基于BP神经网络的储层渗透率预测及质量评价方法

Method for reservoir permeability prediction and qualityevaluation based on BP neural network

储层渗透率预测和评价是油气藏勘探与开发急需突破的瓶颈技术之一,BP神经网络预测储层渗透率的研究在行业中已有一定的应用,但受限于数据规模、参数调整及模型评价方法,该方法预测结果不稳定,且不能准确给出全井段储层的连续渗透率的预测质量,在油田现场并未大规模推广应用.本文针对传统BP神经网络预测储层渗透率方法中存在的问题,在对机器学习的数据处理、参数选择系统考察的基础上,定量分析了不同输入曲线、网络结构、样本大小对渗透率预测模型精度的影响,总结了BP神经网络预测渗透率模型的参数优选方案;并提出了一种基于模型森林的预测曲线质量逐点评价方法,实现了对全井段渗透率预测的质量评价.实际应用表明,本研究提出的储层渗透率预测及质量评价方法与实际岩心渗透率吻合度高,推广应用前景良好.

Reservoir permeability prediction and evaluation is one of the core technologies that need to be brokenin the exploration and development of oil and gas reservoirs,the research of BP neural network has been applied to predict permeability in the industry,but due to dataset,parameter adjustment andmodel evaluation method,the prediction result of this method is uncertain,and the predictive quality of continuous permeability of the whole well cannot be given accurately,it is seldom used in oilfield.In view of the problems existing in the traditional BP neural network method for predicting permeability,based on the systematic investigation of data processing and parameter selection of the machine learning,this paper quantitatively analyzes the influence of different input curves,network structure and sample size on the accuracy of permeability prediction model,summarizes the parameter selection scheme of BP neural network for permeability prediction model;and puts forward a pointbypoint evaluation method of prediction curve quality based on model forest,which achieves the quality evaluation of whole well permeability prediction.The practical application shows that the reservoir permeability prediction and quality evaluation method proposed in this paper is highly consistent with the core permeability,and has a good application prospect.