The invention discloses a
tail gas
sulfur-containing substance concentration real-time prediction method based on a stack-type auto-
encoder, and aims to solve the problem of online prediction of
sulfur-containing substance concentration in
tail gas by applying the stack-type auto-
encoder and deeply consider the unsupervised characteristic of the stack-type auto-
encoder and the
information loss problem of layer-by-layer
feature extraction. Specifically, according to the method, in the training process, output data of all
layers of encoders of the stack type auto-encoder are set as
sulfur substance concentration data, so that the unsupervised problem is solved, in addition, input data of all
layers of auto-encoders contain flow data capable of being measured in real time at the same time, and therefore the
information loss problem is avoided. Compared with a traditional method, the method has the advantages that the output
estimation value of each layer of auto-encoder is fully utilized,the least square regression is used for further improving the precision of soft measurement, and the superiority of the method relative to the traditional method is verified through a specific application case.