Study On Optimization Of Prediction Model Of Wax Deposition Rate Of Waxy Crude Oil Pipeline

Pipeline transportation is the most important way to transport waxy crude oil in China,and the wax deposition phenomenon generated during pipeline transportation is a major problem affecting the safe and efficient operation of pipelines.Predicting the rate of wax deposition and developing a reasonable pigging cycle is an effective measure to solve this problem.But the wax deposition rate is affected by many factors and there is a complex non-linear relationship between the factors and between the factors and the rate,and the limitations of previous wax deposition rate prediction models make it difficult to achieve accurate and stable predictions.Therefore,it is important to optimize the wax deposition rate prediction model.This study is oriented towards model optimization.On the basis of combing a large amount of literature,firstly,a theoretical review of wax deposition mechanism,factors,models and experimental devices was presented.It was conducted to summarize the existing wax deposition rate prediction methods and their respective advantages and limitations,so as to establish the foundation for the subsequent analysis and judgement in model optimization.Secondly,a system containing seven factors which are influencing the wax deposition rate was established.The wax deposition rate dataset was obtained,after which the correlation between factors and rates was determined using EW-GRA.And then,the principal component features of the dataset were subsequently extracted using KPCA to eliminate redundant information between the data and improve data quality.Again,RELM was used as an artificial neural network for wax deposition rate prediction,and four strategies were used to improve SAPSO.The ISAPSO-RELM wax deposition rate prediction model was constructed using ISAPSO to optimize the input weights and implied layer node thresholds for RELM.By example,The predictive performance of the ISAPSO-RELM model was verified to be superior.Preliminary optimization of the wax deposition rate prediction model was achieved.Finally,to address the problem of large deviation in some prediction results of a single ISAPSO-RELM model,the idea of ensemble learning was adopted.The ISAPSO-RELM model was used as a weak predictor and the Adaboost algorithm was used for ensemble to further optimize the prediction accuracy and strengthen the stability of the model.The results of the study show that: The final optimised ISAPSO-RELM-Adaboost ensemble model for wax deposition rate prediction has high prediction accuracy and a smaller error range in the test.RMSE,MRE/% and Theil IC of the model are 0.04774,0.42906% and 0.00216,which can effectively reflect the development of wax deposition in crude oil pipelines.This study provides a new methods for decision making and management of crude oil pipeline,and also provides a reference for future studies on prediction model for wax deposition rate.However,there are still gaps in research on multi-model linkage prediction,which can be used as a direction for future study.