基于SVM的CO_2驱油藏输油管道脆弱性评价研究

Study on vulnerability evaluation of oil pipeline for CO_2 flooding reservoir based on SVM

CO2驱油在全国范围内的广泛开展导致内外扰动对输油管道的威胁大大增加,为指导企业发现输油管道的薄弱点从而预防事故发生,提出CO2驱油藏输油管道脆弱性概念及研究思路。将脆弱性分为5个等级并确定各级脆弱性的取值范围。深入分析脆弱性要素,从致灾因子、承灾体和灾害响应3个方面建立脆弱性评价指标体系,并确定各等级脆弱性对应的指标范围。利用MATLAB R2013a的SVM回归方法,构建脆弱性评价模型并进行实例应用。结果表明:模型训练的输出与期望输出拟合较好,均方误差为9.98052×10-7;训练好的SVM模型具有较强的泛化能力和较高的准确性,其对检验样本脆弱性进行预测的最大相对误差为0.027。利用模型得到研究区域某输油管道的脆弱性值为0.381,其脆弱性程度为不太脆弱。

The extensive CO2 flooding projects in the country brings increasing threat of internal and external perturbations to oil pipeline. In order to guide enterprises to find the weak points of the pipeline and take appropriate measures to prevent accidents, the concept of oil pipeline vulnerability for CO2 flooding reservoir and the research thoughts were proposed. The vulnerability was divided to five grades, and the value range of each grade was deter- mined. By analyzing the vulnerability factors in depth, the index system of vulnerability evaluation was established from three respects including the hazard factors, hazard bearing body and hazard response, and the span of every index corresponding to each vulnerability grade was determined. By using SVM regression method in MATLAB R2013a, the vulnerability evaluation model was built, and the case application was conducted. The results showed that the output of the model and the expected output fitted well, and the mean square error was 9. 98052 × 10^-7 The trained SVM model had strong generalization ability and high accuracy, the maximum relative error between the model-evaluated value and the expected output in the confirmatory experiment was only 0. 027. By using the trained SVM model, the vulnerability of a certain oil pipeline was obtained as 0. 381, and the vulnerability level was not too vulnerable.