基于OB-PI-LSTM模型的发电机组再热蒸汽温度预测研究OA

To predict the temperature trend of boiler reheat steam,an Observation network-enhanced Physics-Informed Long Short-Term Memory(OB-PI-LSTM)network was designed for real-time correction.The model was trained using actual plant production data.The influence of fuel,air volume,and steam dynamic characteristics on the model's predictive capability was ana-lyzed,and the effect of load variation intervals on prediction performance was investigated.The re-sults indicate that the OB-PI-LSTM model exhibits superior performance.Compared with the stan-dard Long Short-Term Memory(LSTM)model and the Physics-Informed Long Short-Term Memory(PI-LSTM)model,the Mean Absolute Error(MAE)is reduced by 51.34%and 13.90%,respec-tively.The model structure was enhanced for multi-step prediction,enabling accurate forecasting of boiler reheat steam temperature and demonstrating good generalization performance even during load variation intervals.This framework contributes to improving the control efficiency of steam systems and can be extended to prediction applications in other domains such as power and gas systems.