Aiming at solving the agglomeration and blockage problem occurred during biomass gasification in dual fluidized bed system

a fault diagnosis model based on long short-term memory network(LSTM) and soft voting is proposed. Firstly

the agglomeration and clogging faults are simulated on the dual fluidized bed cold state experimental system

and the pressure fluctuation signals corresponding to the riser and the gasification chamber are collected. Secondly

the original single is decomposed into small samples with suitable length

and used as the input of the LSTM model for training and test. Finally

the soft voting method is used to make decisions on the output of the test samples to determine the fault category. Compared with many common models

the experimental results show that the proposed model has higher fault diagnosis accuracy(91.25%)

which has certain practical application value.