Processed Wind Power and Weather Time Series for Machine Learning Applications in Brazil

This dataset contains time series data for wind power generation modeling in two Brazilian regions: Serra (Espírito Santo) and Santa Vitória do Palmar (Rio Grande do Sul). The data were prepared to support supervised machine learning models for short-term wind power forecasting. The original datasets combine simulated wind power, obtained from a mathematical model of wind turbines, with real meteorological measurements collected from weather stations. Both datasets were aligned in time and standardized to fixed temporal resolutions (15 minutes for ES and 60 minutes for RS). A preprocessing pipeline was applied to ensure data quality and temporal consistency. This included: - merging date and time information into a unified timestamp, - removal of duplicate records, - filtering of inconsistent measurements (e.g., zero power output under wind speeds above cut-in threshold), - reindexing to a complete and continuous time grid, - identification of missing data through binary flags and gap length categorization. The final datasets (dfes_reindexed and dfrs_reindexed) include both cleaned measurements and metadata describing missing data patterns, enabling robust time series analysis and machine learning applications. These datasets can be used for: (i) wind power forecasting, (ii) missing data analysis and imputation studies, (iii) evaluation of machine learning models in renewable energy applications.