Performance of Hybrid Neuro-Fuzzy Model for Solar Radiation Simulation at Abuja, Nigeria: A Correlation Based Input Selection Technique

Abstract

Solar Radiation (Rs) simulations for specific locations are critical for guiding decisions about the design and operation of solar energy conversion devices. The expensive instruments required to make high-resolution Rs measurements, as well as the rigorous maintenance procedures connected with such devices, limit Rs measurements. As a result, the ability to simulate Rs using easily observed environmental data is essential (such as temperature, humidity, cloud cover, etc.). This study looks at how well a machine learning model called Adaptive Neuro-Fuzzy Inference System (ANFIS) performs in estimating Rs in Abuja, Nigeria. Monthly maximum and minimum temperatures, relative and specific humidity, precipitation, and surface pressure data were collected. Four different statistical metrics (R², R, MSE, RMSE) are considered to evaluate the performance of this model. Best results were produced from a parameter combination of minimum temperature, precipitation, and surface pressure with R² of 0.8914 and RMSE of 0.0550 in the training phase and R² of 0.9744, and RMSE of 0.0444 in the testing phase. The results show that the hybrid model, ANFIS, is highly efficient in forecasting Rs in Abuja.