Synthetic minority oversampling technique for synthetic meteorological data generation*

This study investigates different approaches for synthetic data generation of time series. The main focus of this study is the Synthetic Minority Oversampling Technique (SMOTE) model, which is widely applied in several situations due to its particular ability to deal with unbalanced datasets. In order to assess the effectiveness and accuracy of different models, a comparison between them is performed. In particular, the SMOTE model is applied using two datasets from the Pinhão Region and Quinta de Santa Bárbara, being applied to the analysis and generation of several different variables, namely, Average temperature, Wind direction, Average wind speed, Maximum instantaneous wind speed and Solar radiation. Through the use of SMOTE it was possible to generate missing data in Quinta de Santa Bárbara in the period from September to December 2022, resulting from this process an analysis of the results and a comparison with the results obtained with the Linear Regression Method and the Generative Adversarial Networks (GANs) model. Thus, with this study, the SMOTE model demonstrated good results in the synthetic generation of the missing data, making it a viable option for the function.