The University of Melbourne
UNIMELB_PERERA-Maneesha_VYT-LOCAL-2021.mp4 (69.94 MB)

Deep Learning for Accurate Distributed Solar Power Forecasting

Download (69.94 MB)
posted on 2021-08-26, 06:21 authored by Maneesha PereraManeesha Perera

Rooftop solar photovoltaic (PV) systems have seen enormous growth in recent years due to the high demand for solar power as a clean and renewable energy source. While these systems have helped to reduce power bills for households, they have also introduced significant challenges in managing our electrical networks and markets, such as over-generation of electricity during the mid-day and steep ramping needs during the evening. Accurate solar power forecasts are important to overcome many of these issues. Artificial Intelligence techniques, specifically deep learning methods such as Convolutional Neural Networks, Long Short-Term Memory Networks, have shown promising results in the forecasting literature. However, the full potential of these methods in solar power forecasting has to be extensively investigated. So, my research investigates state-of-the-art deep learning methods and proposes novel techniques based on Convolutional Neural Networks to predict future solar generation accurately.