy to manage low-voltage distribution systems. The proposed approach relies on an aggregate of the flexibility at the residential unit level into a three-dimensional chart that represents the injected active and reactive power, and the flexibility cost. First, this method solves a multiparametric optimization problem offline at the residential unit level to aggregate the flexibility of the assets. Then, a semi-explicit model predictive control problem is solved to account for forecasts. By combining the results of these problems with measurements, the method generates the desired flexibility chart. The proposed approach is compatible with realtime control requirements, as heavy computations are performed offline locally, making it naturally parallelizable. By linking realtime flexibility assessment with energy scheduling, our approach enables efficient, low-cost, and privacy-preserving management of low-voltage distribution systems. We validate this method on a low-voltage network of 5 buses by comparing it with an ideal technique.

Figures (6)

• Figure 1: Simplified representation of the different stages of the proposed control strategy
• Figure 2: Example of piecewise affine value function with respect to the SOC at the end of the ongoing market period
• Figure 3: Timeline of the multiparametric optimization problem, separated into 2 intervals and incorporating a view of the future through $\Pi_{SOC}$
• Figure 4: Approximation of the PV flexibility chart depending on $P_{max}$Powertech2025
• Figure 5: Example of approximation of the BESS flexibility chart