The study focuses on enhancing time-frequency data management and diagnostics in power grid systems using big data technologies. Despite stable communication via power line carrier technology, existing metering devices face synchronization issues. To address this, a time-frequency network based on carrier technology is proposed for real-time transmission and monitoring. The study employs Bayesian improved decision tree and Markov Monte Carlo chain (MCMC) methods for real-time anomaly diagnosis of time-frequency networks and nodes. These algorithms analyze large-scale time-frequency data to identify and predict anomalies, thereby improving diagnostic accuracy and efficiency. The Bayesian decision tree enhances classification by incorporating Bayesian inference during node splitting, while MCMC efficiently approximates complex probability distributions through random sampling. Key contributions include quick fault identification, optimized synchronization strategies, and ensuring stable power system operation. The integration of AI and big data technologies significantly improves the operating efficiency and reliability of carrier-based time-frequency networks, providing robust technical support for new power systems.