Gap Filling in Solar Radiation Data using Artificial Neural Network for Nine Stations in Pakistan

Tahir, Zia ul Rehman, Abdullah, Muhammad, Usman, Rahim, Raza, Agha Haider, Adil, Muhammad Taimoor, Asim, Muhammad, Akbar, Zahoor, Abbas, Sabtain, Nawaz, Ibtasam, Hadi, Ali, Ahmad, Sajeer, Tahir, Zia ul Rehman, Zafar, Usama, Hassan, Ahmad

EuroSun 2022 · Kassel, Germany · 2022-09-25
Published by International Solar Energy Society (ISES)
DOI: 10.18086/eurosun.2022.15.10

Abstract

This study presents an approach to fill the gap of ground measured solar data for different stations in Pakistan using machine learning. The Artificial Neural Network (ANN) technique was used on high-quality data measured by ESMAP at nine meteorological stations. A total of 33 non-linear regressive models with different combinations of nine available input parameters were evaluated based on their statistical importance. The evaluation was performed using statistical parameters; AIC, aAIC, BIC, t-stat, p-value, rMBE, rMAE, rRMSE and correlation coefficient. Three parameters (periodicity factor, pressure, wind speed) were removed based on statistical insignificance and a final model with six input parameters and ten neurons was selected for solar data gap filling for all stations. The estimated data shows best (least) agreement with ground measured data at Karachi (Lahore) station with a correlation of 0.966 (0.933). The model proposed in this study can be used by researchers for solar data gap filling in their respective regions with reasonable accuracy.

← All proceedings papers