SOLETE Platform v3.0

SOLETE

Author: Daniel Vázquez Pombo - Contact: [email protected]

LinkedIn: https://www.linkedin.com/in/dvp/

ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository used to be complementary material to its twin "Data in Brief" article [1], and a series of papers covering Solar PV power forecasting [2, 3, 4]. The objective is to increase the transparency of my work, which is one of the main limitations of Machine Learning in general.
However, as it sometimes happens, the project has grown life by itself and has now become a platform to experiment on time-series forecasting based on Machine Learning.
I included a number of functions that can be used by beginners to kickstart their projects with solar power, machine learning, forecasting, or simply python.

Long Live Open Science!

The papers were developed under the PhD thesis Operation and Planning of Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).
Version v1.0 was released during the PhD thus, Copyright 2021 Technical University of Denmark.
Version v2.0 and subsequents were released after finalizing my employment at DTU, therefore, Copyright belongs to me (yeah baby!).

What is new?

Major bugs corrected from the previous versions. These were related to the way data was preprocess before being fed to the machine learning models, but also to the way RMSE was computed in the postprocessing. Furthermore, the build-up of the results matrix and the postprocessing itself has been completely reworked, resulting in better solving speed. The gain is specially evident for the highest sampling rates.

Dependencies

1. Python 3.9.12
2. Pandas 1.5.0
3. Numpy 1.23.1
4. Matplotlib 3.6.0
5. Scikit-Learn 1.1.2
6. Keras 2.10.0
7. TensorFlow 2.10.0
8. CoolProp 6.4.3
9. The SOLETE dataset [1] -> https://doi.org/10.11583/DTU.17040767

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and sneak a peek at its contents.
3. Open the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • The main objective is to introduce the SOLETE dataset and help people learning basics of time series forecasting based on Machine Learning
   • You can basically replicate most of the methodology from [2, 3, 4] and build on top.
   • I included some error messages to debug what I expect are the most common errors when running stuff.
   • Let me know if you like it or what needs to be fixed.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

Note that the latest version of the SOLETE dataset includes a 1sec resolution version. The file is quite large, which might meant that you PC is not able to open it. Please consider only reading part of it if you really want to play with that resolution. Alternatively, drop it in an HPC and enjoy yourself. :D

Notes for MATLAB users

I have been reached out by several people complaining that hdf5 can't be imported in MATLAB. That is not true, they weren't doing properly. Nevertheless, worry not dear user. Your peers have asked and I answer:

1. Open the file RunMe_matlab.m in MATLAB and hit F5. That will import SOLETE as a table.
2. Alternatively, you can run the Python scripts from MATLAB, which I find a bit weird... but hey! You do you baby!

*I coded this using 2021b, so anything newer should work, but I haven't actually checked with older versions.

How to cite this:

Technically, you should cite the repository itself, however I don't get those citations captured where it matters, so please cite [1] like this:

@Article{pombo2022solete,
title={SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions},
author={Pombo, Daniel Vazquez and Gehrke, Oliver and Bindner, Henrik W},
journal={Data in Brief},
volume={42},
pages={108046},
year={2022},
publisher={Elsevier}
}

Nontheless, here is the citation for the git itself:
D. V. Pombo, The SOLETE platform (March, 2023).doi:10.11583/DTU.17040626.URL https://data.dtu.dk/articles/software/TheSOLETEplatform/17040626

@Article{SOLETE2021Code,
author = "Daniel Vazquez Pombo",
title = "{The SOLETE platform}",
year = "2023",
month = "Mar,",
url = "https://data.dtu.dk/articles/software/The_SOLETE_platform/17040626",
doi = "10.11583/DTU.17040626"
}

References

[1] Pombo, D. V., Gehrke, O., & Bindner, H. W. (2022). SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions. Data in Brief, 42, 108046.
    
[2] Pombo, D. V., Bindner, H. W., Spataru, S. V., Sørensen, P. E., & Bacher, P. (2022). Increasing the accuracy of hourly multi-output solar power forecast with physics-informed machine learning. Sensors, 22(3), 749.

[3] Pombo, D. V., Bacher, P., Ziras, C., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Benchmarking physics-informed machine learning-based short term PV-power forecasting tools. Energy Reports, 8, 6512-6520.

[4] Pombo, D. V., Rincón, M. J., Bacher, P., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Assessing stacked physics-informed machine learning models for co-located wind–solar power forecasting. Sustainable Energy, Grids and Networks, 32, 100943.

