The invention discloses an
aluminium smelting process furnace box temperature prediction method based on a
deep belief network. The method comprises the following steps that 1, a plurality of sets oforiginal data are acquired; 2, for the
original data acquired in the step 1, abnormal data is eliminated, and
noise is removed to obtain normal data; 3, for the normal data obtained in the step 2, thedeep belief network is used for
feature extraction to obtain feature vectors; 4, the sets of feature vectors are divided into a
training set and a testing set, and a prediction model is built, wherein through the sets of feature vectors in the
training set, the prediction model is trained constantly to obtain a trained prediction model; 5, each set of feature vectors in the testing set is used for testing the trained prediction model, if the testing stability is good, the prediction model can be used for predicting a furnace box temperature, and if the testing stability is not good, the step3 is returned to. According to the
aluminium smelting process furnace box temperature prediction method based on the
deep belief network, the furnace box temperature can be predicted through other indexes easy to detect, components and parts are not prone to be damaged, and the
economic benefits are good.