Advanced Warehouse Energy Storage System Control Using Deep Supervised- and Reinforcement Learning

Sven Opalic will defend his PhD thesis Advanced Warehouse Energy Storage System Control Using Deep Supervised- and Reinforcement Learning 10 November 2023.

Summary of the thesis:

As we continue to build increasingly energy efficient buildings the complexity of the more ambitious projects also increases rapidly. The energy systems in these new buildings are now so complex, with the possibility for local production and storage of electrical and thermal energy, that the construction companies are not able to plan and construct these buildings in an optimal way. During operation these very complex building-integrated energy systems are often operated by unqualified personnel who do not have the necessary knowledge to optimize the buildings energy usage. The same personnel are also expected to manage several other facility management tasks, unrelated to energy-efficient operation. In this doctoral thesis, we approach these challenges with the exciting possibilities within the fields of machine learning and artificial intelligence (A.I.) for the intelligent operation of building integrated energy systems. The thesis explores moving away from predetermined programming of how the energy system should react to all possible scenarios and explores possibilities within "Deep Reinforcement Learning". The A.I. will over time be able to learn how to manage the energy system in the most efficient way within strict limitations set in place to provide operational security and a satisfactory indoor climate. The software will be implemented and tested on the new Rema 1000 main distribution warehouses located at Vagle outside Sandnes and at Vinterbro outside Oslo. The projects have received Enova funding for highly complex and efficient energy systems, containing large solar power plants, thermal and electrical energy storage, energy-efficient and environmentally friendly cooling installations, and automated logistics systems.

Find more information about time and place for the doctoral defense.