Corrigendum

Versions of the SOLETE Platform up to and including v2.3 contained two major bugs.

1. When spliting the data in training, validation, and testing sets. This was directly affecting accuracy.
2. In the postprocessing of results, when calculating RMSE. This was affecting evaluation quality.

I can only apologize for these mistakes, which have been corrected in versions v3.0 and upwards.

SOLETE Platform v2.3

SOLETE Platform v2.3

Author: Daniel Vázquez Pombo - Contact: [email protected]
LinkedIn: https://www.linkedin.com/in/dvp/
ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository used to be complementary material to its twin "Data in Brief" article [1], and a series of papers covering Solar PV power forecasting [2, 3, 4]. The objective is to increase the transparency of my work, which is one of the main limitations of Machine Learning in general.
However, as it sometimes happens, the project has grown life by itself and has now become a platform to experiment on time-series forecasting based on Machine Learning.
I included a number of functions that can be used by beginners to kickstart their projects with solar power, machine learning, forecasting, or simply python.

Long Live Open Science!

The papers were developed under the PhD thesis Operation and Planning of Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).
Version v1.0 was released during the PhD thus, Copyright 2021 Technical University of Denmark.
Version v2.0 was released months after finalising my employment at DTU, therefore, Copyright belongs to me (yeah baby!).

What is new?

Minor bug fixed. It prevented importing the 1sec and 1 min resolutions. This was a protection inherited from the first version of the dataset.

Dependencies

1. Python 3.9.12
2. Pandas 1.5.0
3. Numpy 1.23.1
4. Matplotlib 3.6.0
5. Scikit-Learn 1.1.2
6. Keras 2.10.0
7. TensorFlow 2.10.0
8. CoolProp 6.4.3
9. The SOLETE dataset [1] -> https://doi.org/10.11583/DTU.17040767

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and sneak a peek at its contents.
3. Open the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • The main objective is to introduce the SOLETE dataset and help people learning basics of time series forecasting based on Machine Learning
   • You can basically replicate most of the methodology from [2, 3, 4] and build on top.
   • I included some error messages to debug what I expect are the most common errors when running stuff.
   • Let me know if you like it or what needs to be fixed.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

Note that the latest version of the SOLETE dataset includes a 1sec resolution version. The file is quite large, which might meant that you PC is not able to open it. Please consider only reading part of it if you really want to play with that resolution. Alternatively, drop it in an HPC and enjoy yourself. :D

Notes for MATLAB users

I have been reached out by several people complaining that hdf5 can't be imported in MATLAB. That is not true, they weren't doing properly. Nevertheless, worry not dear user. Your peers have asked and I answer:

1. Open the file RunMe_matlab.m in MATLAB and hit F5. That will import SOLETE as a table.
2. Alternatively, you can run the Python scripts from MATLAB, which I find a bit weird... but hey! You do you baby!

*I coded this using 2021b, so anything newer should work, but I haven't actually checked with older versions.

How to cite this:

Technically, you should cite the repository itself, however I don't get those citations captured where it matters, so please cite [1] like this:

@Article{pombo2022solete,
title={SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions},
author={Pombo, Daniel Vazquez and Gehrke, Oliver and Bindner, Henrik W},
journal={Data in Brief},
volume={42},
pages={108046},
year={2022},
publisher={Elsevier}
}

Nontheless, here is the citation for the git itself:
D. V. Pombo, The SOLETE platform (March, 2023).doi:10.11583/DTU.17040626.URL https://data.dtu.dk/articles/software/TheSOLETEplatform/17040626

@Article{SOLETE2021Code,
author = "Daniel Vazquez Pombo",
title = "{The SOLETE platform}",
year = "2023",
month = "Mar,",
url = "https://data.dtu.dk/articles/software/The_SOLETE_platform/17040626",
doi = "10.11583/DTU.17040626"
}

References

[1] Pombo, D. V., Gehrke, O., & Bindner, H. W. (2022). SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions. Data in Brief, 42, 108046.
    
[2] Pombo, D. V., Bindner, H. W., Spataru, S. V., Sørensen, P. E., & Bacher, P. (2022). Increasing the accuracy of hourly multi-output solar power forecast with physics-informed machine learning. Sensors, 22(3), 749.

[3] Pombo, D. V., Bacher, P., Ziras, C., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Benchmarking physics-informed machine learning-based short term PV-power forecasting tools. Energy Reports, 8, 6512-6520.

