Paper: arXiv 2404.02270
Authors: Arkadiusz Lipiecki, Bartosz Uniejewski, Rafał Weron
Abstract
Operational decisions relying on predictive distributions of electricity prices can result in significantly higher profits compared to those based solely on point forecasts. However, the majority of models developed in both academic and industrial settings provide only point predictions. To address this, we examine three postprocessing methods for converting point forecasts of day-ahead electricity prices into probabilistic ones: Quantile Regression Averaging, Conformal Prediction, and the recently introduced Isotonic Distributional Regression. We find that while the latter demonstrates the most varied behavior, it contributes the most to the ensemble of the three predictive distributions, as measured by Shapley values. Remarkably, the performance of the combination is superior to that of state-of-the-art Distributional Deep Neural Networks over two 4.5-year test periods from the German and Spanish power markets, spanning the COVID pandemic and the war in Ukraine.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced statistical methods (quantile regression, conformal prediction, isotonic regression, Shapley values) with significant mathematical formulation, while also demonstrating strong empirical rigor through extensive backtesting on real market data over long periods, including major crises, and comparing against state-of-the-art benchmarks.
Research Flowchart
flowchart TD
A["Research Goal<br>Enhance probabilistic forecasting of<br>day-ahead electricity prices"] --> B["Data: German & Spanish Power Markets<br>4.5-year periods<br>COVID & War in Ukraine"]
B --> C["Input: Point Forecasts"]
C --> D["Postprocessing Methods"]
subgraph D
direction LR
D1["Quantile Regression Averaging"]
D2["Conformal Prediction"]
D3["Isotonic Distributional Regression"]
end
D --> E["Ensemble of Predictive Distributions"]
E --> F["Key Findings"]
subgraph F
F1["Isotonic Distributional Regression<br>contributes most via Shapley values"]
F2["Combination outperforms<br>state-of-the-art Deep Neural Networks"]
F3["Operational decisions based on<br>predictive distributions yield higher profits"]
end
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes converting point forecasts of day-ahead electricity prices into probabilistic forecasts using three post‑processing methods and combines them into an ensemble called LEAR‑Ave.
Source sentence
We examine three postprocessing methods for converting point forecasts of day-ahead electricity prices into probabilistic ones: Quantile Regression Averaging, Conformal Prediction, and the recently introduced Isotonic Distributional Regression.
Assumes 2
- Conformal Prediction quantiles are derived under the assumption of symmetrically distributed errors. statistical
Source sentence
However, the estimated PIs are centered on the point forecast, so obtaining quantile forecasts from CP is only possible under the assumption of symmetrically distributed errors.
- The innovations of the time‑series models are Gaussian (N(0,σ)). statistical
Source sentence
A common assumption underlying time series models is that the innovations are Gaussian.
Data
- ENTSO‑E day‑ahead market data market-prices public
Source sentence
The data we use is publicly available and has been downloaded from ENTSO‑E (https://transparency.entsoe.eu; day-ahead prices, day-ahead load forecasts, day-ahead onshore/offshore wind and solar generation forecasts)
- Investing.com fundamental data fundamentals public
Source sentence
and Investing.com (https://www.investing.com/; carbon emission, ...
Costs
Not applicable
Finds 3
- The LEAR‑Ave ensemble yields the lowest CRPS across both German and Spanish markets and outperforms state‑of‑the‑art Distributional Deep Neural Networks over the full 4.5‑year test periods. performance
Source sentence
The resulting LEAR‑Ave combination outperforms state-of-the-art Distributional Deep Neural Networks of Marcjasz et al. (2023) over two 4.5‑year test periods from the German and Spanish power markets
- Isotonic Distributional Regression contributes the most to the ensemble according to Shapley values. predictive-accuracy
Source sentence
we find that while the latter demonstrates the most varied behavior, it contributes the most to the ensemble of the three predictive distributions, as measured by Shapley values
- LEAR‑Ave ensemble yields the lowest CRPS across both markets and all test subperiods, except in a few isolated subperiods where other models are comparable. performance
Source sentence
This time, the LEAR‑Ave ensemble yields the lowest CRPS across both markets and all test subperiods, except the first sub‑period labeled ‘2000†’ in Germany
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 12 related papers). Coverage is partial, so this is neither replication nor refutation.
