Dynamic Hybrid Model for 24-Hour ahead Solar Forecasting

Abstract

One of the most popular ways of generating clean electric power nowadays is through the use of solar panels. However, there are a couple of problems that arise when considering solar panels, such as the fact that the efficiency of the panels decreases as the temperature increases and other problems that are weather-dependent. With the increasing integration of photovoltaic (PV) panels into grid networks, the complexity of grid management is increasing because of the fluctuating nature of solar energy. This project aims to develop a predictive model for the global horizontal irradiance (GHI) of PV panels using Python. GHI is an important parameter that influences the energy output of solar panels, and accurate GHI predictions can help optimize grid network energy management. The projects involve gathering historical weather data from GHI, preprocessing the data to extract the GHI values and other relevant features, training a hybrid model, integrating deep learning methods with ensemble methods, and evaluating its performance. The model will be tested on the test set of the data to assess its R square error (R^2), Root Mean Square Error (RMSE), Mean Squared Error (MSE), Normalized Root Mean Square Error (nRMSE) and Mean Absolute Error (MAE). The project aims to provide a useful tool for grid energy management to optimize energy production and reduce costs.

Date of Award	22 Jul 2024
Original language	American English
Supervisor	KHALIFA ALHOSANI (Supervisor)