Main Steam Temperature System Modeling Based on PCA and Neural Networks

In this paper, a neural network modeling method using field data is developed for superheated steam model to improve the traditional modeling method based on the mechanism modeling. It is not necessary to get all the transfer functions and mathematical models, but only some main factors which affect superheated steam temperature and the influencing delay time. They could be confirmed by analyzing the super heater mechanism. After training with a large amount of field data, the model output agree with the real system well. Then some of new data are taken into a simulation experiment to prove the feasibility of this modeling method. Based on the new method, the field data will be processed by using principal component analysis (PCA). And according to this process, it is easy to reduce the input data and the size of network and optimize the modeling method for its application. The result shows that, using this method, the calculate time is almost halved without essential decrease of precision.