Surrogate Models in Bidirectional Optimization of Coupled Microgrids

TL;DR

This paper proposes replacing iterative optimization in coupled microgrid control with surrogate models like neural networks or radial basis functions, demonstrating efficiency and well-posedness through real-world data simulations.

Contribution

It introduces a novel approach using surrogate models for microgrid optimization, reducing communication rounds and computational effort in coupled microgrid systems.

Findings

Surrogate models achieve comparable control performance to traditional methods.

The approach reduces the number of communication rounds needed.

Numerical simulations validate the method's efficiency with real data.

Abstract

The energy transition entails a rapid uptake of renewable energy sources. Besides physical changes within the grid infrastructure, energy storage devices and their smart operation are key measures to master the resulting challenges like, e.g., a highly fluctuating power generation. For the latter, optimization based control has demonstrated its potential on a microgrid level. However, if a network of coupled microgrids is considered, iterative optimization schemes including several communication rounds are typically used. Here, we propose to replace the optimization on the microgrid level by using surrogate models either derived from radial basis functions or neural networks to avoid this iterative procedure. We prove well-posedness of our approach and demonstrate its efficiency by numerical simulations based on real data provided by an Australian grid operator.