Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process

Zhouhe Zhang; Haochen Hua; Xingying Chen; Jia Shao; Jin Zheng; Bo Wang; Lei Gan; Kun Yu

doi:10.1109/EEM64765.2025.11050153

Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process

Zhouhe Zhang
, Haochen Hua^*
, Xingying Chen
, Jia Shao
, Jin Zheng
, Bo Wang
, Lei Gan
, Kun Yu

^*Corresponding author for this work

Mathematics

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

33 Downloads (Pure)

Abstract

Spot market electricity price fluctuations can expose market participants to substantial financial risks if not accurately forecasted. Traditional statistical models (e.g., ARIMA) can capture linear trends but struggle with complex nonlinear relationships, while pure neural network models (e.g., Transformer) are insufficiently sensitive to random fluctuations. To address both persistent price trends and transient volatility, this paper combines Transformer with stochastic differential equations driven by Ornstein-Uhlenbeck process. Compared to ARIMA, the proposed model achieves an average MAE reduction of approximately 45 %, along with comparable improvements in RMSE and MAPE and a significant boost in R2. Against a standalone Transformer, it also exhibits substantial performance gains across all key metrics.

Original language	English
Title of host publication	2025 21st International Conference on the European Energy Market (EEM)
Publisher	IEEE
Number of pages	6
ISBN (Electronic)	9798331512781
ISBN (Print)	9798331512798 (PoD)
DOIs	https://doi.org/10.1109/EEM64765.2025.11050153
Publication status	Published - 10 Jul 2025
Event	21st International Conference on the European Energy Market, EEM 2025 - Lisbon, Portugal Duration: 27 May 2025 → 29 May 2025

Publication series

Name	International Conference on the European Energy Market
Publisher	IEEE
ISSN (Print)	2165-4077
ISSN (Electronic)	2165-4093

Conference

Conference	21st International Conference on the European Energy Market, EEM 2025
Country/Territory	Portugal
City	Lisbon
Period	27/05/25 → 29/05/25

Bibliographical note

Publisher Copyright:
© 2025 IEEE.

Keywords

electricity price forecast
spot market
stochastic differential equations
Transformer

ASJC Scopus subject areas

Marketing
Energy Engineering and Power Technology
Fuel Technology

Access to Document

10.1109/EEM64765.2025.11050153

ZhangZ2025SpotAccepted author manuscript, 995 KBLicence: Creative Commons: Attribution (CC BY)

Cite this

Zhang, Z., Hua, H., Chen, X., Shao, J., Zheng, J., Wang, B., Gan, L., & Yu, K. (2025). Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process. In 2025 21st International Conference on the European Energy Market (EEM) Article 11050153 (International Conference on the European Energy Market). IEEE. https://doi.org/10.1109/EEM64765.2025.11050153

@inproceedings{0fbe4c0086564f77847fe50472d5d1f4,

title = "Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process",

abstract = "Spot market electricity price fluctuations can expose market participants to substantial financial risks if not accurately forecasted. Traditional statistical models (e.g., ARIMA) can capture linear trends but struggle with complex nonlinear relationships, while pure neural network models (e.g., Transformer) are insufficiently sensitive to random fluctuations. To address both persistent price trends and transient volatility, this paper combines Transformer with stochastic differential equations driven by Ornstein-Uhlenbeck process. Compared to ARIMA, the proposed model achieves an average MAE reduction of approximately 45 \%, along with comparable improvements in RMSE and MAPE and a significant boost in R2. Against a standalone Transformer, it also exhibits substantial performance gains across all key metrics.",

keywords = "electricity price forecast, spot market, stochastic differential equations, Transformer",

author = "Zhouhe Zhang and Haochen Hua and Xingying Chen and Jia Shao and Jin Zheng and Bo Wang and Lei Gan and Kun Yu",

note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; 21st International Conference on the European Energy Market, EEM 2025 ; Conference date: 27-05-2025 Through 29-05-2025",

year = "2025",

month = jul,

day = "10",

doi = "10.1109/EEM64765.2025.11050153",

language = "English",

isbn = "9798331512798 (PoD)",

series = "International Conference on the European Energy Market",

publisher = "IEEE",

booktitle = "2025 21st International Conference on the European Energy Market (EEM)",

}

Zhang, Z, Hua, H, Chen, X, Shao, J, Zheng, J, Wang, B, Gan, L & Yu, K 2025, Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process. in 2025 21st International Conference on the European Energy Market (EEM)., 11050153, International Conference on the European Energy Market, IEEE, 21st International Conference on the European Energy Market, EEM 2025, Lisbon, Portugal, 27/05/25. https://doi.org/10.1109/EEM64765.2025.11050153

Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process. / Zhang, Zhouhe; Hua, Haochen; Chen, Xingying et al.
2025 21st International Conference on the European Energy Market (EEM). IEEE, 2025. 11050153 (International Conference on the European Energy Market).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process

AU - Zhang, Zhouhe

AU - Hua, Haochen

AU - Chen, Xingying

AU - Shao, Jia

AU - Zheng, Jin

AU - Wang, Bo

AU - Gan, Lei

AU - Yu, Kun

PY - 2025/7/10

Y1 - 2025/7/10

N2 - Spot market electricity price fluctuations can expose market participants to substantial financial risks if not accurately forecasted. Traditional statistical models (e.g., ARIMA) can capture linear trends but struggle with complex nonlinear relationships, while pure neural network models (e.g., Transformer) are insufficiently sensitive to random fluctuations. To address both persistent price trends and transient volatility, this paper combines Transformer with stochastic differential equations driven by Ornstein-Uhlenbeck process. Compared to ARIMA, the proposed model achieves an average MAE reduction of approximately 45 %, along with comparable improvements in RMSE and MAPE and a significant boost in R2. Against a standalone Transformer, it also exhibits substantial performance gains across all key metrics.

AB - Spot market electricity price fluctuations can expose market participants to substantial financial risks if not accurately forecasted. Traditional statistical models (e.g., ARIMA) can capture linear trends but struggle with complex nonlinear relationships, while pure neural network models (e.g., Transformer) are insufficiently sensitive to random fluctuations. To address both persistent price trends and transient volatility, this paper combines Transformer with stochastic differential equations driven by Ornstein-Uhlenbeck process. Compared to ARIMA, the proposed model achieves an average MAE reduction of approximately 45 %, along with comparable improvements in RMSE and MAPE and a significant boost in R2. Against a standalone Transformer, it also exhibits substantial performance gains across all key metrics.

KW - electricity price forecast

KW - spot market

KW - stochastic differential equations

KW - Transformer

UR - https://www.scopus.com/pages/publications/105011074228

U2 - 10.1109/EEM64765.2025.11050153

DO - 10.1109/EEM64765.2025.11050153

M3 - Conference contribution

AN - SCOPUS:105011074228

SN - 9798331512798 (PoD)

T3 - International Conference on the European Energy Market

BT - 2025 21st International Conference on the European Energy Market (EEM)

PB - IEEE

T2 - 21st International Conference on the European Energy Market, EEM 2025

Y2 - 27 May 2025 through 29 May 2025

ER -

Spot Market Electricity Price Forecast via the Combination of Transformer and Ornstein-Uhlenbeck Process

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this