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

Abstract

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