Prediction of water softening capacity in tobacco production based on spatiotemporal attention LSTM
Water softening processing capacity prediction is important to improve the quality of soft water in cigarette production and reduce the energy consumption of soft water system.To ad-dress the challenges of low prediction accuracy and difficulty in accurately setting the operational runtime of the water softening system in cigarette production,this paper proposes a method for predicting the water softening processing capacity based on a spatial-temporal attention long short-term memory network(STLSTM).The grey relational analysis method is employed to calculate the correlation values of various influencing factors on the water softening processing capacity,thereby identifying key factors affecting water softening processing capacity.To address the differ-ences in the impact of different input features on the prediction of water softening processing capac-ity and the variations in prediction performance due to different sequence lengths,a prediction model for water softening runtime and water softening processing quantity based on the spatiotem-poral attention mechanism of LSTM is constructed.This model aims to achieve precise predictions of the operational runtime and processing quantity of water softening.Comparative experiments are conducted using the water softening production system at the Yinkou Cigarette Factory of Hongta Liaoning Tobacco Co.,Ltd.The results indicate that,compared to other prediction methods,STLSTM reduces the prediction deviations of water softening operational runtime and processing quantity with reductions of around 1.3％and 11.9％respectively.
Water softening systemSpatial-temporal-attention LSTMGrey relational analy-sisPrediction modelWater softening capacity
NETL
NSTL
万方数据
