Auto-encoder based modeling of combustion system for circulating fluidized bed boiler

Deep learning attract the interests of many researchers. Multidimensional algorithms require large data storage space. This paper proposes a modeling of the combustion system used for Circulating Fluidized Bed Boiler (CFBB), which is based on the method of auto-encoder of deep learning. The 20 dimensional input samples set is the input layer, and then the units of hidden layer are calculated. The data dimension is reduced through the auto-encoder, further, these data are as input of the RBF network. The modeling is carried out by the Radical Basis Function (RBF) neutral network. Compared with traditional methods, the auto-encoder is suitable for modeling. The samples are greatly reduced for the subsequent work. Numerical results provided in this paper validate the proposed model and method, as well as the validity of the conversion from the auto-encoder strategy.