LGPMI - Data-Driven Solar Panels Power Prediction: Model Comparison

Conference paper

Younes Benazzouz, Djilalia Guendouz
The 1st International Conference on Applications and Technologies of Renewable Energy Systems (ICATRES2024), 2024

Read

Cite

APA Click to copy
Benazzouz, Y., & Guendouz, D. (2024). Data-Driven Solar Panels Power Prediction: Model Comparison Including Hyperparameter Tuning Process. In The 1st International Conference on Applications and Technologies of Renewable Energy Systems (ICATRES2024).

Chicago/Turabian Click to copy
Benazzouz, Younes, and Djilalia Guendouz. “Data-Driven Solar Panels Power Prediction: Model Comparison Including Hyperparameter Tuning Process.” In The 1st International Conference on Applications and Technologies of Renewable Energy Systems (ICATRES2024), 2024.

MLA Click to copy
Benazzouz, Younes, and Djilalia Guendouz. “Data-Driven Solar Panels Power Prediction: Model Comparison Including Hyperparameter Tuning Process.” The 1st International Conference on Applications and Technologies of Renewable Energy Systems (ICATRES2024), 2024.

BibTeX Click to copy

@inproceedings{benazzouz2024a,
  title = {Data-Driven Solar Panels Power Prediction: Model Comparison Including Hyperparameter Tuning Process},
  year = {2024},
  author = {Benazzouz, Younes and Guendouz, Djilalia},
  booktitle = {The 1st International Conference on Applications and Technologies of Renewable Energy Systems (ICATRES2024)}
}

Abstract

This research aims to compare and choose the suitable machine learning model based on specific data to predict the output of a solar panel system, which is the power energy. The focus was on a real-world solar power dataset with features like infrarouge radiation IRR, module temperature, and AC/DC current measurements. Different models used in this study, such as Linear Regression, Decision Trees, Support Vector Machines (SVR), K-Nearest Neighbors, a baseline Random Forest, and a Random Forest model with tuned hyperparameters using the grid search technique, are evaluated using principal evaluation metrics like the coefficient of determination,Mean Absolute Error, and Root Mean Squared Error. As a result, we found that the Random Forest baseline model was the accurate model for solar power forecasting based on the metrics evaluations.