论文部分内容阅读
通过不同储层原油色谱指纹特征的差异,采用线性模型可准确计算2层合采的分层产能贡献,但对于3层及以上的合采井计算结果误差较大。以陆梁油田为例,在实验室进行3层原油配方实验,并进行原油全烃气相色谱分析,确定单层和配比原油指纹,建立特征指纹数据库,用BP网络算法建立分层产能模型。结果表明,BP神经网络具有极强的非线性映射能力,计算结果与实际配比误差较小,能为油田利用地球化学方法进行多层合采的单层产能配比计算提供一个经济适用的途径。
According to the difference of chromatographic fingerprints of crude oil in different reservoirs, the linear model can accurately calculate the contribution of stratified production capacity of 2-layer co-production, but the error of calculation results of co-production wells of 3 layers and above is larger. Taking Luliang Oilfield as an example, a three-layer crude oil formulation experiment was conducted in the laboratory. Total petroleum gas chromatographic analysis of crude oil was carried out. Fingerprints of single-stranded and stoichiometric crude oil were determined. Fingerprinting database was built and hierarchical productivity model was established by BP network algorithm. The results show that BP neural network possesses strong nonlinear mapping ability, and the calculated result has little error with the actual ratio, which can provide an economical and feasible way for oil field stratum to calculate single stratum production mix proportion by geochemical method .