[4] Pombo, D. V., Rincón, M. J., Bacher, P., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Assessing stacked physics-informed machine learning models for co-located wind–solar power forecasting. Sustainable Energy, Grids and Networks, 32, 100943.

SOLETE Platform v2.2

SOLETE Platform v2.2

Author: Daniel Vázquez Pombo - Contact: [email protected]
LinkedIn: https://www.linkedin.com/in/dvp/
ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository used to be complementary material to its twin "Data in Brief" article [1], and a series of papers covering Solar PV power forecasting [2, 3, 4]. The objective is to increase the transparency of my work, which is one of the main limitations of Machine Learning in general.
However, as it sometimes happens, the project has grown life by itself and has now become a platform to experiment on time-series forecasting based on Machine Learning.
I included a number of functions that can be used by beginners to kickstart their projects with solar power, machine learning, forecasting, or simply python.

Long Live Open Science!

The papers were developed under the PhD thesis Operation and Planning of Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).
Version v1.0 was released during the PhD thus, Copyright 2021 Technical University of Denmark.
Version v2.0 was released months after finalising my employment at DTU, therefore, Copyright belongs to me (yeah baby!).

What is new?

A very small thing honestly. There appeared to be a bug affecting some Matlab versions in which SOLETE was imported with its rows in random order. This was reported and corrected by Mr Jon Martinez-Rico. A big thank you to him.

Nonetheless, we still recommend using Python.

Dependencies

1. Python 3.9.12
2. Pandas 1.5.0
3. Numpy 1.23.1
4. Matplotlib 3.6.0
5. Scikit-Learn 1.1.2
6. Keras 2.10.0
7. TensorFlow 2.10.0
8. CoolProp 6.4.3
9. The SOLETE dataset [1] -> https://doi.org/10.11583/DTU.17040767

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and sneak a peek at its contents.
3. Open the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • The main objective is to introduce the SOLETE dataset and help people learning basics of time series forecasting based on Machine Learning
   • You can basically replicate most of the methodology from [2, 3, 4] and build on top.
   • I included some error messages to debug what I expect are the most common errors when running stuff.
   • Let me know if you like it or what needs to be fixed.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

Note that the latest version of the SOLETE dataset includes a 1sec resolution version. The file is quite large, which might meant that you PC is not able to open it. Please consider only reading part of it if you really want to play with that resolution. Alternatively, drop it in an HPC and enjoy yourself. :D

Notes for MATLAB users

I have been reached out by several people complaining that hdf5 can't be imported in MATLAB. That is not true, they weren't doing properly. Nevertheless, worry not dear user. Your peers have asked and I answer:

1. Open the file RunMe_matlab.m in MATLAB and hit F5. That will import SOLETE as a table.
2. Alternatively, you can run the Python scripts from MATLAB, which I find a bit weird... but hey! You do you baby!

*I coded this using 2021b, so anything newer should work, but I haven't actually checked with older versions.

How to cite this:

Technically, you should cite the repository itself, however I don't get those citations captured where it matters, so please cite [1] like this:

@Article{pombo2022solete,
title={SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions},
author={Pombo, Daniel Vazquez and Gehrke, Oliver and Bindner, Henrik W},
journal={Data in Brief},
volume={42},
pages={108046},
year={2022},
publisher={Elsevier}
}

Nontheless, here is the citation for the git itself:
D. V. Pombo, The SOLETE platform (March, 2023).doi:10.11583/DTU.17040626.URL https://data.dtu.dk/articles/software/TheSOLETEplatform/17040626

@Article{SOLETE2021Code,
author = "Daniel Vazquez Pombo",
title = "{The SOLETE platform}",
year = "2023",
month = "Mar,",
url = "https://data.dtu.dk/articles/software/The_SOLETE_platform/17040626",
doi = "10.11583/DTU.17040626"
}

References

[1] Pombo, D. V., Gehrke, O., & Bindner, H. W. (2022). SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions. Data in Brief, 42, 108046.
    
[2] Pombo, D. V., Bindner, H. W., Spataru, S. V., Sørensen, P. E., & Bacher, P. (2022). Increasing the accuracy of hourly multi-output solar power forecast with physics-informed machine learning. Sensors, 22(3), 749.

[3] Pombo, D. V., Bacher, P., Ziras, C., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Benchmarking physics-informed machine learning-based short term PV-power forecasting tools. Energy Reports, 8, 6512-6520.

[4] Pombo, D. V., Rincón, M. J., Bacher, P., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Assessing stacked physics-informed machine learning models for co-located wind–solar power forecasting. Sustainable Energy, Grids and Networks, 32, 100943.

