A BP neural network model for estimating the drilling and completion investment is built based on the BP neural network method with 86 representative offshore oilfields in West Africa and Asia-Pacific as samples. The model uses five factors, including oil price, water depth, well number, well depth and geologic condition, as the input parameters, and outputs the drilling and completion investment parameters. Comparison of the model with a regression analysis model shows that the established model is reasonable and valuable because the BP neural network is an active learning process, able to effectively describe the non-linear relationship between variables and solve complicated problems. The established BP neural network model has high fitting accuracy and the errors of most samples are within 30%, satisfying the requirements for engineering development, and are much smaller than that of regression analysis.%以西非及亚太地区86个具有代表性的油田作为样本,以BP神经网络方法为基础,将油价、水深、井数、井深、地质条件这5个影响因素作为输入层参数,钻井完井投资额作为输出层参数,构建海上油田钻井完井投资的BP神经网络模型,并与回归分析模型进行比较分析.结果表明,由于BP神经网络方法是一个主动学习的过程,可以有效地描述变量之间的非线性关系,体现其解决复杂问题的优势,构建的模型具有合理性和实际参考价值.构建的钻井完井投资BP神经网络估算模型具有很好的拟合精度,大部分样本的预测误差在30％以内,基本满足工程开发所要求的误差精度,而且误差水平远远低于回归分析模型.
展开▼
