| Summary: | In recent years, the increased of the world's electricity consumption has resulted in a significant rise in carbon dioxide emissions from electricity generation, contributing heavily to global warming. To combat this, renewable energy sources like solar power are being increasingly adopted to reduce reliance on fossil fuels and slow down global warming. However, to effectively use solar power, accurate forecasting of solar irradiance is crucial for predicting the output of solar PV systems and optimizing their operation for better system reliability. This study aims to investigate various short-term forecasting methods for solar irradiance and evaluate their prediction accuracy using mean absolute error and root mean squared error measures. Specifically, the report focuses on implementing three machine learning methods which are Long Short-Term Memory (LSTM), Feedforward Neural Network (FNN), and Autoregressive Integrated Moving Average (ARIMA). Based on the research, extracting results and comparison, Long Short-Term Memory has proven to be the most effective and easiest method to implement compared to Feedforward Neural Network (FNN) and Autoregressive Integrated Moving Average (ARIMA).
|
|---|