Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Day-Ahead Electricity Price Forecasting for Volatile Markets Using Foundation Models with Regularization Strategy

Kritchanat Ponyuenyong, Pengyu Tu, Jia Wei Tan, Wei Soon Cheong, Jamie Ng Suat Ling, Lianlian Jiang

TL;DR

The paper addresses day-ahead electricity price forecasting in volatile markets, where spikes and nonlinear dynamics hinder traditional models. It proposes a spike-regularization framework STL-KF and evaluates a broad set of Time Series Foundation Models (TSFMs) against ARIMA, LSTM, CNN-LSTM, and PatchTST using Singapore half-hourly data with exogenous weather and calendar features. Results show TSFMs consistently outperform baselines, with TTMs in multivariate fine-tuning delivering the best accuracy and regularization further boosting robustness. The work provides practical guidance for deploying TSFMs in high-volatility EPF settings and suggests directions for extending the approach to other markets and probabilistic forecasting.

Abstract

Electricity price forecasting (EPF) is essential for energy markets stakeholders (e.g. grid operators, energy traders, policymakers) but remains challenging due to the inherent volatility and nonlinearity of price signals. Traditional statistical and deep learning (DL) models often struggle to capture complex temporal dependencies and integrate heterogeneous data effectively. While time series foundation models (TSFMs) have shown strong performance in general time series forecasting tasks, such as traffic forecasting and weather forecasting. However, their effectiveness in day-ahead EPF, particularly in volatile markets, remains underexplored. This paper presents a spike regularization strategy and evaluates a wide range of TSFMs, including Tiny Time Mixers (TTMs), MOIRAI, MOMENT, and TimesFM, against traditional statistical and DL models such as Autoregressive Integrated Moving Average (ARIMA), Long-short Term Memory (LSTM), and Convolutional Neural Network - LSTM (CNN-LSTM) using half-hourly wholesale market data with volatile trends in Singapore. Exogenous factors (e.g. weather and calendar variables) are also incorporated into models where applicable. Results demonstrate that TSFMs consistently outperform traditional approaches, achieving up to 37.4% improvement in MAPE across various evaluation settings. The findings offer practical guidance for improving forecast accuracy and decision-making in volatile electricity markets.

Day-Ahead Electricity Price Forecasting for Volatile Markets Using Foundation Models with Regularization Strategy

TL;DR

The paper addresses day-ahead electricity price forecasting in volatile markets, where spikes and nonlinear dynamics hinder traditional models. It proposes a spike-regularization framework STL-KF and evaluates a broad set of Time Series Foundation Models (TSFMs) against ARIMA, LSTM, CNN-LSTM, and PatchTST using Singapore half-hourly data with exogenous weather and calendar features. Results show TSFMs consistently outperform baselines, with TTMs in multivariate fine-tuning delivering the best accuracy and regularization further boosting robustness. The work provides practical guidance for deploying TSFMs in high-volatility EPF settings and suggests directions for extending the approach to other markets and probabilistic forecasting.

Abstract

Electricity price forecasting (EPF) is essential for energy markets stakeholders (e.g. grid operators, energy traders, policymakers) but remains challenging due to the inherent volatility and nonlinearity of price signals. Traditional statistical and deep learning (DL) models often struggle to capture complex temporal dependencies and integrate heterogeneous data effectively. While time series foundation models (TSFMs) have shown strong performance in general time series forecasting tasks, such as traffic forecasting and weather forecasting. However, their effectiveness in day-ahead EPF, particularly in volatile markets, remains underexplored. This paper presents a spike regularization strategy and evaluates a wide range of TSFMs, including Tiny Time Mixers (TTMs), MOIRAI, MOMENT, and TimesFM, against traditional statistical and DL models such as Autoregressive Integrated Moving Average (ARIMA), Long-short Term Memory (LSTM), and Convolutional Neural Network - LSTM (CNN-LSTM) using half-hourly wholesale market data with volatile trends in Singapore. Exogenous factors (e.g. weather and calendar variables) are also incorporated into models where applicable. Results demonstrate that TSFMs consistently outperform traditional approaches, achieving up to 37.4% improvement in MAPE across various evaluation settings. The findings offer practical guidance for improving forecast accuracy and decision-making in volatile electricity markets.
Paper Structure (17 sections, 5 equations, 8 figures, 3 tables)

This paper contains 17 sections, 5 equations, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Proposed Method
  3. Forecasting Framework
  4. Data Input
  5. Data Preprocessing: Extreme Price Spike Detection and Regularization
  6. Feature Engineering
  7. Model Configurations
  8. Results and Discussion
  9. Evaluation Metrics
  10. Analysis and Discussion
  11. Discussion
  12. Impact of STL-KF based regularization
  13. Impact of log transformation and features
  14. Comparison of foundation models zero-shot and univariate fine-tuned results
  15. Comparison of foundation models zero-shot and multivariate fine-tuned results
  16. ...and 2 more sections

Figures (8)

  • Figure 1: Overall forecasting framework.
  • Figure 2: Workflow of spikes detection and regularizing strategy STL-KF on raw electricity price data.
  • Figure 3: Comparison of original electricity price data with detected extreme price spikes in red (top), and the regularized price data after anomaly removal (bottom).
  • Figure 4: A closer view of the identified spikes and regularized signals in a randomly selected window.
  • Figure 5: Best MAPE performance of each type of model, ranked from lowest to highest error.
  • ...and 3 more figures