Year: 2024
Type: conference-paper
Abstract: The transition in energy systems aims for efficiency improvements at a higher level, targeting the reduction of climate change impacts. Investing in solar energy, endorsed by the global scientific community, is essential. One of the core obstacles hindering the seamless integration of photovoltaic (PV) systems into power grids is the associated uncertainty. This paper focuses on machine learning forecasting algorithms for solar power generation, specifically using the Long Short-Term Memory (LSTM) algorithm, a type of recurrent neural network (RNN). Accurate prediction models are crucial for maximising the efficiency and reliability of solar energy systems, especially with high PV penetration. The LSTM architecture's ability to capture temporal dependencies makes it well-suited for time series forecasting tasks such as solar irradiance prediction. Using ensemble LSTM improves output accuracy and reduces mean square error in solar energy plant production. Temporal variations in solar power production, analysed using SARIMA/ ARIMA models, highlight the volatility of PV power generation, which causes issues such as frequency instability, dispatch difficulties, and surges in current/voltage on the grid.
Authors: Sahar Daebes, Mohamed Darwish, Chun Sing
Institution: Brunel University of London
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4407938793