Exploring the Interpretability of Forecasting Models for Energy Balancing Market

Paper: arXiv 2602.00049

Authors: Oskar Våle, Shiliang Zhang, Sabita Maharjan, Gro Klæboe

Abstract

The balancing market in the energy sector plays a critical role in physically and financially balancing the supply and demand. Modeling dynamics in the balancing market can provide valuable insights and prognosis for power grid stability and secure energy supply. While complex machine learning models can achieve high accuracy, their black-box nature severely limits the model interpretability. In this paper, we explore the trade-off between model accuracy and interpretability for the energy balancing market. Particularly, we take the example of forecasting manual frequency restoration reserve (mFRR) activation price in the balancing market using real market data from different energy price zones. We explore the interpretability of mFRR forecasting using two models: extreme gradient boosting (XGBoost) machine and explainable boosting machine (EBM). We also integrate the two models, and we benchmark all the models against a baseline naive model. Our results show that EBM provides forecasting accuracy comparable to XGBoost while yielding a considerable level of interpretability. Our analysis also underscores the challenge of accurately predicting the mFRR price for the instances when the activation price deviates significantly from the spot price. Importantly, EBM’s interpretability features reveal insights into non-linear mFRR price drivers and regional market dynamics. Our study demonstrates that EBM is a viable and valuable interpretable alternative to complex black-box AI models in the forecast for the balancing market.

Complexity vs Empirical Score

• Math Complexity: 3.5/10
• Empirical Rigor: 6.5/10
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper applies standard gradient boosting algorithms (XGBoost and EBM) to forecast mFRR activation prices in Norwegian energy markets. While the mathematical formulations are standard implementations of GAMs and boosted trees, the empirical focus on interpretability versus accuracy in balancing market dynamics offers practical insights.

Research Flowchart

  flowchart TD
    A["Research Goal:<br>Evaluate Accuracy vs. Interpretability Trade-off in mFRR Price Forecasting"] --> B["Data Inputs:<br>Real energy balancing market data across price zones"]
    
    B --> C["Model Training & Benchmarking"]
    
    subgraph C ["Model Training & Benchmarking"]
        C1["Naive Baseline Model"]
        C2["XGBoost (Black-Box)"]
        C3["EBM (Glass-Box)"]
        C4["Ensemble (XGBoost + EBM)"]
    end
    
    C --> D["Evaluation:<br>Forecasting Accuracy & Global/Local Feature Interpretability"]
    
    D --> E["Key Outcomes:<br>1. EBM matches XGBoost accuracy with high interpretability<br>2. Identifies non-linear price drivers & regional market dynamics<br>3. Highlights challenge in predicting extreme mFRR price spikes"]

Cited References (90)

1. Impacts of large-scale deployment of vertical bifacial photovoltaics on European electricity market dynamics (2024)
2. The impact of electricity market reform on renewable energy production (2024) — link
3. Electricity balancing challenges for markets with high variable renewable generation (2024) — link
4. The impact of variable renewable energy technologies on electricity markets: An analysis of the Turkish balancing market (2021) — link
5. Short-term imbalance forecasting with machine learning for proactive mFRR controlling (2024) — link
6. Comprehensive review of transmission system operators–Distribution system operators collaboration for flexible grid operations (2024) — link
7. Transforming the electrical grid: the role of AI in advancing smart, sustainable, and secure energy systems (2025) — link
8. Understanding the Black Box: A Deep Empirical Dive into Shapley Value Approximations for Tabular Data (2025) — link

0 of 90 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.