基于SVR的组合预测模型及其应用  

Combination predication model based on SVR and its application

作　　者：刘显德[1] 高泓[1] 

机构地区：[1]大庆石油学院计算机与信息技术学院,黑龙江大庆163318

出　　处：《计算机工程与设计》2009年第19期4506-4508,4523,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(60473051);黑龙江省自然科学基金项目(ZA2006-11);黑龙江省科技攻关基金项目(GZ07A103)

摘　　要：针对油田开发单一产量预测模型泛化能力低、中长期预测准确性差等问题,提出了一种基于支持向量回归机(SVR)的组合预测模型。该模型可基于小样本建模并能综合不同单一预测模型的适用条件和优势,具有较强的泛化能力,对只可获得少量实验数据的油田产量预测问题具有较好的适应性。给出了SVR组合预测模型的结构设计和实现算法,对油田实际产量数据进行处理,取得了较精确的预测结果,验证了模型和方法的有效性。Aimed at the problems that single-production prediction models of oilfield development have low generalization ability and poor accuracy for median or long term prediction, a combination predication model based on support vector regression （SVR） is presented. The model is built based on small samples and has powerful generalization ability, which combines the applicable conditions and advantages of different single-prediction models, and is applied to predict oilfield production perfectly that obtains just a few experimental data. The structure designing and algorithm implementing of a combination predication model based on SVR is given, and is applied to process the data of oilfield actual production, which obtained accurate predication results and verified the effectiveness of the model and method.

关 键 词：组合预测 支持向量回归机 算法 油田产量预测 核函数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]