Deep Neural Energy Price Forecasting for the Hydrogen Industry

Research output: ThesisMaster's Thesis

64 Downloads (Pure)

Abstract

The HyCentA Research GmbH analyses, among other things, the technical and economic design of hydrogen plants. Since hydrogen plants are almost exclusively operated electrically, the price of electricity plays a major role in operating costs. Based on transformer models, the electricity price should be predicted for various scenarios. These scenarios consist of the electricity mix (what percentage of the electricity comes from which source), own generation and the gas price. The gas price was added to the electricity generation data because it has a major influence on the electricity price due to the merit order system. This could be observed, among other things within the Ukraine crisis (2022). In addition, these transformer models were used to identify electricity price trends depending on the type of electricity generation. As the energy system in Europe is moving towards more renewable energies, the electricity mix is also changing towards these. A strong positive trend towards lower electricity prices was observed for wind energy and biomass in particular. The opposite trend was observed for solar power generation: Electricity prices rose as solar power generation increased. However, it was observed that even small amounts of solar power in the electricity mix reduce the price of electricity. This means that the electricity price initially starts at a lower price and only rises later. The later increase could be explained by the increased need for balancing energy due to the volatile nature of solar power generation. However, this effect still is subject to further investigation.
Translated title of the contributionTiefe neuronale Energiepreisprognose für die Wasserstoffindustrie
Original languageEnglish
QualificationDipl.-Ing.
Awarding Institution
  • Montanuniversität
Supervisors/Advisors
  • Rückert, Elmar, Supervisor (internal)
  • Dave, Vedant, Co-Supervisor (internal)
Award date22 Mar 2024
DOIs
Publication statusPublished - 2024

Bibliographical note

no embargo

Keywords

  • Energy
  • Price
  • Forecasting
  • Transformer
  • LSTM
  • Long Short Term-Memory Network
  • Scenarioanalysis
  • ENTSO-E
  • Neural Network
  • Artificial Intelligence
  • Energy market
  • Day Ahead Price
  • Energy mix
  • Gas price

Cite this