An Early Warning Method for the Operating Status of Turbine Sliding Pin System Based on Multi-Step Time Series Prediction

Real-time monitoring of the sliding pin system is essential for ensuring the safe and reliable operation of the turbine generator set, especially under peak loads. This study begins by examining the operational mechanism of the sliding pin system, identifying key monitoring parameters reflecting its operating status, gathering time series data via the distributed control system (DCS), and conducting multi-step predictions using the Long Short-Term Memory (LSTM) model to facilitate indirect early fault detection. The model demonstrates strong performance according to a comprehensive four-dimensional evaluation index assessing the accuracy of the LSTM model in multi-step predictions of individual time series data. These findings hold significant engineering and practical value for steam turbine operation and maintenance practices.