基于遗传算法优化BP神经网络的地层破裂压力预测方法

Prediction method of formation fracture pressure based on BP neural network optimized by genetic algorithm( GA)

针对地层破裂压力现有预测方法适用性差、误差较大等问题,提出了遗传算法优化BP神经网络(GABP)预测地层破裂压力的方法。分析了地层破裂压力的影响因素;以地层深度、地层孔隙压力当量密度和岩石密度为输入变量,以地层破裂压力当量密度为输出变量,建立了GABP预测地层破裂压力模型,并利用塔里木盆地YB1井的数据进行神经网络学习和结果验证。GABP模型的预测结果误差约3.5%,精度远高于Eaton法。该方法不受地质构造条件影响,且具有精度高、计算速度快等特点。

The existing formation fracture pressure prediction methods is of poor adaptability or great error,for this reason,a formation fracture pressure prediction method based on BP neural network optimized by GA( GABP) is proposed. The influencing factors of formation fracture pressure are analyzed,and the formation fracture pressure prediction model is established taking formation depth,equivalent density of pore pressure and rock density as input variables and equivalent density of formation fracture pressure as output variable. Based on the data of well YB1 in the Tarim Basin,The result predicted using GABP is compared with that using conventional method,and the error is analyzed. The result shows that,the result predicted using GABP is more close to actual formation fracture pressure,the error is 3. 5%. The prediction accuracy of GABP is higher than that of Eaton method. The prediction result of GABP method is not influenced by geological conditions,and is of high calculation accuracy and speed.