Research On Intelligent Prediction Method For Reservoir Scaling Of Strong Base ASP Flooding

ASP flooding technology is a means that may significantly increase oil recovery.The field test has shown that compared with water flooding,the strong base ASP flooding can significantly contribute to oil increment and water control,and increase the oil recovery by more than 20%.With the injection of strong base ASP flooding into the formation,it reacts with the formation fluid and rock minerals,breaking the physicochemical equilibrium state between the original fluid and rock minerals,causing changes in the composition and content of ions in the underground fluid,resulting in Scaling phenomenon leads to blockage of some pores of the reservoir,which affects the sweep efficiency and displacement efficiency,and reduces the recovery rate.At the same time,with the migration of fouling formation fluids,the screw pump rod breakage,pump leakage,and frequent pump jamming of the pumping pump often occur during the production and operation of the production well lifting equipment,which seriously threatens the normal operation of crude oil extraction.Therefore,predicting the scaling type and scaling trend of the production well is the guarantee for effective scale cleaning and prevention.At present,the application and promotion of Scaling prediction methods based on physical and chemical simulation and intelligent prediction are not applicable.The main reasons are that the prediction involves too many uncertain factors and poor regularity,and it is difficult to use traditional or manual prediction methods.The second is that part of the training process using intelligent prediction methods is complex,requires high environmental requirements,weak generalization ability,poor portability,and low accuracy of prediction results.In this thesis,the development zone in the north of Xingshugang Oilfield was taken as the test area.By analyzing the scaling mechanism and laws in the test area,we studied the intelligent diagnosis method and related technologies of solving key scientific problems and researched the intelligent scaling prediction method.All these efforts help solve the problems of low prediction accuracy,poor portability,poor dynamic updating ability and lack of time series warning of the existing method.The key research contents are shown as follows:1.Build the data mining-based scaling prediction model(SASP-DMSP)In order to overcome the shortcomings of the scale prediction knowledge base,such as poor transplantation,weak dynamic updating ability,and lack of time series prediction knowledge,a scale prediction model based on data mining(SASP-DMSP)was designed as an overall scheme to solve the problem of intelligent prediction,so as to improve the accuracy of scale prediction and the reasoning ability of the scale prediction knowledge base.The SASP-DMSP model framework includes three layers: knowledge acquisition layer,knowledge modeling layer and knowledge inference layer.The knowledge acquisition layer is the basic layer of the model,and it mainly relates to the manual acquisition of scaling prediction knowledge and the intelligent training of the knowledge base.The knowledge modeling layer is the middle layer of the model,and it provides a set of standard domain common ontology and knowledge organization system using the ontology modeling technology.The knowledge inference layer is the application layer of the model.The three layers collaborate and finally complete the scaling prediction through inference and expression.2.Study the ontology-based knowledge modeling method for scaling predictionAiming at the problems of weak semantic expression ability and relatively poor portability of the scaling prediction model,we studied the ontology-based knowledge modeling method for scaling prediction.Through the analysis and abstraction of the scaling prediction domain system and mechanism,and the expert experience and knowledge,the core ontology and standard semantics of the domain knowledge model for scaling prediction were established,to provide the standard specifications of the required data and knowledge structure for the work of scaling prediction.By using Protégé as the ontology modeling tool,and taking the scaling prediction in the test area as an example,we constructed the knowledge model and knowledge base of scaling prediction and carried out inference and diagnosis to verify the effectiveness of the proposed method.3.Study data mining-based knowledge base training method for scaling predictionIn order to fill in the loss of empirical knowledge,the historical data related to scaling prediction was trained by using the intelligent mining technology,and the new scaling prediction knowledge discovered in the learning training was added to the knowledge base to realize the dynamic update of the knowledge base.As to incomplete description of scaling prediction rules,inaccurate threshold setting,and low prediction accuracy caused by single scaling prediction model under reservoir physical property differences,we designed a combined classification pattern mining method,which mainly consists of physical property classification and pattern mining.The physical property classification adopts the block division method to achieve data segmentation of a given sample set and generate independent training sample subsets;the pattern mining derives the classification knowledge and updates the knowledge base by obtaining the sample subset training to realize the scaling prediction of the oilfield under different physical properties.In order to adapt to the fuzzy,mixed and incomplete characteristics of training data during pattern mining,a feature selection method based on variable threshold IFDS and DPSO was designed to improve the utilization rate of historical data and the accuracy of feature selection.In view of the lack of knowledge of time series trend prediction of scaling phenomenon and the periodicity and chaos of time-series historical data of the scaling prediction domain,we proposed a time series prediction method for scaling of production wells.Taking the representative six ion test data as an example,the characteristics of time series data were analyzed;a time series prediction method that can deal with chaotic time series was designed by using ESN technology to provide knowledge of time series prediction for the knowledge base and realize the early warning of scaling trend.4.Design the ASP flooding scaling prediction systemTaking the SASP-DMSP model as the theoretical guide,we designed the ASP flooding scaling prediction system.The system consists of three subsystems: ontology-based scaling prediction knowledge management system,scaling prediction data integration system,scaling prediction and scale removal and prevention management system.The collaborative operation of the subsystems realized the intelligent scaling prediction.The actual data test and analysis in the real environment shows that the application of the system improves the intelligent level of scaling prediction in the test area,increases the accuracy of scaling prediction.The results show that the intelligent prediction method for scaling of strong base ASP flooding production wells based on data mining can improve the accuracy of scaling prediction,the updating ability of scaling knowledge and the portability of the prediction system.At the same time,through extended research and analysis,this method provides solutions for dealing with such business application problems and knowledge inference problems.