Solar Irradiance Forecast using Feed Forward Neural Network: A Case Study of Zaria Town

• Ismaila MAHMUD

  Department of Electrical Engineering, Ahmadu Bello University, Zaria, Nigeria

  Author

• Mahmud MUSTAPHA

  Department of Electrical/Electronic Engineering, Nuhu Bamalli Polytechnic, Zaria, Kaduna State, Nigeria

  Author

• Sulaiman H. SULAIMAN

  Department of Electrical Engineering, Ahmadu Bello University, Zaria, Nigeria

  Author

• Ibrahim ABDULWAHAB

  Department of Electrical Engineering, Ahmadu Bello University, Zaria, Nigeria

  Author

• Ibrahim A. SHEHU

  Department of Electrical Engineering, Ahmadu Bello University, Zaria, Nigeria

  Author

• Aminu J. ALIYU

  Department of Electrical Engineering, Ahmadu Bello University, Zaria, Nigeria

  Author

• Yusuf S. ABU

  Department of Electrical/Electronic Engineering, Federal University Dutsin-Ma, Katsina State, Nigeria

  Author

• Nuraddeen A. ILIYASU

  Department of Electrical Engineering, Ahmadu Bello University, Zaria, Nigeria

  Author

This study aims to forecast solar irradiation using Artificial Neural Network (ANN), with the goal of developing a high-performance prediction model based on real meteorological data. Lack of sufficient meteorological data in Nigeria necessitate the development of model to forecast solar irradiance for optimal utilization. The model is designed to predict daily solar irradiation for Zaria town, providing valuable insights to the utilities managing solar energy generation and monitoring systems. Feed forward Neural Network (FFNN) was applied to perform day-ahead solar irradiance forecasting. We employ a day-ahead persistence model as a baseline, a commonly used method in solar irradiance forecasting research. It operates under the assumption that current conditions will persist over the forecast horizon. Specifically, it uses the irradiance values from the previous day as the predictions for the following day. The findings highlight the significance of meteorological factors (such as minimum humidity, maximum temperature, day, month, and wind direction) in the FFNN model training. Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) were used to evaluate the performance of the model. The RMSE of 4.46 W/m² and MAE of 2.52 W/m² obtained indicate an excellent performance of the FFNN model. The model outperformed the Persistence model in predicting daily solar irradiance, indicating its superiority solar irradiance forecast. The results show the ability of the model to forecast day – ahead solar irradiance in Zaria town which can address the issue of non-recorded meteorological data.

• FJET_12_23_16

Copyright (c) 2025 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.