SOLETE Platform v2.1

SOLETE Platform v2.1

Author: Daniel Vázquez Pombo - Contact: [email protected]
LinkedIn: https://www.linkedin.com/in/dvp/
ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository used to be complementary material to its twin "Data in Brief" article [1], and a series of papers covering Solar PV power forecasting [2, 3, 4]. The objective is to increase the transparency of my work, which is one of the main limitations of Machine Learning in general.
However, as it sometimes happens, the project has grown life by itself and has now become a platform to experiment on time-series forecasting based on Machine Learning.
I included a number of functions that can be used by beginners to kickstart their projects with solar power, machine learning, forecasting, or simply python.

Long Live Open Science!

The papers were developed under the PhD thesis Operation and Planning of Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).
Version v1.0 was released during the PhD thus, Copyright 2021 Technical University of Denmark.
Version v2.0 was released months after finalising my employment at DTU, therefore, Copyright belongs to me (yeah baby!).

Dependencies

1. Python 3.9.12
2. Pandas 1.5.0
3. Numpy 1.23.1
4. Matplotlib 3.6.0
5. Scikit-Learn 1.1.2
6. Keras 2.10.0
7. TensorFlow 2.10.0
8. CoolProp 6.4.3
9. The SOLETE dataset [1] -> https://doi.org/10.11583/DTU.17040767

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and sneak a peek at its contents.
3. Open the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • The main objective is to introduce the SOLETE dataset and help people learning basics of time series forecasting based on Machine Learning
   • You can basically replicate most of the methodology from [2, 3, 4] and build on top.
   • I included some error messages to debug what I expect are the most common errors when running stuff.
   • Let me know if you like it or what needs to be fixed.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

Note that the latest version of the SOLETE dataset includes a 1sec resolution version. The file is quite large, which might meant that you PC is not able to open it. Please consider only reading part of it if you really want to play with that resolution. Alternatively, drop it in an HPC and enjoy yourself. :D

Notes for MATLAB users

I have been reached out by several people complaining that hdf5 can't be imported in MATLAB. That is not true, they weren't doing properly. Nevertheless, worry not dear user. Your peers have asked and I answer:

1. Open the file RunMe_matlab.m in MATLAB and hit F5. That will import SOLETE as a table.
2. Alternatively, you can run the Python scripts from MATLAB, which I find a bit weird... but hey! You do you baby!

*I coded this using 2021b, so anything newer should work, but I haven't actually checked with older versions.

How to cite this:

Technically, you should cite the repository itself, however I don't get those citations captured where it matters, so please cite [1] like this:

@Article{pombo2022solete,
title={SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions},
author={Pombo, Daniel Vazquez and Gehrke, Oliver and Bindner, Henrik W},
journal={Data in Brief},
volume={42},
pages={108046},
year={2022},
publisher={Elsevier}
}

Nontheless, here is the citation for the git itself:
D. V. Pombo, The SOLETE platform (March, 2023).doi:10.11583/DTU.17040626.URL https://data.dtu.dk/articles/software/TheSOLETEplatform/17040626

@Article{SOLETE2021Code,
author = "Daniel Vazquez Pombo",
title = "{The SOLETE platform}",
year = "2023",
month = "Mar,",
url = "https://data.dtu.dk/articles/software/The_SOLETE_platform/17040626",
doi = "10.11583/DTU.17040626"
}

References

[1] Pombo, D. V., Gehrke, O., & Bindner, H. W. (2022). SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions. Data in Brief, 42, 108046.
    
[2] Pombo, D. V., Bindner, H. W., Spataru, S. V., Sørensen, P. E., & Bacher, P. (2022). Increasing the accuracy of hourly multi-output solar power forecast with physics-informed machine learning. Sensors, 22(3), 749.

[3] Pombo, D. V., Bacher, P., Ziras, C., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Benchmarking physics-informed machine learning-based short term PV-power forecasting tools. Energy Reports, 8, 6512-6520.

[4] Pombo, D. V., Rincón, M. J., Bacher, P., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Assessing stacked physics-informed machine learning models for co-located wind–solar power forecasting. Sustainable Energy, Grids and Networks, 32, 100943.

