作　　者： ; ; ; ; ; ; (李琪深）;

机构地区： 西南石油学院

出　　处： 《油气井测试》 1999年第2期12-16,75,共6页

摘　　要： 针对专家系统技术和人工神经网络方法的应用 ,提出了一种基于句法模式识别的试井模型识别新技术。它克服了现有技术在曲线形态识别与模型诊断推理方面的困难 ,将识别过程分五步处理 :曲线平滑与分割、特征基元抽取、曲线形态跟踪与流动期划分、解释模型初步推断和解释模型的确认 ,以层次化方式分步完成识别任务。利用该技术开发的识别程序已能正确识别数十种复杂的试井模型。在此基础上 ,可进一步开展模型参数初值的自动估计研究 ,结合最小二乘法 ,最终实现自动试井解释 。 In accordance with the application of expert system and artificial nerve network method,this paper advances anew recognition technique for well testing models based on the recognition of syntactic pattern,which overcomesthe difficulties of existing techniques in curve pattern recognition and inference of model diagnosis. The techniquetreats the recognition process into five steps: curve smoothing and cut apart, extraction of elementary cell ofcharateristics,tracing of curve patterns and division of flow periods, preliminary inference and confirmation ofinterpretation models, it finishes recognition tasks step by step by the way of sequence. The recognitionprogramme developed using the technique is able to correctly identify several dozens of complicated well testingmodels. Based on that, research on automatic estimation of initial value of model parameters may be furtherdeveloped,combined with least square method, automatic well testing interpretation will be finally realized toincrease quality of well testing interpretation.

关 键 词： 试井解释 模型识别 人工智能

领　　域： [石油与天然气工程]