Solar Photovoltaic Power Potential and Generation Forecasting in Nigeria using Machine Learning

Abstract 
Nigeria possesses significant solar energy 
potential, yet uncertainty in generation 
forecasts limits large-scale adoption. This 
study applies ML models to forecast solar PV 
output using satellite- derived meteorological 
data. 
The increasing trend of using solar 
photovoltaic (PV) energy as an alternative to 
fossil fuels has increased the demand for 
precise forecasting tools, particularly in 
developing nations such as Nigeria, where grid 
instability and load mismatch are prevalent. 
This paper examines the capability of 
Artificial Neural Network (ANN) models for 
forecasting solar PV energy in grid-connected 
systems in the Nigerian energy sector. Based 
on the historical meteorological and load 
demand data from 2020 to 2025, an ANN 
model was designed, trained, and simulated 
using MATLAB R2022a software. The model 
included essential parameters such as solar 
irradiance, temperature, and time variables to 
forecast solar power generation. 
The simulated values were validated against 
the actual output to determine the accuracy of 
the model using parameters such as Mean 
Absolute Percentage Error (MAPE), Root 
Mean Square Error (RMSE), and the 
coefficient of determination (R²). The ANN 
model yielded a MAPE of 6.83%, an RMSE 
of 12.47 kW, and an R² of 0.95, indicating 
excellent forecasting accuracy and 
adaptability to the non-linear solar output 
variations. In addition, the study presents 
graphical results, such as predicted vs. actual 
output graphs, error distribution histograms, 
and regression plots, which verify the 
robustness of the model. 
These findings affirm the viability of using 
ANN for the forecasting of solar PV and 
underscore its promise to improve the energy 
planning, stability, and dispatch of energy in 
Nigeria. Moreover, the findings of this study 
encourage the adoption of AI-based 
forecasting tools to improve the  

energy management of Nigeria to maximize 
the benefits of renewable energy. This paper 
contributes to the existing knowledge on 
intelligent forecasting for smart grid 
applications and presents a model that can 
be replicated in other developing countries.