6 papers in the archive : Online Multivariate Regularized Distributional Regression … ; Probabilistic Forecasting for Day-ahead Electricity Prices, …
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (46)
- Combining predictive distributions for the statistical post-processing of ensemble forecasts (2018)
- CRPS learning (2023)
- Forecasting electricity prices with expert, linear, and nonlinear models (2023)
- Generative machine learning methods for multivariate ensemble postprocessing (2024)
- A probabilistic forecast methodology for volatile electricity prices in the Australian National Electricity Market (2024)
- Least angle regression (2004)
- Regularization paths for generalized linear models via coordinate descent (2010)
- Additive models and robust aggregation for GEFCom2014 probabilistic electric load and electricity price forecasting (2016)
and 38 more
- Tests of conditional predictive ability (2006)
- Probabilistic solar forecasting: Benchmarks, post-processing, verification (2023)
- Strictly Proper Scoring Rules, Prediction, and Estimation (2007)
- Combining prediction intervals in the M4 competition (2020)
- Quantile evaluation, sensitivity to bracketing, and sharing business payoffs (2017)
- Statistical Learning with Sparsity: The Lasso and Generalizations (2015)
- Accelerating the Pool-Adjacent-Violators Algorithm for Isotonic Distributional Regression (2022)
- Isotonic distributional regression (2021)
- Conformal prediction interval estimation and applications to day-ahead and intraday power markets (2021)
- Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark (2021)
- Is It Better to Average Probabilities or Quantiles? (2013)
- Probabilistic load forecasting via Q uantile R egression A veraging on sister forecasts (2017)
- From local explanations to global understanding with explainable AI for trees (2020)
- Understanding global feature contributions with additive importance measures (2020)
- A hybrid model for GEFCom2014 probabilistic electricity price forecasting (2016)
- Forecasting Electricity Prices (2023)
- Forecasting electricity prices using deep neural networks: A robust hyper-parameter selection scheme (2020)
- Distributional neural networks for electricity price forecasting (2023)
- Probabilistic electricity price forecasting with NARX networks: Combine point or probabilistic forecasts? (2020)
- Econometric modelling and forecasting of intraday electricity prices (2020)
- Combining predictive distributions of electricity prices. Does minimizing the CRPS lead to optimal decisions in day-ahead bidding? — → 2308.15443
- Computing electricity spot price prediction intervals using quantile regression and forecast averaging (2015)
- Recent advances in electricity price forecasting: A review of probabilistic forecasting (2018)
- Neural basis expansion analysis with exogenous variables: F orecasting electricity prices with NBEATSx (2023)
- Neural Networks for Postprocessing Ensemble Weather Forecasts (2018)
- A tutorial on conformal prediction (2008)
- On the importance of the long-term seasonal component in day-ahead electricity price forecasting: Part II – P robabilistic forecasting (2019)
- Regularized Quantile Regression Averaging for probabilistic electricity price forecasting (2021)
- Variance Stabilizing Transformations for Electricity Spot Price Forecasting (2018)
- Statistical postprocessing for weather forecasts: Review, challenges, and avenues in a big data world (2021)
- Easy Uncertainty Quantification ( EasyUQ ): Generating Predictive Distributions from Single-Valued Model Output (2024)
- Combining Probabilistic Load Forecasts (2019)
- Electricity price forecasting: A review of the state-of-the-art with a look into the future (2014)
- Combining quantiles of calibrated solar forecasts from ensemble numerical weather prediction (2023)
- Adaptive Conformal Predictions for Time Series (2022)
- Probabilistic mid- and long-term electricity price forecasting (2018)
- Day-ahead electricity price forecasting with high-dimensional structures: U nivariate vs. multivariate modeling frameworks (2018)
- PostForecasts.jl – J ulia package for postprocessing forecasts (2024) — link
1 of 46 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.