This thesis explores the enhancement of photovoltaic (PV) system reliability through the development and optimisation of machine learning algorithms for proactive fault detection. Utilising data from the Desert Knowledge Australia Solar Centre, the study focuses on inverter failures and intermittent issues, employing advanced machine learning techniques for predictive analysis. The research introduces novel adaptive methods for fault prediction and differentiation, significantly contributing to predictive maintenance strategies. These findings pave the way for future investigations into robust predictive maintenance frameworks, aiming to improve PV system efficiency and support the transition towards sustainable energy solutions.