Alternative Models for Calibrating Commodity Forward Curves

Summary

The document asks how to model energy forward curves using a seasonal component plus a smooth deviation, and whether alternatives exist to the Lucia–Schwartz seasonality specification. It names the Gabillon two-factor model as one alternative, describing it at a high level as a spot-price model with a mean-reverting Gaussian factor and a lognormal long-term level.

The answer is brief and points elsewhere for details. It does not compare calibration procedures, explain how to construct the seasonal function, or give empirical evidence about model performance. As a result, it offers a starting reference rather than a practical evaluation of competing forward-curve methods. Readers would need further sources to assess suitability for particular energy products or data.

Key ideas

• A commodity forward curve can be represented as a seasonal component plus a smooth deviation.
• The question concerns alternatives to the Lucia–Schwartz seasonal specification.
• The Gabillon two-factor model is identified as an alternative for oil futures.
• The answer provides only a high-level model description and no comparative evidence.

Tags

# Commodities forward curve


# Commodities forward curve












I'm dealing with the calibration of the forward curve for energy products. I found an approach proposed by Benth et al., in which the forward curve is parameterized as $f(t) = s(t) + \epsilon(t)$ where $s$ is the seasonality and $\epsilon$ is a smooth curve used to quantify the deviation from the seasonality. In the paper, $s$ has been chosen according to Lucia and Schwartz.

My questions are:

- Which alternatives to Benth et al. are used?

- Do exist other models other than Lucia and Schwartz for the deterministic function $s$ useful in this setting?

## Answer by equanimity (score 1)

https://quant.stackexchange.com/a/69639

To answer your first question, an alternative approach is the Gabillion two-factor model (which was originally proposed for oil futures). At a high-level, Gabillon models the spot price as a single-factor Gaussian process that means reverts to a lognormal long-term rate. See here for additional details.

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.