基于生产资料特征分析的气井分类方法

Gas well classification method based on production data characteristic analysis

科学有效的气井分类有利于厘清气井生产状况、掌握其生产特征,从而有针对性地制定单井精细管理策略。为进一步指导气井管理策略的实施并提升分类工作的时效性,从气井的生产状况评价及管理策略制定出发,以大量的现场气井生产数据为基础,引入LDA算法(Linear Discriminant Analysis,线性判别分析),建立了基于生产资料分析的气井分类方法。该方法以排液能力和产液强度两个分类项目对气井类型进行描述,并以易获取的生产数据为基础,分析其特征提出了两个分类项目各自的评价指标,建立了气井分类特征指标体系;基于大量现场气井生产资料及管理经验,构建了带有先验性分类结果的气井分析样本集;引入LDA算法对分析样本数据进行挖掘和处理,实现了对分析样本原始高维数据的降维处理并获得了分析样本在新的低维子空间的分布;最终基于不同类型的分析样本在低维子空间中的分布特征确定了气井分类界限,完成了基于生产资料分析的气井多指标分类。选取分析样本外的20个评估样本实施分类以检验分类方法的性能,获得的分类结果均符合现场实践认识。该方法能够基于对现场生产资料的分析挖掘,实现即时高效的气井分类,同时为制定气井管理策略提供更具针对性的指导,为气井分类提供了一种新思路,对气井的分类研究工作具有一定的指导意义。

Scientific and effective classification of gas well is beneficial to figure out its production situations and clarify its production characteristics, so as to specifically prepare single-well fine management strategies. In order to further guide the implementation of gas well management strategies and improve the efficiency of gas well classification, this paper took the production situation evaluation and management strategy preparation of gas well as the beginning point to establish a gas well classification method based on production data analysis by introducing linear discriminant analysis algorithm(LDA), based on abundant production data of gas wells. In this method, two classification items of drainage capacity and liquid producing intensity are adopted to describe gas well types. Based on the easily accessible production data, data characteristics are analyzed, evaluation indexes of two classification items are put forward, and the characteristic index system of gas well classification is established. The gas well analysis sample set with prior classification result is constructed based on a large amount of gas well production data and management experience. LDA algorithm is introduced to mine and process the analysis sample data, so as to realize the dimension reduction processing of initial high-dimensional data of analysis samples and recognize the distribution of analysis samples in the new lowdimensional subspace. Finally, the boundary of gas well classification is determined based on the distribution characteristics of various analysis samples in the low-dimensional subspace, and multi-index gas well classification based on production data analysis is completed. Twenty evaluation samples beyond the analysis samples were selected and classified to verify the performance of the classification method, and the classification result is in line with the practical field recognition. In conclusion, based on the analysis and mining of field production data, this method can realize instant and efficient gas well classification while providing specific guidance for the preparation of gas well management strategies, providing a new way of thinking for gas well classification, and playing a certain role in guiding gas well classification.