Paper: arXiv 2407.07795

Authors: Katarzyna Maciejowska, Weronika Nitka

Abstract

In this article, a multiple split method is proposed that enables construction of multidimensional probabilistic forecasts of a selected set of variables. The method uses repeated resampling to estimate uncertainty of simultaneous multivariate predictions. This nonparametric approach links the gap between point and probabilistic predictions and can be combined with different point forecasting methods. The performance of the method is evaluated with data describing the German short-term electricity market. The results show that the proposed approach provides highly accurate predictions. The gains from multidimensional forecasting are the largest when functions of variables, such as price spread or residual load, are considered. Finally, the method is used to support a decision process of a moderate generation utility that produces electricity from wind energy and sells it on either a day-ahead or an intraday market. The company makes decisions under high uncertainty because it knows neither the future production level nor the prices. We show that joint forecasting of both market prices and fundamentals can be used to predict the distribution of a profit, and hence helps to design a strategy that balances a level of income and a trading risk.

Complexity vs Empirical Score

  • Math Complexity: 6.0/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces a nonparametric resampling method (multiple split approach) requiring significant statistical reasoning and extension to multivariate distributions, but the rigorous evaluation on German electricity market data with established metrics (CRPS, coverage) and a practical trading strategy application demonstrates high empirical implementation depth.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Multidimensional Probabilistic<br>Forecasting of Electricity Markets"] --> B["Data: German Short-Term<br>Electricity Market Data"]
  
  B --> C{"Methodology:<br>Multiple Split Approach"}
  
  C --> D["Computational Process:<br>Repeated Resampling<br>Nonparametric Uncertainty Estimation"]
  
  D --> E["Key Findings / Outcomes"]
  
  E --> F["Highly Accurate<br>Multivariate Predictions<br>of Prices & Fundamentals"]
  
  E --> G["Profit Distribution Prediction<br>for Wind Energy Generator<br>Optimizing Day-Ahead vs<br>Intraday Market Decisions"]
  
  E --> H["Enhanced Value from<br>Functions of Variables<br>e.g., Price Spread, Residual Load"]

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 a multiple split method for constructing multidimensional probabilistic forecasts that can be combined with any unidimensional point‑forecasting scheme.

Source sentence
In this article, we propose a multiple split method to construct probabilistic forecasts of both one- and multidimensional random variables.

Assumes 6

  • Operational and maintenance costs are constant per MWh of production. execution
    Source sentence
    For simplicity, these costs are kept constant and expressed in EUR per 1 MWh of production.
  • The company stops generation when a selected profit quantile is less than zero. execution
    Source sentence
    We assume that the generation is stopped when a selected quantile of the profit distribution is less than zero.
  • The training sample is split into two disjoint sets: estimation and calibration. model
    Source sentence
    The approach splits the training sample into two disjoint sets: estimation and calibration.
  • Forecast errors from the training window are used to build an ensemble rather than to estimate distribution quantiles. statistical
    Source sentence
    Second, forecast errors from the training window are used to construct an ensemble of forecasts rather than to estimate the distribution quantiles.
Show 2 more
  • The number of random splits varies between 1 and 20, reducing computational complexity. model
    Source sentence
    MS uses a small number of splits: here, the number varies between 1 and 20. This significantly decreases the computational complexity.
  • The dataset covers four calendar years with hourly resolution, roughly equally divided into training and test periods. data
    Source sentence
    The data span four calendar years, between October 1, 2015 and September 30, 2019, with an hourly resolution. They are roughly equally divided into training and test periods.

Data

Backtest sample: German short-term electricity market · commodities · intraday

  • EPEX SPOT German electricity market · EPEX SPOT market-prices public
    2015-10-01 to 2019-09-30 · German day‑ahead and intraday electricity prices, total load, RES generation
    Source sentence
    In this article, we use a dataset describing the EPEX SPOT market in Germany. The data span four calendar years, between October 1, 2015 and September 30, 2019, with an hourly resolution.

Costs

Modeled Costs are part of the reported headline results.

Operation & Management costs are set to 10 Euro per MWh of production and are included in profit calculations.

Source sentence
Operation & Management costs are 10 Euro/MWh.

Finds 4

  • The multiple split approach yields highly accurate multidimensional probabilistic forecasts, especially for functions such as price spread or residual load. predictive-accuracy
    Source sentence
    The results show that the proposed approach provides highly accurate predictions. The gains from multidimensional forecasting are the largest when functions of variables, such as price spread or residual load, are considered.
  • Joint forecasting of market prices and fundamentals enables prediction of profit distribution, supporting a strategy that balances income and trading risk. performance
    Source sentence
    We show that joint forecasting of both market prices and fundamentals can be used to predict the distribution of future profit, and hence helps to design a trading strategy that balances the level of income and the risk.
  • Using an ensemble of split forecasts avoids the need for aggregating quantiles and reduces variability of results. statistical
    Source sentence
    Second, forecast errors from the training window are used to construct an ensemble of forecasts rather than to estimate the distribution quantiles.
  • The multiple split method reduces computational complexity compared with delete‑d jackknife, jackknife+, or leave‑k‑out methods. theoretical
    Source sentence
    MS uses a small number of splits: here, the number varies between 1 and 20. This significantly decreases the computational complexity.

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.

Cited by 4 papers in the archive (4 by independent authors): The Evolution of Probabilistic Price Forecasting Techniques: … (uses its method or data)

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.