Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Forecasting Electricity Prices in the German Energy Market: The Influence of Renewable Energy and Market Events

Assar, Ornella Yasmine LU and Egilsdóttir, Berglind LU (2024) DABN01 20241
Department of Economics
Department of Statistics
Abstract
This research paper aims to identify the most effective algorithms for predicting electricity prices in the German day-ahead market, with a specific emphasis on how renewable energy sources, particularly wind and solar, influence hourly price pro- files. These profiles, or characteristic patterns, depict how prices fluctuate within a day. Different machine learning regression models and methods are implemented to predict future hourly profiles in terms of renewable generation information us- ing data from 2018 to 2023. These models contribute to the shaping vector of the Hourly Price Forward Curve (HPFC) in the German electricity market, which relies heavily on wind and solar energy. This shaping vector accounts for all periodicities such... (More)
This research paper aims to identify the most effective algorithms for predicting electricity prices in the German day-ahead market, with a specific emphasis on how renewable energy sources, particularly wind and solar, influence hourly price pro- files. These profiles, or characteristic patterns, depict how prices fluctuate within a day. Different machine learning regression models and methods are implemented to predict future hourly profiles in terms of renewable generation information us- ing data from 2018 to 2023. These models contribute to the shaping vector of the Hourly Price Forward Curve (HPFC) in the German electricity market, which relies heavily on wind and solar energy. This shaping vector accounts for all periodicities such as the hourly profiles, capturing variations across different times of the day and various types of days to accurately reflect demand and supply dynamics in the electricity market. The analysis incorporates periods of high volatility, specifically the COVID-19 lockdowns and the Russia-Ukraine conflict, as well as considering public holiday data in Germany.
Three types of algorithms are employed in the analysis: Linear Regression, Gradi- ent Boosting (an XGBoost implementation) and a Neural Network. Among these, the XGBoost model demonstrated the best performance, achieving an R2 score of 0.92. The analysis revealed that price lags and onshore wind significantly enhanced predictive accuracy, whereas specific periods such as the COVID-19 pandemic and the Russia-Ukraine conflict, as well as most weekday variables, proved less effective for predictions. These findings can improve hourly electricity price forecasting and set a benchmark for future research employing different methodologies. (Less)
Please use this url to cite or link to this publication:
author
Assar, Ornella Yasmine LU and Egilsdóttir, Berglind LU
supervisor
organization
course
DABN01 20241
year
type
H1 - Master's Degree (One Year)
subject
keywords
Electricity Price Forecasting, German Energy Market, Renewable Energy Impact, Machine Learning Models, Market Volatility
language
English
id
9160489
date added to LUP
2024-09-24 08:32:08
date last changed
2024-09-24 08:32:08
@misc{9160489,
  abstract     = {{This research paper aims to identify the most effective algorithms for predicting electricity prices in the German day-ahead market, with a specific emphasis on how renewable energy sources, particularly wind and solar, influence hourly price pro- files. These profiles, or characteristic patterns, depict how prices fluctuate within a day. Different machine learning regression models and methods are implemented to predict future hourly profiles in terms of renewable generation information us- ing data from 2018 to 2023. These models contribute to the shaping vector of the Hourly Price Forward Curve (HPFC) in the German electricity market, which relies heavily on wind and solar energy. This shaping vector accounts for all periodicities such as the hourly profiles, capturing variations across different times of the day and various types of days to accurately reflect demand and supply dynamics in the electricity market. The analysis incorporates periods of high volatility, specifically the COVID-19 lockdowns and the Russia-Ukraine conflict, as well as considering public holiday data in Germany.
Three types of algorithms are employed in the analysis: Linear Regression, Gradi- ent Boosting (an XGBoost implementation) and a Neural Network. Among these, the XGBoost model demonstrated the best performance, achieving an R2 score of 0.92. The analysis revealed that price lags and onshore wind significantly enhanced predictive accuracy, whereas specific periods such as the COVID-19 pandemic and the Russia-Ukraine conflict, as well as most weekday variables, proved less effective for predictions. These findings can improve hourly electricity price forecasting and set a benchmark for future research employing different methodologies.}},
  author       = {{Assar, Ornella Yasmine and Egilsdóttir, Berglind}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Forecasting Electricity Prices in the German Energy Market: The Influence of Renewable Energy and Market Events}},
  year         = {{2024}},
}