SOLETE Platform v2.0

SOLETE Platform v2.0

Author: Daniel Vázquez Pombo - Contact: [email protected]

LinkedIn: https://www.linkedin.com/in/dvp/

ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository used to be complementary material to a series of papers covering Solar PV power forecasting [1, 2] and its twin "Data in Brief" article [3]. The objective is to increase the transparency of [1, 2, 3, 4], which is one of the main limitations of Machine Learning in general.
However, as it sometimes happens, the project has grown life by itself and has now become a platform to experiment on time-series forecasting based on Machine Learning.
I included a number of functions that can be used by beginners to kickstart their projects with solar power, machine learning, forecasting, or simply python.

Long Live Open Science!

The papers were developed under the PhD thesis Operation and Planning of Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).
Version v1.0 was released during the PhD thus, Copyright 2021 Technical University of Denmark.
Version v2.0 was released months after finalising my employment at DTU, therefore, Copyright belongs to me (yeah baby!).

Dependencies

1. Python 3.9.12
2. Pandas 1.5.0
3. Numpy 1.23.1
4. Matplotlib 3.6.0
5. Scikit-Learn 1.1.2
6. Keras 2.10.0
7. TensorFlow 2.10.0
8. CoolProp 6.4.3
9. The SOLETE dataset [1] -> https://doi.org/10.11583/DTU.17040767

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and sneak a peek at its contents.
3. Open the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • The main objective is to introduce the SOLETE dataset and help people learning basics of time series forecasting based on Machine Learning
   • You can basically replicate most of the methodology from [2, 3, 4] and build on top.
   • I included some error messages to debug what I expect are the most common errors when running stuff.
   • Let me know if you like it or what needs to be fixed.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

Note that the latest version of the SOLETE dataset includes a 1sec resolution version. The file is quite large, which might meant that you PC is not able to open it. Please consider only reading part of it if you really want to play with that resolution. Alternatively, drop it in an HPC and enjoy yourself. :D

How to cite this:

Technically, you should cite the repository itself, however I don't get those citations captured where it matters, so please cite [1] like this:

@Article{pombo2022solete,
title={SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions},
author={Pombo, Daniel Vazquez and Gehrke, Oliver and Bindner, Henrik W},
journal={Data in Brief},
volume={42},
pages={108046},
year={2022},
publisher={Elsevier}
}

Nontheless, here is the citation for the git itself:
D. V. Pombo, The SOLETE platform (March, 2023).doi:10.11583/DTU.17040626.URL https://data.dtu.dk/articles/software/TheSOLETEplatform/17040626

@Article{SOLETE2021Code,
author = "Daniel Vazquez Pombo",
title = "{The SOLETE platform}",
year = "2023",
month = "Mar,",
url = "https://data.dtu.dk/articles/software/The_SOLETE_platform/17040626",
doi = "10.11583/DTU.17040626"
}

References

[1] Pombo, D. V., Gehrke, O., & Bindner, H. W. (2022). SOLETE, a 15-month long holistic dataset including: Meteorology, co-located wind and solar PV power from Denmark with various resolutions. Data in Brief, 42, 108046.
    
[2] Pombo, D. V., Bindner, H. W., Spataru, S. V., Sørensen, P. E., & Bacher, P. (2022). Increasing the accuracy of hourly multi-output solar power forecast with physics-informed machine learning. Sensors, 22(3), 749.

[3] Pombo, D. V., Bacher, P., Ziras, C., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Benchmarking physics-informed machine learning-based short term PV-power forecasting tools. Energy Reports, 8, 6512-6520.

[4] Pombo, D. V., Rincón, M. J., Bacher, P., Bindner, H. W., Spataru, S. V., & Sørensen, P. E. (2022). Assessing stacked physics-informed machine learning models for co-located wind–solar power forecasting. Sustainable Energy, Grids and Networks, 32, 100943.

SOLETE Platform v1.0

SOLETE Platform v1.0

Author: Daniel Vázquez Pombo - Contact: [email protected]

LinkedIn: https://www.linkedin.com/in/dvp/

ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository is complementary material to a series of papers covering Solar PV power forecasting [1, 2] and its twin "Data in Brief" article [3]. The objective is to increase the transparency of [1, 2], which is one of the main limitations of Machine Learning in general.
In addition, I included a number of functions that can be use by begginers to kickstart their projects in time series forecasting with different machine learning methods.

This work was developed under the PhD thesis Energy Management Systems for Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).

