出砂预测新方法及应用研究

A new sanding prediction method and application

针对东辛油区出砂严重及其复杂机理,常规出砂预测方法精度差的问题,从储层特点、开发参数出发,系统分析出砂原因,在此基础上提出了基于支持向量机的出砂预测方法。通过选取不同的核函数,对已知出砂情况的井及其影响因素进行学习,建立了支持向量机出砂预测模型。结果表明,该模型采用径向基核函数具有较高的预测精度,同时避免了常规出砂预测方法只考虑静态参数和神经网络方法需要大量样本的局限性。在小样本前提下,支持向量机模型有着自身独特的优势,具有广泛的应用前景和工程价值。

Serious sanding in Dongxin oil district is with complex mechanism,to which the traditional sanding prediction method always gets poor performance.In order to solve this problem,a new vector machine supported method is proposed.The method,in which storage feature and development parameters are taken into consideration,sets up a vector machine supported prediction model by selecting and training different kernel functions for different sanding wells and influence factors.The experimental results showed that the radial kernel function resulted in high prediction accuracy and also circumvented the limitation of the traditional prediction methods,for which only the static parameters are involved and a large number of samples are required.The proposed method has special advantage in small sample pattern recognition and it has extensive application potential and engineering value.