Missing data imputation of solar radiation data under different atmospheric conditions

Area of Science:

• Meteorology and Climatology
• Renewable Energy Systems
• Data Science and Time Series Analysis

Background:

• Global solar broadband irradiance is crucial for weather and energy studies.
• Pyranometer data from meteorological networks often contain missing or erroneous values.
• Data gaps and errors hinder accurate time series analysis of solar radiation.

Purpose of the Study:

• To evaluate the effectiveness of the multivariate imputation by chained equations (MICE) method for ten-minute solar radiation data.
• To compare MICE performance against Inverse Distance Weighting (IDW) and Multiple Linear Regression (MLR).
• To assess the capability of MICE in utilizing network data for sensor data imputation.

Main Methods:

• Utilized ten-minute solar radiation data from nine MeteoGalicia network stations.
• Applied the multivariate imputation by chained equations (MICE) method for data imputation.
• Compared MICE with Inverse Distance Weighting (IDW) and Multiple Linear Regression (MLR) using RMSE.

Main Results:

• MICE achieved an average RMSE of 13.37% for solar radiation data prediction.
• MLR resulted in an average RMSE of 28.19%, and IDW had an average RMSE of 31.68%.
• MICE demonstrated superior performance in imputing missing or incorrect solar radiation sensor data.

Conclusions:

• The multivariate imputation by chained equations (MICE) method is highly effective for imputing missing solar radiation data.
• MICE offers a significant improvement over traditional methods like IDW and MLR for solar radiation time series.
• This imputation technique enhances the reliability of solar radiation data for scientific studies.