83 / 2025-04-07 00:23:53

Sequential Processing-Based Deep State Space Model for Soft Sensing of Thermal Power Plant Generation Output

power generation prediction,soft sensor,sequential filtering,deep state space model

This paper proposes a Deep State Space Model with Sequential Processing (Seq-Dssm) for predicting the generation output of thermal power plants. Compared to conventional deep state-space models, the proposed method avoids matrix inversion operations through sequential processing, thereby improving computational efficiency and numerical stability. Experimental validation using actual operational data from a thermal power plant demonstrates that the proposed model achieves significant improvements in prediction accuracy over traditional methods. Additionally, by providing uncertainty estimates for predictions, it offers more reliable references for plant operations. The results confirm that the method maintains high prediction accuracy while substantially enhancing computational efficiency, indicating strong potential for engineering applications.