PREDICTION MODELING BASED ON RECURREN TNEURAL NETWORKS WITH SELF-TUNING FUNCTION

This paper discusses the architecture and algorithm of recurrent neural networks(RNN) and proposes an approach of prediction modeling for non-linear time-varying system based on the recurrent neural networks with self-tuning function in combination with principal component analysis. The method is applied to predict the top temperature of vacuum distillation column and has been proven to have better performance than other methods.