Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark (2008.08004v2)

Abstract: While the field of electricity price forecasting has benefited from plenty of contributions in the last two decades, it arguably lacks a rigorous approach to evaluating new predictive algorithms. The latter are often compared using unique, not publicly available datasets and across too short and limited to one market test samples. The proposed new methods are rarely benchmarked against well established and well performing simpler models, the accuracy metrics are sometimes inadequate and testing the significance of differences in predictive performance is seldom conducted. Consequently, it is not clear which methods perform well nor what are the best practices when forecasting electricity prices. In this paper, we tackle these issues by performing a literature survey of state-of-the-art models, comparing state-of-the-art statistical and deep learning methods across multiple years and markets, and by putting forward a set of best practices. In addition, we make available the considered datasets, forecasts of the state-of-the-art models, and a specifically designed python toolbox, so that new algorithms can be rigorously evaluated in future studies.

• The paper presents an open-access benchmark dataset and models that standardize the evaluation of day-ahead electricity price forecasting methods.
• The authors implement and compare two accessible forecasting models—a Lasso-based autoregressive approach and a deep neural network—highlighting their efficacy.
• The study advocates for best practices including robust metrics such as relative MAE and ensemble techniques to enhance forecasting accuracy.

Evaluating Models for Day-Ahead Electricity Price Forecasting: A Methodological Study

In the evolving landscape of electricity markets, accurately forecasting day-ahead prices is critical for stakeholders looking to optimize decision-making processes in energy trading. The paper at hand provides an in-depth review of state-of-the-art electricity price forecasting (EPF) algorithms, alongside offering a comprehensive open-access benchmark that aims to establish a rigorous standard for evaluating new predictive methods in the field.

Core Contributions and Findings

The authors identify a need for more systematic evaluations of EPF models, highlighting prevalent issues such as the use of non-public datasets and inadequately short test periods that undermine the reliability and generalizability of results. To counter this, the paper proposes three substantial contributions:

1. Benchmark Dataset: The authors provide an open-access benchmark dataset composed of six years of electricity price and exogenous variable data for five distinct day-ahead electricity markets. This dataset is designed to facilitate comprehensive and fair comparisons.
2. Open-Source Benchmark Models: The authors propose two benchmark methods characterized by their simplicity and accessibility: the Lasso Estimated AutoRegressive (LEAR) model, a parameter-rich autoregressive approach using LASSO for regularization, and a Deep Neural Network (DNN) model. Both models are made available as part of an open-source Python library, providing researchers with accessible, state-of-the-art benchmark methods.
3. Best Practices for EPF: The paper articulates a set of best practices including the use of meaningful evaluation metrics, the importance of statistical testing, and the need for appropriate test dataset lengths. These guidelines are aimed at improving the rigor and reproducibility of future EPF studies.

Evaluation Metrics and Analysis

The paper emphasizes the importance of using appropriate metrics to evaluate forecasting accuracy. It postulates the relative Mean Absolute Error (rMAE) as a superior metric compared to the Mean Absolute Percentage Error (MAPE) due to its reliability and scalability across different datasets and methods. This assertion is supported by extensive empirical tests showing the consistency and advantage of rMAE in representing performance.

The paper also stresses the role of ensemble models in enhancing forecasting accuracy. It presents an empirical analysis demonstrating that ensemble methods combining forecasts from different models significantly outperform their individual constituents.

Implications and Future Directions

The research has practical implications for various stakeholders in electricity markets, including energy traders and utility companies, who can leverage these findings to adopt more accurate forecasting methods. The released benchmark dataset and models encourage adherence to standardized evaluation practices, thereby fostering transparency and comparability across studies.

Theoretically, the paper sets a foundation for future research, prompting questions on the potential of newer machine learning techniques and the impact of incorporating additional exogenous factors. Future studies may also explore the integration of real-time data streams and advanced ensemble learning techniques to further improve forecasting capabilities.

In summary, the paper establishes a robust framework for evaluating EPF models, addressing significant gaps in the literature and offering valuable tools for advancing the state of knowledge in electricity price forecasting.