Research On Prediction Method Of Key Corrosion Control Parameters At The Top System Of Atmospheric Tower

The top system of atmospheric tower is the "leading device" of petroleum refining and chemical industry,and its operation condition is related to the production progress of the whole refining and chemical enterprises.Due to the continuous deterioration of crude oil quality in the world,the corrosion problem of the top system of atmospheric tower has become more and more serious,and the risk of unplanned shutdown caused by corrosion leakage is increasing.At present,various refining and chemical enterprises have established a perfect corrosion monitoring and testing system,and accumulated a lot of corrosion data.Through data analysis,it can master the corrosion condition of the top system of atmospheric tower and help to formulate effective anticorrosion strategy.However,the current data analysis mainly relies on manual completion,a large number of supervision and testing data is not fully utilized,and the analysis work has a certain lag,which may cause the equipment to miss the best maintenance time.Therefore,in this paper,the prediction model for key corrosion control parameters of the top system of atmospheric tower is established based on the monitoring and testing data in order to predict the corrosion development trend and improve the efficiency of corrosion diagnosis.1.In order to accurately predict the variation trend of iron ion concentration in the top system of atmospheric tower,the main corrosion mechanism of the top system of atmospheric tower is analyzed and the main corrosion influencing factors are determined.In order to improve the quality of data samples,outlier test,standardization and dimension-reduction processing are carried out on the data.On this basis,the extreme learning machine model(ELM)is used to predict the iron ion concentration,and the artificial bee colony algorithm(ABC)is used to improve the stability of the prediction results of ELM model.Compared with BP model and ELM model,the combined ABC-ELM model proposed in this paper has the highest prediction accuracy.2.In order to accurately predict the corrosion degree of pipelines in the top system of atmospheric tower,the grey prediction model(GM(1,1))is used to predict the corrosion degree of pipelines based on the wall thickness of pipelines.Aiming at the defects of GM(1,1)model itself,two methods of dynamic generation coefficient reconstruction background value formula and original sequence exponential transformation are used to improve the modeling process of GM(1,1)model,and nonlinear adaptive inertial weight particle swarm optimization(IPSO)is used to solve the global optimal dynamic generation coefficient,thus,the IPSO-GM(1,1)model is established.Compared with the GM(1,1)and the PSO-GM(1,1)model,the IPSO-GM(1,1)combined model proposed in this paper can accurately predict the variation trend of the wall thickness of the key corroded parts.3.The corrosion prediction module of refinery unit has been completed,and the online prediction of iron ion concentration and pipeline corrosion degree in the top system of atmospheric tower has been realized.