BP神经网络在脱气井产出剖面测井解释中的应用

Application of BP Neural Network to Degassed Well Production Profile Log Interpretation

油井开发后期,因流压降低脱气,井筒中的两相流动变为三相流动,原有的测井解释方法已不再适应。为了研究三相流产出剖面相关参数计算方法,在模拟井进行三相流产出剖面测井实验。模拟实验采用由涡轮流量计、电导探针含水率计及压差密度计组成的多参数组合仪进行。实验在三相流模拟井中获得3个原始参数:涡轮转数、持水率及混合密度。应用时间序列统计分析方法在时域中提取压差密度计响应值的6个特征量。用这6个特征量及实验获得的3个原始参数,共9个反映三相流动特性的特征量作为输入向量,运用BP人工神经网络进行训练,对总流量、液相流量及含水率进行预测,效果良好。将该方法应用于现场试验井,取得良好效果。

In the late stage of oil well development,because flowing pressure drop leads to oil well degassed and the two phase flow becomes the three-phase-flow.The original log interpretation methods were no longer avaliable.In order to research the three-phase-flow related parameters computing method,simulation experiments were made by a three-phase-flow multi-parameter tool.The tool was combined with turbine flowmeter,conductivity probe water cut meter and differential pressure densimeter.Experiments were made in the three-phase-flow test facility and obtained three original parameters:turbine revolutions,water holdup and the mixed density Statistical analysis of time series method in the time domain is used to extract six characteristic vectors of differential pressure densimeter response value.These six vectors and three original parameters which are obtained from experiments reflect the characteristics of three-phase flow.Using the nine vectors as the input vectors,we trained the BP artificial neural network to predict the total flow rate,liquid flow rate and the water cut.The method was applied to field test well,and also achieved good effect.