Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Development of Artificial Intelligence Model to Forecast Photovoltaic Power Generation Including Airborne Particulate Matter
Purpose This study aims to suggest an optimal model for predicting photovoltaic (PV) power generation by comparing single and hybrid models that include particulate matter in the atmosphere as input parameters. Methods From December 2016 to December 2019, 1 MW-class PV power generation data in Jindo-gun, Jeollanam-do and meteorological data and particulate matter data from Mokpo were used. Radiation, sunshine time, pressure, temperature, humidity, wind speed, wind direction, snow load, precipitation, PM10, and PM2.5 were used as input parameters. We used single models such as random forest (RF), artificial neural network (ANN), long short-term memory (LSTM), and gate recurrent unit (GRU) and hybrid model such as LSTM-ANN and GRU-ANN. Coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) were used to compare and evaluate the prediction performance of the models. Results and Discussion The variable importance through RF was as follows: radiation (77.66%), day light hours (4.85%), pressure (4.16%), temperature (3.98%), humidity (2.25%), wind speed (2.21%), PM10 (2.72%), PM2.5 (1.65%), wind direction (1.44%), snow cover (0.05%), and precipitation (0.02%). GRU-ANN showed the highest R2 (0.838) among the models and lower epoch (8) than GRU using the early stop. Conclusion The GRU-ANN model was the most suitable for forecasting PV power generation including particulate matter.
Development of Artificial Intelligence Model to Forecast Photovoltaic Power Generation Including Airborne Particulate Matter
Purpose This study aims to suggest an optimal model for predicting photovoltaic (PV) power generation by comparing single and hybrid models that include particulate matter in the atmosphere as input parameters. Methods From December 2016 to December 2019, 1 MW-class PV power generation data in Jindo-gun, Jeollanam-do and meteorological data and particulate matter data from Mokpo were used. Radiation, sunshine time, pressure, temperature, humidity, wind speed, wind direction, snow load, precipitation, PM10, and PM2.5 were used as input parameters. We used single models such as random forest (RF), artificial neural network (ANN), long short-term memory (LSTM), and gate recurrent unit (GRU) and hybrid model such as LSTM-ANN and GRU-ANN. Coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) were used to compare and evaluate the prediction performance of the models. Results and Discussion The variable importance through RF was as follows: radiation (77.66%), day light hours (4.85%), pressure (4.16%), temperature (3.98%), humidity (2.25%), wind speed (2.21%), PM10 (2.72%), PM2.5 (1.65%), wind direction (1.44%), snow cover (0.05%), and precipitation (0.02%). GRU-ANN showed the highest R2 (0.838) among the models and lower epoch (8) than GRU using the early stop. Conclusion The GRU-ANN model was the most suitable for forecasting PV power generation including particulate matter.
Development of Artificial Intelligence Model to Forecast Photovoltaic Power Generation Including Airborne Particulate Matter
Jaeseong Yoon (Autor:in) / Kyung-Min Kim (Autor:in) / Johng-Hwa Ahn (Autor:in)
2022
Aufsatz (Zeitschrift)
Elektronische Ressource
Unbekannt
Metadata by DOAJ is licensed under CC BY-SA 1.0
| Europäisches Patentamt | 2019
Environmental effects of airborne particulate matter
| British Library Conference Proceedings | 1998
Non-health effects of airborne particulate matter
| Elsevier | 1981
Contribution of bioaerosols to airborne particulate matter
| Taylor & Francis Verlag | 2020