Dependencies

1. Python 3.8.10
2. Pandas 1.2.4
3. Numpy 1.19.5
4. Matplotlib 3.4.2
5. Scikit-Learn 0.24.2
6. Keras 2.5.0
7. TensorFlow 2.5.0
8. The SOLETE dataset -> https://data.dtu.dk/articles/dataset/The_SOLETE_dataset/17040767
9. Functions.py

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and take a peak at its contents.
3. Opend the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • You can basically replicate the studies from [1, 2] and build on top.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

How to cite this:

D. V. Pombo, The SOLETE platform (Jan, 2022).doi:10.11583/DTU.17040626.URL https://data.dtu.dk/articles/software/TheSOLETEplatform/17040626

@Article{SOLETE2021Code,
author = "Daniel Vazquez Pombo",
title = "{The SOLETE platform}",
year = "2022",
month = "Jan,",
url = "https://data.dtu.dk/articles/software/The_SOLETE_platform/17040626",
doi = "10.11583/DTU.17040626"
}

References

[1] D. V. Pombo, H. W. Bindner, S. V. Spataru, P. E. Sørensen, P. Bacher, Increasing the Accuracy of Hourly Multi-Output Solar Power Forecast with Physics-Informed Machine Learning, Sensors 22 (3) (2022) 749.

[2] D. V. Pombo, P. Bacher, C. Ziras, H. W. Bindner, S. V. Spataru, P. E. Sørensen, Benchmarking Physics-Informed Machine Learning-based Short Term PV-Power Forecasting Tools, Under Review.

[3] D.V. Pombo, O.G. Gehrke, H.W. Bindner, SOLETE, a 15-month long holistic dataset including: meteorology, co-located wind and solar PV power from Denmark with various resolutions, Data in Brief. In Press.

Copyright 2021 Technical University of Denmark.

A Trailer of SOLETE

The SOLETE platform and its twin dataset enable physics informed Solar PV power forecasting. Since the related publications are still under-review and due to the fear of scooping, this release is nothing but a trailer. Users have a tiny piece of the dataset available and a tiny piece of the code. The whole platform will be made public upon publication.
Cheers!

Here you have a copy paste of the ReadME

SOLETE

Author: Daniel Vázquez Pombo - Contact: [email protected]

LinkedIn: https://www.linkedin.com/in/dvp/

ResearchGate: https://www.researchgate.net/profile/Daniel-Vazquez-Pombo
ORCID: https://orcid.org/0000-0001-5664-9421

This repository is complementary material to a paper covering Solar PV power forecasting [1] and its twin "Data in Brief" article [2]. The objective is to increase the transparency of [1], which is one of the main limitations of Machine Learning in general.

This work was developed under the PhD thesis Energy Management Systems for Isolated Hybrid Power Systems at the Technical University of Denmark (DTU).

Note for reviewers

Whether you are reviewing either one of those papers you should be aware of the following:

1. What you can find here is a "trailer" of what will be provided upon acceptance.
2. The dataset only contains a minimum number of samples
3. There is a file missing called Functions.py which contains the most of the code.
4. Files RunMe.py and MlForecasting.py depend on it to be run, but you can see at the moment what they cover.

Dependencies

1. Python 3.8.10
2. Pandas 1.2.4
3. Numpy 1.19.5
4. Matplotlib 3.4.2
5. Scikit-Learn 0.24.2
6. Keras 2.5.0
7. TensorFlow 2.5.0
8. The SOLETE dataset -> https://figshare.com/s/2c9d8167531e6d611ff3
9. Functions.py -> A script that will be added to this git uppon publication of the related papers.

How to use

1. Store the SOLETE dataset in the same folder as the scripts from this repository
2. Open the RunMe.py file. This allows you to load SOLETE and take a peak at its contents.
3. Opend the MLForecasting.py file. This allows you to configure Random Forest (RF), Support Vector Machine (SVM), and three kinds of Artificial Neuronal Networks: Convolutional Neuronal Network (CNN), Long-Short Term Memory (LSTM), and a Hybrid (CNN-LSTM).
   • The file itself contains notes explaining how to use it.
   • You can basically replicate the study from [1] and build on top.
4. Have Fun!

You can of course use your own dataset, you will have to adapt things here and there, but you will be able to reuse most of the code.

How to cite this:

• under construction -

References

[1] D.V. Pombo, H.W. Bindner, S.V. Spataru, P. Sørensen, P. Bacher, Increasing the Accuracy of Hourly Multi-Output Solar Power Forecast with Physics-Informed Machine Learning, Solar Energy. In Press.

[2] D.V. Pombo, O.G. Gehrke, H.W. Bindner, Solete, a 15-month long holistic dataset including: meteorology, co-located wind and solar PV power from Denmark with hourly resolution, Data in Brief. In Press.