Solar Irradiation and Temperature Prediction Using LSTM Neural Network and Solar Energy Potential in Northern Cameroon for Solar Electric Vehicle Application

Abstract

The fight against global warming is becoming increasingly important because of the race to find new sources of energy. This work presents a method for estimating solar energy (SE) potential by applying LSTM neural networks to predict the parameters influencing photovoltaic production, namely, solar irradiation and temperature over 24 h. The database comes from MERRA-2 and includes the meteorological parameters’ temperature, relative humidity, atmospheric pressure, wind speed, wind direction, solar irradiation and rainfall for 20 locations in northern Cameroon with hourly time steps over 23 years, from 2000 to 2022. The method consists of designing an LSTM neural model to obtain the lowest deviation between the real and predicted data. The performance criteria used to validate the model that predicts the two parameters are an accuracy of 92.45% and an RMSE of 20.9. The model was tested in two localities, Makary and Banyo, with the most important values out of the 20 localities. The database was used to estimate that the average SE potential is 2.193 MWh/m2/year in Makary, and the lowest potential is 1.949 MWh/m2/year in Banyo. This information can be used to select sites for solar power plant installations, solar photovoltaic energy management, solar electric vehicle (EV) fleet management and to select sites for the construction of EV recharging stations.