Building energy systems (BESs) are essential for modern infrastructure but face significant challenges in equipment diagnosis, energy consumption prediction, and operational control. The complexity of BESs, coupled with the increasing integration of renewable energy sources, presents difficulties in fault detection, accurate energy forecasting, and dynamic system optimisation. Traditional control strategies struggle with low efficiency, slow response times, and limited adaptability, making it difficult to ensure reliable operation and optimal energy management. To address these issues, researchers have increasingly turned to machine learning (ML) techniques, which offer promising solutions for improving fault diagnosis, energy scheduling, and real-time control in BESs. This review provides a comprehensive analysis of ML techniques applied to fault diagnosis, energy consumption prediction, energy scheduling, and operational control. According to the results of analysis and literature review, supervised learning methods, such as support vector machines and random forest, demonstrate high classification accuracy for fault detection but require extensive labelled datasets. Unsupervised learning approaches, including principal component analysis and clustering algorithms, offer robust fault identification capabilities without labelled data but may struggle with complex nonlinear patterns. Deep learning techniques, particularly convolutional neural networks and long short-term memory models, exhibit superior accuracy in energy consumption forecasting and real-time system optimisation. Reinforcement learning further enhances energy management by dynamically adjusting system parameters to maximise efficiency and cost savings. Despite these advancements, challenges remain in terms of data availability, computational costs, and model interpretability. Future research should focus on improving hybrid ML models, integrating explainable AI techniques, and enhancing real-time adaptability to evolving energy demands. This review also highlights the transformative potential of ML in BESs and outlines future directions for sustainable and intelligent building energy management.