Powering the future: AI innovations in the energy sector

By Jan Helsen (VUB – Accoustics & Vibrations Research Group)

The share of renewable energy production in the electricity mix is increasing fast. One of the thriving factors is the significant cost reduction of wind and solar generated electricity. However, these sources are intermittent in nature. Furthermore, the energy generation happens in a distributed way since wind farms and solar plants are spread around geographically. These two challenges require coordination between the renewable and conventional electricity production units to meet overall grid demand. Energy storage solutions such as batteries or hydrogen production units can support this by creating energy buffers. AI can play a substantial role in this complex ecosystem. This talk focusses on different aspects: advanced weather forecasts, asset health assessment, AI-supported control of production units (e.g., offshore wind farms) focusing on multi-objective optimization and finally optimized AI-supported collaboration between energy production and storage units to minimize grid impact as well as optimize storage unit capacity.

By Chris Develder (UGent)

The AI for Energy (AI4E) team at IDLab, Ghent University, focuses on developing new data analytics and machine learning algorithms to support the energy transition. The latter faces challenges coping with on the one hand the still growing deployment of renewable sources in the electricity system, and on the other the increasing electrification (e.g., heat pumps, electric vehicles). Aside from diagnostic and predictive support tools for grid operators, the main focus of AI4E is on applications of reinforcement learning (RL) to develop innovative control algorithms for flexible assets such as heat pumps, batteries, and electric vehicle chargers. This talk will highlight some recent works of the team in the realm of (1) data-efficient learning of high-quality control policies, and (2) explainable AI. For (1), we will highlight the benefit of using physics-informed RL for the case of heat pump control. For (2), we explore the distillation of an interpretable decision tree controller derived from a black box RL policy, again for heat pump control. Another case study for (2) will illustrate a policy correction framework for battery control.

By Daneel Geysen (RENEWI)

This session will showcase applied AI use cases at RENEWI, such as detecting gas canisters and sorting waste, highlighting some real-world AI integrations. They will also share lessons learned from adopting AI, focusing on the challenges faced during integration. Additionally, the session will explore the role of onsite renewable energy, including wind and solar projects aimed at optimizing energy management and addressing their future energy challenges.