摘要
新疆焉耆盆地博湖坳陷宝浪油田开发初期气测特征明显,但经过10年的开发,随着注水量增大,气测异常发生了很大变化,主要表现在C1相对含量呈上升趋势,而重烃组分与全烃的比值呈下降趋势。在这种情况下沿用原来的气测录井解释方法已很难准确划分油气水层。为此,提出了人工神经网络油气水层解释模型,阐述了模型特征参数的优化及确定方式、模型的产生过程和权系数的确定途径,经过反复训练与试验,实际输出与期望输出最大误差小于0.01。应用表明,采用人工神经网络油气水层解释模型更有利于准确评价油气水层,其解释符合率可达到86.7%。
Baolang oilfield lies in Bohu depression of Xinjiang Yanqi Basin, during its initial stage of development, gas logging characteristics were clear, but through ten years development and water injection rate increasing, great changes have happened in the gas logging abnormalities, which mainly displayed that methane relative content was in an upward trend and the ratio between heavy hydrocarbon and total hydrocarbon was in downward trend. In this case, continuing to use the original gas logging interpretation methods were very difficult to divide oil-gas-water layer accurately, so, the paper provided an artificial nerve network oil-gas-water layer interpretation model, set forth model characteristic parameter optimization and determination, establishing process and weight coefficient determination. Through repeatedly training and testing this model, the maximum error between actual output and expected output was less than 0.01. The applied results indicated that adopting artificial nerve network oil-gas-water layer interpretation model was beneficial to evaluate oil-gas-water layer accurately and its interpretation coincidence rate is up to 86.7 per cent.
出处
《录井工程》
2006年第1期21-24,共4页
Mud Logging Engineering
关键词
气测录井
组分
变化
规律
神经网络
解释模型
gas logging, component, variation, rule, nerve network, interpretation model
