This study proposes a predictive maintenance and fault monitoring method for smart distribution networks based on the Internet of Things and machine learning, aiming to address the challenges of distribution network operation complexity and fault risk. The real-time status of the distribution network is monitored through the Internet of Things technology, and key parameters such as voltage, current, and temperature are captured to ensure the accuracy and timeliness of data collection. Combined with deep learning models (such as CNN and LSTM), this method significantly improves the accuracy of fault detection and prediction, and can detect potential faults in a timely manner and predict fault trends. The experimental verification results show that the system performs well under different load conditions, and the fault detection model responds quickly and accurately, meeting the high standards of the smart distribution network. The fault prediction model provides sufficient warning time, effectively reducing the losses caused by faults. The system’s integrated sensor network, data processing module and user interface form a fully functional intelligent monitoring and maintenance platform. Data visualization and user-friendly interface design improve operational efficiency and user experience. The research results provide an effective technical solution for the operation and maintenance of smart distribution networks, demonstrating its wide applicability and potential value.