ReModels: Quantile Regression Averaging models

TL;DR

The paper tackles the challenge of probabilistic electricity price forecasting and reproducibility by focusing on Quantile Regression Averaging (QRA), a leading approach for predicting price quantiles. It introduces ReModels, a Python package that implements nine QRA variants, data acquisition from ENTSO-E, variance-stabilizing transformations, point-forecast pipelines, and standardized probabilistic forecast evaluation. Mathematically, QRA estimates conditional quantiles by solving $\beta_k = \arg\min_{\beta} E[\rho_k(Y - X \beta_k)]$ for each quantile $k$, where $X$ is the matrix of point forecasts and $Y$ is the price; ReModels operationalizes this workflow end-to-end. The package emphasizes reproducibility and extensibility, enabling fair cross-market comparisons and facilitating future extensions to other probabilistic forecasting methods and domains beyond energy markets.

Abstract

Electricity price forecasts play a crucial role in making key business decisions within the electricity markets. A focal point in this domain are probabilistic predictions, which delineate future price values in a more comprehensive manner than simple point forecasts. The golden standard in probabilistic approaches to predict energy prices is the Quantile Regression Averaging (QRA) method. In this paper, we present a Python package that encompasses the implementation of QRA, along with modifications of this approach that have appeared in the literature over the past few years. The proposed package also facilitates the acquisition and preparation of data related to electricity markets, as well as the evaluation of model predictions.

Figures (2)

• Figure 1: A summary of QRA variants. Point predictions block refers to matrix of forecasts of electricity prices obtained in any chosen way, e.g. with autoregressive model trained on historical prices. Multiple series of point predictions can be prepared. Averaging involves computing the average of rows in the point predictions matrix. This average serves as the input for the next step. In the PCA block, Principal Components of the point predictions matrix are computed, with the number of utilized components being a hyper-parameter. Standardization entails scaling the rows in the point predictions matrix w.r.t. their mean and standard deviation.
• Figure 2: The impact of selected transformations on electricity prices on the German market. The plots present the price over time (left) and the corresponding density (right).