Optimal Allocation of ESSs in Active Distribution Networks to achieve their Dispatchability

TL;DR

This paper introduces a convex optimization-based method for optimally siting and sizing energy storage systems in active distribution networks to improve dispatchability, accounting for uncertainties in renewable generation and loads.

Contribution

It proposes a novel two-level optimization framework combining AR-OPF and Benders decomposition for optimal ESS placement and sizing in radial grids.

Findings

Achieves global optimal solutions for ESS siting and sizing.

Reduces dispatch error in active distribution networks.

Validated on a real Swiss network with high PV penetration.

Abstract

This paper presents a method for the optimal siting and sizing of energy storage systems (ESSs) in active distribution networks (ADNs) to achieve their dispatchability. The problem formulation accounts for the uncertainty inherent to the stochastic nature of distributed energy sources and loads. Thanks to the operation of ESSs, the main optimization objective is to minimize the dispatch error, which accounts for the mismatch between the realization and prediction of the power profile at the ADN connecting point to the upper layer grid, while respecting the grid voltages and ampacity constraints. The proposed formulation relies on the so-called Augmented Relaxed Optimal Power Flow (AR-OPF) method: it expresses a convex full AC optimal power flow, which is proven to provide a global optimal and exact solution in the case of radial power grids. The AR-OPF is coupled with the proposed dispatching control resulting in a two-level optimization problem. In the first block, the site and size of the ESSs are decided along with the level of dispatchability that the ADN can achieve. Then, in the second block, the adequacy of the ESS allocations and the feasibility of the grid operating points are verified over operating scenarios using the Benders decomposition technique. Consequently, the optimal size and site of the ESSs are adjusted. To validate the proposed method, simulations are conducted on a real Swiss ADN hosting a large amount of stochastic Photovoltaic (PV) generation.