Emilia Sindila

Solar Power Generation Prediction Using Neural Networks

Name: Emilia Sindila
Primary Supervisor: Prof. Andy Morse
Year: 2
Discipline: Geography
Presentation type: Poster
Project Title: Solar Power Generation Prediction Using Neural Networks

Abstract:

Renewable energy sources are crucial in combating climate change by reducing greenhouse gas emissions from fossil fuels. The United Kingdom (UK) has significantly increased its share of renewable energy in electricity generation, aiming for net-zero emissions by 2050.

This project chapter analyses solar energy and various weather variables (wind speed, solar irradiation, humidity, and temperature) around the University of Liverpool campus. It utilises two weather stations located on different parts of the campus, as well as solar panels installed on buildings around campus.

Neural Networks (NN) have been utilised to develop prediction models that enable the identification of three different time steps: 24 hours, 48 hours and 1 week. A starting NN model using dense layers, 64 input layers, 32 hidden layers and 1 linear output layer, trained for 50 epochs, achieving different MSE and RMSE values depending on different activation functions tested. The target variable is solar power generation forecasted for three different time steps.