A Comparative Evaluation of Deep Neural Networks for Electricity Price Forecasting

Md Readion Islam Razon, Md Tanjim, Sayed Mahmudul Haque, Md Palash Uddin and Mahmudul Hasan ()
Additional contact information
Md Readion Islam Razon: Hajee Mohammad Danesh Science and Technology University
Md Tanjim: Hajee Mohammad Danesh Science and Technology University
Sayed Mahmudul Haque: Hajee Mohammad Danesh Science and Technology University
Md Palash Uddin: Deakin University
Mahmudul Hasan: Deakin University

A chapter in Machine Learning Technologies on Energy Economics and Finance, 2025, pp 27-50 from Springer

Abstract: Abstract Electricity price forecasting is crucial for effective energy management, particularly in markets characterized by high volatility. This research focuses on a comparative evaluation of Deep Learning (DL) algorithms to identify the most suitable models for predicting electricity prices. After handling missing values and normalization, we employ Principle Component Analysis to reduce the dimensionality of the dataset to make it more trainable. It leverages empirical data from Spain, encompassing electrical consumption, generation, pricing, and weather data. Six existing DL models are employed, alongside the development of two hybrid DL models stacked LSTM and CNN-LSTM aimed at improving forecasting accuracy. The results demonstrate that the Encoder-Decoder model exhibits superiority over other models, achieving an MAE of 1.6581, MSE of 4.7675, RMSE of 2.1834, and R 2 $$R^2$$ of 93%. CNN-LSTM also demonstrates strong predictive capabilities, achieving similar R 2 $$R^2$$ of 93% but with slightly higher error metrics. In contrast, traditional RNN and GRU models perform less effectively, with R 2 $$R^2$$ values of 89% and 90%, respectively, and consistently higher error metrics. The study concludes that complex architectures like Encoder-Decoder and hybrid models such as CNN-LSTM offer superior forecasting accuracy compared to other models. This comprehensive analysis underscores the significant potential of DL models in electricity markets. Future research could explore ensemble approaches that combine multiple model architectures and incorporate additional exogenous variables to further enhance predictive performance. Furthermore, real-time implementation and scalability of these models remain pivotal for supporting dynamic electricity markets effectively.

Keywords: Electricity price forecasting; Deep learning; Energy price; Models’ comparative evaluation (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:isochp:978-3-031-95099-5_2

Ordering information: This item can be ordered from
http://www.springer.com/9783031950995

DOI: 10.1007/978-3-031-95099-5_2

Access Statistics for this chapter

More chapters in International Series in Operations Research & Management Science from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().