Time-Series Prediction Algorithm for Boiler Power Generation Steam Temperature

In boiler power generation scenario, multi-modal data could be collected from the system, where main steam temperature could be the most important factor as it influence the whole power generation system's efficiency and safety. In this paper a time series prediction algorithm SCEM is proposed to predict future main steam temperature based on it's history data, combines time sequential feature extraction and contextual feature extraction by cross attention mechanism, outputting a more accurate prediction on main steam temperature than traditional models, providing valuable numerical references for more industrial applications.