Neural network quantification for solar radiation prediction: An approach for low power devices
Accurate solar radiation prediction leverages various machine learning techniques, with artificial neural networks (ANN) being the most common and precise due to their ability to detect and learn relationships between meteorological variables and solar radiation. Traditionally, training and deployi...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universidad de Santander
2025-01-01
|
| Series: | AiBi Revista de Investigación, Administración e Ingeniería |
| Subjects: | |
| Online Access: | https://revistas.udes.edu.co/aibi/article/view/4107 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841564203879497728 |
|---|---|
| author | Brenda Alejandra Villamizar-Medina Angelo Joseph Soto Vergel Byron Medina-Delgado Darwin Orlando Cardozo-Sarmiento Dinael Guevara-Ibarra Oriana Alexandra Lopez-Bustamante |
| author_facet | Brenda Alejandra Villamizar-Medina Angelo Joseph Soto Vergel Byron Medina-Delgado Darwin Orlando Cardozo-Sarmiento Dinael Guevara-Ibarra Oriana Alexandra Lopez-Bustamante |
| author_sort | Brenda Alejandra Villamizar-Medina |
| collection | DOAJ |
| description |
Accurate solar radiation prediction leverages various machine learning techniques, with artificial neural networks (ANN) being the most common and precise due to their ability to detect and learn relationships between meteorological variables and solar radiation. Traditionally, training and deploying these models require high-capacity computers. However, the proliferation of low-power smart devices, such as embedded systems and mobile devices, necessitates exploring methodologies for implementing ANN on systems with limited computational resources. This paper proposes a quantized neural network model for solar radiation prediction, considering the hardware limitations of low-power devices like the Raspberry Pi RP2040 microcontroller. The methodology involves five stages: hardware and software selection, neural network development and quantization, microcontroller implementation, model validation, and result analysis. Experimental design allows detailed performance evaluation of quantized neural networks, demonstrating that the TensorFlow Lite Quantized Aware model is suitable for solar radiation prediction. Metrics such as root mean square error (RMSE) of 44.24 and R² of 0.96 indicate that the selected quantized model differs from the original non-quantized model by less than 0.5% in RMSE and 0.04% in R². The study concludes that implementing quantized ANN models on microcontrollers is a technically and economically viable solution for solar radiation prediction. Quantization enables complex predictive models to run on low-cost, energy-efficient devices, thereby democratizing advanced prediction technologies for critical applications like solar energy generation.
|
| format | Article |
| id | doaj-art-d59a552db3774edfa0e72ffb51c4baab |
| institution | Kabale University |
| issn | 2346-030X |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Universidad de Santander |
| record_format | Article |
| series | AiBi Revista de Investigación, Administración e Ingeniería |
| spelling | doaj-art-d59a552db3774edfa0e72ffb51c4baab2025-01-02T23:06:38ZengUniversidad de SantanderAiBi Revista de Investigación, Administración e Ingeniería2346-030X2025-01-0113110.15649/2346030X.4107Neural network quantification for solar radiation prediction: An approach for low power devicesBrenda Alejandra Villamizar-Medina0https://orcid.org/0000-0002-0946-0762Angelo Joseph Soto Vergel1https://orcid.org/0000-0001-5093-0183Byron Medina-Delgado2https://orcid.org/0000-0003-0754-8629Darwin Orlando Cardozo-Sarmiento3https://orcid.org/0000-0003-3177-3893Dinael Guevara-Ibarra4https://orcid.org/0000-0003-3007-8354Oriana Alexandra Lopez-Bustamante5https://orcid.org/0000-0003-4601-1111Universidad Francisco de Paula Santander - Cúcuta, ColombiaUniversity at Buffalo - Buffalo, United StatesUniversidad Francisco de Paula Santander - Cúcuta, ColombiaUniversidad Francisco de Paula Santander - Cúcuta, ColombiaUniversidad Francisco de Paula Santander - Cúcuta, ColombiaUniversidad Francisco de Paula Santander - Cúcuta, Colombia Accurate solar radiation prediction leverages various machine learning techniques, with artificial neural networks (ANN) being the most common and precise due to their ability to detect and learn relationships between meteorological variables and solar radiation. Traditionally, training and deploying these models require high-capacity computers. However, the proliferation of low-power smart devices, such as embedded systems and mobile devices, necessitates exploring methodologies for implementing ANN on systems with limited computational resources. This paper proposes a quantized neural network model for solar radiation prediction, considering the hardware limitations of low-power devices like the Raspberry Pi RP2040 microcontroller. The methodology involves five stages: hardware and software selection, neural network development and quantization, microcontroller implementation, model validation, and result analysis. Experimental design allows detailed performance evaluation of quantized neural networks, demonstrating that the TensorFlow Lite Quantized Aware model is suitable for solar radiation prediction. Metrics such as root mean square error (RMSE) of 44.24 and R² of 0.96 indicate that the selected quantized model differs from the original non-quantized model by less than 0.5% in RMSE and 0.04% in R². The study concludes that implementing quantized ANN models on microcontrollers is a technically and economically viable solution for solar radiation prediction. Quantization enables complex predictive models to run on low-cost, energy-efficient devices, thereby democratizing advanced prediction technologies for critical applications like solar energy generation. https://revistas.udes.edu.co/aibi/article/view/4107quantized neural networksolar radiation predictionmicrocontroller |
| spellingShingle | Brenda Alejandra Villamizar-Medina Angelo Joseph Soto Vergel Byron Medina-Delgado Darwin Orlando Cardozo-Sarmiento Dinael Guevara-Ibarra Oriana Alexandra Lopez-Bustamante Neural network quantification for solar radiation prediction: An approach for low power devices AiBi Revista de Investigación, Administración e Ingeniería quantized neural network solar radiation prediction microcontroller |
| title | Neural network quantification for solar radiation prediction: An approach for low power devices |
| title_full | Neural network quantification for solar radiation prediction: An approach for low power devices |
| title_fullStr | Neural network quantification for solar radiation prediction: An approach for low power devices |
| title_full_unstemmed | Neural network quantification for solar radiation prediction: An approach for low power devices |
| title_short | Neural network quantification for solar radiation prediction: An approach for low power devices |
| title_sort | neural network quantification for solar radiation prediction an approach for low power devices |
| topic | quantized neural network solar radiation prediction microcontroller |
| url | https://revistas.udes.edu.co/aibi/article/view/4107 |
| work_keys_str_mv | AT brendaalejandravillamizarmedina neuralnetworkquantificationforsolarradiationpredictionanapproachforlowpowerdevices AT angelojosephsotovergel neuralnetworkquantificationforsolarradiationpredictionanapproachforlowpowerdevices AT byronmedinadelgado neuralnetworkquantificationforsolarradiationpredictionanapproachforlowpowerdevices AT darwinorlandocardozosarmiento neuralnetworkquantificationforsolarradiationpredictionanapproachforlowpowerdevices AT dinaelguevaraibarra neuralnetworkquantificationforsolarradiationpredictionanapproachforlowpowerdevices AT orianaalexandralopezbustamante neuralnetworkquantificationforsolarradiationpredictionanapproachforlowpowerdevices |