Paper: arXiv 2510.12685
Authors: Runyao Yu, Ruochen Wu, Yongsheng Han, Jochen L. Cremer
Abstract
Accurate probabilistic forecasting of intraday electricity prices is critical for market participants to inform trading decisions. Existing studies rely on specific domain features, such as Volume-Weighted Average Price (VWAP) and the last price. However, the rich information in the orderbook remains underexplored. Furthermore, these approaches are often developed within a single country and product type, making it unclear whether the approaches are generalizable. In this paper, we extract 384 features from the orderbook and identify a set of powerful features via feature selection. Based on selected features, we present a comprehensive benchmark using classical statistical models, tree-based ensembles, and deep learning models across two countries (Germany and Austria) and two product types (60-min and 15-min). We further perform a systematic generalization study across countries and product types, from which we reveal an asymmetric generalization phenomenon.
Complexity vs Empirical Score
- Math Complexity: 5.0/10
- Empirical Rigor: 8.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper presents advanced statistical modeling with specific loss functions and regularization for quantile forecasting, but its core contribution is a rigorous, multi-country/multi-product empirical benchmark with systematic generalization tests using real market data and comprehensive hyperparameter tuning.
Research Flowchart
flowchart TD A["Research Goal<br>Assess orderbook features & generalization<br>in intraday electricity forecasting"] --> B["Data Collection & Feature Extraction<br>Germany & Austria: 60-min & 15-min products"] B --> C["Feature Selection<br>Identify top 384 orderbook features"] C --> D["Model Benchmarking<br>Classical ML, Tree-based, Deep Learning"] D --> E["Generalization Study<br>Train on Country A / Test on Country B"] E --> F["Key Findings<br>1. Orderbook features significantly improve accuracy<br>2. Asymmetric generalization observed<br>3. 60-min > 15-min generalization"]
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 extracting a large set of order‑book features and using feature selection to build probabilistic intraday electricity price forecasting models, and studies asymmetric cross‑market generalization.
Source sentence
we extract 384 features from the orderbook and identify a set of powerful features via feature selection.
Assumes 3
- Weak‑form market efficiency is assumed when using only the last price as a predictor. market
Source sentence
Previous works even argue that the last price already reflects past information, assuming weak‑form efficiency
- Trade‑count is used as a proxy for market liquidity. data
Source sentence
We use trade-count as a simple liquidity proxy
- The extracted order‑book features are assumed to capture the relevant information for price forecasting. model
Source sentence
the rich information in the orderbook remains underexplored
Data
- Intraday order‑book data order-book unknown
Source sentence
across two countries (Germany and Austria) and two product types (60‑min and 15‑min)
Costs
Not applicable
Finds 3
- Models trained on more liquid markets or products transfer well to less liquid ones, while the reverse transfer leads to substantial performance degradation. performance
Source sentence
models trained on more liquid markets or products transfer well to less liquid ones, whereas the reverse transfer leads to substantial performance degradation
- A comprehensive benchmark of classical statistical models, tree‑based ensembles, and deep learning models is provided using the selected order‑book features. other
Source sentence
we present a comprehensive benchmark using classical statistical models, tree‑based ensembles, and deep learning models
- Quantile MLPs (QMLPs) emerge as a strong baseline for probabilistic forecasting when using domain features. performance
Source sentence
QMLPs emerge as a strong baseline for probabilistic forecasting when using domain features
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 14 related papers). Coverage is partial, so this is neither replication nor refutation.
6 papers in the archive
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (26)
- Ensemble forecasting for intraday electricity prices: Simulating trajectories (2020)
- Efficient capital markets (1970)
- Econometric modelling and forecasting of intraday electricity prices (2020)
- Short-term price forecasting models based on artificial neural networks for intraday sessions in the Iberian electricity market (2016)
- Probabilistic price forecasting for day-ahead and intraday markets: Beyond the statistical model (2017)
- Forecasting the price distribution of continuous intraday electricity trading (2019)
- Simulation-based forecasting for intraday power markets: Modelling fundamental drivers for location, shape and scale of the price distribution (2024)
- Multivariate probabilistic forecasting of intraday electricity prices using normalizing flows (2023)
and 18 more
- Multivariate simulation-based forecasting for intraday power markets: Modeling cross-product price effects (2024)
- Towards the Prediction of Electricity Prices at the Intraday Market Using Shallow and Deep-Learning Methods (2021)
- Quantile and Probability Curves Without Crossing (2010) — link
- OrderFusion: Encoding Orderbook for End-to-End Probabilistic Intraday Electricity Price Forecasting — → 2502.06830
- Understanding intraday electricity markets: Variable selection and very short-term price forecasting using LASSO (2019) — link
- Trading on short-term path forecasts of intraday electricity prices (2022)
- Winner’s Curse? On Pace, Progress, and Empirical Rigor (2018) — link
- A Troubling Analysis of Reproducibility and Progress in Recommender Systems Research (2021) — link
- Optuna: A Next-generation Hyperparameter Optimization Framework (2019) — link
- Comparing predictive accuracy (2002)
- Kernel and nearest-neighbor estimation of a conditional quantile (1990)
- A tutorial on support vector regression (2004)
- Regression quantiles (1978)
- LightGBM: A Highly Efficient Gradient Boosting Decision Tree (2017) — link
- XGBoost: A Scalable Tree Boosting System (2016) — link
- Neural Networks for Pattern Recognition (1995)
- KAN : Kolmogorov \textendash Arnold Networks (2025) — link
- Deep Learning for Electricity Price Forecasting: A Review of Day-Ahead, Intraday, and Balancing Electricity Markets — → 2602.10071
2 of 26 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.