Forecasting solar power output in Ibadan: A machine learning approach leveraging weather data and system specifications

2025

Read Forecasting solar power output in Ibadan: A machine learning approach leveraging weather data and system specifications (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-07462.txt.

Introduction

Obarotu Peter Urhuerhi, Christopher Udomboso, and Caston Sigauke forecast solar power output for Ibadan, Nigeria, from weather data and panel specs, in a paper from 2025. Sun gauges are costly, so a weather based forecast can aid grid plans and off grid design.

Their two stage method first calls clear sky sun levels from weather alone, then folds in cloud type for real sky levels. Three model types were trained on past weather, and a forest model won. Dry season calls beat wet season calls on error. A sun library then turned light levels into power output for set panels.

Catalog record

YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-07462

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Forecasting solar power output in Ibadan: A machine learning approach leveraging weather data and system specifications Corresponding author: Email address: [email protected], Phone No: +27 15 962 8135 (C Sigauke) Abstract This study predicts hourly solar irradiance components, Global Horizontal Irradiance (GHI), Direct Normal Irradiance (DNI), and Diffuse Horizontal Irradiance (DHI) using meteorological data to forecast solar energy output in Ibadan, Nigeria. The forecasting process follows a two-stage approach: first, clear-sky irradiance values are predicted using weather variables only (e.g., temperature, humidity, wind speed); second, actual (cloudy-sky) irradiance values are forecasted by integrating the predicted clear-sky irradiance with weather variables and cloud type. Historical meteorological data were preprocessed and used to train Random Forest, Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) models, with Random Forest demonstrating the best performance. Models were developed for annual and seasonal forecasting, capturing variations between the wet and dry seasons. The annual Random Forest model’s normalised Root Mean Square Error (nRMSE) values were 0.22 for DHI, 0.33 for DNI, and 0.19 for GHI. For seasonal forecasts, wet season nRMSE values were 0.27 for DHI, 0.50 for DNI, and 0.27 for GHI, while dry season nRMSE values were 0.15 for DHI, 0.22 for DNI, and 0.12 for GHI. The predicted actual irradiance values were combined with solar system specifications (e.g., maximum power (Pmax), open-circuit voltage (Voc), short-circuit current (Isc), and AC power (Pac)) using PVLib Python to estimate the final energy output. This methodology provides a cost-effective alternative to pyranometer-based measurements, enhances grid stability for solar energy integration, and supports efficient planning for off-grid and grid-connected photovoltaic systems. Keywords: Neural Networks, Random Forest, Renewable energy, Solar power forecasting, Weather data. 1 Introduction 1.1 Background and motivation In the context of Nigeria, the energy landscape presents its unique challenges. The country’s electricity supply is unreliable, with frequent power outages and an inadequate grid infrastructure failing to meet rising energy demands [ 35 , 30 ] . This situation has led businesses, individuals, and government agencies to explore alternative energy sources, particularly renewable energy. Among these alternatives, solar energy, especially solar photovoltaic (PV) systems, has gained considerable attention as a reliable solution [ 41 ] . Across the country, solar energy is widely available and offers a reliable, clean power source less prone to outages. As illustrated in Figure 1 , the Global Horizontal Irradiance (GHI) in Nigeria varies from around 3.5 kWh/m²/day in coastal parts to 7.0 kWh/m²/day in northern locations [ 50 , 40 ] . Figure 1: Map of the global horizontal irradiance of Nigeria [ 50 ] . The Nigerian government has aggressively promoted renewable energy as a key component of its plan to address the nation’s energy difficulties after realising the potential of solar power. This involves leveraging solar energy through programs like the National Renewable Energy and Energy Efficiency Policy (NREEEP) and the Nigeria Solar Energy Policy (NSEP) [ 43 ] . Rural electrification is one of the most urgent problems in Nigeria’s energy industry. Many Nigerians, especially those living in rural regions, still do not have access to power. An estimated 70% of these rural areas live in energy poverty since they are not linked to the national electricity grid [ 33 ] . The Nigerian government has started several rural electrification initiatives to solve this issue and provide more access to energy in these neglected areas. As of 2015, eight Power Purchase Agreements (PPAs) had indicated interest in establishing solar power projects, according to a report from the Nigerian Bulk Electricity Trading Company (NBET) [ 32 ] . Deploying off-grid and mini-grid solutions powered by renewable sources is one of the tactics, with solar energy at the forefront. These off-grid solar solutions offer a practical and scalable approach to bringing electricity to remote and rural communities where extending the national grid may not be economically or logistically feasible. By harnessing solar energy, these rural electrification projects aim to provide reliable, clean, and sustainable power to millions of Nigerians without access to electricity [ 8 , 17 ] . However, intermittent solar energy presents a problem for off-grid and grid-connected applications. It is challenging to forecast and control the energy supply because solar system output mostly relies on meteorological factors [ 5 , 6 ] . This fluctuation emphasises the importance of creating precise forecasting models to estimate solar energy output. Whether for rural electrification initiatives or as a supplement to the national grid, these models are essential for guaranteeing the dependability of the power supply [ 38 , 6 , 53 , 55 ] . Thus, this study aims to investigate advanced techniques for predicting solar energy production in Nigeria, enhancing the precision and dependability of these projections. By doing this, it hopes to encourage Nigeria’s wider adoption of solar energy, aid in attempts to electrify both urban and rural areas and assist in providing sustainable energy to fulfil the nation’s expanding energy needs. While there has been research on forecasting Global Horizontal Irradiance (GHI) in Nigeria, there is a need for continued efforts to improve accuracy. This research will focus on predicting GHI, direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI), as both DNI and DHI are crucial for accurately calculating the output energy of solar systems. Given that solar energy generation is highly site-dependent, this study will concentrate on Ibadan, using the Center of Petroleum, Energy Economics and Law (CPEEL), University of Ibadan (Lat 7.4515 o N, Lon 3.8899 o E) to develop a more precise predictive model tailored to the local environment. 1.2 Literature Review Machine learning (ML) and artificial neural networks (ANN) have emerged as essential tools in energy modelling, proving their adaptability across renewable and non-renewable energy sectors. In the oil and gas sector, for instance, these techniques have been successfully applied to predict oilfield scale formation, forecast gas production trends, and model crude oil prices [ 21 , 19 , 20 ] . These applications highlight the ability of ML and ANN to handle complex datasets and provide accurate predictions, thereby enhancing operational efficiency and economic analysis in energy systems. Building on this foundation, the current review focuses on applying traditional statistical learning models and artificial neural networks (ANN) in solar energy research. This section begins by examining significant global contributions to the field. From this worldwide perspective, it explores solar energy research undertaken in Nigeria, tackling the unique challenges and opportunities relevant to this setting. Lastly, it refines its focus to studies centred on Ibadan, Nigeria, which acts as the case study for this research. David et al. [ 16 ] draw comparisons between solar time series and financial data by probabilistically forecasting solar irradiance on a global scale using a combination of ARMA and GARCH models. With fewer parameters and faster implementation, the recursive ARMA-GARCH hybrid exhibits accuracy equivalent to machine learning methods like Support Vector Machines and Gaussian Process. The hybrid model’s biggest advantage is its ability to provide real-time forecasts without requiring a training period. However, forecast accuracy decreases with increasing lead time and under variable sky conditions. Sites with stable weather, like Desert Rock, showed better forecast accuracy, while more variable locations, such as Oahu and Fouillole, presented greater challenges in predictability. This study highlights the importance

Read the complete text