Prediction Method Of Tight Well Productivity Based On GA-BP Neural Network

In the process of tight oil production and development,the accurate prediction of tight well productivity can provide the basis for oil field production optimization and improve the economic benefit of oil field.Tight well productivity prediction is to achieve the purpose of productivity prediction by analyzing the internal relationship between geological parameters and productivity of oil Wells.Due to the complex formation environment in which tight oil is located,the main productivity prediction methods for tight wells can be roughly divided into three categories: analytical method,semi-analytical method and numerical simulation.They mainly analyze the flow relationships between tight wells,build models and summarize production prediction formulas.The disadvantages of the analytical and semi-analytical methods are that the model is ideal,not suitable for all types of tight wells,and the accuracy of the production prediction formula is low.The numerical simulation model is more flexible,but the parameters in the model need to be adjusted manually according to experience,so the prediction accuracy is not guaranteed and the efficiency is low.Productivity prediction using BP neural network can solve these problems.The model is suitable for all tight wells,with high prediction accuracy and no need to manually adjust parameters.However,BP neural network itself is prone to fall into local optimum,and the accuracy of prediction results is unstable.In this paper,a tight well productivity prediction method based on GA-BP neural network is proposed.This method is applicable to all tight wells without manual adjustment of parameters,which ensures the stability of the prediction results and improves the accuracy of the prediction results.The main work of this paper is as follows:(1)A tight well productivity prediction method based on GA-BP neural network was proposed.The nonlinear relationship between productivity parameters and productivity can be analyzed based on GA-BP neural network,which is suitable for tight wells in all geological environments.The real coded genetic algorithm is used to replace the gradient method to update the parameters of BP neural network,which ensures the diversity of network parameters.It can not only avoid the network falling into local optimum,but also improve the accuracy of prediction results.(2)Based on the background of tight oil field in Ordos Basin,this paper presents a tight well productivity prediction method based on GA-BP neural network.The production data of tight Wells are processed by grey correlation analysis method,and the appropriate productivity parameters are extracted as the input of the neural network,and the productivity is taken as the output.The results of tight well productivity prediction based on GA-BP neural network are taken as reference,and the results of BP neural network productivity prediction are compared with it.The results show that the former is more accurate and has better stability.