Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Exploring the Interpretability of Forecasting Models for Energy Balancing Market

Oskar Våle, Shiliang Zhang, Sabita Maharjan, Gro Klæboe

TL;DR

The paper investigates the accuracy-interpretability trade-off in forecasting mFRR activation prices in Norway's energy balancing market. It compares a high-performing black-box model (XGBoost) with an inherently interpretable Explainable Boosting Machine (EBM) and a stacked EBM+XGBoost ensemble across five price zones using Volue Insight data. Results show that EBM achieves forecasting accuracy comparable to XGBoost while providing additive, transparent explanations of price drivers and regional dynamics. The study also highlights the difficulty of predicting activation price deviations from the day-ahead spot price and suggests enriching feature sets with real-time grid information in future work.

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.

Exploring the Interpretability of Forecasting Models for Energy Balancing Market

TL;DR

The paper investigates the accuracy-interpretability trade-off in forecasting mFRR activation prices in Norway's energy balancing market. It compares a high-performing black-box model (XGBoost) with an inherently interpretable Explainable Boosting Machine (EBM) and a stacked EBM+XGBoost ensemble across five price zones using Volue Insight data. Results show that EBM achieves forecasting accuracy comparable to XGBoost while providing additive, transparent explanations of price drivers and regional dynamics. The study also highlights the difficulty of predicting activation price deviations from the day-ahead spot price and suggests enriching feature sets with real-time grid information in future work.

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.
Paper Structure (12 sections, 13 equations, 4 figures, 3 tables)

This paper contains 12 sections, 13 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Activation price prediction and explainability
  3. Models for mFRR activation price forecasting
  4. Naïve model
  5. XGBoost
  6. Explainable boosting machine (EBM)
  7. Stacked EBM+XGBoost model
  8. Simulations and results
  9. Dataset
  10. Evaluation of model forecast accuracy
  11. Result analysis and interpretability
  12. Conclusions

Figures (4)

  • Figure 1: Learned EBM feature shape $f(x)$ for heating demand in upwards mFRR across price zones.
  • Figure 2: Global feature importance plots for each price zone for the upward mFRR price by the EBM model. "Month (cos/sine)" represents cosine/sine component of the cyclical encoding for the month of the year, and "Hour (cos/sin) indicates cosine/sine component of the cyclical encoding for the hour of the day.
  • Figure 3: Learned EBM feature shape $f(x)$ for hydropower production in upwards mFRR across price zones.
  • Figure 4: Global feature importance plots for each price zone for the downward mFRR price by the EBM model.