基于支持向量机的压裂井层优选

压裂是油田重要的增产措施，合理选择压裂井层是一项十分复杂的工作。在对影响压裂效果的各种因素综合分析的基础上，提出了基于支持向量机技术的压裂效果预测方法。采用该方法，利用油田开发动、静态数据，构建了压裂井层优选的支持向量机模型，对大庆油田采油八厂样本进行处理，符合率达89％以上，并在不同数目学习样本的情况下同模糊神经网络作了比较，性能远优于模糊神经网络，可很好地克服过学习问题。

Optimized Determination of Fractured Well Layers Based on Support Vector Machine

Fracturation is an important measure to increase oil production, and it is complicated to select a well layer to be fractured. A new method for predicting the fracturing effect with support vector machine is developed on the basis of a comprehensive analysis on various factors affecting fracturing effect, Using the method, an optimized model based on support vector machine for determinating fractured well layers is constructed by oil exploiting dynamic and static data, Samples of No.8 Oil Production Factory of Daqing Oilfield are processed, with the coincidence above 89%. At the same time, the performance comparison with fuzzy neural network under different sample scales shows that SVM is much better than FNN and the overfitting problem is overcomed perfectly.