Artificial intelligence in Energinet’s energy system simulations

PhD student Stefanie Buchholz, DTU Management, has significantly improved the methodologies for solving the capacity expansion problem. The capacity expansion problem is concerned with investing in VRE, generation units etc. to meet demand such that socio economy in the energy system is optimized. Energinet’s software Sifre/Adapt solves the capacity expansion problem and is used to investigate the socio-economic potential of e.g. Power to X facilities in Denmark.

The research of Stefanie Buchholz consists of artificial intelligent methods to select representative days from a given time horizon and then make investment decisions based on these days only. It is implemented in Sifre/Adapt and has reduced simulation times with 90% while maintaining 90% solution quality.

“This is a huge improvement of our Sifre/Adapt Software”, Mette Gamst from Energinet explains. “Given the large uncertainties in the input data, we are not too concerned with sub-optimal solutions. Reducing our simulation times from up to 8 hours to less than one hour, enables us to perform more analyses on more scenarios and thus improve the quality of our work”.

Implementation of the methods in Adapt was made possible through a close cooperation between Energinet and Stefanie Buchholz and David Pisinger, WP5 lead, from DTU Management.

“Working together with Stefanie and David has inspired us to continue the development of our simulation tools”, says Mette Gamst. “We look forward to testing more of Stefanie’s promising research results in our software, such as her AI methods for investigating the robustness of investments.”

 

Leave a Reply

Your email address will not be published. Required fields are marked *