IEET: DOSA

DOSA

Decentralised optimisation of existing energy supply infrastructure through self-learning agent systems

Abstract

The expansion of decentralized renewable power plants and the ongoing electrification of the heating and transport sectors are leading to significant fluctuations in feed-in, load and energy flows within low-voltage distribution networks. These networks originally were not designed for bidirectional energy transfer or high simultaneous loads, meaning that grid overloads and voltage band violations are expected to become more frequent in the future. Whilst conventional grid expansion can mitigate this, it involves condisderable planning, technical and financial effort.

The aim of the DOSA project is therefore to develop and simulate an autonomous, decentralized and self-organizing control concept for the efficient, stable and climate-friendly operation of low-voltage grids. The concept is based on a multi-agent system that uses artificial intelligence (AI) and machine learning tools to enable adaptive coordination between generators and consumers within local grid segments.

The system architecture comprises two levels: at household level, local agents optimize self-consumption from photovoltaic systems and controllable loads. At grid level, the connected units coordinate automatically to maximize local energy utilization whilst adhering to technical constraints. This is based on AI-supported forecasts of generation and consumption, which enable proactive and flexible load control.

By reducing the need for active intervention by grid operators, efficiency is improved whilst also promoting the integration of climate-friendly technologies such as photovoltaics, heat pumps and electric vehicles. The concept thus strengthens acceptance of the energy transition and accelerates the integration of renewable energies into existing distribution networks. Furthermore, the system increases the resilience of energy supply, as existing assets can be utilized optimally and costly grid expansion measures can be reduced or postponed.