Topology-free optimal power dispatch for distribution network considering security constraints and flexible building thermal inertia

TL;DR

This paper proposes a topology-free optimal power dispatch method for distribution networks that leverages building thermal inertia and machine learning to ensure security constraints without needing detailed network topology.

Contribution

It introduces a novel topology-free dispatch framework using thermal inertia and ML-based security constraint reformulation, reducing reliance on network topology data.

Findings

Achieves feasible and optimal dispatch without topology information

Utilizes building thermal inertia as a flexible resource

Replaces explicit security constraints with ML-based mixed-integer reformulation

Abstract

With the increasing integration of distributed PV generation, the distribution network requires more and more flexibility to achieve the security-constrained optimal power dispatch. However, the conventional flexibility sources usually require additional investment cost for equipment. Moreover, involving the security constraints is very challenging due to the requirements of accurate network model that may be unavailable in practice. This paper addresses the aforementioned challenge by proposing a topology-free optimal power dispatch framework for distribution networks. It utilizes building thermal inertia to provide flexibility to avoid additional investment. To guarantee the operation safety, a multi-layer perception (MLP) is trained based on historical operational data and then reformulated as mixed-integer constraints to replace the inexplicit original security constraints. Numerical results confirm that the proposed framework can derive a feasible and optimal strategy without any topology information.