This research focuses on predicting rooftop photovoltaic (PV) power generation in composite climates using machine learning (ML) techniques, including Random Forests, Support Vector Machines (SVMs), and Long Short-Term Memory (LSTM). The study aims to optimise renewable energy system dependability by evaluating and comparing these algorithms, exploring ensemble modelling benefits, and integrating geographical components. The research contributes to the literature on energy forecasting, emphasising the importance of accurate predictions in composite climates. The methodology involves comprehensive data gathering, preprocessing, and model training, with results demonstrating the superiority of the Ensemble Model. The study concludes that the Ensemble Model offers a robust solution for accurate rooftop PV power generation predictions in diverse climatic conditions.