Monitoring Turbine Generator Sets with a Hybrid Unsupervised Learning Strategy

Gas and steam turbines play a critical role in ensuring power grid stability, particularly as the share of variable renewable energy sources like wind and solar continues to increase. As essential components of modern thermal power systems, gas and steam turbines offer significant operational flexibility, allowing them to rapidly adjust output in response to real-time fluctuations in grid demand. Several industrial applications have demonstrated that Self-Organizing Maps (SOMs) provide substantial benefits in anomaly detection, as well as in identifying and visualizing process phases, thus facilitating comprehensive monitoring and improved understanding of technical processes. To address the limitations associated with SOMs, particularly when dealing with imbalanced datasets and strongly correlated process variables, this article evaluates a hybrid unsupervised learning strategy (HULS) for monitoring a turbine generator set.