Curated Solar Weather and PV AC Power Output Dataset for Explainable Solar Power Forecasting
Description
This dataset contains a curated, processed solar-weather and photovoltaic AC power output dataset prepared for machine-learning-based solar power forecasting. The dataset was developed by integrating publicly available PV generation data and solar-weather data originally obtained from Kaggle. The PV generation component was derived from Plant_1_Generation_Data.csv, which contains timestamped photovoltaic power generation records. The solar-weather component was derived from SolarPrediction.csv, which contains solar radiation and meteorological variables, including temperature, relative humidity, wind speed, wind direction, and related atmospheric measurements. The final processed dataset is provided as Solar_Weather_Data_Processed.csv. It contains 2,777 continuous hourly samples from 1 September 2016 10:00 to 26 December 2016 02:00. The dataset includes global horizontal irradiance, temperature, relative humidity, wind speed, wind direction, hour, day, month, solar zenith angle, solar azimuth angle, and AC output. The target variable is AC Output, representing photovoltaic AC power output used for supervised regression modeling. The dataset was cleaned and prepared by converting timestamps, resampling hourly, renaming features, converting units, performing feature engineering, calculating solar angles, and final formatting for machine learning experiments. This dataset supports reproducibility of research on explainable and uncertainty-aware solar power forecasting for smart grid applications.
Files
Steps to reproduce
1. Download the final processed dataset file Solar_Weather_Data_Processed.csv from this Mendeley Data record. 2. Download or inspect the accompanying preprocessing notebook/script to understand the construction of the final dataset. 3. The PV generation source file was processed by converting DATE_TIME into a datetime format, resampling the AC power records to an hourly resolution, and renaming AC_POWER as AC Output. 4. The solar-weather source file was processed by converting UNIXTime into timestamp format, resampling the meteorological variables to hourly resolution, and standardizing the feature names. 5. Temperature was converted from degrees Fahrenheit to degrees Celsius, and wind speed was converted from miles per hour to meters per second. 6. Time-derived features, including hour, day, and month, were extracted from the timestamp. 7. Solar zenith angle and solar azimuth angle were calculated from timestamp and location information. 8. The final dataset was cleaned to remove missing values and duplicate records. 9. The model-ready dataset contains 2,777 continuous hourly samples and can be used for supervised regression with AC Output as the target variable. 10. For model evaluation, the dataset should be split chronologically, using the earlier observations for training and the later observations for testing.
Institutions
- Rajshahi University of Engineering and TechnologyRajshahi Division, Rajshahi