Wind Offshore Power Forecasting - A comparative analysis of forecasting accuracy between Machine Learning and classical Time Series models
(2024) DABN01 20241Department of Economics
Department of Statistics
- Abstract
- This thesis investigated the prediction accuracy of offshore wind power using machine learning and classical time series methods. We used Support Vector Regression (SVR) with a Radial Basis Function (RBF) kernel along the Autoregressive (AR) model. Time series cross-validation with data from the Lillgrund Wind Farm and hyperparameter tuning were used to enhance the performance of both models. Throughout the forecasting horizons that were examined, the SVR model consistently outperformed the AR model as indicated by lower Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) values. This was because offshore wind power data has inherent non-linearities. These results demonstrate how... (More)
- This thesis investigated the prediction accuracy of offshore wind power using machine learning and classical time series methods. We used Support Vector Regression (SVR) with a Radial Basis Function (RBF) kernel along the Autoregressive (AR) model. Time series cross-validation with data from the Lillgrund Wind Farm and hyperparameter tuning were used to enhance the performance of both models. Throughout the forecasting horizons that were examined, the SVR model consistently outperformed the AR model as indicated by lower Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) values. This was because offshore wind power data has inherent non-linearities. These results demonstrate how well-sophisticated machine learning approaches can identify complex patterns in offshore wind energy data, which might have a big impact on improving the reliability of wind power integration into energy grids. Future research should focus on adding more predictor variables, improving model robustness through various data sources, and investigating other machine learning techniques to further improve prediction accuracy. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9166111
- author
- Jankunas, Adomas LU and Poultourtzidou, Zografa LU
- supervisor
- organization
- course
- DABN01 20241
- year
- 2024
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Offshore Wind Power Forecasting, Machine Learning, Support Vector Regression, Autoregressive Model, Time Series Analysis
- language
- English
- id
- 9166111
- date added to LUP
- 2024-09-24 08:34:06
- date last changed
- 2024-09-24 08:34:06
@misc{9166111, abstract = {{This thesis investigated the prediction accuracy of offshore wind power using machine learning and classical time series methods. We used Support Vector Regression (SVR) with a Radial Basis Function (RBF) kernel along the Autoregressive (AR) model. Time series cross-validation with data from the Lillgrund Wind Farm and hyperparameter tuning were used to enhance the performance of both models. Throughout the forecasting horizons that were examined, the SVR model consistently outperformed the AR model as indicated by lower Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) values. This was because offshore wind power data has inherent non-linearities. These results demonstrate how well-sophisticated machine learning approaches can identify complex patterns in offshore wind energy data, which might have a big impact on improving the reliability of wind power integration into energy grids. Future research should focus on adding more predictor variables, improving model robustness through various data sources, and investigating other machine learning techniques to further improve prediction accuracy.}}, author = {{Jankunas, Adomas and Poultourtzidou, Zografa}}, language = {{eng}}, note = {{Student Paper}}, title = {{Wind Offshore Power Forecasting - A comparative analysis of forecasting accuracy between Machine Learning and classical Time Series models}}, year = {{2024}}, }