Simulering av Elektricitetspriser med Volatiliteter beroende av Dolda Markovprocesser

The dynamics of electricity prices are complex, and many actors in the energy industry aim to mitigate financial risks associated with price fluctuations. The increasing interest in renewable energy can also lead to sharper price increases, influenced by current energy supply and consumption levels. These dynamics can manifest itself as volatility shifts across different economic regimes, but further research on their predictive power is needed. This thesis analyzes and compares models for daily electricity prices from the Nord Pool market, spanning April 2018 to January 2025. Specifically, it compares an Ornstein-Uhlenbeck (OU) process where volatilities depend on a hidden Markov process with a standard OU model without Markov-modulated volatility. The method proposed by Elliott et al. (2003) was adapted to estimate Markov-modulated volatilities, while other parameters were estimated using maximum likelihood. Volatility forecasts were generated and the properties of the white noise and confidence intervals were further investigated. The results demonstrate a major reduction in forecast errors when hidden Markov processes were incorporated and the negative logarithmic likelihood exhibited better performance under volatile market conditions. This suggests that incorporating economic regimes into volatility modeling is a valuable approach to capture abrupt changes in volatility. However, the assumption of normally distributed white noise appears to be inadequate and should be replaced with distributions that better account for heavy tails.