Traceability analysis of wastewater in coal to ethylene glycol process based on dynamic simulation and deep learning

Abstract

Due to the operating condition fluctuations in coal to ethylene glycol (CtEG) plant, a stream of wastewater with high and unstable chemical oxygen demand (COD) can affect the normal running of CtEG wastewater treatment process. In this paper, a novel traceability method is therefore developed to explore causes of unstable wastewater for optimizing CtEG process, which is beneficial to promote the cleaner production of the whole CtEG. The steady-state CtEG process is simulated firstly, followed by dynamic simulation to generate datasets under six conditions. Subsequently, a convolutional neural network (CNN) is trained based on above datasets to classify the causes of high COD. The granger causality test (GCT), attention mechanism (AM), and gate recurrent unit (GRU) are integrated to accurately localize the cause from part to unit and then to calculate influence weights of variables towards COD. Based on the above research, the causal traceability network for COD in CtEG wastewater is used for process optimization with effective wastewater reduction. As a result, the application of traceability method can effectively reduce treatment costs of chemical enterprises.