应用支持向量机方法预测砾石充填防砂井产能

影响砾石充填防砂井产能的因素很多。关系非常复杂．常规理论方法难以建立准确、适用的预测模型。为此，对防砂井产能的主要影响因素进行分析。引入支持向量机方法，与自然产能比方法相结合。建立了防砂井产能预测模型。该模型通过有限经验数据的学习。能够导出防砂前后采油指数与其影响因素的非线性关系。分别使用支持向量机模型和BP神经网络模型对砾石充填防砂井产能进行预测对比结果表明，支持向量机模型有着更高的预测精度．在小样本的模式识别方面，有着自身独特的优势。

Application of Support Vector Machine for Productivity Prediction in Gravel Packed and Sand Controlled Wells

There exist many factors influencing the productivity in gravel packed and sand controlled wells, its correlation is complex, a precision and practical model for prediction is hard to be established by using conventional theory. Thus the major factors influencing its productivity are analyzed, the method of support vector machine is introduced and it is combined with the method of natural productivity ratio and a productivity prediction model is established for the sand controlled wells. Through learning the finite empirical data, the model can be used to derive the nonlinear correlation between the productivity index and its influencing factors both before and after sand control. Therefore the models of support vector machine and BP neural network are respectively used for predicting the productivity in the gravel packed and sand controlled wells. The result of comparison shows that the model of support vector machine has higher prediction precision, it has a special advantage in small sample pattern recognition.