Urban regions struggle with worsening energy management issues from the rate of demographic expansion and rising energy requirements and building integration of renewable power sources. Sanitation requires efficient energy management techniques because it ensures environmental sustainability and reduces both operational expenses and greenhouse gas emissions. The inability of traditional energy management systems to react to urban environment changes constitutes their primary drawback. The application of machine learning (ML) provides strong analytics which helps forecast energy requirements while optimizing grid management and controlling distributed resources alongside enhancing real-time data-driven decision-making processes. The paper investigates how machine learning techniques implement urban energy management through demand forecasting, energy efficiency optimization and load balancing and anomaly detection procedures. A diverse set of ML models including artificial neural networks, support vector machines, ensemble models as well as deep learning approaches are evaluated for their specific use cases and their effectiveness. Current urban infrastructure along with smart cities demonstrate how effective this concept can be through case study analysis. Moreover, the paper investigates obstacles relating to data protection together with difficulty understanding machine learning models and issues of merging them into current electrical systems. The paper discusses ML-developed power systems through the lens of smart grids together with edge computing and adaptive learning strategies for the future design direction. The paper delivers an extended discussion about how machine learning systems transform modern urban energy management strategies.