Analysis of Day-Ahead Prices and Influencing Factors in the Nordic Power Market
Author
Term
4. term
Education
Publication year
2023
Submitted on
2023-10-06
Abstract
This thesis analyses the performance of machine learning methods including Lasso, neural networks and random forests, on forecasting day-ahead prices in the Nordic power market. To obtain this, two benchmark models is presented for comparison to the machine learning methods, namely the ARIMA and the naïve model. The models compare against each other on 15 time series of day-ahead prices from the Nordic power market ranging from 2015 throughout September 2023. Furthermore, the machine learning models is fed with wind and solar power production forecasts to improve on the effectiveness. The models is empirically compared by using MAE, MSE, MAPE and RMSE as performance metrics. Some of the main results is that neural networks perform very well, when it is tuned to not overfit the data. Furthermore Lasso gives stable results without extensive supervision.
Documents