An Energy-Efficiency Prediction Method in Crude Distillation Process Based on Long Short-Term Memory Network

The petrochemical industry is a pillar for the development of national economy affecting people’s daily living standards. Crude distillation process core and leading unit industry. Its energy consumption accounts more than 20% total whole plant, which highest link. A model based on long short-term memory network (LSTM) proposed in this paper to predict analyze efficiency. This extracts complex relationship between many variables predicts efficiency crude process. Firstly, simulation carried out. By adding random disturbance, data set different working conditions obtained, difference expressed with distance-coded heat map. Secondly, Savitzky–Golay (SG) filter used smooth data, preserves original characteristics improves prediction effect. Finally, LSTM products under conditions. MAE, MSE, MAPE results test are lower 1.3872%, 0.0307%, 0.2555%, respectively. Therefore, can be considered perfect set, have high reliability accurately