Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts
Integrating new renewable energy resources requires robust and reliable forecasts to ensure a stable electrical grid and avoid blackouts. Sophisticated representation learning techniques, such as autoencoders, play an essential role, as they allow for the extraction of latent features to forecast th...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Energies |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1996-1073/15/21/8062 |
| _version_ | 1797468414938185728 |
|---|---|
| author | Jens Schreiber Bernhard Sick |
| author_facet | Jens Schreiber Bernhard Sick |
| author_sort | Jens Schreiber |
| collection | DOAJ |
| description | Integrating new renewable energy resources requires robust and reliable forecasts to ensure a stable electrical grid and avoid blackouts. Sophisticated representation learning techniques, such as autoencoders, play an essential role, as they allow for the extraction of latent features to forecast the expected generated wind and photovoltaic power for the next seconds up to days. Thereby, autoencoders reduce the required training time and the time spent in manual feature engineering and often improve the forecast error. However, most current renewable energy forecasting research on autoencoders focuses on smaller forecast horizons for the following seconds and hours based on meteorological measurements. At the same time, larger forecast horizons, such as day-ahead power forecasts based on numerical weather predictions, are crucial for planning loads and demands within the electrical grid to prevent power failures. There is little evidence on the ability of autoencoders and their respective forecasting models to improve through multi-task learning and time series autoencoders for day-ahead power forecasts. We can close these gaps by proposing a multi-task learning autoencoder based on the recently introduced temporal convolution network. This approach reduces the number of trainable parameters by 38 for photovoltaic data and 202 for wind data while having the best reconstruction error compared to nine other representation learning techniques. At the same time, this model decreases the day-ahead forecast error up to <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>18.3</mn></mrow></semantics></math></inline-formula>% for photovoltaic parks and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>1.5</mn></mrow></semantics></math></inline-formula>% for wind parks. We round off these results by analyzing the influences of the latent size and the number of layers to fine-tune the encoder for wind and photovoltaic power forecasts. |
| first_indexed | 2024-03-09T19:06:09Z |
| format | Article |
| id | doaj.art-1ce54d7f31eb4420857891c20e64251e |
| institution | Directory Open Access Journal |
| issn | 1996-1073 |
| language | English |
| last_indexed | 2024-03-09T19:06:09Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Energies |
| spelling | doaj.art-1ce54d7f31eb4420857891c20e64251e2023-11-24T04:31:15ZengMDPI AGEnergies1996-10732022-10-011521806210.3390/en15218062Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power ForecastsJens Schreiber0Bernhard Sick1Intelligent Embedded System, University of Kassel, Wilhelmshöher Allee 71, 34121 Kassel, GermanyIntelligent Embedded System, University of Kassel, Wilhelmshöher Allee 71, 34121 Kassel, GermanyIntegrating new renewable energy resources requires robust and reliable forecasts to ensure a stable electrical grid and avoid blackouts. Sophisticated representation learning techniques, such as autoencoders, play an essential role, as they allow for the extraction of latent features to forecast the expected generated wind and photovoltaic power for the next seconds up to days. Thereby, autoencoders reduce the required training time and the time spent in manual feature engineering and often improve the forecast error. However, most current renewable energy forecasting research on autoencoders focuses on smaller forecast horizons for the following seconds and hours based on meteorological measurements. At the same time, larger forecast horizons, such as day-ahead power forecasts based on numerical weather predictions, are crucial for planning loads and demands within the electrical grid to prevent power failures. There is little evidence on the ability of autoencoders and their respective forecasting models to improve through multi-task learning and time series autoencoders for day-ahead power forecasts. We can close these gaps by proposing a multi-task learning autoencoder based on the recently introduced temporal convolution network. This approach reduces the number of trainable parameters by 38 for photovoltaic data and 202 for wind data while having the best reconstruction error compared to nine other representation learning techniques. At the same time, this model decreases the day-ahead forecast error up to <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>18.3</mn></mrow></semantics></math></inline-formula>% for photovoltaic parks and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>1.5</mn></mrow></semantics></math></inline-formula>% for wind parks. We round off these results by analyzing the influences of the latent size and the number of layers to fine-tune the encoder for wind and photovoltaic power forecasts.https://www.mdpi.com/1996-1073/15/21/8062transfer learningwind powerphotovolatic powerautoencodersdeep learningtime series |
| spellingShingle | Jens Schreiber Bernhard Sick Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts Energies transfer learning wind power photovolatic power autoencoders deep learning time series |
| title | Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts |
| title_full | Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts |
| title_fullStr | Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts |
| title_full_unstemmed | Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts |
| title_short | Multi-Task Autoencoders and Transfer Learning for Day-Ahead Wind and Photovoltaic Power Forecasts |
| title_sort | multi task autoencoders and transfer learning for day ahead wind and photovoltaic power forecasts |
| topic | transfer learning wind power photovolatic power autoencoders deep learning time series |
| url | https://www.mdpi.com/1996-1073/15/21/8062 |
| work_keys_str_mv | AT jensschreiber multitaskautoencodersandtransferlearningfordayaheadwindandphotovoltaicpowerforecasts AT bernhardsick multitaskautoencodersandtransferlearningfordayaheadwindandphotovoltaicpowerforecasts |