BP神经网络模型在元素-矿物转换中的应用

Application of BP neural network model in element-mineral conversion

XRF元素录井与XRD全岩矿物分析录井在岩性识别、潜山界面卡取、储层精确评价等油气勘探过程中发挥着越来越重要的作用。XRD全岩矿物分析录井可直接获得矿物含量,但是样品制作相对复杂、周期较长,故在实际勘探钻井中XRF元素录井应用更为广泛,但是仅依靠XRF元素录井得不到单矿物的含量,不能精准得到目的层矿物的类型。利用测井中的氧闭合模型思路,将XRF元素录井得到的元素转化为氧化物,依据氧闭合模型对XRF元素录井数据进行校正。据此利用渤海某区块的XRF元素录井和XRD全岩矿物分析录井数据,根据XRF元素录井测量的元素含量计算氧化物含量,依据氧闭合模型对元素氧化物含量进行校正,并将校正后的元素数据作为输入,XRD全岩矿物分析录井测量的矿物含量作为输出,用BP神经网络模型多次训练,建立了该地区元素和矿物之间的转换关系,用于对实际数据进行分析,元素恢复矿物符合率达到80%以上。

XRF element logging(XRF for short)and XRD whole rock minerl analysis logging(XRD for short)play increasingly important roles in oil and gas exploration process such as lithology identification buried hill interface capture,accurate reservoir evaluation and so on.XRD can directly obtain mineral content,but sample preparation is relatively complex and the cycle is long,so XRF is widely used in actual exploration drilling.However,only relying on XRF can not get the content of single mineral,and can not accurately get the mineral type of target layer.Using the idea of oxygen closure model in well logging,the elements obtained from XRF are transformed into oxides,and the XRF data can be corrected according to oxygen closure model.In this paper,the data of XRF and XRD in a block of Bohai Sea are used:calculating the oxide content according to the element content measured by XRF,correcting the element oxide content according to the oxygen closure model,taking the corrected element data as input and the mineral content measured by XRD as output.The conversion relationship between elements and minerals in this area is established by training with BP neural network model for many times,which is used to analyze the actual data.The coincidence rate of elements recovering minerals is over 80%.