Research On Inversion Method Of Water Intake Profile Of Water Injection Well Based On Machine Learning And DTS Monitoring

Water injection profile interpretation based on temperature logging data is an important means to understand reservoir water absorption.Distributed optical fiber temperature sensor(DTS)can record the whole wellbore temperature profile in real time,provide transient temperature change information in oil and gas wells,and provide a more reliable new way for water injection profile interpretation.In this paper,the DTS data of water injection wells are analyzed and inversed by machine learning method to obtain the water injection profile,which provides method support for the evaluation of water injection effect.The traditional temperature logging interpretation mainly focuses on qualitative analysis.In this paper,the micro thermal effect of reservoir temperature field of injection well is fully considered,and the prediction model of single-phase liquid coupling transient temperature profile is established to study the sensitivity of temperature distribution of injection well to key parameters such as permeability and injection volume.Then,the training sample set is constructed by using the temperature profile prediction model,and the inversion model is established based on three machine learning algorithms:random forest,XGBoost and Light GBM to obtain the water absorption profile along the wellbore.Finally,the most effective inversion model of water injection profile is optimized and applied to the correlation analysis of field example wells.The results show that the temperature profile prediction model based on DTS temperature monitoring data can realize the prediction and fitting analysis of temperature profiles of vertical wells and highly deviated injection wells;According to the optimization of evaluation parameters of machine learning model,XGBoost model is the optimal inversion model of water absorption profile,which has high accuracy and good generalization ability;The XGBoost water injection profile inversion model can be applied to the DTS data inversion of field example wells to obtain the water injection profile of water injection wells.This paper provides a fast and accurate method for the inversion of water injection profile of water injection wells,and has guiding effect on subsequent measures such as profile control,pressure reduction and injection increase.