Choosing Electricity Price Models for Seasonality, Jumps, and Negative Prices

Summary

The document asks how to model electricity prices, which can feature strong seasonal patterns, jumps, and negative values that make a basic geometric Brownian motion unsuitable for some purposes. The response stresses that model choice depends on the intended use: short-term forecasting, medium-term risk analysis, and long-term investment decisions may call for different approaches.

For short- and medium-term electricity price modelling, the respondent highlights Markov regime-switching and jump-diffusion models as interesting possibilities. The answer points to published research as a broader source for surveying model families, but it does not compare calibration methods, forecast performance, or the handling of seasonality and negative prices in specific implementations. The question’s suggestion that a Variance-Gamma process may permit negative values is not directly evaluated, so the document is guidance on framing the modelling problem rather than a definitive model selection.

Key ideas

• Electricity price models should be selected according to whether the aim is forecasting, risk analysis, or investment analysis.
• Strong seasonality, price jumps, and negative prices can challenge a basic geometric Brownian motion specification.
• Markov regime-switching and jump-diffusion models are suggested for short- to medium-term modelling.
• The response does not assess model performance or directly resolve the question about Variance-Gamma processes.

Tags

# Modelling Power Prices


# Modelling Power Prices












Since electricity prices involve strong seasonality, jump components as well as negative prices and can not be modelled by the GBM, what models/distributions exist, which would allow for modelling these processes. If I remember correctly the Variance-Gamma Process allows for negative prices?

## Answer by Daniel Wahlström (score 1)

https://quant.stackexchange.com/a/61056

This question should be defined a little more clearly. What purpose is the model to have? Are you modelling the prices for short-term (predictive basis), medium-term (risk basis) or long-term (investment basis)?

There are plenty of types of models for all purposes. Raphael Weron has plenty of research published on all this - even including some papers centered around summary of recent research. These should be accessible through e.g. Google Scholar.

Personally, I find Markov Regime-Switching models and jump-diffusion models fascinating for short-term and medium-term electricity price modelling.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.