基于BP和LSTM神经网络的顺北油田5号断裂带地层孔隙压力智能预测方法

顺北油田断裂发育，地质构造复杂，储集层埋深达8 000 m，具有高温高压、窄钻井液密度窗口等特征，地层孔隙压力的预测精度难以满足工程需求。为了提高地层孔隙压力的预测精度，利用人工智能方法在处理复杂非线性问题上的优势，采用反向传播神经网络BP和长短期记忆循环神经网络LSTM这2种人工智能算法，基于顺北油田5号断裂带上3口井的声波时差、自然电位和自然伽马等11种特征数据以及经实测校正的地层孔隙压力标签数据，建立了顺北油田5号断裂带地层孔隙压力智能预测模型，BP神经网络模型的预测误差为3.927%，LSTM神经网络模型预测误差为2.864%。测试结果表明，LSTM神经网络模型具有更好的预测效果，满足现场地层孔隙压力的预测精度，为保障顺北油田5号断裂带钻井安全提供数据参考。

Intelligent method for predicting formation pore pressure in No. 5 fault zone in Shunbei oilfield based on BP and LSTM neural network

Faults are developed in the Shunbei Oilfield with complex geological structures, and the reservoirs are buried as deep as 8 000 m, which are characterized by high temperature, high pressure, and narrow drilling fluid density window, and the prediction accuracy of formation pore pressure cannot meet the engineering requirements. In order to improve the prediction accuracy of formation pore pressure, by taking advantage of artificial intelligence methods in dealing with complex nonlinear problems, two artificial intelligence algorithms, namely back-propagation neural network BP and long-short-term memory cycle neural network LSTM, were adopted. Based on the data (11 kinds of characteristic data such as acoustic time difference, spontaneous potential and natural gamma, and the formation pore pressure label data corrected by actual measurement) from 3 wells in the No. 5 fault zone in Shunbei Oilfield, an intelligent model for predicting the formation pore pressure in the No. 5 fault zone in Shunbei Oilfield was established. The prediction error of BP neural network model is 3.927%, and the prediction error of LSTM neural network model is 2.864%. The testing results show that the LSTM neural network model has a better prediction effect, and meets the prediction accuracy of the formation pore pressure on site, which can provide a data reference for ensuring the drilling safety in the No. 5 fault zone .